Home

CARBWARE_Manual_March_2012

Contents

1.
2. Field Name Description Units PlotID Plot ID None Area_ha Area of Plot default 0 05 ha County County lookup ID user defined None Ownership Ownership lookup ID user defined None LandUseField Land use lookup ID user defined None Soil_EF_Field Soil EF ID lookup ID user defined None Forest_Category_ID Forest category lookup ID user defined None Stump biomass tO from inventory or other Stump_C_ tO model tC plot Deadlog biomass t0 from inventory or other Deadlog C tO model tC plot Standing deadwood_t0O From AB standing deadwood tC plot Deadwood_t01 Stump_C_t0 Standing tO and Deadlog_C_tO tC plot HR_logs t01 harvest residue from AB TH and CF tC plot HR_root_t01 harvest residue from RB TH CF and mortality tC plot Standing _deadwood_t01 deadwood from mortality model t1 tC plot Deadwood_in_t01 Stump_C_t1 standing t1 and Deadlog_C_t1 tC plot Deadwood_out_t01 decomposition of cumulative deadwood t0 and t1 tC plot Net_deadwood_t01 Deadwood_in_t01 minus deadwood out_t01 tC plot Deadwood_t02 deadwood t1 t0 decomposition tC plot HR_logs t02 harvest residue from AB TH and CF tC plot HR_root_t02 harvest residue from RB TH CF and mortality tC plot Standing _deadwood_t02 deadwood from mortality model t2 tC plot Deadwood _in_t02 Stump_C_t2 standing t2 and Deadlog_C_ t2 tC plot Deadwood_out_t02 decomposition of cumulative deadwood t2 tC plot Net_deadwood_t02 Deadwood_in_t02 minus deadwood out_t02 tC plot Table 10 Descr
3. AC pw Lp M imber t L mort BB a SOD ee D e rene ay Gee Ter eee CST 13 timber A small amount approximately 4 percent of harvested timber is assumed to be left on site following harvest and this is used to estimate Ly 102 L 5 L imber The deadwood input from natural mortality Miimper is derived from allometric equations applied to the DBH and H of dead trees after mortality iterations see Appendix A amp D while Lionas and Lyrroot are known from the analysis for the litter pool in the previous section above The decomposition losses from the new input deadwood carbon pool eq 13 existing decaying logs DLog and decaying stumps DS qi are calculated using equation 15 based on decomposition factors of 0 095 for stumps and 0 076 for roots Tobin et al 2007 Ds xt Diog AP Dow a 1 5 L M imber gt DL 14 5 oF gt L nort BB gt Lurroot gt DS oia Pee See eee 1 5 The carbon stocks in the decaying existing deadwood pools DL gg and DS gq are input by the user defined in section 4 lookup table In the case of decaying logs Soil Carbon Stock Change Project_soils table In the current version of the software and the example lookup table soils are classified into three major groups mineral peat and peaty mineral soils Peat soils are organic soils with a depth greater that 30 cm and peaty mineral soils are a continuum between the peat and mineral categories Current research informatio
4. Show Report Cancel and Exit Reporting Report results are stored in CAP rogram Files X86 COFORD Carbware C P rogram Files x86 COFORD Carbware incrementDB ndb Project_1_2006_CARBWARE_RESULTS Table e The location of the report table in the IncrementDB is shown in blue text at the bottom of the screen e To view the report select the Show Report button E Report Report for Project 1 for the year 2006 Forest Soi ADJ Area ar cat lou loetey AGGan lc Loss AGNet BGGan lec Loss BGNet LmarNet OWNet Soils Net e Ggc Coa 0 1 1726709 6042841 6328574 285733 1190690 301276 889422 3753759 1343705 1018806 4682340 17168609 o 1 336934 4445023 2565289 7010312 117 6181 210821 138 7002 1346622 58 0964 9 3598 656 3326 2406 553 O 1 3 2270 227670 2440774 2660452 66762 210061 277623 1251023 050753 0000 044299 303 5764 0 2 1 82657 262542 1583470 1320928 596253 2796445 220092 5049591 1038762 546676 81063 237232 0 2 2 84234 18 6586 13 0987 55599 4848 0000 4848 91279 0 000 2 4849 17 0509 62 520 0 2 3 12635 77 0685 67 2923 105763 16 5908 2326 12 3582 32 3981 259052 0000 812378 2978719 oO 5 1 4 11828 27884 39M2 03475 0000 43475 11914 0000 24849 56122 205779 0o 2 0 000 0 000 0 000 0 000 0 000 0 000 0 000 0 000 0 000 0 000 0 000 0 000 0 3 84234 26808 343014 369822 07875 02071 0996 136669 121127 0000 121971 447229 0 6 1 0 000 0 000 0 000 0 000 0 000 0000 0 000 0
5. 6 1 3 Selecting the required Stand Modification Events e Before initiating Growth Simulation and Stand Modification routines you can select whether or not to apply all or some stand modification events using the Stand Modification controls within the Project Parameters screen Click on the Tick Box controls next to Natural Mortality Clear Fell and Thinning to activate or deactivate these stand modification events for your CARBWARE project If Natural Mortality is activated the CARBWARE model will perform annual mortality events during each annual growth simulation cycle as described in Section 3 2 The species cohort specific probability threshold values used in CARBWARE s natural mortality model can 37 be defined by the user in the P_Death_Thresholds table located in the _Carbware_Param database which is installed as part of the CARBWARE installation E P_Death Thresholds na adj_lookup Table PI spruces z Pines Larches Other_Conii Fast_Growir Slow_Growi Date Created 08 07 2010 1 Date Modified 26 08 2010 0 02 0 05 0 05 0 05 P_Death_Thresholds Table 0 0 0 0 0 Date Created 24 08 2009 1 Date Modified 24 08 2009 Paste Errors Table Record 4 1of1 gt PL bD XK No Filte Search Date Created 24 09 2008 1 Date Modified 24 09 2008 tbiProj Table Date Created 12 09 2008 1 Date Modified 07 05 2011 If Clear Fell and or Thinning
6. Bias 2 pred obs n RMSE X pred obs n p Where p is the number of parameters in the model Results Fitted model parameters 85 Table 5 shows the partial coefficients for each species and productivity class for the Chapman Richards H Age functions Table 5 Spruce cohort HI range YCeq Precision RMSE Bias ay a2 a3 gt 1 2 gt 24 1 02 5 59 2 04 1 8 4 69 0 32 1 1 2 24 1 05 7 05 2 32 1 42 4 23 0 23 0 8 1 22 0 76 5 98 1 63 1 33 3 21 0 11 0 6 0 8 20 0 66 5 51 1 33 0 66 2 55 0 56 0 5 0 6 18 0 57 5 26 1 12 0 89 1 69 0 45 0 4 0 5 16 0 53 5 35 1 47 1 11 3 66 0 32 0 3 0 4 14 0 48 5 32 0 54 0 74 3 54 0 62 0 2 0 3 12 0 44 6 59 2 20 1 53 4 53 0 24 0 1 0 2 10 0 35 6 93 2 27 0 69 1 77 0 43 lt 0 1 lt 10 0 28 8 02 0 35 1 9 4 23 0 7 86 Appendix 1D CARBWARE stand modification functions The NFI permanent plots structure is modified at each growth cycle iteration to simulate the losses associated with natural mortality and harvest This section discusses the development of the CARBWARE modification functions from draft versions for submission to International peer reviewed Scientific Journals I Mortality models Introduction In the general context of forest growth models and at the most basic level the tree mortality module s role at each iteration is to classify a particular tree in the dataset as being either dead or alive This paper approaches this problem in
7. Carbware Database in use C Documents and Settings Projects Carbon E Project Co2 Allocation Database About Carbware Select Database Co2 Allocation Co2 Reporting Choose Project Close Program e You will see the following Pre Processing control screen Pre Processor Database in use C Documents and Settings Projects Carbon_Project Carbwini2005 mdb x Database Get Plots with Small Trees i z List Small Trees Fill Carbware table Populate Small Trees Trees table Info IndividualTrees table Info IDPlots___ ID______ TreeNumbei IDPlots 300 ID____ ProcessCode Species Height_m Diameter MM Age Origin _ Close Pre Pro dow Note The database in use is always referenced at the top of this screen 29 If you do not wish to proceed with pre processing your inventory database click on the Close Pre Process Window button and you will return to the Main Menu screen To proceed with pre processing your inventory database click on the Get Plots with Small Trees button A drop down listing of all inventory plots with small trees will be shown Pre Processor Database in use C Documents and Settings Projects Carbon_Project Carbwini2005 mdb x Database Get Plots with Small Trees e a List Small Trees SL Fill Carbware table Populate Small Trees Trees table Info a IndividualTrees table Info m Close Pre Process Window Note Refer to Section 3 2 for a detailed description of the CARBWARE pr
8. Introduction Height growth for small trees is a driving developmental force as trees compete for light and vertical growing space Because of this the small tree portion of CARBWARE is a height growth driven model height growth is estimated first and then diameter growth is predicted from height growth see section B of this appendix E3 Equations used to predict small tree height increment vary by species variant silvicultural practice and site type Most single tree based models for young growth generally use the same the same predictors as described for DBH increment models However the NFI data set provides little or no information on predictors for young tree height The development of a H growth model for trees less than 1 3 m to a maximum H of 2 3 to 5 1 m i e the diameter at breast H DBH is described here The model uses a empirical Chapman Richards approach for different species with sub models for different height index ratios i e mean H over age as proxy s for young stand productivity and site factors Methodology Modelling framework The model uses a empirical Chapman Richards approach for different species with sub models for different height index ratios i e mean H over age as proxy s for young stand productivity and site factors ntl a xHinc 1 exp lx PE T a a a a raa a e E Ee 1 en al a where xH is mean height of all trees in the NFI plot for the jth species and jth H index rati
9. and Sterba 1997 who log transformed the response where we log transform the expected value of Dinc and didn t model autocorrelation Where fitting was unsatisfactory i e because of parameter instability or data sparseness a submodel was selected A criteria of model selection was that the parameters should be qualitatively similar to those estimated by Monserud and Sterba 1997 In this respect the fitting of the increment models is better described as model calibration than model selection The parameters for the fitted models were FGB E Dinc EXP 2 8528 LN DBH 1 1729 0 00012 DBH LN CR 0 8241 0 000015 CCF Larch E Dinc EXP 2 2969 LN DBH 0 6338 0 00096 CCF oc E Dinc EXP 1 4191 LN DBH 0 554 0 00025 DBH LN CR 0 5549 0 00052 CCF 0 00646 BAL Pine E Dinc EXP 1 3466 LN DBH 0 741 0 001 DBH LN CR 0 998 0 00066 CCF 0 00417 BAL SGB 80 E Dinc EXP 2 5897 LN DBH 0 7534 0 00068 DBH 0 0006 CCF 0 00979 BAL Spruce E Dinc EXP 1 8628 LN DBH 0 9456 0 0005 DBH LN CR 1 1639 0 000638 CCF 0 00273 BAL Uncertainty In this section we look at various measures of the performance for the different models discussed above The performance measures quoted give rough ideas about how the models perform It should be noted that performance can be improved somewhat by including plot and site effects but since these are problematic for extrapolation from PSP
10. classification codes codes user defined EF_soil Emission factor for soils M Lookup Numeric negative value represents an uptake of C positive values represent an emission e g peat soils Values in t C ha Equil_Time Time taken yeas for soilto O Lookup Numeric reach equilibrium 19 Data Description Requirement Table Units Format Net_Soil_DCtn_1 Soil emission factor at the M Lookup EF_soil at plot area plot level e g 0 034 tC for 0 05 ha positive values represent emissions negative values are removals MineralNO2_EF Mineral soil N20 emission O Lookup Numeric in tCO2 following drainage eq expressed on the size of the plot Organic_poor_NO2_ Poor nutrient poor organic O Lookup Numeric in tCO2 eq EF soil N20 emission following expressed on the drainage e g peats size of the plot Organic_rich_ NO2_ Nutrient rich organic soil Lookup Numeric in tCO2 eq EF N20 emission following expressed on the drainage size of the plot NO2T Total soil N20 emission O Lookup Numeric in tCO2 eq following drainage expressed on the size of the plot PlotID ID of plot must be the same M tblEvent Numeric as those used in trees individual trees afor_criteria and lookup tables Year Year thinning or clearfell is M tblEvent Numeric e g 2007 planned Event Thinning or cleafell M tblEvent Numeric code 100 clearfell 200 thinning Reduction Amount of basal area perha M tblEvent Num
11. gains due to litterfall L as given by equation 7 harvest residue litter input Lug in equation 6 mortality litter inputs Mj and losses associated with decomposition of the litter pool Laecomp BO Err he tM Lapa oe a 9 where Mvis the input to the litter pool from natural mortality i e all aboveground dead material with a diameter less than 7 cm This is derived from the Linorjag Minus the timber fraction of the new dead pool L mort tim Ma a ie Cronin ee ed ee ee a The decomposition losses of the new input litter Luecomp and existing litter pool Loa are calculated using decomposition factors of 0 14 taken from national research Saiz et al 2007 Black et al 2009b D xt Pins CSI eee ey gan Oe PIS OTTAA EEIE AEE 11 D xt 1 Hes SO Ga Bie Me ese dence tate ttaate cman tate ee 12 where Lini is the initial litter pool estimate input by the user in the lookup table in the Carbw mdb The remaining litter from the newly input litter harvest residue and mortality pools from the previous years n 1 n 2 etc were accumulated following decomposition Deadwood Carbon Stock Change Project_deadwood table Net deadwood stock changes ACpw were derived from carbon inputs associated with timber extraction residue L timber from mortality Miimber dead roots from mortality Lone roots from harvest Lyproos and carbon loss due to decomposition of the new and previously existing deadwood pool Dow
12. 0 13 5 1 14 5723300409226 6 98310561081454 15 825436650979 7 93467381694266 5 2 470 15 0 10 4 1 11 2289807437818 6 15551928365813 12 2257140172344 7 00905111672825 54 5 470 15 0 13 3 1 14 5331735830111 7 0182693504201 15 7730228875948 7 9635028687432 gj 54 8 470 15 0 14 8 1 15 9900709719204 7 11257639982557 17 3829087149634 8 22749988651247 5 1 470 15 0 7 5 1 8 3060604389705 5 14678188654834 9 08355204624202 5 86787731169718 54 9 470 15 0 10 3 1 11 4085488056554 6 11148357168705 12 4411698667485 6 95341207185485 5 6 470 15 0 14 3 1 15 3395499671895 7 12324447604233 16 6533230335891 8 13715540662636 54 900101 470 15 0 5 2 1 5 71640869517508 4 15412735139133 6 27054605181266 4 77980501015061 v Search Record n 10f2787 gt M 7 2 2 Carbware Tables for each Annual Growth Cycle e As noted in Section 5 3 the Carbware Table forms the basis of all growth simulation stand modification and carbon allocation routines to be performed on the inventory database While the original Carbware Table associated with a CARBWARE project remains in the selected inventory database an adapted copy of it is created in the IncrementsDB mdb database file for use in annual growth cycle simulations This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_AnnualGrowthCycleNumber In this example the project name is Project_1 and so the first project specific CARBWARE Table is named Project _1_00 e P
13. 000 0 000 0 000 0000 0 000 0 6 2 0000 0000 0000 0 000 0000 0000 0 000 0000 0 000 0000 0000 0 000 0 6 3 84234 25 0451 1048516 798064 73572 205M2 13144 346946 65 3165 0 000 70807 258893 O 7 1 547519 3460993 3205746 255247 731224 24926 481964 1759433 888128 32303 3061737 11226367 6 2 84234 2 6505 0264 2816 05281 0 000 05261 02271 0000 41415 72571 26 6093 0 3 16 8467 145486 3623265 347 7779 0 3325 62623 829555 135 1142 245 2186 0 000 50 4006 1848023 0 8 1 42117 47680 265939 218252 1 4009 0000 1 4003 19 0862 00003 20192 36576 13 4113 0 8 2 84234 13044 802005 78891 0 3832 1993 16098 592674 65666 2899 17581 4 4637L sea aeatsoresenton 61 e The results table can be exported to Excel by selecting the green Copy results to Clipboard button e Apop up window confirming the export is shown click OK BS yy muas slein Copy Report Copy to Clipboard complete E e Open excel and a new worksheet and paste the copies cells Save in you project record with a relevant file name e Explanation of field headers and results are shown in Table x below Field Name Description Unit Category All categories are aggregated none Forest category as specified in Lookup table see Table1 none Forest Cat section 4 Soil category as specified in Lookup table under Soil_EF none category see Table1 section 4 Note that these can be user Soil Cat defi
14. C LT_C representative LT biomass LT Exp factor kgC Volume representative tree volume m3 Table 5 Description of Clearfell Events Table Fields in the IncrementsDB mdb database 8 2 3 Thinning Events Table The Thinning Events Table within the IncrementsDB mdb database file lists all records trees within your inventory database plots that have been thinned during annual growth cycles This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_Thinning_ Table In this example the project name is Project_1 and so the project specific Thinning Events Table is named Project_1_ Thinning Table This events table lists all of the trees within your inventory database that have been thinned during any of your CARBWARE project s annual growth cycles In this example there have been three annual growth cycles for Project_1 2005 to 2006 2006 to 2007 and 2007 to 2008 noted by the fact that there are four Carbware Tables Project_1_00 Project_1_01 Project_1_02 and Project_1_03 The year of each thinning event is noted in the ModificationYear field in this table The EventCode field notes the stand modification event type that has modified a record tree This will always be 200 Thinning In this example there has been a total of 23 trees thinned over the three annual growth cycles for Project_1 All of the events were in 2008 and only 1 Plot was modified 46 Projec
15. Height M Individual Numeric in m to mandatory for small trees Trees Trees one decimal point random sample for big trees in Trees table were DBH is measured CrownBase_m Height at base of living O M Trees Numeric in m to 1 crown when available decimal point recommended but not mandatory DeadCrBase_m Height at base of dead O M Trees Numeric in m to 1 crown decimal point Tree_Length_m Length of tree in m for O M Trees Numeric in m to 1 slanted trees mandatory if decimal point crown_base_m is reported Numeric Crown_Length_m Length of crown in m i e O M Trees Numeric in m to 1 Tree length Crown_base decimal point mandatory if Crown_base is Numeric reported NewOrMissingTree Missing or new tree ID O Trees Specific NFI data not currently used DBH1_mm Alternative DBH O Trees Specific NFI data measurement not currently used Species Species code M Individual Numeric user Trees Tree defined lookup codes see species look up and cohorts DBH2_mm Alternative DBH O Trees Specific NFI data measurement not currently used StandLayer Stand layer code O Trees Specific NFI data not currently used Age Tree age in years M Individual Numeric no Trees Trees decimal point 14 Data Description Requirement Table Units Format BreastHeight_cm Height at which DBH is measured in cm O Trees NUMERIC NO DECIMAL PLACE Diameter03h_mm Diameter at 1 3 of stem length Tr
16. Ln CCF a H a bDBH DBH SGB 1 477 0 005 0 017 0 578 ICR a a BAL a H a ee BAL BAL is the sum of the basal area of all individual trees larger than the subject tree m per ha CCF is the crown competition factor which is a measure of the crown areas of the subject tree relative to a open grown tree that would not be subjected to crown competition taken from Hassenhaur see section B of this appendix DBH is the diameter at breast height cm H is height m form actual or predicted H estimates Table 1 72 Fitted and Actual Height rm 10 20 30 40 Actual Height m Figure 2 Fitted and actual height plotted all cohorts model 2 against actual height Residual rm 10 20 20 40 Predicted height rm Figure 3 Raw residuals from the fitted model plotted against the fitted height value 73 External validation Based on the data presented above model 2 was selected for validation against external data sets In this section we compare model predictions against data from PSP non research plots 30 25 20 Fitted Data Height rm a 10 10 20 30 40 50 DBHicm Figure 4 Estimated and observed validation heights versus DBH Generalised and plot specific models In this section we discuss the implications of using a generalised DBH H model i e one whose parameters are fitted to the entire dataset with a plot specific model i
17. Ownership lookup ID user defined None LandUseField Land use lookup ID user defined None Soil_EF_Field Soil EF ID lookup ID user defined None Forest_Category_ID Forest catgory lookup ID user defined None AB_deadtree_C t01 Dead above ground biomass AB time t1 tC plot AB_Living C_t01 Living AB biomass t1 tC plot Mortality AB t01 AB mortality t1 tC plot Thin_AB t01 AB thinnings t 1 tC plot CF_AB t01 AB clearfells t1 tC plot AB_C t01 AB form t1 tC plot Litterfall_C_t01 Litterfall t1 tC plot SB_from_thin_C_t01 Stem biomass SB from thinning t1 tC plot SB_from_cf_C_t01 Stem biomass SB from clearfell t1 tC plot HWP_t0O1 Timber biomass to harvested wood product HWP pools tC plot t1 Note Harvested timber is assumed to be immediate oxidised following harvest but timber is allocated to the HWP table is further analysis is required HR_t01 Timber harvest residue let of site default assumed a tC plot constant value of 4 of timber harvested AB_C_t01 minus AB living plus mortality Thin and cleafell AB_C_gain_t01 t01 negative value represents a gain tC plot AB_C_loss_tn01 sum of AB mortality litter SB from CF and Thin t01 tC plot Net_AB_t01 sum AB_C_gain and AB_C_loss tC plot AB_deadtree_C_t02 dead tree AB from mortality model tC plot AB_Living_C_t02 Living AB biomass t2 tC plot Mortality _AB_t02 AB mortality t2 tC plot Thin_AB_t02 AB thinnings t 2 tC plot CF_AB_t02 AB clearfells t2 tC plot AB_C_t02 AB form t2 tC plot Litte
18. Pine H a a BAL a BA 1 exp b BAL 16 905 0 083 0 0803 0 042 Larch H a a BAL a BA 1 exp b BAL 32 59 0 1052 0 1229 0 023 Conifers H a a BAL a BA 1 exp b DBH 23 226 0 1381 0 0703 0 027 1 1021 FGB H a a BAL a BA 1 exp b DBH 14 661 0 1167 0 0187 0 076 SGB H a a BAL 1 exp b DBH 29 677 0 1034 0 044 0 7813 BAL is the sum of the basal area of all individual trees larger than the subject tree m per ha BA is the basal area of all trees in the plot normalised to a ha DBH is the diameter at breast height cm 71 Table 2 CR models used in CARBWARE for the 6 different cohorts If dependent variables had no significant influence on the H model prediction these variables were excluded from the model The CR model takes the form of _ exp CR 1 exp CR where CR is derived from the non linear equations which may vary for different cohorts Cohort Model CR variations al a2 a3 a4 a5 b c Spruce H _ 4 8705 0 017 0 397 0 119 0 296 0 0003 2 ICR a a BAL a Ln CCF a H a HA bDBH Pine ICR a a BAL a Ln CCF a H bDBH 3 8478 0 024 0 213 0 137 0 0002 2 Larch ICR a a BAL a Ln CCF a H 5 8306 0 018 0 794 0 039 Conifers ICR a a BAL a Ln CCF a H bDBH 4 1759 0 019 0 394 0 965 0 0004 2 FGB H 2 4539 0 009 0 145 0 045 0 591 0 0001 2 ICR a a BAL a
19. Section 5 2 click on the Fill Carbware table button in the Pre Processing control screen Note The Fill Carbware table button will remain deactivated until all Small Trees have been populated into the Trees Table using the pre processing routine outlined in Section 5 2 Pre Processor Database in use C Documents and Settings Projects Carbon_Project Carbwini2005 mdb E3 Database Get Plots with Small Trees z f y fi i j i List Small Trees Fill Carbware table Populate Small Trees PAEA eae A IL IEEE j Trees table Info IndividualTrees table Info Close Pre Process Window The final stage of the pre processing routine will begin and a status progress bar will appear The time taken to perform this pre processing will depend on the size of your inventory database When the pre processing routine is complete you will see the message Pre Processing completed Click on the Close Pre Process Window button and you will return to the Main Menu screen Pre Processor Database in use Documents and Setting Get Pots rath Smal Trees rj ee a E cc Close Pre Process Window Close Pre Process Window 32 Note Once complete results of this pre processing routine can be seen in the populated Carbware Table within your inventory database The Carbware Table forms the basis of all growth simulation stand modification and carbon allocation routines to be performed on the inventory database It
20. The field properties can be viewed from Access using the design view option See the example Carbwini 2005 database on instillation disc for reference 25 4 2 Stand modification options Stand modifications include mortality and harvest The extent of these modifications can be set in the P_death threshold table mortality or tblEvent thinning and clearfell see 4 1 1 4 Alternatively the project can be run by not selecting any one of the modification options see section 8 This is useful if one needs to establish the maximum increment or basal area if no mortality or harvests occur 4 2 1 Mortality probability threshold options Mortality is modelled based on a binary indicator of mortality arbitrarily 1 tree dead at time of DBH measurement 0 tree alive The default probability threshold p of a tree being dead was derived from analysis shown in appendix 1D However the user can change the p value for the 6 different cohorts in the P_Death_thresholds table in _Carbware parameter database refer to appendix 1D for guidance It is recommended that the minimum and maximum range 0 no mortality 0 2 or max mortality should not be exceeded 5 Pre processing of Forest Inventory Data dS Please refer to the example database provided on the Installation CD Carbwini2005 db for all software steps Remember to always save a copy of your original inventory database file e g Carbwini2005_Raw_Data db before pre processing is
21. are activated the CARBWARE model will perform clearfell events and or thinning events after each annual growth simulation cycle as described in Section 3 2 Activated clearfell and thinning events are determined by the tblEvent table located in the selected inventory database All plots PlotID with a prescribed harvest event clearfell and or thinning must be listed in this table along with the Year of the event the Event code 100 Clearfell 200 Thinning and the basal area Reduction factor refer to Section 4 1 1 4 for details relating to the tblEvent table Tables Afor_Criteria Ej tblevent x P pPlotD Year Event Reduction amp 117 2007 100 70 234097091 andiidustirces SE 189 2008 200 18 922626704 Lookup 225 2016 100 75 85305089 Carbware 253 2014 100 41 574880843 378 2012 100 28 067530235 510 2018 100 71 117915697 834 2009 100 58 470003699 al 992 2012 100 47 289908646 Record 4 s 10of19 gt mH W NoFilter Search Trees H TreesDeleted900x 6 1 4 Saving a New Carbware Project e Once you have set all of the required parameters for your new CARBWARE project you must save the set parameters by clicking the Save Settings button on the Project Parameters screen e You can now begin annual Growth Simulation and Stand Modification routines see Section 7 or return to the Main Menu screen by clicking the Close button on the Project Parameters screen 6 2 Selecting a Saved Carbwar
22. based on the Yang Weibull funtion cf Table 1 Green red and black are data estimates and single standard error of prediction respectively Competition denotes a generalised model with competition covariates cf Model 2 Table 2 Random denotes a plot specific random asymptote cf Model 3 Table 2 Generalised denotes parameters are estimated from the entire dataset The smallest standard error of prediction is associated with the Plot specific model followed by the Comp Random model Average s e p for these models are 1 39 and 1 25 respectively Thinning effects All observations in the dataset were categorised by us as thinned or non thinned depending on the general management regime for the plot We estimated the following model to test for a residual thinning effect having conditioned on other effects H U a a BAL a DENS a BA a I Thinned exp b DBH where I Thinned is an indicator function valued 1 if the plot was thinned and 0 otherwise The BIC of this model was 45037 and the Wald test for the a parameter p 0 08 indicated that the thinning effect was not statistically significant at the 5 level The a estimate was greater than zero perhaps reflecting the longer tail in the height distribution for trees in thinned plots Figure 7 75 oo a Height irri 20 Figure 7 Strip plot of Heights in the calibration dataset 1 Thinned 0 Unthinned Discussion We have sh
23. database 47 8 2 3 Mortality Events Table The Mortality Events Table within the IncrementsDB mdb database file lists all records trees within your inventory database plots that have died as a result of the CARBWARE natural mortality function during annual growth cycles This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_Mortality_Table In this example the project name is Project_1 and so the project specific Mortality Events Table is named Project_1_ Mortality_Table This events table lists all of the trees within your inventory database that have died during any of your CARBWARE project s annual growth cycles In this example there have been three annual growth cycles for Project_1 2005 to 2006 2006 to 2007 and 2007 to 2008 noted by the fact that there are four Carbware Tables Project_1_00 Project_1_01 Project_1_02 and Project_1_03 The year of each mortality event is noted in the ModificationYear field in this table The EventCode field notes the stand modification event type that has modified a record tree This will always be 300 Natural Mortality In this example a total of 441 trees have died over the three annual growth cycles for Project_1 Mortality events were in 2006 2007 and 2008 Tables Project_1_00 H Date Created 15 09 2011 12 33 19 Date Modified 15 09 2011 12 33 19 Project 101 Date Created 15 09 2011 12 33 21 Date Modified 15 09
24. e one whose parameters are estimated for each plot separately We compare a mixed effects model and a plot specific model The former is plot specific by the inclusion of a random residual plot effect In what follows by mixed model we mean the random asymptote model Table 2 Model 3 To get an idea of the difference between plot specific and mixed model results we extract a plot from the dataset that exhibits a wide range of DBH and H values and then compare the models for that plot This makes sense because the context of the comparison is how well a given model will perform for a given plot primarily In particular we will compare the standard error of prediction for a new tree height for both models In the case of the mixed model this standard error of prediction is derived as conditional on the estimated random plot effect A plot specific Yang Weibull model gives a smaller standard error of prediction than the same model estimated from the entire dataset because residual variability for any given model will always increase from a subset of the data plot specific to the entire dataset generalised In other words the generalised model predictions are less precise than the plot specific predictions for any given plot and the model mean estimate tends towards the overall mean and away from the plot specific mean 74 Random effect Height rn OBH tenn Figure 5 Model predictions for a single plot with various models all
25. functions Height Diameter And Crown Ratio Modelling For Six Species Cohorts It is common among forestry datasets that tree height H or crown ratio CR is not measured on every tree This creates interest in estimating the height of such trees A common forest inventory approach used to derive missing H and CR values involves the use of single parameter DBH models based on species and plot specific predictions NFI 2007 Wykoff et al 1982 However is has been suggested that these Chapman Richards functions or derivations there of are problematic because the function approaches the asymptote too rapidly particularly when there is a weak relationship between DBH and H in larger trees In addition individual plot DBH H data is sometimes too sparse to parameterise plot specific functions Generalised DBH H functions avoid the need to parameterise relationship for every stand Since the relationship between DBH and H is influenced by the relative competitive position of trees within a stand and management interventions site level stand density information is often incorporated Temesgen and Gadow 2004 Taking their results as a starting point we address here several issues that arise in the context of our modelling dataset These include the application of nonlinear mixed effects models which successfully borrow strength across all permanent plots thereby facilitating imputation in plots where data is sparse or unevenly distributed The perm
26. needle biomass LB and the foliage turn over rates F Tobin et al 2006 IRENE E22 S ER ERE AAE A A E A EER 7 101 Allometric equations and coefficients used for the calculation of LB for different species cohorts with either AG or DBH as dependent variables are shown in Appendix A The F rate was assumed to be 5 years i e F 0 2 for conifer crops and 1 year for broadleaf crops Tobin et al 2006 The mortality of trees is based on nationally derived single tree mortality models Appendix D The above ground biomass loss from mortality Lmorag was Calculated using DBH and H as dependent variables in biomass algorithms Appendix A Biomass carbon losses from the below ground biomass pool AC gq were calculated as the sum of losses due to death of roots after harvest Lugroot Natural mortality of roots Linonjes and root death following fire Lire IN OPEN OE LENE RE EA RA RUT ONT te 8 Lurroot is the root biomass transferred to the deadwood pool following harvest All roots are assumed to die and decompose following harvest The mortality of roots is assumed to follow that for trees as estimated from nationally derived single tree mortality models Appendix D The below ground biomass loss from mortality Lmores was Calculated using above ground and total biomass algorithms Appendix A Litter Carbon Stock Change Project_Litter tables Net litter stock change AC was calculated based on litter inputs
27. of the plot Table 1 Model 3 The competition covariates are plot basal area BA m ha basal area in larger trees BAL m ha which is the integral of the empirical frequency distribution of the BA variable from the subject tree to the 69 largest diameter tree in the plot and plot density DENS trees ha Models were fitted in NLMixed procedure in SAS using the Trust Region algorithm Grids were specified as starting values for parameters where sensible Height m 30 40 20 10 0 20 40 60 80 DBH cm Figure 1 Model 2 Height estimates red and actual heights black The estimates presented here depict a cloud because they are conditioned on covariates that vary between trees BAL and plots Density Basal Area and over time BAL Density BA 70 Number Model 2l BIC 1 H a 1 exp b DBH 65185 65223 2 H a a BAL a BA 1 exp b DBH 58341 58417 3 44980 H Ui a a BAL a DENS a BA 1 exp b DBH 1 28 45034 Table 1 Likelihood statistics for different forms of the DBH H model Model 2 is the model used in CARBWARE for the 6 different cohorts If dependent variables had no significant influence on the H model prediction these variables were excluded from the model Cohort Model 2 variation al a2 a3 b c1 c2 Spruce H a a BAL a BA 1 exp b DBH 84 33 69 0 274 0 1603 0 024 0 8846 0 0064
28. of tree per ha form DBH and plot area Density Calculated Tree per ha RepreBA Calculated BAha Expansion factor BAL Sum of all RepreBA in a plot smaller than object tree BAL_ties_adjusted3 Adjusted BAL to account identical DBH values HT Measured H from Trees and individual trees tables CalcHeight_m NFI generalised DBH H estimate from Trees table Estim_Ht Calculated H using individual tree models Logit_CR Log of Crown ratio CR_Invent Measured crown ration 999 is null CR_Calc Calculated crown ratio OGCD Calculated max radius threshold for open grown tree OGCA Calculated max diameter threshold for open grown tree EP_OGCA Adjusted competition factor for partial sample CCF Calculated crown competition factor InCCF Ln CFF DBH_Increment Calculated DBH increment HT_Increment Calculated H increment for larger trees Cohort_Code Species COHORT ID see ID field in look up file on instillation disc file name model_cohort_lookup xls From Mean_H Mean H of small trees 33 H_Index H index H meanH small trees plotba_m2_ha Plot basal area Pred_H_inc Predicted H increment for small trees Adj_Ht Adjusted value for small tree H increment Adj_Ht_Err Erorr note 0 no error DeadTree Dead tree flag 100 alive 200 recently dead 300 dead logit_pmort Mortality algorithm pmort Probability of tree dying Weight Weight factor u
29. related to a tree which may not have been measured in a given inventory cycle because it does not meet the diameter threshold in the particle concentric plot If the tree were to grow and the diameter now meets the required threshold then the tree should be measured and include in calculations However CARBWARE does not facilitate the detection of the tree due to in growth The errors associated with omission due to in growth are yet to be quantified so a complete sample of trees in a plot is recommended 4 1 1 2 The IndividualTrees table As mentioned previously this table contains information on trees with a DBH of less than 7cm This is required because additional information is captured in this case In this table all entered trees must have a measured tree height and DBH if present These tables are also used to create virtual plots from establishment The IndividualTrees table should contain the following headers with specified information also see Table 1 a IDPlots This must correspond with the IDPlot number in the Trees table b ID small tree ID starting with number 1 each tree for a given species should be consecutively numbered If there are mixed species in a plot then should be c Edit_date the date and time the data was entered or modified e g 16 07 2005 12 34 56 d Edit_user user who input or changes the data SYSTEM should be used as a default e Species These are codes specifying the species It is importan
30. the context of an individual tree model of mortality that is both age and distance independent The specific modelling framework within which the mortality module will be applied is a framework similar to the PrognAus framework with the goal of estimating annual forest dynamics for Ireland Literature review There are two areas of interest for the literature context of this paper tree mortality modelling and threshold based classification Note that this paper is not focussed on a survival analysis as one might perhaps expect because such models are time dependent 1 Mortality modelling in Forest Succession Wunder et al 2006a compared the use of classical stress thresholds in mortality modules of forest succession gap models They conclude that logistica1 regression based models are superior to stress threshold models with regard to predicting time of tree death Baesens et al 2003 review threshold based classifiers in the context of credit scoring They examine logistic regression discriminant analysis k nearest neighbour neural networks and decision trees advanced kernel based classification algorithms such as support vector machines and least squares support vector machines LS SVM They assess performance using the classification accuracy and the area under the receiver operating characteristic curve They found that both the LS SVM and neural network classifiers yield a very good performance but also simple classifiers
31. the minimum DBH ranges suitable for the DBH increment model The transition from the H to DBH increment model is based on the threshold H value which corresponds to the minimum allowable DBH value to be used in the DBH increment model Table 4 If a tree has a larger corresponding DBH than the threshold value it is grown using the DBH increment model Table 4 Threshold minimum DBH values suitable for use din DBH increment model and corresponding cut off H values used for H growth in small trees Cohort DBH threshold cm Corresponding H m Spruce 4 2 7 Pine 4 5 1 Larch 2 3 6 Other conifers 4 3 1 Slow growing Broadleaves SGB 2 4 2 Fast growing Broadleaves FGB 2 3 2 Datasets and measure of goodness of fit We used both the Coillte PSP and NFI individual tree data base to develop H age curves range 0 1 to 12 m Data operations were concerned with assembling datasets of the variables used in the H model insofar as was feasible We looked at the performance of the various models H Age for different cohort for the combined datasets Some measures we could have used that are used by Thurig et al 2005 for example are accuracy precision and excess error calculated as follows Accuracy 2 predicted observed n 100 m Where m is E obs and nis the number of observations Precision SD pred obs Empirical Excess error measures could not be performed because there was no external validation data set Thirig et al 2005
32. the threshold exceeds the modelled probability Also they derive a total correct classification accuracy of 86 Materials and Methods We fitted logistic regression models to the growth dataset We investigated model performance in the case of separate models for each cohort Principal issue here was the lack of data for some cohorts The response variable was a binary indicator of mortality arbitrarily 1 tree dead at time of DBH measurement 0 tree alive We only included trees whose cause of death was natural mortality e g such causes as windblown diseased were excluded Explanatory variables were as such that were selected by Monserud and Sterba 1999 DBH and transformations thereof CR BAL CCF but we also investigated relative growth indicators that Bigler and Bugmann 2004 noted as being useful correlates Site and plot effects were modelled as random and consecutive observations on the same tree were modelled as being correlated Conditional on this correlation structure the fixed effects parameters were selected by backward selection starting with the candidate set of covariates just listed Models were fitted by maximum likelihood and individual fixed effects were identified as non significant on the basis of asymptotic Wald tests This was done for each cohort separately Performance of candidate models was then evaluated by cross validation and external validation comparing fitted to observed mortality in NFI dataset and w
33. to comply with these formats will result in error warnings which are logged in a text error files located in the destination folder where the software is installed 12 Table 1 The required information filed names properties units required and field order in tables located in the Carbwini 2005 mdb database Data Description Requirement Table Units Format IDPlots Plot ID number M Individual Numeric e g 1 to Trees and 100000 does not Trees have to be continuous or in any order xX_m X co ordinate of tree in m O Trees Numeric positive from local origin point or negative Y_m Y co ordinate of tree in m O Trees Numeric positive from local origin point or negative IDSmallTrees Small tree ID M Individual Numeric always 1 Trees ID Tree ID M Individual Numeric In Trees Trees Trees table small individual trees should be listed as 900X where X is the Small tree ID from Individual tree table and corresponding plot ProcessCode Processing notification key O Individual QC of post data Trees capture processing Trees functions in Irish NFI FieldStatus Processing notification key O Individual QA QC of post data Trees capture processing Trees functions in Irish NFI Edit_date Date captured data was M Individual QA QC of post data edited Trees capture processing Trees functions in Irish NFI Edit_user User who performed edit M Individual QA QC of post data Trees capture processing Trees f
34. variability is partly a function of data resources i e the number of cases and the size of the validation 83 sample as a proportion of the number of cases The low variability of Pine and Spruce the cohorts with by far the most number of cases reflects this In Figure 8 the better performance of PSP versus NFI is partly a result of including such blocking effects as site and plot This idea is also illustrated with more detail in the document on DBH H modelling From Figure 8 bias levels are low for both NFI and PSP Pine and Spruce the most important cohorts are among the top performers This partly reflects the better data resources for those cohorts Taken together these results can inform uncertainty sensitivity analyses to be completed in 2011 References Robert A Monserud Hubert Sterba 1999 Modeling individual tree mortality for Austrian forest species Forest Ecology and Management 113 109 123 Hubert Hasenauer 1997 Dimensional relationships of open grown trees in Austria Forest Ecology and Management 96 197 206 Robert A Monserud Hubert Sterba 1996 A basal area increment model for individual trees growing in even and uneven aged forest stands in Austria Forest Ecology and Management 57 80 Thurig E Kaufmann E Frisullo R and Bugmann H 2005 Evaluation of the growth function of an empirical forest scenario model Forest Ecology and Management 204 51 66 ID Modelling height increments for small trees
35. 0 00449 x DBH Spruce cohort 1 Pmort IL 6 8976 0 0912 x BAL 21 3795 x CR 0 8287 x DBH 49 15 x aan 0 008 x DBH Where 0 lt Pmort lt 1 is the probability the tree is dead We map then this estimated probabiilty onto the binary Dead Alive outcome using a cutoff which may differ between cohorts More details on this is give elsewhere IL is the inverse logit e g IL x exp x 1 exp x Choosing the operational cut off To identify a cut off level to use for the mortality probability estimate we plotted the True positive rate TPR and FPR on the same axis versus the cut off e g Figure 15 In forest mortality the number of positive cases dead trees is usually greatly outnumbered by the number of negative cases This suggests that all mis classification costs being equal the cut off should be chosen with a view to keeping as small as feasible the rate of false positives predicted by the resulting rule even though the rate of true positives is reduced as an unavoidable consequence When combining individual cohort results to make an aggregate prediction the issue of false positive rate is of particular importance for large cohorts because they have a greater weight in the aggregate estimate In Figure 15 we represent an FPR of not greater than 0 001 with a blue vertical line and an FPR of not greater than 0 01 with a green vertical line to illustrate the trade off involved in each particular case Ta
36. 16 AB H gt 3 8m ax DBH cxH4 0 022 2 73 0 19 2 06 0 96 0 46 1 008 ii iii 17 AB H lt 3 8m axH xc 0 005 1 58 1 12 0 86 0 28 1 02 i ii 18 TB exp Ln a bx Ln AG 1 59 0 96 0 99 0 28 1 005 ii iii 4 BB TB AB 5 FB ABxa bxexp cx AB 0 025 0 089 0 003 0 68 3 4 0 98 i ii 19 SB exp Ln a bx Ln AG 0 89 0 96 0 98 0 57 1 055 ii iii 7 Lur AB SB Slow growing broadleaves 20 H gt 3 0m bx DBH 0 08 25000 2 5 246872 iv a DBH 246872 21 AB H lt 3 0m axH 0 031 1 72 0 84 0 88 0 91 i ii 22 BB expl a Ln DBH b 1 509 0 284 iv 23 FB DBH gt 10cm ax DBH x10 0 009 1 47 0 96 v 24 FB DBH lt 10cm ABx0 3 0 78 1 2 0 79 i ii 25 SB DBH gt 19cm ax DBH x10 0 0002 2 5 0 97 v 26 SB DBH lt 9cm AB BB BEF 1 4 7 Lur AB SB Slow growing broadleaves 20 AB H gt 3 0m bx DBH 0 06 25000 2 5 246872 iv a DBH 246872 67 Function Equation Coefficients ir RMSE Slope Source 21 AB H lt 3 0m axH 0 031 1 72 0 84 10 88 0 91 i i 22 BB exp a Ln DBH b 1 509 0 284 iv 27 FB DBH gt 3cm a bx DBH 0 375 0 0024 2 517 0 90 vi 28 FB DBH lt 3cm ABx0 3 0 78 12 0 79 ii 29 ISB DBH gt 35cm 4x DBH 0 0001 2 535 0 97 V 30 ISB DBH lt 9cm AB BB BEF vii 1 4 7 Liik AB SB i National research harvested tree database COFORD funded project CARBiFOR ii Black et al Biomass equations for modelling C dynamics in Ir
37. 20 4 2 Stand Modification Options cccccceececceceeeeeceeecaeeeeceeeeeeeaeeeecaeaeeeecaeeeeseaaeeecaaeeeeseaeeeesecaeeeeieeeeeenaeeess 26 4 2 1 Mortality probability threshold Options eee ec eenee eter ater eeeeeeteae ease teases eeeeeeseneeeeeeeete 26 5 Pre processing of Forest Inventory Datta ceccecececeececeeeeeseeeeeeeeeeeeeeceeeesaeeeseeeseeeeeeeeeseeeeseeeeneeeseeeeseaes 26 5 1 Selecting the required Forest Inventory Database ceecceeeeeceeeeeeeeeceneeceneeeeeeeeaeeeeeeeteeseeeeeanees 27 5 2 Populating Trees Table with Small Tree Records from ceccesecceseeceeeceeeeeeseeeeeeeeeseneeseeeeseeeeeenees 29 Individual Trees Table csiective stk cS reiencceeaee cb ae a a a a aaa o tie RRA 29 5 3 Creating the Carbware Table orero annro ane en E E RRE AE EE AREA E AETAT EEEE 32 6 Creating or Selecting a Carbware Project ceccceececeececeeeeeeeeeceeeeeeeeeeceeeeseeeeeeeeeseeeeeeeeseeeeseeeeseneseeeereaes 34 6 1 Creating a New Carbware Project cceeceeseeeceeeeceneeeeeeeeeeeeaeeeeaeeeeaeeseaeesaeessaeeeseesaeesiaeesieeeeeees 34 6 1 1 Selecting the required Forest Inventory Database ecceeeceeeeceeeeeceeeeceaeeeeaeeeeeeeeaaeeenaeeeaas 36 6 1 2 Filtering the Forest Inventory Database by Plot or Date Criteria 0 0 eeeeeeeeeeteeeeteeeereeeenees 37 6 1 3 Selecting the required Stand Modification EVent cecceeeceseeeeseeeeseeeseeeeseeeeseneeseeeeseeeenenees 37 6 1 4 Saving a New Car
38. 2011 12 33 21 Project_1_02 E Project_1_Mortality Table HHY Date Created 15 09 2011 23 16 46 al Date Modified 15 09 2011 23 16 46 s EA Project 1 03 Date Created 22 09 2011 15 19 21 Date Modified 22 09 2011 15 19 21 ES Project_1_Clearfell_Table Date Created 15 09 2011 12 34 07 Date Modified 22 09 2011 14 36 52 Project_1_Increments Date Created 15 09 2011 12 33 19 Date Modified 22 09 2011 15 19 21 Project_1_ModifRecords Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Mortality Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Thining_ Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 tblTempo Date Created 15 09 2011 12 18 16 Date Modified 15 09 2011 12 18 16 The Mortality Events Table contains all of the necessary information for performing annual Carbon Allocation routines on the results of annual growth and stand modification simulations specific to prescribed thinning events The Mortality Events Table fields are exactly the same as the Clearfell Events Table fields described in Table 6 see Section 8 2 2 with one exception The final field is named standing deadwood and replaces the Volume field in the Clearfell Events Table and Thinning Events Table see Table 7 below 48
39. 33207401 16 5132852421127 20 9279742080938 15 4408033052125 22 1688469537186 20 5777755248994 25 4878197844968 20 3448015769485 17 6527118934627 15 8589113513347 v gt e The Thinning Events Table contains all of the necessary information for performing annual Carbon Allocation routines on the results of annual growth and stand modification simulations specific to prescribed thinning events The Thinning Events Table fields are described in Table 6 Field Code Description Unit ModificationYear Year of Thinning None EventCode Code 200 thinning None PlotID Plot ID number None TreelD Tree ID harvested None CohortCode Cohort see cohort look up None Estim_Height Height estimated m DBH Diameter at breast height cm Age Age of harvested tree years ExpansionFactor Expansion factor for partial sample plot None DeadTree tree status 100 alive 200 dead None AG Aboveground biomass kg C TB Total biomass kg C NB Needle leaf biomass kg C LTR Harvest stumpage residue kg C SB Stem biomass kg C RB Root biomass kg C LT lop and top kg C representative AG biomass AG Exp AB_C factor kg C SB_C representative SB biomass SB Exp factor kg C RB_C representative RB biomass RB Exp factor kg C LT_C representative LT biomass LT Exp factor kgC Volume represenatrive tree volume m3 Table 6 Description of Thinning Events Table Fields in the IncrementsDB mdb
40. 52 0 04531 187E3 52 86 0 0001 DBH 0 8127 0 007225 187E3 112 49 0 0001 BAL 0 08083 0 000999 187E3 80 91 0 0001 RelDiamInc 23 0015 0 3995 187E3 57 57 0 0001 Slow growing broadleaves 92 Parameter Estimate s e df Wald statistic Wald p value Intercept 29 6029 7 1305 1027 4 15 0 0001 DBH 2 1970 0 4873 1027 4 51 0 0001 BAL 0 1225 0 01754 1027 6 98 0 0001 RelDiamInc 2199 90 521 36 1027 4 22 0 0001 Spruce cohort Parameter Estimate s e df Wald statistic Wald p value Intercept 1 2286 0 02747 298E3 44 72 0 0001 DBH 0 6640 0 003840 298E3 172 93 0 0001 BAL 0 05051 0 000529 298E3 95 57 0 0001 RelDiamInc 13 0524 0 2544 298E3 51 30 0 0001 Candidate Model 2 The fixed effects in Candidate model 2 were those in Monserud and Sterba 1999 and diameter increment as a proportion of diameter RelDiamlnc Cross validation and deployment performance We performed plot wise and case wise leave k out cross validation of the chosen models The case wise deletion algorithm was very slow for the Pine and Spruce cohorts in which case we opted to use only plot wise deletion The algorithm selected plots for deletion from the fitting dataset using a Bernoulli mechanism with parameter p which we sometimes changed depending on the number of plots in the cohort dataset Details are provided with each graphical representation of the results in Figures below Twenty leave outs were performed and the variability in these twenty runs is represented by
41. 94 338 a 3 cs cs When fitting models to the NFI data we used backward elimination starting with the parameters in the Monserud and Sterba 1999 model Relative diameter was not used because the dataset is cross sectional In Figure 14 we present an example of the out of sample performance i e their performance in predicting NFI data of the two PSP calibrated models and the in sample performance of the NFI calibrated model Fast growing broadieave cohort NFI validation Fast growing broadieaves Candidate 2 NFivalidation Pee i iy ee in R True postive rte o4 25 gT 04 a6 04 25 02 uF Dr 00 02 04 06 os 10 00 02 o4 06 08 10 a b Fast growing broadhaves Candidate 2 NFIfit os os True poetive tate Figure 14 The Receiver operating characteristic curve for Fast growing broadleaves cohort Candidate models 1 and 2 fitted on PSP and for the NFI fitted model The selected CARBWARE models based on NFI data fits Fast growing broadleaves cohort 95 1 Pmort IL 12 93 0 068 x BAL 2 868 x CR 0 962 x DBH 72 28 x DRH 0 009 x DBH Larch cohort Pmort IL 4 9266 0 04273 x DBH Other conifers Pmort IL 4 5226 0 067 x BAL 6 05 x CR 0 066 x DBH Pine cohort Pmort IL 2 395 0 0408 x BAL 3 0036 x CR 0 2263 x DBH 24 21 x Slow growing broadleaves Pmort IL 15 78 0 0109 x BAL 2 2807 x CR 0 771 x DBH 94 002 x ir
42. Box 1 the plot is 0 05 ha c County these are user defined lookup codes must be numeric used to agreegating final data in to regional categories This does not have to be counties but the field header should not be changed d Ownership A user defied lookup code for describing forest owner categories e Forest_category this can be user defined to categorise plots under different forest types For example the Irish UNFCCC reporting system uses the following categories 23 Table 2 Default forest category codes used Forest_Category_Code Forest_Category_Description 101 to 115 O ee Ss 8 7 o 9 P 101 to 115 pF 200 Spruce Pure Mainly Sitka and Norway spruce Pine Pure Prodominantly Scots and lodgepole pine Larch Pure Other conifers Pure Fast growing broadleaves Pure such as ask Alder Sycamore Birch Slow growing broadleaves Pure such as Oak and Beech onifer mixes oadleaf mix onifer Boradleaf mix pen areas including biodiversity areas roads within the forest boundary own areas subjected to windthrow crub felled or failed areas planted and unplanted ew afforestation after 2006 atural succession and regenreating land arvested areas E g 101 are harvested spruce areas urned areas w 200 Forest stands were considered to be pure if one species represents 80 or more of the canopy h LanduseField not mandatory Litter_C_tn_1 This is the initial litter carbon pool expressed per unit area of the plo
43. CARBWARE Forest GHG Inventory Software User Manual version1 5 were TTT Date 30 March 2012 Authors Mark Tarleton and Kevin Black The development of CARBWARE was funded by COFORD under the NDP 2007 to 2012 COFOR Table of Contents aaie o LULO i oT a TAEAE AE be cube dent te ong EE tages aa sad dees teagt A A nets denteecate hens teens eee ley 4 PANE E ES Stam E E AEE E EE E E ENEE E VE T E E E 5 2 1 The Intended USer aane e e ea o eaae ra e a ae aaa aaa as raaa aas etter ea eee 5 2 2 Hardware requirements ccecccccececeeeeeceeceeeeeeseeeneeeeeeseneaeee nnet ceeneaceeeeeseeeaeeeeeeeeeeeeeeceseeeeeaeeeeeeeenees 5 2 3 Installing Carbware ON youl PC cceceecececeneeeeneeeeneeceaeeesaeeeeaeeeeaeeesaeeeeaeeeseeesaeeesaeeesieeeseesenestieestneees 5 3 Overview of CARBWARE functionality ccccceeceeceeeeeneeceneeeeneeeeaeeeeaeeecaeeesaeeeeaeeesaeeesaeeeseeesaeeseeesieeetieeees 8 3 1 User quick start flow Gia Qraim eececcecesceceseceeeeceseeeeseeeeaeeneaeeseaeeesaeeseaeesaaeeenaeesaaesaaeseaaeseaeeneaeeneaeenaas 8 3 2 Detailed functionality aasa S einai a el eae ee 8 4 Data Input requirements sksins ee ave ceed devi nap ae eaaa need uate agente abt N R snes ae 11 4 1 Forest inventory database i e Carbw 200X ccecccceseeeeeeceeneeceeeeceeeeeeeeeseaeeseaeseeaeeeeateseaeeeeateeeaeeeas 11 4 1 1 Required Inventory Database Structure amp File Naming Conventions ccseesteeseeeesteeees
44. Field Code Description Unit ModificationYear Year of mortality None EventCode Code 300 mortality None PlotID Plot ID number None TreelD Tree ID dead None CohortCode Cohort see cohort look up None Estim_Height Height estimated m DBH Diameter at breast height cm Age Age of harvested tree years ExpansionFactor Expansion factor for partial sample plot None DeadTree tree status 100 alive 200 dead 300 None damaged tree AG Aboveground biomass kg C TB Total biomass kg C NB Needle leaf biomass kg C LTR Harvest stumpage residue kg C SB Stem biomass kg C RB Root biomass kg C LT lop and top kg C AB_C representative AG biomass AG Exp kg C factor SB_C representative SB biomass SB Exp factor kg C RB_C representative RB biomass RB Exp factor kg C LT_C representative LT biomass LT Exp factor kg C Standing representative dead tree total biomass kg C deadwood Table 7 Description of Mortality Events Table Fields in the IncrementsDB mdb database 9 Carbon Allocation Once you have completed the growth and stand modification event simulations for your CARBWARE project you must complete the Carbon Allocation process before you can generate any CARBWARE results As described in Section 3 2 the CARBWARE Carbon Allocation process uses the intermediate growth output tables described in Sections 7 and 8 to generate carbon stock estimates for 6 major carbon pools These are Above Ground Below Ground Deadwood Lit
45. I data not currently used RootDamageDegree Trees Specific NFI data not currently used RootDamageAge Oo oO OF OF o AQ AQ OF OF OF BD AQ ol AQ oO Trees Specific NFI data 15 Data Description Requirement Table Units Format OtherDamage O Trees Specific NFI data not currently used HeightSampleTree O Trees Specific NFI data not currently used DO3hSampleTree O Trees Specific NFI data not currently used VitalitySampleTree O Trees Specific NFI data not currently used BaseDBH_mm O Trees Specific NFI data not currently used CalcHeight_m Calculated tree height from O Trees Numeric independent models Vol_BFC_m3 BFC volume of tree O Trees Specific NFI data not currently used Adjust_Vol_BFC_m3 Adjusted volume O Trees Specific NFI data not currently used Repre_Vol_BFC_m3 Representative volume O Trees Specific NFI data not currently used TreeNumber Number of trees measured M Trees Numeric no default 1 decimal places BasalArea_m2 Individual tree basal area M Trees Numeric up to 15 decimal places in m2 AdjustBasalArea NFI adjustment default M Trees Numeric up to 15 BasalArea_m2 decimal places in m2 RepreBasalArea_m2 BasalArea_m2 M Trees ExpansionFactor see below TimberHeight_m Height of timber O Trees Numeric in m RepreTreeNumber Number of trees the M Trees Numeric up
46. Land Cover Habitat and Soils Mapping Modelling EPA project report EPA Forest Service 2003 Forestry Schemes Manual Department of Communications Marine and Natural Resources Stationery Office Dublin Gardiner M J and Radford T 1980 Soil Associations of Ireland and their land use potential Explanatory bulletin to the soil map of Ireland 1980 Soil Survey Bulletin No 36 Teagasc formerly An Foras Taluntais Oak Park Carlow Gallagher G Hendrick E Byrne K 2004 Preliminary estimates of biomass stock changes in managed forests in the Republic of Ireland over the period 1900 2000 Irish Forestry 61 16 35 Hargreaves KL R Milne and M G R Cannell 2003 Carbon balance of afforested peatland in Scotland Forestry 76 3 299 317 NFI 2007a National Forest Inventory Republic of Ireland Results Forest Service Department of Agriculture Fisheries and Food Johnstown Castle Estate Co Wexford NFI 2007b National Forest Inventory Republic of Ireland Methodology Forest Service Department of Agriculture Fisheries and Food Johnstown Castle Estate Co Wexford Saiz G Black K Reidy B 2007 Assessment of soil CO2 efflux and its components using a process based model in a young temperate forest site Goederma 139 79 89 Tobin B Black K Osborne B 2006 Assessment of allometric algorithms for estimating leaf biomass leaf area index and litter fall in different aged Sitka spruce forests Forestry 79 453 465 104
47. N Graham Areas beneath the relative operating characteristics ROC and relative operating levels ROL curves Statistical significance and interpretation Quarterly Journal of the Royal Meteorological Society 128 2145 2166 2002 R Monserud Simulation of forest tree mortality Forest Science 22 438 444 1976 R A Monserud and H Sterba Modeling individual tree mortality for Austrian forest species Forest Ecology and Management 113 109 123 1999 N A Obuchowski Nonparametric analysis of clustered ROC curve data Biometrics 53 567 578 1997 T Sing O Sander N Beerenwinkel and T Lengauer ROCR Visualizing classifier performance in R Bioinformatics 21 20 3940 3941 2005 J Wunder C Bigler R Bjrn L Fahsse and H Bugmann Optimisation of tree mortality models based on growth patterns Ecological Modelling 197 196 206 2006a J Wunder B Reineking C Bigler and H Bugmann Predicting tree mortality from growth data how virtual ecologists can help real ecologists Journal of Ecology 96 174 187 2006b J Wunder B Brzeziecki H Zybura B Reineking C Bigler and H Bug mann Growth mortality relationships as indicators of life history strategies a comparison of nine tree speceis in unmanaged european forests Oikos 117 815 828 2008 II Other modifications in the growth simulator Thinning Harvest We assume that all thinning occur randomly Random thinning can be implemented on an individual plot level T
48. WARE project are stored in the tblProj Table located in the _Carbware_Param database which is installed as part of the EPES Co2_Done o v v v 0 m Record 4 1of1 bombo G la m To create a new CARBWARE project click on the Choose Project button in the Main Menu screen 34 Carbware Database in use C Documents and Settings Projects Carbon 3 Project Co2 Allocation Database About Carbware Select Database Co2 Allocation Pre Processing Co2 Reporting Close Program You will see the following Project Selection screen Project Selecction New Project Note If you have already created a CARBWARE project it will be listed on the right hand side of this screen see Section 6 2 for details on how to select an existing CARBWARE project To return to the Main Menu screen without creating or selecting a CARBWARE project click on the Cancel button To proceed with creating a new CARBWARE project click on the New Project button You will see the following Project Parameters screen 35 Carbware Project Parameters Project Name Select Database B Filter Baseline Inventory Database Stand Modification C by Date Criteria Natural Mortality Clear Fell M Thinnning You have done 0 cycles This screen allows you to set the parameters for your CARBWARE project s Growth Simulation and Stand Modification
49. _NO2_EF Not included in current version Organic_rich_NO2_EF Not included in current version NO2T Not included in current version 4 1 1 4 The tblEvents table The Events table instructs the software when to harvest specific plots and specifies the basal areas to be removed See table 1 for details a b c d PlotID must be the same as ID plot in other tables Year is the year when the harvest occurs the harvest takes place after a growth cycle at the end of the year Event defines a thinning 200 or clear fell 100 harvest Reduction is the basal area per ha removed For clear fells all the trees are removed regardless of the target basal area For thinning trees are randomly removed until the threshold basal area is reached Note that a thinning will not be executed if the thinning basal area is greater than the standing basal area We recommend that preliminary runs are performed to check if the correct basal area reduction is applied at thinning 4 1 1 5 The Afor_Criteria table The afor_criteria table is used to pre sample the database when creating a project This allows for quick selection of plots that fall into the UNFCCC reporting categories under articles 3 3 post 1990 and 3 4 pre 1990 forests under the Kyoto protocol Field names and data formats The screen shot of the Individual trees table below shows the precise field names order of field and data formats which must be adhered to
50. _litter t1 minus decomposition tC plot Litter_input_t02 deadwood t1 t0 Deadwood_t01 tC plot Mortality_LT_t02 Litter and branch from mortality t1 tC plot Thin_AB_t02 Litter top leaf and branch from thinning t1 tC plot CF_AB t02 Litter top leaf and branch from clearfell t2 tC plot T_Litter_in_t02 Mort LT Thin_AB CF_AB_Litter input t2 tC plot Decomposition of accumulated T_litter in t01 t02 and T_Litter_out_t02 net_litter_tO tC plot Net_Litter_t02 T Litter in minus T litter out t1 tC plot Litter_Stock_t02 net_Litter tO and T_litter t1 t2 minus decomposition tC plot Table 11 Description of Litter Carbon Table Fields in the IncrementsDB mdb database 57 9 2 5 Soil Carbon Table e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_Soil In this example the project name is Project_1 and so the project specific Soil Carbon Table is named Project_1_ Soil Field Name Decription Unit PlotID Plot ID None Area_ha Area of Plot default 0 05 Ha County County lookup ID user defined None Ownership Ownership lookup ID user defined None LandUseField Land use lookup ID user defined None Soil_EF_Field Soil EF ID lookup ID user defined None Forest_Category_ID Forest category lookup ID user defined None CARBWARE_soil Soil code from Lookup see soil lookup and soiltype_lookup xls None Soil_ EF_Cate
51. able not set line 380 in Sub adiGrowth CarbwareTo_0 variable or block variable not set line in Sub mdlGrowth CarbwareTo_0 variable or block variable not set line in Sub mdlGrowth CarbwareTo_0 variable or with block variable not set line in Sub sdiGrowth CarbwareTo_0 variable or with block variable not set line in Sub mdlGrowth CarbwareTo_0 O1 11 2 variable or with block variable not set in line in Sub mdlGrowth CarbwareTo_O O1 11 2 _ Q 01 11 37 01 11 2 01 11 2 01 11 23 01 11 2 01 11 27 0141 11 2 sesesesesessesesese In this case the input table Afor_Criteria was not included in the Carbwxxx file or there was an Field name error as specified by the object variable or with block variablr not set error Errors can occur for many reasons but the most common reasons include e Exclusion of required input field or tables e Incorrect entry of field header names e Entry of incorrect numeric values e g DBH 1234 cm 65 Appendix 1 Appendix 1A Allometric biomass equations used in allocation module Table 1 Allometric equations used to calculate biomass component for individual trees kg d wt tree Similar species are grouped into 6 different cohorts based on available research information Spruces Pines Larches Other conifers fast growing broadleaves and slow growing broadleaves Abbreviations AB above ground TB total biomass BB below ground FB foliage SB stem i e timber gt 7cm diamete
52. aggregated to the plot level using information in the lookup table located in the Carbw mdb database created by the user see section 4 Detailed calculation steps for the estimation of the biomass pool changes are shown in Appendix E equations 1 to 9 The losses from the biomass pool due to harvest obtained from the Project_thinning and clearfell tables are calculated using biomass algorithms These values are scaled up to the plot level using the scaling factor in CARWBARE table and transferred to the Project_HWP table Harvest residue left on site is transferred to the Project_deadwood or Project_litter tables Biomass from mortality is either transferred to the deadwood or litter pool Project_deadwood or Project_litter tables depending on the size DBH of the trees see Appendix E vi The litter pool stock changes are obtained from the initial input values lookup table and functions in the module described using equations 9 12 in appendix E vii The deadwood pools are calculated using the initial data inputs from the user lookup table see section 4 and functions described in equations 13 15 appendix E viii The soil pools are calculated using plot area soil classification and emission accumulation factors described by the user in the look up table see section 4 and appendix E ix The HWP pool is derived using the biomass of harvested trees allocated from the biomass pools This module does not calculate HWP because the current r
53. alled on your PC as part of the CARBWARE software installation process see Section 2 3 A series of new tables are created for each of the annual growth cycles These are described in Section 7 2 below To perform another growth cycle click on the Grow for 1 Year button again Note As each annual growth cycle is completed the Grow for 1 Year button is modified to inform the user of the cycle number and the date the next growth cycle will run to In this example above the button reads Grow for 1 Year to Year 2 2007 To return to the Main Menu screen without performing another growth cycle click on the Close button 41 Note You can return to the Project Parameters screen to perform additional annual growth cycles at any time 7 2 Increment Database As noted in Section 7 1 above results of annual growth cycles for all CARBWARE projects are stored in the database file IncrementsDB mdb which is installed on your PC as part of the CARBWARE software installation process see Section 2 3 A series of new tables are created for each of the annual growth cycles CARBWARE uses these tables to perform additional annual growth cycles see Section 7 1 and to perform carbon allocation routines see Section 9 Tables within the IncrementsDB mdb database file relating to annual growth cycles are described in Sections 7 2 1 and 7 2 2 below Tables within the IncrementsDB mdb database file relating to stand modification events within the an
54. ancel KPAF2010 re 1 Accept Delete e Select Delete e A pop up warning will confirm that this Project will be deleted and promts the use to select OK or Cancel if this is the wrong selection e Click OK and all files output files in the IncremetnDB related to the project will be deleted e Note the input database will not be effected e g Carbwini2005 11 2 Compacting and reparing databases If databases in Access are modified it is good proactrive to compact and repair the database to avoud redundant disc space usage and minimise database interoperability issues This can be done in Access e Open the database IncrementDB in this example which was modified because Project2 was deleted e Select the Database Tools menu and then the Compact and repair ribbon in Access2010 or Compact and repair drop down for Access 2007 and earlier e The Database with be repaired and compacted you will see the disc allocation space has reduced 63 189 2007 CARBWARE RESULTS 185 2008 CARBWARE RESULTS Table Con gt Date Mo 30 189 2009 CARBWARE RESULTS Table Date Created 2206 2013 15292 11 3 Error log If there is run time error or variable violation during the running of CARBWARE a error log file will be created in the program file directory WK oa B SES HE AaBbi AaBbC aaBbCel AaBbC AaBeCcD AaBdCcO AaBbCeL AaBb AaBbCcE AaBbCeO ia EE a v ew a s oe Theadng YHeadeg THwadngs Theadegs INermat TNo Spac
55. anent sample plot data taken from a range of spacing and thinning experiments used in this study is well suited albeit not arising by design to evaluate these stand density parameters to describe variations in H and CR across different silvicultural conditions Materials and methods Data Data used were obtained from Coillte Teoranta s the Irish Forestry Board state commercial forestry company permanent sample plot record system The dataset contains records from many silvicultural and thinning trials established during the period 1963 to 2001 The trials were initially established as replicated experimental designs with repeated measurements typically undertaken every five years The dataset is described in Broad and Lynch 2007 Incorporating competition covariates The modelling here follows Temesgen and Gadow 2004 who based their work on Yang et al 1978 and incorporated competition covariates into the Yang Weibull function Table 1 Model 2 We evaluate that model and also use test for differences between management regimes conditional on the DBH H model by incorporating dummy indicator variables in the linear regression models of the model parameters Our aim in this section was to test if the inclusion of certain covariates typically relating to the competition in a forest stand plot improved the baseline DBH H model Table 1 Model 1 We also investigated whether the model was improved by including random effects on the level
56. anner However it should be noted that the empirical tree growth and 1 Tiers refer to methodological rankings used as set out by the IPCC good practice guidance Tier 1 refers to default methods higher tiers use country specific and more complex modelling approaches mortality models have been developed using Irish permanent sample plot data see Appendices Users outside the Ireland and the UK wishing to extend the use of CARBWARE should consider developing eco regional specific growth and mortality models and regional species specific site factors such as decomposition rates Tree growth and mortality is based on distance independent single tree growth models which use basic individual tree data such as DBH Height and crown ratio and stand information such as stocking basal area and soil characteristics Duffy et al 2001 Hawkins et al in press The input data and archiving systems requirements are outlined in section 4 The functionality of the various components of the software Growth simulator stand modifier CO allocation and reporter are outlined in section 3 All intermediate and final output files from a set of projections i e a project are stored on Microsoft Access files see system requirements section 2 which controlled by the user The Output data is collated in a format compatible with the Common Reported Format CRF tables used in UNFCCC and Kyoto reporting of the LULUCF sector 2 Getting Started 2 1 The Int
57. ansion factor we use and that used by the NFI assumes a random distribution for tree diameter in the plot Because of that assumption the weight assigned to a tree in the ith diameter class is where Ri denotes the radius of the concentric circle associated with the ith diameter class In practice the expansion factor or weight is used to estimate plot level features e g basal area In such calculations we estimate the number of trees of the ith diameter class that were not included in the 2 sample by oe ni where ni is the number of trees of the ith class that are included in the sample The expansion factor therefore defines the relationship between each included tree and the estimated number of trees of the same class that were not included Equation 2 w aceasta cep ara can erneyg te tice 2 ij where n _ EF is the product of the expansion factor for the j th tree in the th class and Nj is the corresponding estimate In the terminology of the NFI the RHS of Equation 2 is the representative tree number With minor and obvious changes to the equation we can calculate other tree level estimates including representative basal area and individual tree estimates can be aggregated for the entire plot to give plot level estimates including representative density 89 sub circles qualified trees Sub circle radius m Sub circle area m Treshold diameter mm Figure 10 The NFI co
58. approaches Similarly a generalised model might perform well on plots that are nearer the centre of the sample space than plots where management conditions are more atypical for a given dataset In conclusion we adopt the use of generalised competition based models in the CARBWARE software because this performs better across all data See Table 1 References Broad L and Lynch T 2006 Growth models for Sitka spruce in Ireland Irish Forestry 63 1 2 2006a 76 Temesgen H and von Gadow K 2004 Generalized height diameter models An application for major tree species in complex stands of interior British Columbia European Journal of Forest Research 123 45 51 Yang R C Kozak A and Smith J H G 1978 The potential of Weibull type functions as flexible growth curves Canadian Journal of Forest Research 8 424 431 Zeide Boris 1993 Analysis of growth equations Forest Science TT Appendix 1C Growth modelling I Modelling diameter increments in Irish Forests Introduction The modelling approach adapted in this version of CARBWARE v5 is the use of diameter increment models for all trees with a DBH greater that 5cm This model in a distance independent individual tree growth model parameterised on Coillte permanent plot data recorded every 4 to 6 year since 1954 to 2003 Theses include pure and mixed species stands at establishment planting densities of 5000 to 1000 trees per ha and with different thinning tr
59. associated with lreland s National Forest Inventory will take several hours to process The Project Parameters screen displays a progress bar during an annual growth cycle informing the user of progress and noting each of the stand modification events as they are simulated Natural Mortality Clearfell amp Thinning 40 Carbware Project Parameters Project Name Inventory Year 2005 Select Database __ C4Documents and Settings Projects Carbon_Project Carbwini2006 mdb Filter Baseline Inventory Database Stand Modification e G r Report on g Vv c oO You haye done 0 cycles Natural Mortality Plot 189 Processing plot Info Plot 189 When the first annual growth cycle is completed a message will be displayed at the bottom of the Project Parameters screen noting the number of annual growth cycles performed to date and the location of the results of the latest growth cycle Carbware Project Parameters Project Name Inventory Year 2005 Select Database PtPecuments and Settings Projects Carbon_Project Carbwini2005 mdb Filter Baseline Inventory Database Stand Modification p Al Reporton Grow for 1 Year e d I to x Year2 2007 c F You have done 1 cycle Results in CAProgram FilestCOFOR D CarbwareilnerementDB mdb In table Project_1_Increments Note Results of annual growth cycles for all CARBWARE projects are stored in the database file IncrementsDB mdb which is inst
60. ave been measured in the inventory plot s see appendix 1B for description of functionality DBH should be measured for every tree above a height of 1 3m Heights should be measured for trees smaller than 1 3m iii Classification of different species into the 6 cohorts for growth modelling iv Derivation of required parameters used in the growth and mortality modules see appendices for description of models These include parameters such as basal area stocking basal area of the larger trees BAL probability values for the likelihood that a tree may die in the next growth phase and crown competition factors v Calculation of scaling up expansion factors in the case where all trees may not be measured in a plot and where systematic stratified tree sampling is performed see Section 4 These values are used to adjust the stand level values such as basal area and BAL vi Classification and categorisation of tree status and age for input in to the growth and mortality model in the Growth simulator module b Growth simulator This module performs the growth simulations on an annual cycle up to 15 cycles The growth models are based on distance independent single tree models using DBH as the growth output see appendix 1C Growth of trees with a DBH of less than 5 cm is modelled using a simple Chapman Richards growth function with height as the growth output Appendix C The stand modifier modifies individual tree and stand values after eac
61. ble 8 Formulae for some standard performance measures used in the text Note TP TN FP and FN are the numbers of true positives true negatives false positives and false negatives which are tallied by comparing the predictions with the data Performance Measure Formula ee TP4TN Accuracy TP FP IN FN TP4LEP Rate of positive predictions TPLFPLINGFN TP TN FP FN y PP FN TN FP TP FP TN FN Correlation Coefficient Figures 16 illustrate some other considerations for choosing cut off points accuracy rate of positive predictions and a correlation coefficient are plotted for a range of cut offs cf Table 8 for definitions of terms The graphs illustrate why the accuracy measure should not be used in isolation when choosing a cut off For example in Figure 16 a high accuracy is obtained despite the correlation coefficient indicating that the correlation between correct predictions and the data is worse than random i e a negative correlation coefficient Some performance measure formulas are given in Table 8 These measures and others are described in Sing et al 2005 96 Other conifers NFI fit TPR and FPR versus cutoff TRP FPR Figure 15 TPR Black and FPR Red versus cut off for Fast growing broadleaves The vertical green line shows the cut of where FPR lt 0 01 the blue vertical line shows the cut off where FPR lt 0 001 os Accuracy 06 04 02 0 0 a 97 Rate of p
62. bscised needles and leaves The dead wood pool included all lying and standing deadwood dead roots and stumps with a diameter greater than 7cm organic and mineral organic soils are reported see section 11 3 1 2 Biomass stock change Biomass carbon stock changes are calculated using a tier 3 gain and loss method corresponding to the process based approach given by equation 2 4 in Chapter 2 of the 2006 IPCC guidelines which gives the net carbon stock change as the difference between carbon gains and carbon losses DOING PENG aaa pia tae eae Rtas shee eae Shs 2 The biomass carbon gains ACg for both above ground biomass AB and below ground biomass BG are calculated for each forest category i listed in table 11 2 using ING AKO TOTALS OTS fess at teak ate ae tat an i ala ca 3 L where A is the area of the forest category GTOTAL is the biomass change t dm ha yr in area A and CF is the carbon fraction of biomass dry matter which is taken as 50 percent for all carbon pools Black et al 2007 GTOTAL is derived from the sum of all living individual tree components i e AB or BB within the plot for example CTO SAT AC RECTOR ET TEOTT ee RON OT Tn Nene 4 where n is the year of inventory The GTOTAL value for each NFI permanent sample plot is multiplied by a factor e g 20 for 0 05ha plots to scale the plot measurement up to 1 ha specified in the lookup table The AG and BG of all trees were calculated using biomass a
63. bware Project cccceseceeeeeeeeeeceneeeeaeeeeaeeeeaeeesaeeesaeeeeaeeeseeeseeeseeeeneeesneetanens 38 6 2 Selecting a Saved Carbware Project cecceseceeceeeeeneeeeeeeeeneeeeaeeeaaeeeaaeessaeesaaeesaeeseeesieeessaeesieeseeess 38 T Growth Simulation 0 cise eed ons colar een Veet ees eens CEA E ENA E nein E EE 39 7 1 Running Annual Growth Cycles esscr aaa 39 2 Increment Database cc wt ici eens beth oat ste aie eta a ha ed i aa 42 7 2 1 Summary of Annual Growth Cycle Increments IncrementDB Table ccceeeeseeeeeeeeeeneees 42 7 2 2 Carbware Tables for each Annual Growth Cycle ecceeccceeceeseeceeeeeeeeeeseeeeseneeseneeseeeeseeeenenees 43 8 Sand MOCIICAUOM gc8 cis x95 ats chee a a agent Heng harden steed a a a We Moca ee eaten gti 44 8 1 Stand Modification Events during Annual Growth CyCles eccceseceeeeeeceeeeceneeeeneeeseeeeieeesieeeeaeees 44 8 2 Stand Modification Event Tables cccceceeceeeeeeeeeeeeeeeneeeeeeeeaeeeeaeeesaeesaeeessaeesaeeseeesieeesseesneeeeeees 44 8 2 1 Modified Records Table vrastare eitan Se eie e eSEE E EE ee aAa AATEC 45 9 22 Cleartell Events able sso ceca e E R EE dene ER deans 45 8 2 3 Thinning Events Table 3c ii canes yer ete ccoess ovens dag aaa AA aaae Adaa evens a adar aas Aaa paei 46 8 2 3 Mortality Events Tablete dived oae eeens device nv arte tava a ai aa a aaas ead ae aae ehe Aia 48 9 CarbomAllOGation i eia e EA E e E A aE E TE 49 9 1 Running the Carb
64. categories Reporter LULUC F_CRF Figure 2 A schematic representation of the different databases grey boxes containing input tables yellow boxes which are used by software modules black boxes to derive intermediate output tables green boxes and the final common reporting format tables CRF in red The input tables are generated by the users based on standard inventory information clear boxes The CARBWARE software functions represented as all activities in the red broken line border see Figure 2 have the following features 1 Options to select an already created database projects i e Carbw mdb and create delete or save projects i e a series of plot data and scenarios used in a projection 2 Software modules shown in black boxes figure 2 are the application steps which the user follows a Pre processing module This module processes the data into a format that can be used by the growth and modification models The input tables are used to populate a new table called CARBWARE which is written to the created Carbw database project This data is derived from the Trees and individual_trees tables which are created by the user see section 4 Functions performed in this module include i Extraction of and formatting individual tree and plot data from the sample plot inventory data see input requirements section 4 ii Computation of missing height crown ratio crown diameter values which may not h
65. contains many new calculated fields which are described in Table 3 below Refer to Section 3 2 for a description of the CARBWARE pre processing functionality resulting in the Carbware Table E afor_criteria _ o PlotiD TreelD SPP X Age PlantYear DBH dbhcm_ties ExpansionF BA BAha PA RUNDE a z 7 470 15 o 14 2 1 3 2480600631 0 0158367686 0 3167353713 am duanes 54 6 470 15 0 14 3 1 3 2480600631 0 0160606070 0 3212121409 E Lookup 54 3 470 15 0 10 3 1 17 683882566 0 0083322891 0 1666457823 E tblEvent 54 1 470 15 0 7 5 1 17 683882566 0 0044178647 0 0883572934 FA trees 54 8 470 15 0 14 8 1 3 2480600631 0 0172033614 0 3440672274 54 5 470 15 0 13 3 1 3 2480600631 0 0138929081 0 2778581622 E TreesDeleted900x FS Fe ms Sere 54 2 470 15 0 10 4 1 17 683882566 0 0084948665 0 1698973307 54 4 470 15 it 14 8 1 3 2480600631 0 0172033614 0 3440672274 54 3 470 15 0 13 5 1 3 2480600631 0 0143138815 0 2862776306 Field Code Description PlotID Plot ID TreelD Tree ID Species look up ID see ID field in look up file on instillation disc file name SPP model_cohort_lookup xls Age Age of tree PlantYear Year planted DBH Diameter at breast height 1 3 m in cm dbhcm_ties Calculated parameter showing values with the same DBH indicated as 2 ExpansionFactor Representative number of trees see Information Box 1 BA rea Calculated basal area of tree inm BAha Calculated basal area
66. cs model accuracies only when model flexibility was constrained They also provided guidelines for sufficient sampling schemes in real forests In the PrognAus framework Monserud and Sterba 1999 modelled mortality in Austrian forests for six major species based on 5 year re measurements of the permanent plot network of the Austrian National Forest Inventory Their general results varying slightly between species was that inverse of tree diameter crown ratio and BAL were respectively the three most closely correlated factors in their model with 5 year mortality rates They compared mortality rates across tree diameter class thereby identifying a classic U shape in mortality rates as diameter class increased They modelled mortality rates rather than individual tree mortality probability and validated the model with the chi square statistic calculated between observed and estimated Because the explanatory variables in their model were measured on the scale of the individual tree they were able to calculate the classification success rate using the complement of the overall proportion of mortality i e approx 93 although it is not clear from the text as the threshold On this basis their model correctly classifieded between 81 and 92 of live trees and between 25 and 44 of dead trees However their treatment of the threshold is very brief and may not be a typical interpretation e g in their interpretation a tree is classified as dead if
67. cut off has a defined purpose in a physical system Note that a logistic regression with a single explanatory variable can be made to fit such a schema In fact it might be possible to define a convex hull of the multiple explanatory variables to take the place of single variable classifier in that schema Also some variables might be better at defining the threshold than others and this can also be examined A convex hull defined by cut off points in each explanatory variable might be envisaged to play the role of a kind of syncretized cut off point In such an instance it would be relevant to assess the cross correlations among the explanatory variables Conclusions We set out to determine a logistic regression model of mortality that could be used to describe mortality in the NFI data This was the ultimate goal of the model We investigated the possibility of calibrating this model on the permanent sample plot longitudinal data but found that we could improve the result be simply calibrating the parameters on the NFI data alone In the absence of a mis classification cost function we chose the cut off for transforming predictions on the logit scale to the binary dead alive scale based on the false positive rate the rate at which the model predicted mortality incorrectly Specifically we chose the cut off to keep this as small as reasonably possible References B Baesens T van Gestel S Viaene M Stepanova J Suykens and J Van thi
68. d mortality on CR a variable that was not measured on every tree in our dataset There are several points of interest to the results of this model fitting RelDiamInc 1 The characteristics of the parameters 2 The cross validation exercise 3 The out of sample deployment performance E g how well the model described NFI mortality 91 Estimated parameters Candidate Model 1 Used in CARBWARE models The fitted parameters and their standard errors are presented in Table 4 We supply parameter estimates for cohort wise fits and the fit to the entire dataset with no cohort effect parameter Table 4 Candidate model 1 parameters Fast growing broadleaves cohort Parameter Estimate s e df Wald statistic Wald p value Intercept 2 9295 0 1510 11784 19 41 0 0001 DBH 0 4307 0 02508 11784 17 17 0 0001 BAL 0 06816 0 004384 11784 15 55 0 0001 RelDiamInc 1 6783 1 2147 11784 1 38 0 1671 Larch cohort Parameter Estimate s e df Wald statistic Wald p value Intercept 3 0526 0 1691 6544 18 06 0 0001 DBH 0 4373 0 01276 6544 34 27 0 0001 BAL 0 05688 0 003066 6544 18 56 0 0001 RelDiamInc 14 7793 2 5794 6544 5 73 0 0001 Other conifers Parameter Estimate s e df Wald statistic Wald p value Intercept 4 3636 0 1090 21239 40 02 0 0001 DBH 0 8384 0 01447 21239 57 95 0 0001 BAL 0 05970 0 002078 21239 28 72 0 0001 RelDiam ne 29 2957 1 0322 21239 28 38 0 0001 Pine cohort Parameter Estimate 8 e df Wald statistic Wald p value Intercept 2 39
69. deadlog worksheet on instillation disc CARBWARE_soil Soils are sub categories according to soil types In the example database there are 7 categories Soil_EF_category Soils are sub categories according to soils emission factors applied to different soils groups In the example database there are 3 categories mineral organo mineral and peat soils Drainage pre defined 100 excessive drainage 200 well drained 300 moderate to well drained 400 imperfectly drained 500 poorly 600 very poorly 700 undefined limestone pavement These are used for N20 emission calculations Soil_depth_cm Soil depth in m Peat_depth_cm Soil depth in m is used to calculate the emissions from organo mineral soils In the example database soils with a peat depth less than 60 cm are defined as organo mineral and the emission factor is reduced proportionally according to peat depth EF_soil this is the emission factor in t C per ha positive for emissions or soil sequestration for mineral soils negative for sequestration if available The current models does not apply emissions or reductions for mineral soils because current science suggest there is no significant change over a 20 year transition 24 Equilibrium time This is the transition time for the application of soil EF not functional in this version Net_Soil_DCtn_1 EF x the area of the plot x Peat_depth_cm 60 only if peat depth is lt 60 cm MineralNO2_EF Not included in current version Organic_poor
70. dernessRatioCa O Trees Specific NFI data tegory not currently used AgeClass5_60 O Trees Specific NFI data not currently used DBH_cm DBH_mm 10 M Trees Numeric 1 decimal place BranchinessJR O Trees Specific NFI data not currently used RotationTypeJR O Trees Specific NFI data not currently used ThinStatusJR O Trees Specific NFI data not currently used GrowthStageJR O Trees Specific NFI data not currently used DiameterClass1 O Trees Specific NFI data not currently used AgeCalss15_25 O Trees Specific NFI data not currently used AgeClass10_60JR O Trees Specific NFI data not currently used AboveBiom_t O Trees Specific NFI data not currently used Adjust_AboveBiom_t O Trees Specific NFI data not currently used Repre_AboveBiom_t O Trees Specific NFI data not currently used BelowBiom_t O Trees Specific NFI data not currently used Adjust_BelowBiom_t O Trees Specific NFI data not currently used Repre_BelowBiom_t O Trees Specific NFI data not currently used TotalBiom_t O Trees Specific NFI data not currently used Adjust_TotalBiom_t O Trees Specific NFI data 17 Data Description Requirement Table Units Format Repre_TotalBiom_t O Trees Specific NFI data not currently used AboveCarbon_t O Trees Specific NFI data not currently used Adjust_AboveCarbo O Trees Specific NFI data nt not currently used Repre_AboveCarbo O Trees Specific NFI data nt not currently used BelowCarb
71. dilGrowth CarbwareTo_0 3 01 11 2011 15 56 36 Error Subscript out of range in line 190 in Sub mdlGrowth CarbwareTo OC 01 11 2011 15 56 36 Error Subscript out of range in line 230 in Sub mdlGrowth carteaceTo OO 01 11 2011 15 56 36 Error Obj variable or with block variable not set in line 240 in Sub mdlGrowth CarbwareTo_i Error variable or with block variable not set line in Sub mdlGrowth CarbwareTo_ Error Error Error Error Error Error Error Error Error Error Error Error Error Error Error Error Error 01 11 72 O1 11 2 O1 11 2 O1 11 2 O1 11 2 O1 11 2 O1 11 2 01 11 27 01 11 27 variable or with block variable not set line in Sub diGrowth CarbwareTo_ 2 01 11 2 l 0 o variable or block variable not set in line in Sub mdlGrowth CarbwareTo_0 variable or block variable not set line in Sub mdlGrowth CarbwareTo 0 variable or block variable not set line 300 in Sub adiGrowth CarbwareTo_0 variable or with block variable not set line in Sub sdlGrowth CarbwareTo_0 variable or block variable not set line in Sub sdlGrowth CarbwareTo_0 variable or block variable not set line in Sub mdlGrowth CarbwareTo_0 variable or block variable not set line in Sub md Growth CarbwareTo_0 variable or block variable not set line in Sub dlGrowth CarbwareTo_0 variable or with block variable not set line in Sub mdlGrowth CarbwareTo_0 variable or block variable not set line in Sub d Growth CarbwareTo_0 variable or block vari
72. e Project e To select and open a saved CARBWARE project click on the Choose Project button in the Main Menu screen 38 Carbware Database in use C Documents and Settings Projects Carbon Project Co2Allocation Database About Carbware Select Database Co2 Allocation Pre Processing Co2 Reporting Close Program New Project Accept Cancel Delete Note All previously saved CARBWARE projects that have not yet been deleted will be listed on the right hand side of this screen e To return to the Main Menu screen without selecting a CARBWARE project click on the Cancel button To select a CARBWARE project click on one of the Project Names listed on the right hand side of this screen and then click on the Accept button The Project Parameters Growth Simulation screen will open and the saved parameters associated with the selected CARBWARE project will be displayed e You are now ready to perform annual Growth Simulation and Stand Modification routines on your selected inventory database see Section 7 Note To delete an existing CARBWARE project listed on the right hand side of this screen click on the project name it will become highlighted and then click on the Delete button 7 Growth Simulation 7 1 Running Annual Growth Cycles e Following guidelines outlined in Section 6 select and open the required CARBWARE project 39 You will see the following Project Parameters screen
73. e of plot area BAL A function for each plot that M ha takes as its argument any tree s rank in the diameter distribution ordered from smallest to largest and returns the combined basal area of all trees with higher rank BA Plot basal area M ha Annualised DBH t 1 DBH t t 1 t cm diameter increment DBH t stands for DBH on the Dinc occasion of the t measurement Since measurment intervals vary this implies that t 1 t 1 is not necessarily true hence the use of the term annualised Open grown crown width cw is an intermediary varible in the calculation CCF We estimated cw using equations derived by Hasenauer 1997 These equations return open grown crown width in metres Hasenauer 1997 derived species specific equations that we apply in approximation to cohorts Spruce cw exp 0 3232 DBH Other conifers cw exp 0 092 DBH Pine cw exp 0 1797 DBH Larch cw exp 0 3396 DBH Slow growing broadleaves cw exp 0 3973 DBH Fast growing broadleaves cw exp 0 1366 DBH where a circumflex denotes exponentiation Open grown crown area m 0 25 3 141593 cw NFI and PSP datasets differed primarily in the fact that PSP plots were fully enumerated whereas NFI plots were sampled The sampling method in conjunction with an assumption of homogeneous spatial d
74. e processing module functions that process standardised inventory data into a format that can be used by the CARBWARE growth and modification models The user must repeat a routine of listing and then populating the Small Trees within each selected plot This is done as follows Select a Plot using the Drop Down Box In this example Plot 54 is selected Click on the List Small Trees button All small trees within Plot 54 will be listed In this example there are 4 small trees 30 Pre Processor Database in use C Documents and Settings Projects Carbon_Project Carbwini2005 mdb x Database Get Plots with Small Trees 54 Populate Small Trees Trees table Info IndividualTrees table Info Fill Carbware table mMM Close Pre Process Window Click on the Populate Small Trees button All small trees within the selected plot will be added to the Trees Table within your inventory database In this example there are 4 small trees within Plot 54 and 4 new records will be added to the Trees Table Note Results of this pre processing routine can be seen in the Trees Table within your inventory database in the form of new records In this example see below the processing of small trees in plot 54 has resulted in 4 new records being created in the Trees Table Newly created records for small trees will always have an ID beginning with 9001 and CARBWARE will always be assigned to the Edit_user field to identify the fact that the Tre
75. eatments The advantage of using a single tree growth model and the nature of the parameterisation data set is that different silvicultural regimes and species mixtures can be handled by one generalised modelling framework In addition the application data set i e the data from which models will be run does not contain explicit complete longitudinal data representing stand variable which are used in conventional growth models Data operations Two datasets are referred to Coillte permanent sample plot PSP and NFI Some of the data operations referred to below differ between these because the former has complete enumeration on a plot and is longitudinal the latter samples from the plot and is cross sectional In general the modelling framework that we base our work on PrognAus see various references below informed the types of data operations required The framework involves using their terminology site competition and size variables Our focus was on the latter variables and site or plot effects were accounted for using mixed model methods whereby plot or site effects are random blocking effects rather than effects whose levels have physical dimension In any case site or plot effects are not a feature of the growth simulator Furthermore incomplete enumeration of certain independent variables meant that random effects were difficult to estimate because of the sparse data We can illustrate that elsewhere but such detail is not rel
76. ed 15 09 2011 12 34 07 Project_1_Mortality Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Thining_Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 tbiTempo Date Created 15 09 2011 12 18 16 Date Modified 15 09 2011 12 18 16 R LE PP E 2 e The Clearfell Events Table contains all of the necessary information for performing annual Carbon Allocation routines on the results of annual growth and stand modification simulations specific to prescribed clearfell events The Clearfell Events Table fields are described in Table 5 below Field Code Description Unit 45 ModificationYear Year of Clearfell None EventCode Code 100 clearfell None PlotID Plot ID number None TreelD Tree ID harvested None CohortCode Cohort see cohort look up None Estim_Height Height estimated m DBH Diameter at breast height cm Age Age of harvested tree years ExpansionFactor Expansion factor for partial sample plot None DeadTree tree status 100 alive 200 dead None AG Aboveground biomass kg C TB Total biomass kg C NB Needle leaf biomass kg C LTR Harvest stumpage residue kg C SB Stem biomass kg C RB Root biomass kg C LT lop and top kg C representative AG biomass AG Exp AB_C factor kg C SB_C representative SB biomass SB Exp factor kg C RB_C representative RB biomass RB Exp factor kg
77. ees Specific NFI data not currently used HDiamter03h_m Height in m at Diamter03h Trees Specific NFI data not currently used DeadTree Tree dead or alive O O M Trees Numeric pre defined 100 for alive 200 for recently dead 300 for dead longer than one year TreeBreak StemRot Tree break Tree breakage Trees Trees User defined Numeric lookup table describing stem break User defined Numeric lookup table describing stem rot ForkTree Tree forking Trees User defined Numeric lookup table describing forked trees SocialStatus Dominance suppression etc Trees Specific NFI data not currently used IUFROheight Trees Specific NFI data not currently used IUFROvitality Trees Specific NFI data not currently used IUFROgrowth Trees Specific NFI data not currently used CrownShape Trees Specific NFI data not currently used StemStraightness Trees Specific NFI data not currently used Pruning Trees Specific NFI data not currently used Branchiness Trees Specific NFI data not currently used Shapping Trees Specific NFI data not currently used StemDamageDegre e Trees Specific NFI data not currently used StemDamageAge Trees Specific NFI data not currently used PeelingDegree Trees Specific NFI data not currently used PeelingAge Trees Specific NF
78. ended User The software is designed for users interested in assessing stand level or regional carbon stock changes in forests for Ireland and the UK It is assumed that the user has grounding in forest inventory and mensuration and a good understanding of database management particularly using MS access The software is very easy to run but a good knowledge of your specific inventory procedure and measurements taken is required to compile input data base files The primary target user for the current version of CARBWARE is the Irish national forest inventory NFI staff or national inventory compilers for reporting of forest emissions and removal to the UNFCCC 2 2 Hardware requirements The software is designed to operate on all current Microsoft Windows operating system versions XP Vista and Windows 7 which have Microsoft Access installed The minimum hard disk space requirements will depend on the size of the forest inventory databases used For example if the current NFI database is run for 15 years the required hard disk space is approximately 100 MB 2 3 Installing Carbware on your PC You will receive your version of Carbware on an installation CD e Using Windows Explorer double click the Carbware Setup Launcher Icon Carbware 1324 Install Setup Launcher 1 Carbware A Important For users with Microsoft Windows Vista or Windows 7 operating systems right click the Carbware Setup Launcher Icon and select Run as Administrator This
79. enen Benchmarking state of the art classi_cation algorithms for credit scoring Journal of the Operational Research Society 54 627 635 2003 C Bigler and H Bugmann Predicting the time of tree death using dendrochronological data Ecological Applications 14 3 902 914 June 2004 L Broad and T Lynch Growth models for Sitka spruce in Ireland Irish Forestry 63 1 2 2006a L Broad and T Lynch Panel data validation using cross sectional methods Irish Forestry 63 1 2 2006b 99 E R DeLong D M DeLong and D L Clarke Pearson Comparing the areas under two or more correlated receiver operating characteristic curves A nonparametric approach Biometrics 44 837 845 1988 T Fawcett An introduction to ROC analysis Pattern Recognition Letters 27 861 874 2006 P J Heagerty T Lumley and M S Pepe Time dependent ROC curves for censored survival data and a diagnostic marker Biometrics 56 337 344 2000 S A Macskassy F Provost and S Rosset ROC con_dence bands Anempirical evaluation In Proceedings of the 22nd International Conference on Machine Learning Bonn Germany 2005 P Martin Davila J Fortun C Gutierrez P Marti Belda A Candelas A Honrubia R Barcena A Martinez A Puente E de Vicente and S Moreno Analysis of a quantitative PCR assay for CMV infection in liver transplant recipients an intent to find the optimal cut off value Journal of Clinical Virology 33 138 144 2005 S Mason and
80. equired The aggregation fields county and ownership are defined in the Lookup table in the project data base e g Carbwini2005 o All report all fields aggregated so country and ownership categories are not sub categorised o County The default values for county are shown in the county lookup xls file on the instillation disc The user can modify this lookup value according to level of aggregation required This does not necessarily have to be counties although the field header will not change if other categories are defined Selection of the county category will provide aggregated values for forest categories and soil types as specified in the Lookup table for each county o Ownership The default values for ownership are shown in the ownership lookup xls file on the instillation disc The user can modify this lookup value according to level of aggregation required In this example ALL forest and soil categores are agrregated into repective carbon pools see screen shot below P n s Propart Scale Up Value pas z Alorested Area Fire Area l m z kok a i Run Report Show Report y Cancel and Exit Reporting J 10 2 Scale Up Value Option and Afforested amp Fire Area Inputs This function is included to adjust for scaling up of plot information to the final forest area or reported reported Using Carbwini2005 and Project_1 as an example the number of plots representing 400 ha per plot i e plots at a 2 x 2 km samp
81. eric 1 decimal removed point Note plot will not be thinned or clear felled if basal areas exceeds standing basal area at time of harvest 4 1 1 Required Inventory Database Structure amp File Naming Conventions Carbware functionality is strongly dependent on good inventory data The sampling strategy used will depend on national or regional circumstances so CARBWARE has been designed to facilitate such differences as long as the input field are presented in a consistent manner The initialisation process requires 3 major tables in an MS Access database named with the prefix Carbw and suffix must be the year the inventory data was recorded e g Carbwini_20XX the characters in bold are mandatory It is also important that the table field headers are identical to the examples provided otherwise call up functions based on field header titles can not be executed Note if a header is missing or incorrectly entered a warning and error file will be created outlining the nature of the error Data formats All tables in the Access database should not be linked but imported into Access The table and field names and data formats must be adhered to 20 4 1 1 1 The Trees table specific data on mandatory fields This table contains information pertaining to individual trees with a diameter at breast height i e 130cm of 7cm or greater The Trees table should contain the following headers with specified information als
82. es 126 167 178 103 Black KG Farrell EP eds 2006 Carbon Sequestration in Irish Forest Ecosystems Council for Forest Research and Development COFORD Dublin ISBN 1 902696 48 4 Black KG 2008 Scaling up from the stand to regional level an analysis based on the major forest species in Ireland Conference proceedings IIASA GHG Uncertianty workshop Vienna Sept 2008 Black KG O Brien P Redmond J Barrett F and Twomey M 2009a The extent of peatland afforestation in Ireland Irish Forestry 65 61 71 Black KG Byrne KA Mencuccini M Tobin B Nieuwenhuis M Reidy B Bolger T Saiz G Green C Farrell EP and Osborne B 2009b Carbon stock and stock changes across a Sitka spruce chronosequence on surface water gley soils Forestry 85 3 255 271 Black K Hendrick E Gallagher G Farrington P 2012 Establishment of Irelands projected reference level for Forest Management for the period 2013 2020 under Article 3 4 of the Kyoto Protocol Irish Forestry 69 in press Duffy B Hyde E Hanley P O Brien J Ponzi and K Black 2011 National inventory report Greenhouse gas emissions 1990 2008 Reported to the united nations Framework convention On climate change EPA Dublin Edwards PN Christie JM 1981 Yield models for forest management Forestry Commission Booklet No 48 HMSO London Fealy R Loftus M and Meehan G 2006 EPA Soil and Subsoil Mapping Project Summary Methodology Description for Subsoils
83. es Table has been modified by the CARBWARE pre processing module E Trees 2005 13 21 37 SYSTEM 54 3 248 2 224 7 2 54 2 993 5 878 6 2 1 2005 13 20 16 SYSTEM 54 2 196 1 282 9 2 1 2005 12 55 33 SYSTEM 54 0 249 1 608 1 2 2005 12 46 42 SYSTEM 54 0 136 1 739 2 5 2005 13 23 07 SYSTEM 54 0 0 900101 2 3 2011 17 02 24 CARBWARE 54 0 0 900102 2 2011 17 02 24 CARBWARE 54 0 0 900103 2 1 2011 17 02 24 CARBWARE 54 0 0 900104 2 2011 17 02 24 CARBWARE 54 0 989 4 342 5 2 1 2005 13 03 09 SYSTEM 54 1 735 2 019 2 2 2005 13 17 04 SYSTEM 54 2 244 6 2 4 2 2005 13 02 49 SYSTEM 54 5 527 3 858 3 2 1 2005 13 18 56 SYSTEM 60 2 54 0 708 3 2 1 2005 11 15 16 SYSTEM MR 60 1 999 1 345 1 2 2005 11 12 30 SYSTEM J Record M 460f2712 gt M H amp NoFilter Search 4 ect As each plot with small trees is successfully pre processed it will be removed from the drop down listing of inventory plots with small trees The user must repeat the routine of listing and then populating the Small Trees within each plot until all inventory plots with small trees have been 31 pre processed and no more plots are listed The inventory database in now ready to create the Carbware Table which will form the basis of all growth simulation stand modification and carbon allocation routines to be performed on the inventory database see Section 5 3 5 3 Creating the Carbware Table Once all Small Trees have been populated into the Trees Table see
84. eshold space can a priori exhibit a number of different shapes when presented as a trade off curve For example the class ROC trade off curve has a prioria sense in which a classifier is good or bad This is when the majority of the ROC curve lies below the line of equality However the precision recall curve is not so easily understood We know that the best classifier from a group is the one with the largest area between the curve and the line of equality However because the value of the precision at zero threshold is a function of the number of objects in each class to be classified it is possible to have a good classifier for which that area is zero However such a classifier is probably not statistically better than the naive 50 50 classifier We propose that for a classifier to be demonstrably better than the naive classifier it should at the minimum describe a positive region between the curve and the line of equality We conclude that the precision recall curve does not describe a trade off and that in fact a trade off should have a point of equilibrium and the gains and losses should be incurred when the threshold moves from that point in either direction In other words the gains and losses as quantified by the two performance measures should be negatively correlated for the parameterised graph to truly describe a trade off The precision recall performance measures for example are positively correlated both have TP in
85. evant to the CARBWARE software manual The variables described here are those that feature in the diameter increment model that we aim towards calibrating Dinc cm exp ap a INDBH a2DBH a3InCR a4 InCCF as BAL See Table 3 and the text for explanation of symbols Data operations were concerned with assembling datasets of the variables used in the growth model insofar as was feasible Below we describe any substantive data operations that were performed on the variables of interest We exclude from this description any operations related to data cleaning The main data cleaning result was to omit negative diameter increments from the dataset Such omissions were made after such derived variables as BAL BA and plot density were calculated That decision was based on the fact that the omission did not have a significant impact on the results which suggested that no further modelling was necessary to compensate for the omission Also if the trees involved were omitted prior to the calculation of derived variables those variables would have been subject to an even greater bias Table 3 Explanation of some symbols used in the text Variable Formula Scale of measurement CR Crown length height Range 0 1 DBH Diameter at 1 3 m Cm Crown competition The open grown e g if Percent factor CCF every tree had zero competitors crown area of all 78 trees in a plot expressed as a percentag
86. f 15 annual growth cycles 16 carbon allocations can be performed one for the baseline inventory year Year 0 and one 50 for each of the 15 annual growth cycles Co2 Allocation Project_1l C Documents and Settings Projects Carbon Project Carbwini200 To proceed with the Carbon Allocation process for one or more of the annual growth cycles performed during the Growth Simulation stage of your CARBWARE project select one or more of the listed growth cycles In this example two have been selected the baseline inventory year Year 0 and the first annual growth cycles Year 1 The Allocate Co2 button will become active Co2 Allocation Project_1l C Docume Settings Projects Carbon Pro Carbwini200 Click on the Allocate Co2 button CARBWARE will perform the Carbon Allocation process for each of the selected growth cycles Previously unpopulated fields relating to the six carbon pools within the project and growth cycle specific CARBWARE Tables in the IncrementsDB mdb database file see Section 7 2 2 will be populated Note You can run the Allocate Co2 routine for all or some of the listed growth cycles within your CARBWARE project If you choose to run the Allocate Co2 for a sub set of the listed growth cycles you can return to the Co2 Allocation screen to run the unallocated cycles at a later time If you attempt to run the Allocate Co2 routine for growth cycles that have already been allocated you will see the following Error Me
87. gory 1 peats 2 peaty mineral and 3 mineral soils None Drainage poor 100 to good 500 None SoilDepth_cm soil depth in cm cm PeatDepth_cm peat depth in cm cm EPA_SoilGroups NFI and Epa description not used None EF_soil emission factor for peats defualt 0 59 t Cha 1 yr 1 ADJ_EF_SOIL EF_soil PeatDepth 30 if peat depth t Cha 1 yr 1 is less than 30 cm for peaty mineral soils t C plot 1 Net_soil_ emission per plot per year yr 1 Table 12 Description of Soil Carbon Table Fields in the IncrementsDB mdb database 58 9 2 6 Harvested Wood Products Carbon Table e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_HWP In this example the project name is Project_1 and so the project specific HWP Carbon Table is named Project_1_HWP Note this just stores the HWP C form harvests specified in the Events table HWP C storage is not included in the current version of CARBWARE Field Name Decription Unit PlotID Plot ID None Area_ha Area of Plot defualt 0 05 ha County County lookup ID user defined None Ownership Ownership lookup ID user defined None LandUseField Land use lookup ID user defined None Soil_EF_Field Soil EF ID lookup ID user defined None Forest category lookup ID user Forest_Category_ID defined None Harv_Thin01 Harvested C from thinnings t1 tC plot Harv_clearfell01 Harvested C f
88. h growth iteration The drivers for modification of individual trees or plot variables are derived from input tables tb Events see Figure 2 amp section 4 which control the number of trees removed due to harvest or mortality For thinnings trees are randomly removed in until the threshold basal area is reached The virtual removal of individual trees from a list trees in the corresponding plot is controlled by a defined harvest basal area in a given year Note if the same procedure is repeated it may remove different trees if a thinning event or mortality is repeated for the same plot The mortality of trees during a given growth cycle is derived from a probabilistic model see appendix D The probability threshold values can be defined by the user in the P_death_threshold table located in the installed Carbware_parameters data base However the user should refer to appendix D to understand the consequence of adjusting the default P value setting for each species cohort The outputs from the growth and modification modules are stored in the IncrementDB For each project assume file name is Project every growth and stand modification cycle is assigned a table e g Project_t0 or Project t1 These intermediate tables are used for the next iteration if another growth cycle is prompted by the user or an allocation procedure is prompted The IncrementDB also writes tables which summarise any modification events such as thinnings clearfell or m
89. he CARBWARE user sets a basal area BA to be removes as stipulated in the harvest activity data in the Eventstable so thinning of trees are selected at random from the plot until this target BA is achieved The thinned or harvested trees in a given plot are removed from the growth database and populated in a modifier table within the CarwKP_08 database These data are they called up in the allocation module Appendix E2 and section 111 2 3 Although it is common practice that clear felled stands are replanted within 2 years the CARBWARE model does no re populate clear felled plots due to uncertainty of re establishment success and species choice This is a conservative approach and is consistent with the rules applied which differentiate between deforestation and clear fell with re establishment see section 11 4 2 Appendix 1E The allocation module The total carbon stock changes for a given forest category is calculated as the sum of the changes in the above ground biomass AB below ground biomass BB Litter Li deadwood DW and soil So carbon pools Equation 2 3 in Chapter 2 of the 2006 IPCC guidelines 100 ING ENG GONG pA ING TNCs ANG EE T E 1 lu Biomass estimates include biomass for trees only non tree vegetation is assumed to be in steady state following canopy closure Below ground biomass includes all roots up to a diameter of 5cm Litter is defined as deadwood with a diameter of less than 7cm This includes a
90. he height at the base of the dead crown in meters This must be measured if total height in measured Tree_Length_m the total tree length form the stump to the highest point on the crown in m This would be equal to tree height if the tree is not slanted Crown_Length_m This is the Tree_Length_m minus the CrownBase_m Species These are codes specifying the species It is important to use the same codes and classification of species cohorts so the software can allocate a given tree in to the correct cohort where specific algorithms are applied If there is a species not present in the list use a code corresponding to the next most similar species Note The growth models are empirically derived so application of models to trees outside the Republic of Ireland or the UK will introduce model error However CARBWARE developers will parameterise models representative of local ecological and climatic zones if requested and the appropriate calibration data is available For a full list of species lookup codes and corresponding cohorts see model cohort lookup xls on instillation disc Age tree age in years The growth models use are age and distance independent bit this information is used to derive variable such as litter or dead wood pools and to allow aggregation of data in the reporting manual DeadTree These are look up code specifying if the tree id dead and how long ago it died see Table 1 for details This is used for initialising the morta
91. hed before 31 December 1989 18 Data Description Requirement Table Units Format Litter _C_tn_1 Initial Litter C pool M Lookup Numeric mandatory for each plot expressed int C can be derived form per plot area measurement or e g 0 002145 tC independent models per 0 05ha Stump_Ctn Initial dead stump C pool M Lookup Numeric mandatory for each plot expressed int C can be derived form per plot area measurement or e g 0 002145 tC independent models per 0 05ha Deadlog_Ctn Initial dead logs C pool M Lookup Numeric mandatory for each plot expressed int C can be derived form per plot area measurement or e g 0 002145 tC independent models per 0 05ha CARBWARE_soil Soil categories M Lookup Numeric look up mandatory codes user defined Soil_EF_Category Soil emission categories M Lookup Numeric look up Child layer of Siol category codes user defined i e each CARBWARE soil Default category will have a shared 1 peats or individual 2 peaty mineral Soil_EF_category soil 3 mineral soils Drainage Description of site drainage O Lookup Numeric look up codes pre defined 100 excessive drainage 200 well drained 300 moderate to well drained 400 imperfectly drained 500 poorly 600 very poorly 700 undefined limestone pavement SoilDepth_cm Mineral soil depth in cm O Lookup Numeric PeatDepth_cm Peat organic soil depth M Lookup Numeric EPA_SoilGroup EPA soil or other O Lookup Numeric look up
92. iameter distribution informs the calculation of a sampling weight or expansion factor which is used to allow for the possibility that some trees on a given plot were not sampled The expansion factor is inversely proportional to the prior probability of a given tree s inclusion in the sample based on the trees diameter class Each tree in the sample is thus duplicated by a number of times equal to its expansion factor This duplication is allowed for when calculating plot level derived variables e g Density by incorporating the expansion factor into the equations For example the estimated number of trees on a plot with a single sampled tree of 8cm is 12 62 3 See Figure 8 for an explanation 79 sub circles qualified trees Sub circle radius m Sub circle area m Treshold diameter mm Figure 8 The NFI sampling scheme at the plot level The expansion factor for a tree in the ith diameter class is R R Diameter increment The diameter increment model for each cohort was calibrated by fitting to data from the PSP dataset Dinc exp ao a InDBH a DBH a3InCR a InCCF as BAL e Where a i 1 5 are coefficients and e is a residual that was autocorrelated between measurements on the same tree and independent otherwise The fitting was done in the Glimmix procedure in SAS and the model is a GLM with Gaussian variance function and a log link This is slightly different from Monserud
93. icant PC hard disk space Remember to perform CARBWARE database management routines as outlined in Section 11 The Increments Table summarises each of the annual growth cycle DBH and Height increments for each tree within the range of inventory plots included in a CARBWARE project In the example below the inventory database has undergone two simulated annual growth cycles The Increments Table 42 Project_1_Increments shows two sets of DBH and Height data One for the first annual growth cycle Incremented_DBH_01 Incremented_Height_01 and one for the second Incremented_DBH_02 Incremented_Height_02 If additional annual growth cycles are performed for this project see Section 7 1 additional sets of DBH and Height data will be computed and recorded in this table E Project_1_Increments al ia Plot Tree SPP Age PlantYe DBH Cohor plotbe Incremented_DBH_01 Incremented_Height_01 Incremented_DBH_02 Incremented_Height_02 5 4 470 15 0 14 8 1 16 1641129261291 7 1217205223951 17 5747897915641 8 18638112960315 5 7 470 15 0 14 2 1 15 3464401231051 6 99233272999179 16 6918299741026 8 09492658202015 Li 54 900104 470 15 0 5 8 1 6 38565873920623 4 40504284991997 7 00336150176712 5 07247056428759 fai 54 900103 470 15 0 4 7 A 5 18569152853656 3 93972533257765 5 69023732477054 4 52645618062922 F 54 900102 470 15 0 5 4 1 5 93806448617095 4 23007761491416 6 51509347905535 4 87494719533302 fas 5 3 470 15
94. ing AUCs have been developed as a result e g Heagerty et al 2000 The principal complicating factor here is the underlying correlation structure of the comparison which can be influenced by details pertaining to the derivation of the classification forecasts the setup of the calibration datasets or whether the forecasts are clustered in someway e g DeLong et al 1988 Obuchowski 1997 Heagerty et al 2000 Mason and Graham 2002 88 The convex hull of a classifier or group of classifiers in ROC space can be seen as the optimal attainable classification performance Fawcett 2006 notes that candidate classifiers that do not attain the convex hull can be discarded on the grounds that a better classifier in ROC space exists He suggests a method for interpolating between candidate classifiers to better approach the limit of performance estimated by the convex hull based on mis classification costs and the prior class distribution When comparing ROC curves per se a complicating factor when it is of interest to compare different classifiers crops up if the classifiers in question are of a different class e g a probabilistic classifier versus a discrete classifier or more generally comparisons across model classes whose scoring systems are incommensurate Fawcett 2006 Datasets Permanent Sample Plot The mortality model is calibrated on data extracted from the permanent sample plot record system of Coillte Teoranta the Ir
95. ing Madang Title uera othe imp Emphas Organze s clude iniibrary Swewmthe Bum Compatb ty fes New folder J Ator Criteria Table D CarbwKPO8n 2005 D CartowkP09n_2005 Carbw10_2005 Borre CarbwK8_2005 4 v E Gare i P VS DPEN w amp Local Disk E e Open the file to display the error log 64 Fie Ect Format View Help Error in The microsoft Jet database engine cannot find the Inout table or ry Afor_Criteria Make sue Error 91 Object variable or with block variable not set in li in Sub adiGrowth CarbwareTo_0 01 11 2C Error variable or with block variable not set i 30 in Sub adlGrowth CarbwareTo_O 01 11 2C Error variable or with block variable not set 90 in Sub mdlGrowth CarbwareTo OC 01 11 20 Error variable or with block variable not set 100 in Sub mdlGrowth CarbwareTo_ 88 uae a Error variable or with block variable not set 110 in Sub mdlGrowth CarbwareTo_O 01 11 Error variable or with block variable not set ine 120 in Sub mdlGrowth CarbwareTo ot oats Error variable or with block variable ne set 120 in Sub mdlGrowth CarbwareTo_0 01 11 27 Error t variable or with block variable not set in line 120 in Sub mdlGrowth CarbwareTo_0 01 11 72 Error For Soop not initialized in line 130 in s Sub aileron CarbwareTo 00 01711 2011 15 56 36 Error Object variable or with block variable not set in line 160 in Sub mdlGrowth CarbwareTo_O 01 11 27 Error Subscript out of range in line 171 in Sub m
96. ion of the Soil Table these Carbon Allocation Tables will contain a variable number of fields depending on the number of annual growth cycles that have been simulated and allocated within your CARBWARE project The examples presented in Sections 9 2 1 to 9 2 6 below show Carbon Allocation Tables for a project with two annual growth cycles 9 2 1 Above Ground Carbon Table e The Above Ground Carbon Table within the IncrementsDB mdb database file summarises above ground carbon estimates within your inventory database plots for each of the annual growth cycles that have been simulated and allocated within your CARBWARE project e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_AG In this example the project name is Project_1 and so the project specific Above Ground Carbon Table is named Project_1_AG This table lists inventory plot level carbon estimates for the above ground carbon pool e The Above Ground Carbon Table fields are described in Table 8 below TO t1 and t2 in the table refer to the growth cycles 0 1 and 2 representing 2005 2006 and 2007 for Carbwini2005 If the Carbwxxx file is for 2009 then tO will be 2009 53 Field Name Description Units PlotID Plot ID None Area_ha Area of Plot default 0 05 ha County County lookup ID user defined None Ownership
97. iption of Deadwood Carbon Table Fields in the IncrementsDB mdb database 56 9 2 4 Litter Carbon Table e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_Litter In this example the project name is Project_1 and so the project specific Litter Carbon Table is named Project_1_Litter This table lists inventory plot level carbon estimates for the litter carbon pool all wood with diameter less than 7cm Field Name Description Units PlotID Plot ID None Area_ha Area of Plot default 0 05 ha County County lookup ID user defined None Ownership Ownership lookup ID user defined None LandUseField Land use lookup ID user defined None Soil_EF_Field Soil EF ID lookup ID user defined None Forest_Category_ID Forest category lookup ID user defined None net_litter_tO Litter biomass tO from inventory or lookup model tC plot Litter_input_t01 From AG table litterfall t1 tC plot Mortality_LT_t01 Litter and branch from mortality t1 tC plot Thin_AB_t01 Litter top leaf and branch from thinning t1 tC plot CF_AB t01 Litter top leaf and branch from clearfell t2 tC plot T_Litter_in_t01 Mort LT Thin_AB CF_AB_Litter input t1 tC plot Decomposition of accumulated T_litter in t01 and T_Litter_out_t01 net_litter_t0 tC plot Net_Litter_t01 T Litter in minus T litter out t1 tC plot Litter_Stock_t01 net_Litter tO and T
98. ish Forestry Board state commercial forestry company Broad and Lynch 2006b provide details of the dataset in the context of modelling plot volume The database consists of records of many silvicultural and thinning trials These longitudinal trials were established from the 1950s onwards and were initially established as replicated and blocked experimental designs Broad and Lynch 2006a Although there are several categories of disease or mortality causes in the PSP database including Windblown Uprooted Diseased Broken and Dead we modelled only the binary response Dead Alive for the initial model In this way after derived variables basal area plot density etc were calculated only data points that could be classified as Dead Alive were kept in the calibration dataset National Forest Inventory Plot data We validated the ROC curve for the chosen model on the NFI data In the NFI sample the probability that a tree s status as dead or alive will be recorded more generally the probability that any feature of the tree is measured is a function of its diameter class at the time of survey and its distance from the centre of the plot The expansion factor concept is a weight that varies between each tree in the dataset that estimates the prior probability of the tree s inclusion in the dataset Figure 10 shows that trees of three diameter classes are only recorded if they are observed within a certain distance from the plot centre The exp
99. ish forests in prep iii Forest Research pulled tree database Brice Nicholl NRS Forest Research UK iv Brown S 2002 Measuring carbon in forests current status and future challenges Environmental Pollution 116 363 372 v Johansson T Dry matter amounts and increment in 21 to 91 year old common alder and grey alder some practical implicatons Canadian Journal of Forest Research 29 1679 1690 vi Bartelink H H Allometric relationship for biomass and leaf area of beech Fagus sylvatica L Annals of Forest Science 1997 54 p 39 50 vii Black K Tobin B Saiz G Byrne K amp Osborne B 2004 Improved estimates of biomass expansion factors for Sitka spruce Irish Forestry 61 50 65 68 Appendix 1B CARBWARE growth models and pre processing functions The NFI permanent plot sampling procedure does not sample all trees in a plot see Figure 11 4 Therefore it is not possible to derive productivity index information such as Height index or Yield class which can be used to drive conventional stand based productivity models The alternative and most statistically valid procedure adopted was the use of single tree models to simulate tree growth between NFI cycles These models can be cross validated and re parameterised once a repeat NFI cycle is completed This section discussed the development of the CARBWARE growth model from draft versions for submission to International peer reviewed Scientific Journals Pre processing
100. ith threshold based classification tools like the ROC and ROL curves and related measures and hypothesis tests Cross validation was done on a leave k out basis where the data left out was selected at random Up to twenty independent cross validation runs were performed and up to 33 of the data was left out as cross validation data for each run Other performance measures were consulted and the ROC convex hull played a role in our chosen classifier We used threshold averaging to investigate the performance of the classifier in cross validation and bootstrap scenarios We derived confidence bands for the ROC curve of the chosen classifier following the approach of Macskassy et al 2005 Note the authors have also developed techniques for point interval estimation also the reference appearing in that paper Performance measures in ROC space and their role in uncertainty analysis The AUC of the ROC curve is the estimated probability that the classifier will give a higher score to positive cases than negative cases In our application that the estimated probability of mortality is higher for dead trees than live trees We envisage that an uncertainty analysis of the forest growth model of which the mortality classifier is a component part could utilise this probability and its standard error in monte carlo simulation assessments of overall uncertainty and sensitivity The AUC is equivalent to the Mann Whitney U statistic and methods for compar
101. le in your project s inventory database i Note Stand modification events include Natural Mortality Thinning and Clearfelling It is not mandatory to run all or any stand modification events as part of your CARBWARE project You can turn events on or off using the Project Parameter controls when creating a new CARBWARE project see Section 6 1 3 e As noted in Section 7 1 results of annual growth cycles for all CARBWARE projects are stored in the database file IncrementsDB mdb which is installed on your PC as part of the CARBWARE software installation process see Section 2 3 A series of new tables are created for each of the annual growth cycles CARBWARE uses these tables to perform additional annual growth cycles see Section 7 1 and to perform carbon allocation routines see Section 9 Tables within the IncrementsDB mdb database file relating to annual growth cycles are described in Sections 7 2 1 and 7 2 2 below Tables within the IncrementsDB mdb database file relating to stand modification events within the annual growth cycles are described in Sections 8 2 below 8 2 Stand Modification Event Tables All stand modification events performed on your inventory database plots are recorded in a series of new tables stored in the IncrementsDB mdb database file These tables are described in Sections 8 2 1 to 8 2 4 below 44 8 2 1 Modified Records Table e The Modified Records Table within the IncrementsDB mdb database file list
102. lgorithms for different species cohorts based on national research information Appendix A where diameter at breast height DBH and tree height H are used as dependent variables These variables are found intermediate input tables in the IncrementDB e g Project_t0 or Project_T1 n see Figure 2 section 3 The stocking number of trees in a plot is adjusted after every growth simulation cycle using the stand modification module Figure 2 which removes trees based on natural mortality models and harvest activity data Appendix E4 Biomass carbon losses from the above ground biomass pool AC 4g were calculated based on harvest Ltimber harvest residue Lap litter fall LiF and above ground losses due to mortality Lmortag AC aB Limber Lur Lae Lpon AB nei A ets 5 Limber is calculated based on the above ground biomass removed from harvest simulated in the stand modification module Section 3 amp figure 2 The allocation algorithms for timber based on AB H or DBH were derived from national research information see Appendix A Lar includes the harvest residue representing all stems and branches with a DBH less than 7cm and litter left on site after timber is removed ES E ago et tat a AAA 6 L r reflects the transfer of carbon from the AG pool to the litter pool listed in the Project litter table in the IncrementDB This is calculated in the allocation module Section 3 Figure 2 based on nationally derived leaf
103. ling grid is 146 which represents 58400 ha If the inventory has more detailed estimates of area e g remote sensing data of detailed vector data sets then a scale up adjustement can be made e f there are additional areas not included in ther inventory such as fires these can be subtracted from the afforested area In this example 100 ha 0 1 kha were detected using other data sources Note that there are no calculations for carbon stocks these have to be done using user 60 defined methods and added to the final reported after the anlaysis is complete refer to Duffy et al 2011 and the IPCC GPG 2006 for sytems used for Irish reporting of fires e f more accurante data sources suggest that the scaled up area is 58412 ha for example enter the new areas in Kha under the Aforested area 581 412 kha then the scaled up forest area should be 581 312 kha with 0 1 kha entered under fires as shown below Tick the scale up box and enter relevent viaues in clear boxes see screen shot below If the scale up area is the same and no additonal area are added do not select the Scale up Value tick box ony P SE o Scale Up Value Alorested Area Fire Area 5581 312 ona 01 Mate Run Report Show Report Cancel and Exit Reporting The report can now be generated 10 3 Running a Report amp Viewing a Report Table e Select Run Report en O Faan a Scale Up Value s 008 Alorested Area Fire Area wo 581 312 koma PI
104. lity algorithms CalcHeight_m If the inventory uses another method to calculate the height of tree which are not measured this can be entered here TreeNumber This is the number of trees represented for this entry Not to be confused with expansion factor The default is 1 unless trees are listed in the IndividualTrees table So if the tree ID is 9001 with 6 trees in the plot entered in the IndividualTress table then the TreeNumber would be 6 This value is used to validate the data when the Trees and IndividualTrees table is merged when the CARBWARE pre processing table is created ExpansionFactor the number of trees the sampled tree represents see information Box 1 If all trees in a plot are sampled and entered then the Expansion Factor is 1 BasalArea_m2 is the individual tree basal area in m2 DBH_cm DBH_mm divide by 10 RepreBasalArea_m2 is the BasalArea_m2 x ExpansionFactor 21 Box 1 Stratified sampling of trees in a plot The inventory may choose to take a stratified sample of trees within a plot to reduce time and expenses For example the Irish inventory uses the sampling approach shown in Figure 2 In this case all trees with a DBH of less than 70mm are only measured in the 3m inner sub circle Trees with a DBH of 70 to 120 mm are measured within in the second inner circle and all trees with a DBH greater than 200 mm are measured within the circular plot This sampling method in conjunction with an assumption of homogeneous s
105. llShield Wizard Completed The InstallShield Wizard has successfully installed Carbware Click Finish to exit the wizard Using Windows Explorer check that the following three files have been successfully installed in the directory C Program Files COFORD Carbware _Carbware_Param Microsoft Office Access Database IncrementDB Microsoft Office Access Database Carbware Application File Edit View Favorites Tools Help QOx x B PP search By Folders EG Address C Program Fies COFORD Carbware Folders Type Date Modified Microsoft Office Acc 07 05 2011 11 00 Application 07 05 2011 10 39 Microsoft Office Acc 07 05 2011 10 41 A Shortcut Icon will be created automatically on your PC Desktop 3 Overview of CARBWARE functionality 3 1 User quick start flow diagram The flow diagram below illustrates the sequence of events and instructions the user should follow to efficiency utilise the software Create input and control point databases Section 4 Archive database Section4 Run software Select database section 5 l Pre process database section 5 l Create new project section 6 Select plots from database Set stand modifiers Grow and modify forest stands sections 7 and 8 l Allocate Carbon to different pools section 9 Report to output tables section 10 Managing CARBWARE databases section 11 3 2 Detailed functionality The CARBWARE application run
106. model performance is best evaluated by external validation or failing that some cross validation We conduct leave k out cross validation on the Dinc calibration data MAE and RMSE are calculated for each cross validation dataset replicate External validation data was only available for the PSP DBH H model and that is discussed in another document 81 Standard error 0 5 0 0 Crown ratio Height Crown ratio Height Bias Crown ratio Height Crown ratio Height Figure 9 Within sample Precision upper panel and Bias lower panel for imputation Values are plotted for each dataset for cohorts and for models of Height and Crown ratio 82 C 2 5 a Cohort g fyb 3 larch T oc 0 ka pine g sgb 3 spruce ow prt ira 5 10 15 20 Replicate 0 24 D a e eA eean p 5 a Pr ANARAN Avo o 0 20 a GeSesesesoeeSeeeseseay g 3 larch T oc pa pine g 919 sgb Spruce 0 16 ii Por tA peeee 5 10 15 20 Replicate Figure 9 Leave k out crossvalidation results precision top and mean absolute error bottom for the Dinc model The probability of inclusion in the validation dataset is 0 33 20 cross validation replicates are displayed Discussion and conclusions The lines joining the points in Figure 8 are only included to facilitate a comparison between panels The interpolating lines in Figure 9 are indicative of variability between the different cross validation runs This
107. n suggests that mineral soils in Ireland do not represent a source of carbon emissions and therefore soil carbon stock changes are reported only for peats and peaty mineral soils The emission for peat soils given by equation 16 is based on published data from the UK Hargreaves et al 2003 as described for Land Converted to Forest Land in section 7 3 3 of this NIR but information on soil classification and peat depth available from the NFI is also taken into account CERDA 2 Ir enone N E or nee nee eres 16 L The area Aj of the 0 05 ha plots with peat soils is multiplied by 20 to scale the measurement up to 1 ha The EF gi is 4 t C ha yr for the first four years following afforestation and is zero thereafter Emissions from peaty mineral soils are calculated in the same way equation 17 but a soils depth function SD is applied to the emission factor to account for the smaller organic carbon pool available If soil depth is less than 30 cm then BG 2s ACR ESD incl al NE 17 J and EEEE ESEA EE E 18 D depth cm 30cm The soil depth is also defined by the user in the lookup tables Reference List Black K Tobin B Saiz G et al 2004 Improved estimates of biomass expansion factors for Sitka spruce Irish Forestry 61 50 65 Black KG Bolger T Davis P et al 2007 Inventory and Eddy Covariance Based Estimates of Annual Carbon Sequestration in a Sitka spruce Picea sitchensis Bong Carr Forest Ecosystem J Eur For R
108. ncentric plot sampling schema The question we have to address in the current paper is whether we can arrive at a sensible definition of representative mortality At issue is how to derive a binary individual tree level mortality rule based on information in the NFI dataset given the fact that there is missing information due to the sampling scheme With this in mind Figure 2 classifies all dead trees in the PSP database by cohort and describes the empirical distribution of diameter classes conditional on mortality status We have included the diameter class 0 7 for completeness even though there is no equivalent in the NFI dataset Note that the left hand column is very similar to the unconditional distribution of diameter classes so it does not need to be displayed On those grounds a comparison of the columns of Figure 2 shows the dramatic extent to which the chance of mortality declines if a tree does not die while in the lowest diameter class For example the global fraction of trees in the Spruce cohort in the lowest diameter class is very small but this class represents 50 of dead trees in the cohort Similarly for Pine OC and FGB The right hand column of Figure 2 at least for the cohorts with enough observations suggests a way to make the operation of a binary mortality rule more accurate in the context of the NFI sampling scheme The basic idea would be to use the column heights as weights in a finite mixture function whose componen
109. ned in the Lookup table Area within in each category this may be adjusted if Scale kHa ADJ Area Up Value was selected Above ground biomass gain Gg C Note positive is an uptake negative a AG gain loss of C AG net Net above ground biomass loss or gain Gg C BG gain Below ground biomass gain Gg C BG loss Below ground biomass loss Gg C BG net Net below round loss or gain Gg C Litter net gain or loss note includes all litter and wood with Gg C Litter Net diameter less than 7 cm Deadwood net gain or loss note includes all wood with Gg C DW Net diameter greater than 7 cm Soil net gain or loss note default values mineral soils are Gg C Soil Net assumed to be zero but an EF is applied all organic soils Net Gg C Net gain or loss of C from all pools Gg C Net Gg CO Net gain or loss of CO2 Gg CO 62 11 Managing Carbware Databases The IncrementDB stores all intermediate and reported data files These can accumulate and take up a lot of disc space The data can be deleted and databases should be compacted and repaired to avoid accumulation of unwanted data or creation of large memory space requirements for the database 11 1 Deleting projects If you have created an unwanted project all data from these can be deleted from the Assess database e Open Carbware e Select Choose project e Select the project you wish to delete e g Project2 below W Project seiecction 2005val33 New Project KP2010h KP2010n KPO9aft KPOBalt KPO7alt C
110. nual growth cycles are described in Section 8 Project_1_01 Date Created 15 09 2011 12 33 21 Date Modified 15 09 2011 12 33 21 Project_1_Clearfell_Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Increments Date Created 15 09 2011 12 33 19 Date Modified 15 09 2011 12 33 21 Project_i_ModifRecords Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Mortality_Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Thining_Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 tbiTempo Date Created 15 09 2011 12 18 16 Date Modified 15 09 2011 12 18 16 7 2 1 Summary of Annual Growth Cycle Increments IncrementDB Table An Increments Table will be created for each CARBWARE project that has undergone one or more annual growth cycles The Increments Table is created and named by CARBWARE using the naming convention Project Name_Increments and is stored in the database file IncrementsDB mdb In this example the project name is Project_1 and so the Increments Table is named Project_1_Increments A Important If you have created several CARBWARE projects and have run one or more annual growth cycles for each of them there will be several Increments Tables within your IncrementsDB mdb database and named according to your CARBWARE project names These tables can be relatively large and may use up signif
111. o see Table 1 a b a0 LI IDPlots the plot number ID Tree identification number These should be unique starting at 1 not necessarily in numerical order If there are trees within the plot with a DBH less than 7cm then these should be recorded in the IndividualTrees table However all of these trees should also be identified in the Trees table The Tree table ID should be 900x where x is the tree ID in the IndividualTrees table e g if the tree ID in the IndividualTrees table is 1 then the ID in the Trees tae should be 9001 Edit_date the date and time the data was entered or modified e g 16 07 2005 12 34 56 Edit_user user who input or changes the data SYSTEM should be used as a default DBH_mm DBH at 130cm in mm for all trees in the Trees table The growth models have been parameterised using DBH at 130 cm if other conventions are used then this will create bias in the projected C stock changes The software does facilitate a partial sampling system but if a tree is identified in the Trees table the DBH and corresponding data must be entered Height_m the corresponding tree height in m It is not necessary to provide the height of all trees when DBH is measured However a complete list of tree height will reduce modelling error since the software runs algorithms to calculate missing height data CrownBase_m The height at the base of the living crown in meters This must me measured if total height in measured DeadCrBase_m T
112. o at the determined age n 1 The age of the forest n is obtained from the NFI stand attribute data The partial coefficients a for each species and productivity class and goodness of f 84 Once the new mean tree H xH xHinc 1 is computed the individual tree H is recalculated based on a scaling function H H e eH ed ae i ea a E A Oh ie ub Dalal med E sel ct a 2 xH where H 7is the individual H of the tree in the plot in the year following the NFI Hn is the individual H in the year the last NFI was completed 2005 and xH is the mean H of trees in the plot in the year the last NFI was completed The Productivity class H over age categories were defined to match conventional yield class productivity indices YCeq as described by Christy and Edwards 1981 This was derived by comparison of Chapman Richard outputs from each H index ratio HI with static age H tables at ca 10 to 20 year old crops YCeq HI min YCH xH where YC eq is the HI equivalent to YC at the lowest least squares different between the yield table H values YCH and the predicted mean height xH see equation 1 for the th cohort and jth HI Selection of tree for H increment model All trees with no measurable DBH are selected for growth increment using the H model The CARBWARE model also selects eligible trees to be grown using the H growth model based on cohort specific threshold DBH values Table 4 These are derived from analysis of
113. of additional Carbware Table Fields in the IncrementsDB mdb database e Project_1_01 represents an updated version of the project s Carbware Table Project_1_00 following one annual growth cycle simulation including Height amp DBH growth and the results of any stand modification events prescribed by the tblEvent Table located in the selected inventory database This is now the baseline data used by the CARBWARE model for growth simulation and stand modification event simulation from Year 1 the inventory year 1 to Year 2 the inventory year 2 etc For any project there can be up to 16 intermediate Carbware Tables Year 0 to Year 15 Note Any records trees that have been removed due to stand modification events during the first annual growth cycle simulation will no longer be listed in this table Instead they will be listed in one of the Stand Modification Events tables See Section 8 e As well as being used for annual growth cycle simulations these intermediate Carbware Tables are used for Carbon Allocation purposes see Section 9 8 Stand Modification 8 1 Stand Modification Events during Annual Growth Cycles e As described in Sections 4 1 1 4 and 6 1 3 a series of prescribed stand modification routines can be performed on your inventory database plots during annual growth cycles determined by your project s parameter settings see Section 6 1 and the modification events Thinning amp Clearfell prescribed in the tblEvent tab
114. on Allocation Process ccccsececeseeeeeeeeeareceeeeeeeeeaeeeeaeesaaeesaeesaeeesieeesieessineeeaeess 50 9 2 Carbom Allocation Tables ormer antais pe yea iedas EEAS alee pence A EAEAN are AEE a REE dara 53 9 2 1 Above Ground Carbon Table ices reniri re ire ni AE EAE AEE EEA iE 53 92 2 Below Ground Carbon Tables siiperi ian e E E A A E E E E 55 9 2 3 Deadwood Carbon Table ee ereraa EREN ELA ERAEN S EAE RANEREN OENE EE ead 56 9 2 4 Litter Carbon Table eirasiaea iaa a ia aa aaia ea i aa 57 920 Soll Carbon Tabla es aaan SS en a E ERA E aS A E A S E E 58 9 2 6 Harvested Wood Products Carbon Table 0 cccceeceseeeeeeeeeeeeeeeeeeeseeeeseneeseeeseeeeseeeeseneeseneeeenees 59 10 The Reporting Module isinsin a a a E a E A a a 60 10 1 Selecting the Project reporting year and level of aggregation cccesseeeeeeeseeeeneeeeneeeenaeeeaeeees 60 10 2 Scale Up Value Option and Afforested amp Fire Area INputs cccecceceeeeeeeeeeeceeeeeeeeeeeneeeeeenaeeees 60 10 3 Running a Report amp Viewing a Report Table cceecceeseeeeeeeeeeeeeeeeeeeeeeeeeeeeeeaeeeeaeseeaeeeeaeeeeaeeeeaeeeas 61 11 Managing Carbware Databases 0 00 0 ecececeeeneeeneeeeeeeeeaeeeeaeeeaaeeeeaeeeeaeeeaaeeeaaeeeaeeeeaaeseaaeesaeeeeaeseaeeenaeeeas 63 11 1 DOlEtING PROjSCtS E fais taal tees eteanettecedagda la dadagees A E peanehlageuag EA A 63 11 2 Compacting and reparing databases eeceecseeeeeeeeeeeceeeeeeeeeeeeeeeeeeeecaeeceaeeeeaeeeeaee
115. on_t O Trees Specific NFI data not currently used Adjust_BelowCarbon O Trees Specific NFI data _t not currently used Repre_BelowCarbon O Trees Specific NFI data _t not currently used TotalCarbon_t O Trees Specific NFI data not currently used Adjust_TotalCarbon O Trees Specific NFI data _t not currently used Repre_TotalCarbon_ O Trees Specific NFI data t not currently used Diameter _mm Diameter at 1 3m height M Individual Numeric in mm no mandatory if tree is gt 1 3 m Trees decimal points high Origin NFI edit code O Individual Specific NFI data trees not currently used NFI_ID Plot ID the same as ID plots M Afor_Criteria Numeric must be identical numbering Kyoto_ID Forest classification Article M Afor_Criteria Numeric either 3 3 3 3 or 3 4 forests or 3 4 PlotID Plot ID cross referenced to M Lookup Numeric NFI_ID and IDplots Area_ha Area of plot M Lookup Numeric in ha to 2 decimal points County County where plot is located M Lookup Numeric user defied lookup codes Ownership Ownership of forest M Lookup Numeric look up codes user defined Landusefield Landusetype O Lookup Initial dead stump optional C pool mandatory for each plot can be derived form measurement or independent models Forest_category_ID Forest type M Lookup Numeric look up code user defined used to aggregate forest types gt Article 3 3 forests are those established after 31 December 1989 and article 3 4 forests are establis
116. ortality of trees during a growth modification cycle e g Project_Thinning_table The field names and units of values of individual tables are reviewed in sections 7 and 8 3 CO allocation and stratification the allocation module uses the intermediate growth output tables described above Figure 2 to generate carbon stock estimates for the 6 major carbon pools The allocation module performs the following functions i Converts tree measurements DBH and H to biomass components see Appendix A ii Sums all tree biomass components in the plot and normalises the units to 1 ha This is done using the user specified plot sizes located in the lookup table located in the Carbw mdb database 10 iii Allocates the transfer of carbon between the different 6 pools to simulate the flow between biomass AG and BG litter dead wood harvested and soil pools The details of the C flow model are shown in appendix E These outputs are written to the Increment DB using the project name as a prefix iv Some of the initial C pools at the plot level have to be input by the user the lookup table located in the Carbw mdb database These include initial litter and deadwood inputs in addition to the soils emission factors see Section 4 v The biomass stock changes for aboveground Project_AG table and belowground biomass Project _BG are derived from the DBH and H values of individual trees using biomass algorithms Appendix A The biomass pools are
117. ositive predictons Cutoff b hi correlation coeficient 0 04 02 Figure 16 Illustrating some other performance measures of the NFI calibrated model for the Larch cohort across the cut off range and in particular the 0 01 green vertical and 0 001 blue vertical cut off points Discussion In binary classification a common approach is to visualise the parameterised curve described by plotting two performance measures as a parametric curve parameterised by the threshold value Comparing models based on classification and mis classification rate precision recall etc make more sense when there is some hierarchy of misclassification errors That is that we can quantify the relative importance of gains from correct classification and losses from incorrect classification Such a loss function is particularly useful when the number of objects to be classified is not equal because then the trade off curves are much more likely to be nonlinear and the concept of trade off between competing performance measures is not easy to understand The problem is how to specify losses gains in other words how to quantify Trade off how to measure gains and losses in the same units so a net trade off can be calculated Otherwise it is not always clear even for commonly presented parameterised curves in what sense the trade off is occurring particularly when a good classifier e g 98 one that exhibits desirable tendencies in thr
118. own that it is possible to derive a generalised model that performs well and which by its nature deals with the data sparseness issue by estimating the typical parameter value and modifying this value as a function of the plot and tree level characteristics The BIC results and the graphical results suggest that the inclusion of covariates in the model improves the DBH H model i e Model 2 as was shown by Temesgen and von Gadow The inclusion of covariates in the model is a move away from the baseline model which is a generalised approach that presumes that competition as measured on the scale of the plot by DENS and BA and on the scale of the tree by BAL does not affect the allometric relationship between DBH and H over the tree s lifetime when subjected to different competition pressure introduced by spacing or thinning In the next section we address the issue of generalised vs plot specific modelling However our results at this point suggest that the Temesgen and von Gadow model that models plot differences through competition variables is a unified single step approach By contrast the plot specific approach can be seen as a multi step approach whereby the DBH H relationship for each subject is modelled individually and competition effects are at best implicitly described by the plot specific fitted parameters We might suspect that datasets that are heterogeneous across plots might be more accurately modelled using plot specific
119. patial diameter distribution informs the calculation of a sampling weight or expansion factor which is used to allow for the possibility that some trees on a given plot were not sampled The expansion factor is inversely proportional to the prior probability of a given tree s inclusion in the sample based on the trees diameter class Each tree in the sample is thus duplicated by a number of times equal to its expansion factor This duplication is allowed for when calculating plot level derived variables e g Density or Basal area by incorporating the expansion factor into the equations For example the estimated number of trees on a plot with a single sampled tree of 8cm is 12 62 3 17 69 Therefore 1 measured tree with a diameter less than 70mm in the 3m inner circle will represent 17 69 tree and have an Expansion factor of 17 96 Where all trees are measures e g DBH of 200mm then the Expansion factor is 1 See Figure 2 for an explanation sub circles qualified trees Sub circle radius m Sub circle area m Treshold diameter mm Figure 3 The NFI sampling scheme at the plot level The expansion factor for a tree in the ith diameter class is R3 Ri Note if all trees are measured in a plot then an expansion facto of 1 is applied For accurate projection of C balance it is advised to use a sampling regime which measures all trees in the plot since CARBWARE does not perform in growth functions This
120. performed because tables in the database are modified during pre processing and original database formats will be lost 26 5 1 Selecting the required Forest Inventory Database Before pre processing your forest inventory data using Carbware software it is important to conduct some basic file management operations in order to maintain the format of your original inventory database e Using Windows Explorer create a new project folder in a location on your hard drive where you normally store your files e g C Documents and Settings Projects Carbon_Project e Save a copy of your original MS Access inventory database file in this new directory e g C Documents and Settings Projects Carbon_Project Carbwini2005 db amp Carbon_Project File Edit View Favorites Tools Help GO sx Q T S Search e Folders E Address C Documents and Settings Projects Carbon_Project x Folders Carbwini2005_Raw_Data f Carbwini2005 A a Mi ft Office Access Datab 2 Microsoft Office Access Datab E E My Documents a iiij WY My Computer E s Local Disk C A Important Ensure that your inventory database is correctly structured and named in accordance with guidelines presented in Section 4 of this manual e Start the CARBWARE programme by double clicking the Shortcut Icon on your PC Desktop e You will see the following Main Menu screen Carbware No Database in use Project Co2 Allocation Database About Carbwa
121. r L top from harvest residues DBH diameter at breast height 1 3 m in cm H height in m HR lop and Eq Function Range Equation Coefficients Source Spruce 1 AB H gt 4 5m ax DBH cxH 0 23 2 12 5x107 4 99 0 91 0 29 1 01 i ii 2 AB H lt 4 5m axH xc 1 32 1 7 1 38 0 86 0 2 1 1 i ii 3 TB exp Ln a bx Ln AG 1 02 1 033 0 91 0 08 1 03 ii iii 4 BB TB AB 5 FB ABxa bxexp cx AB 0 025 0 089 0 003 0 68 3 4 0 98 i ii 6 SB exp Ln a bx Ln AG 0 405 1 09 0 99 2 99 1 03 ii iii 7 Lur AB SB Pines 8 AB H gt 3 8m ax DBH cxH4 0 07 2 42 0 039 2 51 0 93 0 13 0 94 ii iii 9 AB H lt 3 8m ax H 0 12 3 91 0 95 0 74 0 95 i ii 10 TB exp Ln a bx Ln AG 1 15 1 01 0 96 0 4 1 01 ii iii 4 BB TB AB 5 FB ABxa bxexp cx AB 0 025 0 089 0 003 0 68 3 4 0 98 i ii 11 SB exp Ln a bx Ln AG 0 71 1 005 0 97 0 27 0 96 ii iii 7 Lur AB SB Larch 12 AB H gt 2m ax DBH cxH4 0 11 2 31 0 001 3 29 0 94 0 27 0 94 ii iii 13 AB H lt 2m axH 0 03 1 91 0 67 0 44 1 2 i ii 14 TB exp Ln a bx Ln AG 1 43 0 98 0 99 0 25 0 99 ii iii 4 BB TB AB 66 Function Equation Coefficients Source 5 FB ABxa bxexp cx AB 0 025 0 089 0 003 0 68 3 4 0 98 i ii 15 SB exp Ln a bx Ln AG 0 903 0 972 0 98 0 28 0 96 ii iii 7 Lur AB SB Other conifers
122. r inventory database files correctly Carbware Project Parameters amp Project Name Project_ Inventory Year 2005 Select Database C Documents and Settings Projects Carbon_Project Carbwini2005 mdb Filter Baseline Inventory Database Stand Modification Save Settings by PlotID by Date Criteria F Natural Mortality e M Clear Fell c M Thinnning You have done 0 cycles Close 6 1 2 Filtering the Forest Inventory Database by Plot or Date Criteria e Before initiating Growth Simulation and Stand Modification routines you can select all or a sub set of inventory plots to work with using the Filter Baseline Inventory Database controls within the Project Parameters screen e To select specific plots click the by Plot ID control on the Project Parameters screen and highlight the required plots e To select all plots or plots within a certain age class click the by Date Criteria control on the Project Parameters screen and highlight the required date criteria These are identified from the Afor_Criteria table an essential prerequisite table in database see section 4 1 1 5 o All will select all inventory plots within the selected forest inventory database o Article 3 3 will select all trees within the selected forest inventory database that have a planting year of 1990 or younger o Article 3 4 will select all trees within the selected forest inventory database that have a planting year before 1990
123. re Co2 Allocation Pre Processing Co2 Reporting Choose Project Close Program e Click on the Select Database button and navigate your PC filing system to locate and open the required inventory database e g C Documents and Settings Projects Carbon_Project Carbwini2005 db 27 Look n CEEEMGEES J Oe 2 carbwiniz005 a Carbwini2005_Raw_Data My Computer a File name x Open My Network Files of type Fichiers Carb mdb Cancel aces I Open as read only A Important Always remember to work with a copy of your original MS Access inventory database file as the CARBWARE pre processing function will modify the tables in the database and original database formats will be lost Note Only MS Access database files named with the prefix Carbw will be available for selection see Section 4 1 1 for details e When you have selected the required database you will return to the Main Menu screen and the selected database will be active and ready for pre processing Carbware Database in use C Documents and Settings Projects Carbon Project Co2 Allocation Database About Carbware Select Database Co2 Allocation Pre Processing Co2 Reporting Choose Project Close Program 28 5 2 Populating Trees Table with Small Tree Records from IndividualTrees Table e With the required database selected see Section 5 1 click on the Pre Processing button in the Main Menu screen
124. rfall_C_t02 Litterfall t2 tC plot SB_from_thin_C_t02 Stem biomass SB from thinning t2 tC plot SB_from_cf_C_t02 Stem biomass SB from clearfell t2 tC plot HWP_t02 HWP t2 tC plot HR_t02 HR t2 tC plot AB_C_t02 minus AB_C_t01 plus mortality thinning and AB_C _gain_t02 clearfell t02 negative value represents a gain tC plot AB_C _loss_tn02 sum of AB mortality litter SB from CF and Thin t02 tC plot Net_AB_t02 sum AB_C_gain and AB _C loss tC plot 54 Table 8 Description of Above Ground Carbon Table Fields in the IncrementsDB mdb database 9 2 2 Below Ground Carbon Table e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_BG In this example the project name is Project_1 and so the project specific Below Ground Carbon Table is named Project_1_BG This table lists inventory plot level carbon estimates for the below ground carbon pool Field Name Description Units PlotID Plot ID None Area_ha Area of Plot default 0 05 ha County County lookup ID user defined None Ownership Ownership lookup ID user defined None LandUseField Land use lookup ID user defined None Soil_EF_Field Soil EF ID lookup ID user defined None Forest_Category_ID Forest category lookup ID user defined None RB_deadtree_C_t01 Dead root biomass RB time t1 tC plot RB_Living C_t01 Living RB biomass t1 tC plot Mortali
125. rfell event is noted in the ModificationYear field in this table The EventCode field notes the stand modification event type that has modified a record tree This will always be 100 Clearfell e In this example there has been a total of 142 trees clearfelled over the two annual growth cycles for Project_1 All of the events were in 2007 and 4 Plots were modified Project_1_00 Date Created 15 09 2011 12 33 19 Date Modified 15 09 2011 12 33 19 Project_1_01 Date Created 15 09 2011 12 33 21 Date Modified 15 09 2011 12 33 21 Project_1_02 E Project_1_Clearfell Table Date Created 15 09 2011 23 16 46 Date Modified 15 09 2011 23 16 46 Project_1_Clearfell_Table Date Created 15 09 2011 12 34 07 Date Modified 22 09 2011 14 36 52 ModificationYear EventCode PlotID TreelD CohortCode Estim_Height DBH i 2007 100 15 11 3292584196602 13 6651390981732 2007 100 17 11 4675395578967 14 2987290918098 2007 18 12 0106442684345 16 0461164777161 2007 13 15 5952039019465 27 8414945157853 2007 19 12 5083302122975 17 5212639820589 2007 12 15 4416309221752 26 7495018525976 2007 14 11 6095254183335 14 6652422273164 2007 900101 5 07187703415488 4 06926272493526 2007 9 14 2843070337057 22 5388803866366 ecord 4 4 Lof142 gt Wb K Search lt gt Project 1 Increments Date Created 15 09 2011 12 33 19 Date Modified 15 09 2011 23 16 46 Project_1_ModifRecords Date Created 15 09 2011 12 34 07 Date Modifi
126. roduction CARBWARE is a forest carbon flow model now made assessable as a standalone computer program CARBWARE can be used to report GHG emission reductions or project these forward by simulating changes in forest carbon pools over a 15 year time span and over a landscape spatial scope form 1 plot to 1000s of forest stands The CARBWARE model simulates the flow of carbon within forests and between forests and the atmosphere Figure 1 Carbon dynamics are simulated in aboveground belowground biomass litter dead wood and soils pools The removal of CO from the atmosphere as a function of forest growth is simulated using models parameterised for different species cohorts which exhibit similar growth characteristics Spruce Pines Larch other conifers slow growing and fast growing hardwoods It also includes mortality functions decomposition factors for the deadwood and litter pools and soil emission factors not normally captured by conventional inventory procedures It also accounts for silvicultural management such as those associated with thinning and clearfell Although the current national GHG inventory assumes that all wood products HWP are immediately emitted to the atmosphere at harvest the potential C storage in harvested wood products have not been included in the model This is because HWP C modelling is best done at the national or regional scale using industry data but stand dynamics in the CARBWARE model is simulated in a bo
127. roject_1_00 represents an adapted copy of the project s inventory database Carbware Table at time 0 prior to any annual growth cycle simulations This is the baseline data used by the CARBWARE model for growth simulation and stand modification event simulation from Year 0 the inventory year to Year 1 the inventory year 1 While this table is very similar to the project s original inventory database Carbware Table there are some additional fields required for Carbon Allocation purposes which are described in Table 4 below Field Code Description Unit CF_TH redundant AG Above ground biomass Kg C per tree TB Total biomass Kg C per tree NB Needle leaf biomass Kg C per tree LTR Litter fall rate Kg C per tree SB Stem biomass Kg C per tree RB Root biomass Kg C per tree LT Lop and top biomass Kg C per tree Adj_AB_dead Adjusted dead biomass Standing dead Exp factor Kg C Adjusted live abovegroung biomass Standing dead Kg C Adj_AB_live Exp factor Litterfall Adjusted littefall LTR Exp factor KgC SB_C Adjusted Stembiomass SB Exp factor KgC Dead_RB Adjusted dead root biomass DeadRB Exp factor Kg C RB_C Adjusted root biomass RB Exp factor Kg C 43 deadLT Dead lop and top Kg C LT_C Adjusted dead biomass Standing dead Exp factor Standing Dea Adjusted dead biomass Standing dead Exp factor d_TO Table 4 Description
128. rom clearfell t1 tC plot Tot_Harvest_tC01 Total harvest C for t1 tC plot Thin_Harvest_m301 Harvested volume from thinnings t1 m3 plot CF_Harvest_m301 Harvested volume from clearfell t1 m3 plot Harvest_m301 Total harvest volume for t1 m3 plot Harv_Thin02 Harvested C from thinnings t2 tC plot Harv_clearfell02 Harvested C from clearfell t2 tC plot Tot_Harvest_tC02 Total harvest C for t2 tC plot Thin_Harvest_m302 Harvested volume from thinnings t2 m3 plot CF_Harvest_m302 Harvested volume from clearfell t2 m3 plot Harvest_m302 Total harvest volume for t2 m3 plot Table 13 Description of HWP Carbon Table Fields in the IncrementsDB mdb database 59 10 The Reporting Module The reporting module aggregates the plot level information for each carbon pool see allocation tables in section 9 2 to a national or regional level The reporting format is designed to be interoperable with the UNFCCC common reporting tables http unfccc int national reports annex i ghg inventories reporting requirements items 2759 php 10 1 Selecting the Project reporting year and level of aggregation e Open the reporting module e Select project e g Project_1 from drop down list e Select the reporting year from drop down list see screen shot below the first year from the growth simulation is 2006 in this case e Select the reporting category from drop down list Reporting results table can be aggregated at 3 levels depending on information r
129. routines and to run annual growth simulation amp stand modification cycles for up to 15 years e Enter a Project Name e g Project_1 6 1 1 Selecting the required Forest Inventory Database e To select the required Forest Inventory Database for your new CARBWARE project click on the Select Database button in the Project Parameters screen and navigate your PC filing system to locate and open the required inventory database e g C Documents and Settings Projects Carbon_Project Carbwini2005 db Look in Carbon_Project A T Carbwini2005 TG A Carbwini2005_Raw_Data My Recent 2 Carbwinixxx2005 Documents Desktop File name Carbwini2005 My Network Files of type Fichiers Carb mdb Fiacos I Open as read only a Important You can only work with an inventory database that has been pre processed See Section 5 for details on pre processing your forest inventory data e The selected inventory database will be assigned to the CARBWARE project and you will return to the Project Parameters screen The selected inventory database will be shown on this screen below the Project Name and the Inventory Year will be displayed in the top right hand corner 36 Note Inventory Year is determined by the inventory database file naming convention The last four digits of the inventory database filename determine the Inventory Year for the selected database Refer to Section 4 1 1 for details on how to name you
130. s all records trees within your inventory database plots that have been modified by Natural Mortality Thinning or Clearfell events during annual growth cycles e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_ModifRecords In this example the project name is Project_1 and so the project specific Modified Records Table is named Project_1_ModifRecords e The CF_TH field in this table notes the stand modification event type that has modified a record tree This can be 100 Clearfell 200 Thinning or 300 Natural Mortality 8 2 2 Clearfell Events Table e The Clearfell Events Table within the IncrementsDB mdb database file lists all records trees within your inventory database plots that have been clearfelled during annual growth cycles e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_Clearfell_Table In this example the project name is Project_1 and so the project specific Clearfell Events Table is named Project_1_ Clearfell_Table This events table lists all of the trees within your inventory database that have been clearfelled during any of your CARBWARE project s annual growth cycles In this example there have been two annual growth cycles for Project_1 2005 to 2006 and 2006 to 2007 noted by the fact that there are three Carbware Tables Project_1_00 Project_1_01 and Project_1_02 The year of each clea
131. s off 3 core databases which contain input data tables represented as yellow boxes in Figure 2 intermediate output tables green boxes and archive and control point tables The user must first create the input database The nomenclature of the database must contain the prefix Carbw and be in Microsoft access format e g Carbwini2005 mdb see instillation disc The other 2 databases namely IncrementDB mdb and Carbware_Parameters mdb are downloaded when the software is installed The incrementDB stores intermediate output files green boxes Figure 2 when the growth modification and allocation modules are run after a project is set up The parameters database stores writing directories during the creation of the input Carbw database and assigns mortality probability thresholds which can be changed by the user The instructions for creating compatible input databases for the software are described in section 4 Create database Carbw mdb Litter Deadwood Soils Harvests Disturbances CARBWARE v5 1 IncrementDB mdb Projeci_i6 table Projeci_if table Projecit_tt x table Projeci_clearfell_table Project_ncrements Project_Modirecords Proyeci_mortabiy_tble Project_Sawwng_teble Other GIS data Growth simulator Stand modifier Carbware_Param mdb ThiProj Paste_esrors Project_AG table Project_BG fable Project_Siier table Project_deadwood table Project_sof table Project_HWP table Forest
132. seateseaeeeeaeeenaeeeas 63 IES ENOr OG 2tecscee castes ween teedenccnscete finedecedeceendcense a Ea A aa aa oaa aeo aa cedee tees 64 Appendix T o iase yscucceceedpeetacie cates Sheue Seven e dba cdbsedaden Sedusy an devgaSaeua dence devdesstal Seve aooiee eaaet Sarea EEEa EEEa 66 Appendix 1A Allometric biomass equations used in allocation module ccceeeeceeeeeeeeeeeeteeeees 66 Appendix 1B CARBWARE growth models and pre processing fUNCtIONS ccceeeeeeeeeeeeeeteeeeeees 69 Appendix 1C Growth modelling c cccccceceeceeececeeeceeeeeceeeeeeeaeceeeaaaeeecaeeeesenaeeeeaaeeeeseaeeeesesaeeeeeeseeeeaeeess 78 1 Modelling diameter increments in Irish Forests ecccececeeceeeeeeeeeeeeeeeeeceaeeceaeeeeaeeseaeeseaeeeeaeeseateeas 78 Il Modelling height increments for Small tre S eccceeeceseceeeeeeeeeeeeeeeeeeeeeeeeeeeeeaeeeeaeeeeaeeeeaeeetaeeeeaeeeas 84 Appendix 1D CARBWARE stand modification functions ccecccceececeeeeeeeeeeeceeeteseeeeeesaeceetenaeeteeneeees 87 Is Mortality Models a aa eth dae ae a r a r dessa ehh aai 87 Il Other modifications in the growth simulator sseessessseesreereereittstrtsttsttrsttrstrstersttnstnnsenstenetn nnn 100 Appendix 1E The allocation module ccceecceccceceeeecececeeeeeeeeaaeeeeeeseseaaeeeeessecceeceeeeesesceueeeeeessnaeees 100 Reference Listari naiai a a a a n louis A cite a a welt E adele E Baten toed creed ot 103 1 Int
133. sed for mortality probability Table 3 Description of Carbware Table Fields 6 Creating or Selecting a Carbware Project 6 1 Creating a New Carbware Project Once you have pre processed your inventory database you can now create a CARBWARE project A project sets the parameters for Growth Simulation and Stand Modification routines to be performed on your inventory database prior to running the Carbon Allocation Process see Section 9 The following parameters can be set a The Forest Inventory Database to be used This must be pre processed as per procedures outlined in Section 5 b The range of inventory plots to be analysed within the selected Inventory Database All or a sub set of inventory plots can be selected based on user defined Plot Number Selection or Date Criteria Selection c The Stand Modification routines to be performed on the selected inventory plots CARBWARE installation adj_ lookup Table Date Created 08 07 2010 1 Date Modified 26 08 2010 P_Death_Thresholds Table Date Created 24 08 2009 1 Date Modified 24 08 2009 Paste Errors Table Date Created 24 09 2008 1 Date Modified 24 09 2008 tbIProj Table Date Created 12 09 2008 1 Date Modified 07 05 2011 tbleroj ProjectNam DatabaseName FilterTy DateCrite Cycles Modif_ Modif_ Project_1 0 0 0 1 0 Modif_ Co2Allocate Note Selected parameters for each new CARB
134. ssage 51 Co2 Allocation 1 Carbon has already been allocated for the selected Cyde 1 aaa The final stage of the Carbon Allocation process is Carbon Stratification where CARBWARE stratifies the carbon estimates associated with your forest inventory database into each of the six carbon pools To proceed with the Carbon Stratification process open the Co2 Allocation screen and select the required CARBWARE project If you have performed the Carbon Allocation process for one or more of your project s growth cycles the Stratify Co2 button will be active Co2 Allocation roject Name Database path Iterations Allocate Co2 Stratify Co2 Close Co2 Allocation Note Make sure that you have performed the Carbon Allocation process for each of the project growth cycles you wish to analyse before proceeding with Carbon Stratification Click on the Stratify Co2 button CARBWARE will perform the Carbon Stratification process and will create 6 new tables in the IncrementsDB mdb database file see Section 9 2 These tables are used for all CARBWARE reporting processes see Section 10 dS Important The Carbon Stratification process normalises all inventory plot units to 1 ha by referencing user specified plot sizes recorded in the Lookup table located in your CARBWARE project s Forest Inventory Database e g Carbwini2005 db A Important Some of the initial carbon pools at plot level must be input by the user in the Look
135. such as logistic regression and linear discriminant analysis perform very well for credit scoring Bigler and Bugmann 2004 introduced a new approach to modelling tree mortality based on different growth patterns of entire tree ring series They were interested in predicting time of tree death In their study dendrochronological data from Picea abies Norway spruce in the Swiss Alps were used to calibrate mortality models using logistic regression They introduced a mortality threshold and classified a tree as dead if its modelled mortality probability curve plotted over time went above that threshold They ignored autocorrelation at the modelling stage and applied a jackknife method to correct for the resulting biased variance estimates They found that the most reliable models were those that included relative growth rate and a short term growth trend as explanatory variables Focussing on the role played by life history strategies in determining tree mortality Wunder et al 2008 investigated whether the relationship between growth and mortality divers among tree species and site conditions This carries on from Monserud 1976 who showed that reduced growth generally accompanies a higher mortality risk For each of nine species they modelled mortality probablity as a function of relative basal area increment tree size and site They selected the species specific model with the highest goodness of fit and calculated the area under the receiver opera
136. t tC per plot which is user defined If no data is available then a look up table can be used to derive the value based on forest category table 2 and age of stand see Cpool_lookup_t0 xls litter worksheet on instillation disc Note the look up value is in tC per ha and must be normalised to the area of the plot e g for a 0 05ha plot the look up values must be multiplied by 0 05 Stump_Ctn This is the initial litter carbon pool expressed per unit area of the plot tC per plot which is user defined range 0 to 26 t C per ha The data is derived from measurements of stump volume and decays class where stump carbon volume x wood density 0 4 default x decay constant for different decay classes 0 98 to 0 48 x a C fraction 0 5 default If no data is available then a look up table can be used to derive the value based on forest category table 2 and age of stand see Cpool_lookup_t0 xls stumps worksheet on instillation disc Deadlog_Ctn This is the initial litter carbon pool expressed per unit area of the plot tC per plot which is user defined range 0 to 200 t C per ha The data is derived from measurements of dead logs and decay class data where log carbon volume x wood density 0 4 default x decay constant for different decay classes 0 98 to 0 48 x a C fraction 0 5 default If no data is available then a look up table can be used to derive the value based on forest category table 2 and age of stand see Cpool_lookup_t0 xls
137. t to use the same codes and classification of species cohorts so the software can allocate a given tree in to the correct cohort where specific algorithms are applied If there is a species not present in the list use a code corresponding to the next most similar species Note The growth models are empirically derived so application of models to trees outside the Republic of Ireland or the UK will introduce model error However CARBWARE developers will parameterise models representative of local ecological and climatic zones if requested and the appropriate calibration data is available For a full list of species lookup codes and corresponding cohorts see model cohort lookup xls on instillation disk f Height_m tree height in meters all trees should have a measured height g Diameter_mm diameter of trees at 130cm If tree are less than 139cm high then the cell should be left blank h Age tree age in years 4 1 1 3 The Lookup table The lookup table contains plot specific information one entry per plot such as the size of the sample plot an characteristics used for aggregation and reporting final model outputs Some of these values can be user defined to facilitate different software users However the field header and table format should be identical to the example provided a PlotID this must correspond with the IDPlot number in the Trees and IndividualTrees tables b Area_ha the sample plot area e g using the example in
138. t_1_00 Date Created 15 09 2011 12 33 19 Date Modified 15 09 2011 12 33 19 Project_1_01 Date Created 15 09 2011 12 33 21 Date Modified 15 09 2011 12 33 21 Project_1_02 Date Created 15 09 2011 23 16 46 Date Modified 15 09 2011 23 16 46 Project_1_03 Date Created 22 09 2011 15 19 21 Date Modified 22 09 2011 15 19 21 Project_1_Clearfell Table Date Created 15 09 2011 12 34 07 Date Modified 22 09 2011 14 36 52 Project_1_Increments Date Created 15 09 2011 12 33 19 Date Modified 22 09 2011 15 19 21 Project_1_ModifRecords Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Mortality_Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 Project_1_Thining_Table Date Created 15 09 2011 12 34 07 Date Modified 15 09 2011 12 34 07 tbiTempo Date Created 15 09 2011 12 18 16 Date Modified 15 09 2011 12 18 16 amp Project_1_Thining_Table ModificationYear EventCode PlotID TreelD CohortCode 2008 2008 2008 2008 2008 2008 2008 2008 boos 200 200 200 200 200 200 200 200 200 189 189 189 189 189 189 189 189 189 45 15 30 38 40 2 46 17 42 Record M 49023 noa WK No Fitter Search Estim_Height 16 2349976345477 17 9773544623797 15 7051127426803 18 4173990181126 17 83903880774 18 9198957307508 17 7913684757301 16 7303933692919 15 90406
139. ter Soil Harvested Wood Products 49 9 1 Running the Carbon Allocation Process To run the Carbon Allocation process fora CARBWARE project click on the Co2 Allocation button in the Main Menu screen Carbware Database in use C Documents and Settings Projects Carbon X Project Co2Allocation Database About Select Database Pre Processing Co2 Reporting Choose Project Close Program You will see the following Co2 Allocation screen Co2 Allocation Project Name Database path Number of Iterations Project_1 C Documents and Settings Projects Carbon_Project Carbwini200 Iterations Allocate Co2 Stratify Co2 If you do not wish to proceed with the Carbon Allocation process click on the Close Co2 Allocation button and you will return to the Main Menu screen The Co2 Allocation screen lists all current CARBWARE projects that have yet to undergo the Carbon Allocation Process In this example there is only one project Project_1 To proceed with the Carbon Allocation process select the required project by clicking on the project name The project name will become highlighted and a list of all annual growth cycles Iterations will be shown In this example there have been 3 annual growth cycles and 4 carbon allocations can be performed one for the baseline inventory year Year 0 and one for each of the 3 annual growth cycles Note If your CARBWARE project has been run for the maximum o
140. the dotted curves We estimated the ROC curve for each cohort model s out of sample performance by comparing model predictions with the actual NFI mortality data Figures 13 The cross validation and deployment performance plots are presented pair wise in the Figures below In all cased model candidate outperformed candidate based on false positives and fit For example we show the results for Fast growing cohorts in Figure 13 Note that Slow growing broadleaves cohort did not have enough data for the cross validation to be feasible so the ROC curve for that cohort depicts in sample performance 93 Fast growing broadiewves Cohort Piotwise XVa p 03 Fa growing Dbroadinave cohort NFI validation o i ts i i ii a 7 400 o 04 oo os 1o oo o2 oe oc os 10 Ange iame powlre ete Feme powiver ete a b Fant growing broadioaves Candidate 2 Casewise K Wal p 03 k 20 Fant grow ing broadie aves Cancstate 2 NFI validation n as Tre pote mr cs Average Wve postive rse es oo 02 o4 oe os 10 eo 0 o oe os 10 haag tame posDre ate Fane paise ate c a Figure 13 The Receiver operating characteristic curve for Candidate model 1 panels a b and model 2 panel c d in the Fast growing broadleaves cohort 20 fold cross validation plotwise with average leave out probability p 0 3 Curves for each cross validation run and a threshold averaged curve are shown Models fitted to NFI data
141. the numerator and so their parameterised curve representation does not describe a true trade off situation in every region of threshold space If we overlay the two graphs with precision and recall on the y and y axes and threshold on the x axis we can see more clearly where a true trade off may occur It is likely that should a true trade off occur that the region between the parameterised curve and the line of equality will have to be positive As external corroboration DeLong et al 1988 note that the cost or loss function is essential to deciding the optimal cutpoint threshold for a ROC curve In summary there are therefore two issues comparing classifiers and given a classifier choosing a cut off point This latter can only be done in conjunction with some kind of loss function describing costs of the different types of classification error The kind of classifier we are using based on multiple correlation regression and therefore wholly empirical is easier to select than other types of classifier We can use model selection criteria based on correlation regression or minimization of errors or some other abstract modelling concepts Then the classifier selected we can choose the cut off In what we call mechanistic classifiers such as described in Martin Davila et al 2005 where the classifier is predicated first and foremost on an understood pathway not naive correlation the threshold has a physical dimension and the choice of
142. ting characteristic curve and calibration measures The discriminatory power as measured by AUC ranged from 0 62 to 0 87 They found that most growth mortality relationships differed among species and sites i e there is no universal growth mortality relationship It has been noted that a lack of long term growth mortality data has made it difficult to evaluate the performance of mortality models Wunder et al 2006b adopt a virtual ecology approach to this problem simulating forests with either of two a priori specified growth mortality relationships They simulate different sampling regimes in these virtual forests thereby generating virtual tree ring data forest inventory data or a combination of both They compare eight existing or newly developed models of different structural flexibility by their ability to model the growth mortality relationship in the simulated data and quantify the deviation from the a priori specified growth mortality relationships with the Kullback Leibler distance Of the models they evaluated the highest accuracies were obtained with tree ring based models which required only small approx 60 numbers of dead trees For larger sample sizes approx 87 500 dead trees forest inventory based models were also seen to be accurate They also showed that exible statistical approaches were superior to less flexible models only for large sample sizes totally 2000 trees and that the additional use of Bayesian statisti
143. to 15 individual tree decimal places no measurements represent of trees no units see box 1 stratified sampling ExpansionFactor Number of trees represented M Trees Numeric up to 15 by the measured tree this decimal places applies to systematic partial sampling of inventory plots as above RepreArea_ha Representative area for O Trees Specific NFI data each tree based on volume not currently used the sum of the representative areas is equal to the plot size if trees are present RepreHeight_m Representative height for O Trees Specific NFI data trees used for aggregation in not currently used NFI DiameterGroup O Trees Specific NFI data not currently used DiameterClass5 O Trees Specific NFI data not currently used DimensionClass5 O Trees Specific NFI data 16 Data Description Requirement Table Units Format AgeClass10_150 O Trees Specific NFI data not currently used AgeClass20 O Trees Specific NFI data not currently used SpeciesGroup O Trees Specific NFI data not currently used IDSubcircle O Trees Specific NFI data not currently used GrowthStage O Trees Specific NFI data not currently used ConiferBroadleaf O Trees Specific NFI data not currently used NativeNonNative O Trees Specific NFI data not currently used AgeClass10_60 O Trees Specific NFI data not currently used SlendernessRatio O Trees Specific NFI data not currently used Slen
144. to NFI they were omitted from the Dinc model They were also omitted from within NFI imputation models by which we mean imputation models calibrated on NFI data for similar considerations They were not omitted from PSP specific models We looked at the performance of the various models DBH H CR Dinc for the two datasets Some measures we could have used that are used by Thurig et al 2005 for example are accuracy precision and excess error calculated as follows Accuracy 2 predicted observed n 100 m Where m is E obs and nis the number of observations Precision SD pred obs Empirical Excess error 1 Sec Sei 100 Where Sec is the precision of the calibration data and Sei the precision of the independent data Theoretical Excess error 1 n 2 pred 1 obs 2 pred obs 1 Where pred 1 is the leave one out prediction error Note that empirical excess error is only viable when doing external validation Temesgen and von Gadow 2004 for example use root mean squared error RMSE and Bias to evaluate their models Bias 2 pred obs n RMSE 2 pred obs n p Where p is the number of parameters in the model Another measure is mean absolute error MAE MAE pred obs n A certain amount of model selection was done as noted above when fitting the models to the data in the first place This ensures that the fitted models are the most parsimonious to minimise residual error However
145. ts would be the outcome of the mortality rule Rather than reducing the expansion factor by one unit when death is predicted which we can show can lead to an unrealistically height global mortality rate the actual reduction would be a function of the weight for the given diameter class This method could be stochastic or deterministic Other information might be used to inform the values of the weights including a forester s rule of thumb about global mortality i e _ 6 or information from the NFI or a meta analysis A similar approach would be to mix the outcome of the mortality rule with the diameter class mortality weights It might be possible to iteratively tune the weights and or the rule s cut off parameter 90 0 7 7 42 12 20 20 91 2 i E z Be Diameter class limits ooooos SCNARDO ooooco CNBADO Proportion all i 0 7 7 12 12 20 20 91 2 Figure 11 The empirical distribution in the PSP dataset of diameter classes of dead alive trees classified by cohorts Results Candidate model Number 1Candidate model 1 was a fixed effects model A logistic GLM was fitted in Glimmix The _ fixed effects were DBH BAL and growth t t 2 DBH t Part of the reason for looking at this model was that it was not subject to additional uncertainty due to imputation of missing X data as would have been the case with the model put forward by Monserud and Sterba 1999 which also conditione
146. ttom up basis plot level up using sample plot and conventional forest inventory information Harvested wood products are simulated using another top down model WOODCARB Atmosphere repona Burning j Biomass Cohort 2 thinnings energy p Fi Cohort 1 3 f Biomass 5 5 d Above ground F Prodipe S Ga B Til Branches z 8 Bo Bd Timber Leaves z 2 20 Sti Below ground z 8 So Stumps Roots g amp as ENE LL Landfill Litter dead wood Stable humus soil C Figure 1 Flow diagram of proposed CARBWARE model including all components for UNFCCC and Kyoto reporting wood products are assumed to be a source at harvest The Irish carbon reporting system CARBWARE initially described by Gallagher et al 2004 was implemented to meet reporting requirements to the UNFCCC on national forest sources and sinks Since then CARBWARE has evolved from a tier 2 to tier 3 system using forest inventory data yield models and national research information see Black and Farrell 2006 Gallagher et al 2004 Black et al 2012 Duffy et al 2011 to include activities relating to Articles 3 3 and 3 4 of the Kyoto protocol Although the software was designed to meet Irish national forest reporting requirements under contract to the Irish Council for Forest Research and Development COFORD it can be used globally if input data are formatted in a compatible m
147. ty RB_t01 RB mortality t1 tC plot Thin_RB t01 RB thinnings t 1 tC plot CF_RB_t01 RB clearfells t1 tC plot RB_C t01 RB form t1 tC plot RB_C_gain_t01 RB_C_t01 minus RBliving plus mortality Thin and CF t01 tC plot negative value represents a gain RB_C_loss_tn01 sum of RB mortality RB from CF and Thin tC plot Net_RB_t01 sum RB_C_gain and RB_C_loss tC plot RB_deadtree_C_t02 dead tree RB from mortality model tC plot RB_Living_C_t02 Living RB biomass t2 tC plot Mortality _RB_t02 RB mortality t2 tC plot Thin_RB_t02 RB thinnings t 2 tC plot CF_RB_t02 RB clearfells t2 tC plot RB_C_t02 RB form t2 tC plot RB_C_gain_t02 RB_C_t02 minus RB_C t01 plus mortality Thin and CF tC plot t02 negative value represents a gain RB_C_loss_tn02 sum of RB mortality RB fromCF and Thin t2 tC plot Net_AB_t02 sum RB_C_gain and RB_C loss tC plot Table 9 Description of Below Ground Carbon Table Fields in the IncrementsDB mdb database 55 9 2 3 Deadwood Carbon Table e This project specific CARBWARE Table is created and named by CARBWARE using the naming convention Project Name_DeadWood In this example the project name is Project_1 and so the project specific DeadWood Carbon Table is named Project_1_Deadwood This table lists inventory plot level carbon estimates for the deadwood carbon pool all wood with diameter greater than 7cm and includes roots with diameter less than 5cm
148. ules for accounting have not been finalised However the table does supply the HWP inputs for supplementary analysis using different models or software see section 4 4 Reporter module Final output tables are generated in the reported module using the common reporting format as outlined by the UNFCCC see section 10 4 Data input requirements Input data must be stored in a MS access database in specific tables under specific field headers The MS access database should be named with the prefix Carbw ending with the year of inventory For example Carbwini2005 is the example database from the 2005 inventory provided on the instillation disc The format of headers and data entries should be in the correct format and units for the software to operate properly 4 1 Forest inventory database i e Carbw 200X No specific forest inventory design is required for CARBWARE to function However the inventory should provide mandatory information listed in the table 1 also see example database Carbwini 2005 mdb on 11 instillation disc or web folder Specific details on how the data is obtained are dealt with in the preceding text sections Example databases are provided in the enclosed cd to illustrate data format and layout It is important that the precise field names case sensitive properties of fields and order of fields in the specific tables are identical to the examples provided to ensure that the software run properly Failure
149. unctions in Irish NFI Zm Z azimuth co ordinate of O Trees Numeric positive tree in m from local origin or negative point XMeasurepoint_m Crown measurements O Trees Specific NFI data not currently used YMeasurepoint_m Crown measurements O Trees Specific NFI data not currently used ZMeasurepoint Crown measurements O Trees Specific NFI data M mandatory O optional N is not required at present but fields with names must be incorporated in the table even if the cells are empty 13 Data Description Requirement Table Units Format XCentreCrownProj_m_ Crown measurements O Trees Specific NFI data not currently used YCentreCrownProj m Crown measurements O Trees Specific NFI data SmoothedCrownProj Crown measurements O Trees Specific NFI data not currently used Slant_Azymuth_deg Tree slant O Trees Specific NFI data not currently used Slant_Angle_deg Tree slant angle O Trees Specific NFI data not currently used Stem_volume_m3 Calculated stem volume in O Trees Specific NFI data m3 not currently used CrownProj_Area_m2 Projected crown area O Trees Specific NFI data not currently used Crown_Volume_m3 Crown volume O Trees Specific NFI data not currently used Crown_Surface_m2 Crown surface area O Trees Specific NFI data not currently used DBH_mm Diameter at 1 3m height M Trees Numeric in mm no decimal points Height_m Tree
150. up table located in your CARBWARE project s Forest Inventory Database e g Carbwini2005 db This must be done prior to running the Carbon Stratification process Lookup table inputs include initial litter and deadwood inputs and soils emission factors see Section 4 for further details When you have completed the Carbon Allocation and Stratification process click on the Close Co2 Allocation button and you will return to the Main Menu screen 52 9 2 Carbon Allocation Tables e When the CARBWARE Carbon Allocation and Stratification process is run for a project see Section 9 1 CARBWARE creates a set of six new project specific tables in the IncrementsDB mdb database file for each of the carbon pools All Dates yi Project_1_AG Date Created 23 09 2011 12 37 08 Date Modified 23 09 2011 12 37 08 Project_1_BG Date Created 23 09 2011 12 37 08 Date Modified 23 09 2011 12 37 08 Project_1_Deadwood Date Created 23 09 2011 12 37 08 Date Modified 23 09 2011 12 37 08 Project_1_HWP Date Created 23 09 2011 12 37 08 Date Modified 23 09 2011 12 37 09 Project_1_Litter Date Created 23 09 2011 12 37 08 Date Modified 23 09 2011 12 37 08 Project_1_Soil Date Created 23 09 2011 12 37 08 Date Modified 23 09 2011 12 37 08 e These tables described in Sections 9 2 1 to 9 2 6 below summarise pool specific results of inventory plot level carbon estimates fil Note With the except
151. will overcome the additional security features associated with newer versions Open Run as administrator Scan Carbware1324_Install exe Pin to Start Menu Add to Quick Launch Enter your user details Name amp Organisation i Carbware InstallShield Wizard Customer Information Please enter your information User Name oe Organization pare Install this application for Anyone who uses this computer all users O Only for me Select the required Setup Type select Typical i Carbware InstallShield Wizard Setup Type Choose the setup type that best suits your needs Please select a setup type Typical All program features will be installed Requires the most disk space Minimum required features will be installed Choose which program features you want installed and where they will be installed Recommended for advanced users ee CARBWARE is now ready to be installed on your PC Click Install to complete the installation CARBWARE will be installed in the following directory C Program Files COFORD Carbware i Carbware InstallShield Wizard Ready to Install the Program The wizard is ready to begin installation If you want to review or change any of your installation settings click Back Click Cancel to exit the wizard Current Settings Setup Type Typical Destination Folder C Program Files COFORD Carbware User Information Name MT Insta
152. with the saved parameters associated with your project Carbware Project Parameters Project Name Project_i Inventory Year 2005 Select Database _ 1 OAD ocuments and Settings Projects Carbon_Project Carbwini2005 mdb Filter Baseline Inventory Database Stand Modification eo by Date Criteria Natural Mortality Al Reporton Grow for 1 Year All Clear Fell to C Article 3 3 1990 and younger Year 1 2006 Article 3 4 Before 1990 M Thinnning You have done 0 cycles Note The following parameters are associated with this project Project Name Project_1 Forest Inventory Database Carbwini2005 mdb Inventory Year 2005 as per Forest Inventory Database naming convention Inventory Database Filter All plots selected Stand Modification Options Natural Mortality Clearfell amp Thinning Selected The number of annual growth cycles performed to date will be noted at the bottom of the Project Parameters screen In this example 0 cycles have been performed Up to 15 annual growth cycles can be performed Click on the Grow for 1 Year button to perform the first annual growth cycle CARBWARE will begin an annual growth cycle performing a complex series of growth modelling and stand modification events as described in Section 3 Note The length of time for CARBWARE to complete an annual growth cycle will largely depend on the size of your Forest Inventory Database number of plots A large database such as that

Download Pdf Manuals

Related Search

CARBWARE_Manual_March_2012

Related Contents

User Manual - Projector Central  argus - Rosenbauer  4PRO 1031-A 4PRO 2031-A 4PRO 3031-A 4PRO 5031-A  CABLE FREE WEATHER STATION  Indesit WIB101 Washer User Manual  Seguridad e Higiene  Questionnaire mode d`emploi  Samsung SyncMaster  MANUEL D`UTILISATION  PROGRAMME  

Copyright © All rights reserved. Failed to retrieve file