Home

SPATE-HPC User Manual

Contents

1. 57 25 5 Tree data le ira da A a debe u E Pe 58 5 6 Compartment statistics file ee ee 58 2 Statistics table Mle uoc oer a ak oe we SL EORR 59 Index 61 List of Figures 1 1 3 1 32 3 3 4 1 4 2 43 4 4 4 5 4 6 4 7 4 8 4 9 4 10 The relationships between SPATE GUI SPATE HPC the input and the output files Diagram showing the workflow of asimulation A sample forest area with 38 compartments consisting of 57 polygons The simulated tree distribution is represented by the histogram and the estimated distribution by the solid curve The dashed curve is the shifted curve The general simulation settings 2 ee The Regeneration page showing the model for calculating the expected number of new trees per hectare in the reproduction phase Page showing the reproduction models o The model editor showing the model Y bo 3 Diameter b2 Diameter Local density B3 Epsilon The coefficients Beta 0 through Beta 3 are shown in bold font which indicates that they have been assigned secondary level models 224 251 ea x 8 Qabec hea Sp ay OUI UNSER ee CRT SR oid The smoothing function editor window showing the secondary level model 61 VAF OTS PECES s uo oues ORI Pat Be do hue does ow ox qr AL Ie s The Initial forest page showing example settings for generating an initial stand The For
2. 0 1 0 1 0 1 0 1 lt managementMethod gt rCut container container 0 1 0 1 mmm Figure 5 11 A managementTemplate element defines a set of management rules Low and selective thinning The elements lowThinning and selectiveThinning may contain the following child elements removalPerHectare This mandatory element indicates the volume that is to be removed Its unit is cubic meters per hectare removalTolerance The default removal tolerance is five percent but it can be overridden with this element classWidthInCm The default size class width is 4cm but it can be overridden with this element linearTailRatio If given the tail of the estimated Weibull curve will be linearized according to this ratio secondaryModelValue If given it should contain a numerical value that is subsequently assigned the Management variable in the secondary level model Additionally the selective Thinning element may also contain the following child elements 52 thinningLevel This element specifies the 9o coefficient in formula 3 12 If not given the default value is 0 5 densityCoefficient This element specifies the 6 coefficient in formula 3 12 If not given the default value is 100 Clear cut The element clearCut may contain the following child elements treesLeftPerHectare The number of trees per hectare to spare while the rest are cut forcedTreeDistance The trees that are spared must stand no closer th
3. When running SPATE HPC the paths of externally referenced input files are relative to that of the XML parameter file In other words if the XML parameter file resides in a given directory all other input files such as the polygon files are expected to reside in the same directory if the filenames of the external files are given without a full path For the sake of coherency it is recommended to put externally referenced files in the same directory as the XML parameter input file Output files however are not created in a directory relative to that of the input file Ouput files are created in the current directory the directory from which SPATE HPC is executed or in a directory relative to that If the user provides an output directory name such as output t a directory with that name with t replaced by the current time stamp is created and all output files put into that directory 2 2 1 Running on multiple processors If SPATE HPC has been compiled with MPI support it can be run on a multi processor system in order to reduce total execution time and increase the maximum domain size This guide will de scribe how to run SPATE HPC in parallel using MPICH 2 on a Linux operating system However the exact procedure may vary between different systems and set ups Large supercomputers might use their own batch system to run parallel programs Firstly the MPI background daemon mpd must be running Start it with mpd amp S
4. 56 a polygon that is completely surrounded by another polygon The following is an example of a square polygon with a square cut out region 1 0 0 0 0 100 0 0 0 100 0 100 0 0 0 100 0 0 0 0 0 END 1 40 0 40 0 80 0 40 0 80 0 80 0 40 0 80 0 40 0 40 0 END END 5 3 Empirical data Empirical data can be used to provide a target distribution for marks in the initial stand and the reproduction phase Empirical data are provided as a set of real valued numbers in an external file There should be one value per row and there should not be no empty rows or rows containing any other kinds of data The numbers may appear in any order Example 1 54 14 29 2 93 25 01 19 37 10 45 9 83 5 4 Empirical cumulative distribution An empirical cumulative distribution can be used to provide a target distribution for marks in the initial stand and the reproduction phase An empirical cumulative distribution is given as a list of pairs of numbers with the first number being x and the second number F x where F x is a cumulative distribution function The values of x must appear in increasing order and the values of F x must be monotonically increasing and span from 0 to 1 There should be one pair per row and there should be no empty rows or rows containing any other kinds of data The two values within a pair should be separated by white space i e one or several space or tabular characters Example 57 54 0 0 00 8 0 0 04 11 0 0
5. the management templates that should be implemented in the compartment can be selected Number of initial trees factor If specified the number of trees per hectare in the initial stand is weighted with this factor Age of initial trees factor If specified the age of trees in the initial stand is weighted with this factor Multiple compartments can be combined to form a multi polygon compartment This is done by selecting multiple compartments either in the compartment list or in the area view and clicking the Merge compartments button A multi polygon compartment can subsequently be split by selecting a polygon in the polygon list and clicking the Split button 4 11 Notes page The Notes page contains a text box into which the user may enter arbitrary notes The notes have not effect on the simulation and can only be used for attaching explanatory text to the XML input file 38 Chapter 5 File formats This chapter describes the file formats of the files used as input and output of SPATE HPC and SPATE GUI 5 1 Parameter file The parameter file which is the main input file of SPATE HPC is in the extensible markup lan guage XML format In this section the possible element attributes and values that can be used in the XML input file are described element container Figure 5 1 XML elements are illustrated with these types of boxes XML element are illustrated with boxes connected to each other w
6. 16 14 0 0 40 TIS 0 69 20 0 0 89 23 0 0 98 26 0 1 00 5 5 Tree data file The tree data file is table of tree data in the comma separated values CSV format Each row in the table corresponds to an individual tree and the columns contain specific tree characteristics The columns are the following in order X coordinate The tree s x coordinate The unit is meters Y coordinate The tree s y coordinate The unit is meters DBH Diameter at breast height in centimeters Height Total heigh in meters Initial heterogeneity The Gaussian random number that was assigned the tree as its initial hetero geneity Age The age in years Species The name of the species 5 6 Compartment statistics file The compartment statistics file is a plain text file containing various figures and tables pertaining to a compartment s development All values are averaged over all replications If the user has run several replications some of the values will be accompanied by the lower and upper limits of a 95 Monte Carlo confidence interval A compartment statistics file contains the following sections Simulation summary This section contains general information such as the number of years iteration steps and replications simulated as well as compartment specific information such as compartment area forest type and initial species composition Statistics summary Here tables that show how the stand has developed throughout the simulation are presen
7. 803 x Species 0 568 x Species 2 0 093 x Species 3 0 73 0 235 x Species 0 1125 x Species 2 0 0125 x Species 3 x Diameter 0 0031 0 00622 x Species 0 00275 x Species 2 0 00033333 x Species 3 x Diameter 2 0 0044 0 0004 x Species 0 0002 x Species 2 x Diameter sum 0 1 0 165 x Species 0 1x Species 2 0 015 x Species 3 x Epsilon Edit Clear Figure 4 6 The Initial forest page showing example settings for generating an initial stand 30 E SPATE GUI v 0 0912 svn726 Figure 4 7 The Forest types page showing the forest type spruce_dominated which contains 80 spruce and has diameter transformations defined for both the initial stand and the reproduc tion phase 31 trees of the species to have a smaller or larger diameter than the overall stand one may enter a factor into the input fields Initial tree diameter factor and New tree diameter factor for trees in the initial stand and new trees in the reproduction phase respectively Leaving the fields empty is the same as entering the number 1 Furthermore the tree proportion in the reproduction phase can be separate from the one in the initial stand The reproduction proportion is specified using the Reproduction amount input field Leaving this field empty means that the same proportion is used both in the initial stand and the reproduction phase Ifthe generated t
8. a positive integer Furthermore a com partment element may contain the following child elements forestTypeName This mandatory element should contain the name of the forest type of which the compartment is made up managementTemplateName There may be none or several elements of this type which each con tains the name of a management template that is to be applied to the compartment polygonFile There should be one or several polygonFile elements The elements contain the polygon file s name and must have an id attribute that indicates the polygon to use numberOfInitialTreesFactor If given the number of trees in the initial stand is multiplied by this factor ageOflnitialTreesFactor If given the age of the trees in the initial stand is multiplied by this factor The compartment element and its child elements are shown in figure 5 13 1 many compartment id int gt container Figure 5 13 A compartment element contains compartment specific parameters 5 1 10 Inspection parameters The optional inspectionParameters element enables additional logging and allows one to inspect model calculations and thinning procedures more closely as they are carried out see figure 5 14 The inspectionParameters element may contain the following child elements inspectModel There may be none or multiple inspectModel elements which each may contain one of the following values initial stand growth mortality reproduct
9. a two dimensional Euclidean space Diameter DBH or diameter at breast height 1 3m Height Total height Initial heterogeneity A normally distributed random number assigned to the tree upon its creation Also referred to as tau 7 Age Age in years Species Tree species pine spruce birch etc The characteristics diameter and height are collectively referred to as marks Simulate replication 0 Simulate replication 1 Simulate replication N Start gt Read Combine Write End input files statistics output files A Iterate Generate or load initial stand voran Reproduction p p Figure 3 1 Diagram showing the workflow of a simulation 3 2 Spatial inter tree dependence Growth mortality and reproduction are calculated independently for each tree but the outcome can be made dependent on neighboring trees The dependency is achieved through competition indices and spatially correlated random numbers Competition indices are variables whose values are calculated based on the target tree and its local neighboring trees The spatially correlated random numbers are drawn so that trees within closer proximity are more strongly correlated than trees farther apart see section 3 4 2 When competition indices are calculated only the neighboring trees located within a certain competition distance or competition radius are taken into consideration These neighbori
10. current time stamp standardNormalLimitingFactor Normally distributed random numbers in some parts of the pro gram are limited to a support defined by this factor The support is the value of this element multiplied by the standard deviation Size classes The sizeClasses element contains a list of diameter size classes defined by sizeClass elements Each sizeClass element should contain a name and a minimumDiameter element sizeClass ele ments must be ordered according to minimum diameter and there must be at least one sizeClass element within sizeClasses 40 Jeujejuoo lt ul p jueunieduio2 Jeutrejuo2 lt juewposse gt seuleyuood Duujs euieu ejejduie uewebeuew gt Jeujejuoo lt 8d 1s810J 1eujejuoo seineds 1euiejuoo SJejeuie1e qjueunieduioo J8uigjuo2 SJ8jeuieJeqjueunjosse Jeuiejuo2 S18jeuieJe uewebeuew gt Jeuigjuoo SJ0jeuje Je qad 1se10J Jeuiejuoo lt SJojowesegseneds gt 1Jeuiejuoo SJojeuiejeqjeouoD Jeuiejuoo lt SJajawesequonoedsul gt wesBoid Aq pesn jou 8jou Jeurejuoo uone nuis Figure 5 2 The root element simulation and its direct children 41 Jeuiejuoo sJejeureJequwoJ6 Jeuigjuoo siejeuieje d jeuour Jeuleyuoo lt juewpedwogjeniul gt 4Jeujejuoo Jeuigjuo2 Jeurejuo2 Jeuigjuo2 sieyeuieJequononpoudej ssejoiubieu lt ssejgebe gt lt SSP QAZIS gt Auew Auew
11. from a forest type Secondary model value The value used for the variable Species in the secondary level model see section 3 4 1 Small volume coefficient If a tree is too small for taper curves and a cylinder model is used instead this coefficient is used as a weight Competition distance model A model for calculating the competition distance Taper Curve Parameters Parameters for the taper curve functions used to estimate a trees volume Stump Height Parameters Parameters for the stump height model 4 8 Management Management templates are edited on the Management page see figure 4 9 Management tem plates and methods are described in section 3 9 32 IIl SPATE GUI v 0 0912 svn726 DER General Regeneration Reproduction Growth Mortality Initial forest Forest types Species Management Assortments Compartments Notes gt Species parameters Name Pine Secondary model value 1 Small volume coefficient Competition distance model Y 5 Edit Clear Taper Curve Parameters d d 20h b1 x b2 x 2 b3 x 3 t b5 x 5 b8 x 8 b13 x 13 b21 x 21 b34 x 34 x 1 1 h bi 2 1288 ba osss bx res bs ame b8 1g Foz b34 for Stump Height Parameters stump height c1 dbh 2 h 100 Minimum height m jo 1 Diameter coefficient c1 0 4456 Height coefficient c2 0 09 52 Figure 4 8 The Species page Here the
12. is to stochastically cut trees so that the protruding histogram bars are trimmed down the height of the shifted Weibull curve Figure 3 3 illustrates how the curves are shifted The amount of trees to cut in a given size class is controlled by the thinning rate The thinning rate is a size class dependent value between 0 and 1 and represents the number of trees in the corresponding size class that are to be cut The thinning rate is calculated as the part of the histogram bar protruding above the shifted curve in relation to the whole histogram bar 18 Low thinning Low thinning starts with the smallest size class and proceeds with larger size classes The last size class to be included in the thinning procedure is the one where the intersection point between the two Weibull curves is located Trees are cut stochastically and the thinning probability of a tree being cut is the thinning rate i e all trees in a given size class have the same thinning probability Selective thinning Selective thinning starts with the largest size class and proceeds with smaller size classes As in low thinning the last size class to be included is the one where the intersection point lies In contrary to low thinning selective thinning also considers the local tree density The likelihood of trees being cut in denser neighborhoods is higher than that of trees in sparser neighborhoods Each tree in a size class is assigned its own individual thinning probability
13. of a polygon is hence four which would correspond to triangle It does not matter whether the vertices are given in clockwise or counterclockwise direction Polygons may be conjoined by sharing borders and there may also be regions within the forest area that are not covered by polygons However polygons must not overlap The following is an example of a polygon file containing two polygons a square and a triangle 1 0 0 0 0 100 0 0 0 100 0 100 0 0 0 100 0 0 0 0 0 END 2 100 0 0 0 200 0 0 0 100 0 100 0 100 0 0 0 END END A polygon definition begins with a unique ID number Thereafter follows a list of vertex coor dinates where each line contains the coordinates of one vertex Coordinates are given as a pair of real valued numbers where the first number is the x coordinate and the second one is the y coordinate The numbers must be separated by white space i e one or several space or tabular characters After the polygon s last vertex follows a line with the word END An additional END follows the last polygon 5 2 4 Cut out regions Polygons may contain cut out regions holes Cut out regions are defined by providing an addi tional polygon with the same ID number as a previously defined polygon Instead of becoming a separate polygon the second polygon will constitute a hole in the previously defined polygon One is also free to put another new polygon within the hole In other words it is possible to have
14. of the initial stand is 40 years and the years between clear cuts parameter is 50 the first clear cut takes place the tenth simulated year and second one the sixtieth simulated year etc If the lower basal area or lower compartment volume criterion is given however the clear cut might be postponed which consequently offsets subsequent clear cuts Clearcutting does not fell all trees but preserves some of them This is controlled with the trees left per hectare and forced tree distance parameters The program will try to find and spare the desired number of trees that stand no closer to one another than the forced tree distance This works by first categorizing trees into size classes according to diameter The program then tries to find the trees to spare by analyzing the trees in the largest size class If the desired number of trees is not attained the program tries again by also including the second largest size class in the search and then the third etc Larger trees are hence prioritized in the search for trees to spare Moreover the size of trees to spare can be limited to a certain diameter range Trees larger than the maximum diameter and trees smaller than the minimum diameter will thus not be considered when searching for trees to spare In case the program cannot find as many trees to spare as desired it will accept the reduced number i e the trees found will be spared the rest cut down nonetheless and a pertinent message added to th
15. of trees per hectare to spare after a clear cut Forced tree distance Spared trees may stand no closer than this distance Class width The size class width The default is 4cm Minimum diameter of spared trees Spared trees may not have a diameter smaller than this value Minimum diameter of spared trees Spared trees may not have a diameter larger than this value 35 4 9 Assortments page Assortments are edited on the Assortments page see figure 4 10 If assortments are specified volume statistics on trees removed and on the growing stock for the different assortments are produced after a simulation has finished Assortments are described in section 3 10 E SPATE GUI v 0 0912 svn726 General Regeneration Reproduction Growth Mortality Initial forest Forest types Species Management Assortments Compartments Notes Assortment parameters Name saniog Species minimum diameter cm log length cm The order in which assortments are specified is important the algorithm prioritizes the first assortment Add Remove over the second etc If log length or minimum diameter parameters are not provided the remaining part of the tree is used for the assortment in its entirety regardless of size Figure 4 10 Assortments are defined on the Assortments page Here the assortment sawlog is selected and its size criteria for different species are shown in the panel to the right On the left s
16. species Pine is selected and its settings are shown in the panel to the right 33 E SPATE GUI v 0 0912 svn726 DER General Regeneration Reproduction Growth Mortality Initial forest Forest types Species Management Assortments Compartments Notes Management templates Template parameters Name LowThinning Lower basal area m 2 ha Thinning interval years EE Lower compartment volume m 3 ha po Management method JA Low thinning C Dimension cutting C Energy wood cutting C Selective thinning None Removal per hectare m 3 ha 50 Secondary model value Class width cm Removal tolerance Linear tail ratio Clear cut Use dear cuts Figure 4 9 Management templates are defined on the Management page The selected tem plate LowThinning is set to carry out low thinning removing 50m ha every 13 year if the compartment volume exceeds 200m ha 34 On the left side of the window there is a list of the management templates that are currently defined Selecting a management template in the list brings up its settings on the right side of the window A new management template can be created by clicking New and an existing management template removed by clicking Delete The following is a list of the settings that are configurable for a management template Name The management template must have a unique name which is used to refer to
17. the parameter i e m M1 exp y 3 5 Here mm is a mark value diameter or height of a tree at time step t The growth is calculated separately for each mark 3 5 3 Mortality There are two types of mortality in SPATE HPC regular and irregular random mortality The regular mortality occurs due to deficient growth whereas the irregular mortality occurs randomly but with a probability affected by local competition The regular mortality occurs when a tree s diameter growth in the current iteration is less than a specific mortality threshold A mortality threshold of 1 0 would hence mean that trees are not allowed to shrink There may be a separate mortality threshold for the diameter and the height If no mortality threshold is given the regular mortality is disabled The irregular mortality is modelled with a logistic regression model The probability 7 of an individual tree dying is calculated as 1 where y is obtained from formula 3 2 A uniformly distributed random number u U 0 1 is drawn If u lt 7 the tree dies One may provide a separate irregular mortality model for each mark 3 5 4 Reproduction In the reproduction phase new trees are generated The number of trees to generate is a Poisson random number whose mean is obtained from a linear model called the regeneration model The trees are then placed by a homogeneous Poisson point process The point process is hard core i e new trees cannot be placed insid
18. w which is calcu lated using the logistic regression model 1 SYL exp fo 8i n Ay Aa where Ay is the local density around the target tree and A a is the overall stand density of the com partment The coefficients o and 3 are the thinning level and the density coefficient respectively which are parameters the user may specify The number of trees to cut in a size class is deter mined by a Poisson random number whose mean value is the class s thinning rate multiplied by the total number of trees in the class The program then tries to stochastically cut this amount of trees according to the trees individual thinning probabilities It can however fail to cut the desired number of trees if the trees thinning probabilities are overall low 3 12 Achieving the target volume The user controls the amount of trees to cut with the removal per hectare parameter The objective is to cut trees so that the accumulated volume of trees removed matches the desired removal volume The result of the thinning is controlled by the distance the Weibull curve is shifted However no algorithm is implemented to directly determine the shift distance based on the target removal volume Instead the target volume is achieve through iteration i e by testing different shift values until a sufficiently good one is found The amount the result may deviate by from the target volume is controlled with the removal tolerance parameter whose default
19. A new forest type can be created by clicking New and an existing forest type removed by clicking Delete A forest type must be given a unique name using the Name input field The name is used to refer to the forest type from a compartment The species that the forest type should contain are chosen by adding them to the Species list The relative amount of trees of the species with respect to other species is specified by adjusting the Proportion parameter The proportion does not need to be a percentage the species proportions are only compared to each other relatively If one wants 29 IIl SPATE GUI v 0 0912 svn726 DE General Regeneration Reproduction Growth Mortality Forest types Species Management Assortments Compartments Notes Load initial forest data from file Browse Generate Homogeneous Poisson Homogeneous clustered Homogeneous regular C Inhomogeneous Poisson Inhomogeneous clustered C None C Inhomogeneous regular Average initial age years 100 E Number of marks 2 E Parameters Expected number of points per hectare 4000 a Mark 1 Name Diameter v Enviromental effects Theta0 Thetal Model Y 0 76 0 05 x Species 0 012 x Species 2 0 00034 x Local density 0 02 x Epsilon 0 03 x Xi Edit Clear r Mark 2 Name Height m Enviromental effects Theta0 Y 1 322 0
20. Area If specified the compartment must have basal are larger than this value The value is given in square meters per hectare and the rule applies to all management methods managementMethod Contains one and only one of the following elements lowThinning selec tiveThinning dimensionCutting or energywoodCutting lowerCompartmentVolume If specified the compartment must have a volume larger than this value The value is given in cubic meters per hectare and the rule applies to all management methods yearsBetweenClearCuts This value dictates the minimum number of years between to consecu tive clear cuts This value must be specified if a clear cut is used clearCut If this element is given a clear cut is performed Dimension and energy wood cutting The elements dimensionCutting and energywoodCutting may contain the following child ele ments diameterLimit For dimension cutting trees larger than this diameter are cut whereas for energy wood cutting trees smaller than this diameter are cut This element is mandatory speciesToCut There may be zero one or several elements of this type Each element must contain the name of a species If there are one or more speciesToCut elements only trees of those species are cut secondaryModelValue If given it should contain a numerical value that is subsequently assigned the Management variable in the secondary level model 51 managementTemplate name string gt container
21. Jeuiejuo2 4euigjuo2 4eurgjuo2 Jeuigjuo2 SJojouejeguoneJeuebDo1 sesse jjubieu lt sesse gabe gt lt S9SSB D9ZIS gt L 0 1 0 Jeurejuoo lt SJojouesed jeioueb gt Figure 5 3 The generalParameters element contains assorted simulation parameters 42 Age classes The ageClasses element contains a list of age classes defined by ageClass elements ageClasses is similar to sizeClasses but with more lax cardinalities Each ageClass element may contain a name and should contain a minimumAge element ageClass elements must be ordered according to the minimum age Height classes The heightClasses element contains a list of height classes defined by heightClass elements heightClasses is similar to sizeClasses but with more lax cardinalities Each heightClass ele ment may contain a name and should contain a minimumHeight element heightClass elements must be ordered according to the minimum height Growth mortality and reproduction parameters The models for growth mortality and reproduction are given within the growthParameters mortal ityParameters and reproductionParameters elements respectively The corresponding parameters element may contain zero one or two mark elements The content of a mark element represents the growth mortality or reproduction model for one of the marks diameter or height A mark element must have a name attribute with a value of either Diameter or Height No other name attribu
22. PATE HPC can then be run in parallel by prefixing it with mpirun mpirun n 4 spate hpc input file xml The number following the letter n is the number of processors to run on This number can be any integer between one and the maximum number of processors available on the system However depending on the forest size SPATE HPC limits the number of processors that can be used in a particular simulation The forest area is divided amongst the processors into rectangles of different width and height depending on the shape of the compartments Any such rectangle must not have a width or a height smaller than 100 meters Since the size of the rectangles cannot directly be controlled by the user the number of processors to run on must simply be chosen small enough not create rectangles smaller than 100 meters Load balancing SPATE HPC can dynamically rebalance the load amongst the processors as the simulation is run ning Because the initial domain decomposition is solely based on the area the relative processor load might become uneven due to uneven tree density which can be a result of uneven growth or felling An uneven load balance leads to longer execution times Iteration load balancing coun teracts this by adjusting the inter process boundaries and transferring trees from heavily loaded processors to more lightly loaded processors Compartments are described in section 3 3 Since the rebalancing procedure takes some time to car
23. SPATE HPC User Manual December 20 2009 Kristoffer Paro Department of Information Technologies Abo Akademi University Contents Table of Contents List of Figures 1 Introduction 1 1 Programs input and output files ees 2 Compiling and running the programs 2 1 22 2 3 2 4 Compiling SPATE HPC from source code 0004 Running SPATE HPC 2 ees 2 4 Running on multiple processors lees Compiling SPATE GUI from source code o o Running SPATE GUL orat Sad ds a n 3 The simulator 3 1 3 2 3 3 3 4 3 5 3 6 3 7 3 8 3 9 Jesu ned A Rs Jorge A ASE al Spatial inter tree dependence o o e Compartments n pad a A ar ee eg ee ae a Simulationmodel s s osia a ts sa do a ES 3A Vanables sos sor terass da a A uus 3 4 2 Environmental effects e Simulation phases ius vos ue ip A e a a 3 5 1 Imnitial stand 4 44 sese or Romo ym 3 5 2 TOWER a dede 3 5 37 Mortality ccs Aye epus m bd DR hne Bae ake Eee 3 5 4 Reproduction Mark transformation oaoa SPECIES sitiar ds Bae ah et dk cb eee de eo bete es i EOTESCINDES Fai toe SP a aR No SN all ta Soci Ske AAS Mo ad Management methods 2 2 20 e 3 9 1 Dimension and energy wood cutting 000 3 9 2 Weibull thinning 000 000000 iii OOo ta Bu y 3 9 3 When is thinning carried out lens 20 IDA Clear cut ea ah eee
24. The local correlation factor see section 3 4 2 Iteration load balancing Whether to readjust process boundaries during simulation in order to obtain a better load balance amongst processors This option is only meaningful if SPATE HPC is run in a parallel environment Allow imbalance percentage Since rebalancing takes some time to perform the load balancing procedure will allow a slight imbalance amongst processors which is controlled with this parameter Use calculation inspection Enabling calculation inspection will output very detailed information about calculations performed during a simulation It is not recommended for other than very small simulations when one needs to verify e g growth models since it produces large amounts of log data and slows the simulation process down considerably Use interval One may specify a certain iteration interval when calculation inspection is to be used The interval is inclusive Elements There is one checkbox for each type of model or other calculation one can inspect 4 2 Regeneration page The Regeneration page contains the coefficients of the regeneration model see figure 4 2 which is used to calculate the number of new trees in the reproduction phase Although not explicitly stated the variables units are per hectare The regeneration model is described in section 3 5 4 4 3 Reproduction Growth and Mortality pages The Reproduction Growth and Mortality pages cont
25. The total volume of dead trees per hectare New tree volume The total volume of new trees per hectare Removal The total volume of trees removed per hectare Removal species The total removal volume of the species species per hectare Removal assortment The total removal volume of the assortment assortment per hectare 99 66 Removal species assortment The total removal volume of the species species and assort ment assortment per hectare Final volume The final volume per hectare 59 Growing stock species The final volume of the species species per hectare Growing stock assortment The final volume of the assortment assortment per hectare Growing stock species assortment The final volume of the species species and assortment assortment per hectare Mean diameter The mean diameter in centimeters Mean height The mean height in meters Mean volume The mean trunk volume in liters Number of trees in compartment The total number of trees in the compartment age class The total number of trees in the compartment in age class age class species The total number of trees in the compartment of species species Polygon file The filename of the corresponding polygon file 60 Index ACML 3 allowed imbalance percentage 6 assortment 21 36 calculation inspection 25 Cholesky decomposition 13 clear cut 20 35 command line 4 compartment 9 37 co
26. able can be used several times in different configurations and with different exponents Double clicking on a Beta label brings up a window for editing the corresponding secondary level model or smoothing function A screenshot of this window with a sample secondary level 26 EM SPATE GUI General Regeneration Reproduction Growth Mortality Initial forest Forest types Species Management Assortments Compartments Notes Reproduction parameters I Number of marks 2 pou r Mark 1 Name Diameter z Enviromental effects Theta0 Thetal Model Y 0 7 0 05 x Epsilon Edit Clear Mark 2 Name Height Sd r Enviromental effects Theta0 Thetal m Model 1 Y 0 5 0 4 x Diameter 0 35 x Diameter 2 0 15 x Epsilon Edit Clear Figure 4 3 Page showing the reproduction models 27 New model Beta O x Intercept zu Add variable Betal x Diameter 3 Add variable 2 Beta 2 x Diameter y Local density zu Add variable x Beta3 x Epsilon 1 Add variable Add model term i Cancel Figure 4 4 The model editor showing the model Y By f Diameter Bo Diameter Local density B3 Epsilon The coefficients Beta 0 through Beta 3 are shown in bold font which indicates that they have been assigned secondary level models Smoothing function Beta 1 0 4 Intercept 1 Add var
27. ain the models for calculating reproduc tion growth and mortality respectively The three pages are analogous and are therefore described collectively in this section The Number of marks input field is used to select how many marks zero one or two should be calculated in the corresponding phase If the value is set to zero the phase is disabled altogether The Reproduction page is shown in figure 4 3 25 E SPATE GUI v 0 0912 svn726 General eproduction Growth Mortality Initial forest Forest types Species Management Assortments Compartments Notes Regeneration model 35 Intercept 0 Volume 0 Basal area 0 Trees 0 Epsilon 8500 Inverse Volume 0 Inverse basal area 0 Inverse trees Figure 4 2 The Regeneration page showing the model for calculating the expected number of new trees per hectare in the reproduction phase Each mark has its own separate model which is represented by a by panel labeled Mark within the editor window The Mark panel contains a Name drop down menu and the sub panels Environmental effects and Model The Name menu should be set to the appropriate mark There cannot be two models of the same mark The fields Theta0 and Thetal within the Environmental effects panel represent the coefficients 09 and 0 in equation 3 4 respectively The actual model is shown in the Model panel In the same panel ar
28. an be used in the secondary level model w are the following Intercept The constant 1 Management A value given the management method of the tree s compartment If the compartment is assigned several management methods the value of the first one is used Species A value given the tree s species Age The tree s age in years Age class The tree s age class as an index 0 1 Size class The tree s diameter size class as an index 0 1 Height class The tree s height class as an index 0 1 Indicator variable species Replace species with the name of a species The variable will as sume the value 1 if the species matches that of the target tree and 0 otherwise Epsilon A normally distributed random number 0 1 The size class indices all start from zero In other words if a tree belongs to the first age class the variable Age class will assume the value 0 and if it belongs to the second class the variable will assume the value 1 etc 12 3 4 0 Environmental effects Environmental effects are correlated random numbers that are generated in such a way that random numbers for trees in close proximity are more strongly correlated than those for trees farther apart The covariance between two trees 7 and j is calculated with the following function exp 0o 01rj if Rij lt r Pies ap 3 4 0 otherwise where 0o 01 lt 0 are parameters chosen by the user R is the distance between the trees an
29. an this distance sparedTreesMinDiameter The minimum diameter of trees to spare sparedTreesMaxDiameter The maximum diameter of trees to spare classWidth The width of size classes when classifying trees If not given the default value is 4cm 5 1 8 Assortment parameters The assortmentParameters element may contain one or several assortment elements An assort ment element must have a name element and its value must be unique There may be zero one or several constraints elements which each one specifies constraints for a particular species The species is indicated by the species attribute If no species attribute is given the constraints are applicable to all species Figure 5 12 shows the assortment element and its child elements 0 many assortment container constraints species string gt container Figure 5 12 An assortment element defines a specific assortment The constraints element may contain the minimumDiameter and logLength elements that dic tate the minimum log diameter and the log length respectively If none of the elements are present no constraints will be imposed and the assortment can be used for the top of the tree 53 5 1 9 Compartment parameters The compartmentParameters element must contain one or several compartment elements Each compartment element uniquely identifies a particular compartment A compartment element should have a id attribute that contains a unique identifier which is
30. cation parameters given by the user 15 Empirical data The user provides a set of data points from which a cumulative distribution is built The endpoints of the support are the smallest and the largest data points given The data format is described further in section 5 3 Empirical cumulative distribution The user provides an arbitrary cumulative distribution as a set of x F x pairs The data format is described further in section 5 4 3 7 Species There can be one or several tree species and every tree is of a specific species Species are defined in the parameter input file which means that users are free to define their own species It is possible for different species to have different growth mortality and reproduction behavior and trees of different species may have different stem volumes Each species has a model value The model value is used for the variable Species in the second level models This way different species can be modeled to experience different growth mortality and reproduction Two different methods for estimating a tree s trunk volume is used taper curve functions and a cylinder model The cylinder model is only used when the tree is extremely small whereas taper curves are used in all other cases The taper curve functions used are the ones suggested by Laasasenaho which provide an estimate for the tree s diameter at any arbitrary height position The relationship between the diameter at an arbitra
31. contains parameters for generating trees for the initial stand see figure 5 5 The element should contain one of the elements load or generate or both If a load element is given its text content will be interpreted as a filename of a tree data file in the format described in section 5 5 The file s content will be read and used as the initial stand If a generate element is given trees will be generated based on the parameters provided therein If both a load and a generate element are given data available in the file will be loaded whereas data not available will be generated based on the generation parameters The generate element should contain all of the following three elements pointProcess This is a container element which should contain one of the following elements ho mogeneousPoisson inhomogeneousPoisson homogeneousClustered inhomogeneousClus tered homogeneousRegular or inhomogeneousRegular The child element determines the point process used to place new trees The point process elements and their child elements are listed in section 5 1 4 averagelnitialAge The element s value should be an integer stating the average age of trees in the initial stand initializationParameters This element should contain mark elements that provide the models for 45 calculating mark values of trees in the initial stand Each mark element should contain a set of modelTerm elements and optionally an environmentalEffects container 5 1 3 Mode
32. creating and editing the XML input parameter file the input file editor SPATE GUI can be used SPATE GUI is a separate graphical user interface application that can be run on the user s workstation If the user also runs SPATE HPC on the same computer as SPATE GUI the simulator can use the input files directly from the directory where they were created otherwise the input files must be copied to the computer that runs the simulation Figure 1 1 shows the relationship between the programs and how they read and write files references polygon file s empirical distribution files Simulator parameter file XML Workstation reads and writes SPATE GUI writes Figure 1 1 The relationships between SPATE GUI SPATE HPC the input and the output files y i n Chapter 2 Compiling and running the programs 2 1 Compiling SPATE HPC from source code SPATE HPC can be compiled from source code with the GNU Compiler Collection GCC C compiler or the Portland Group PGI C compiler using the provided makefiles It can be com piled on Unix based systems e g Linux or on Microsoft Windows using Cygwin The following is a list of third party libraries that are dependencies of SPATE HPC Some of them are required whereas others are optional and can be enabled or disabled via conditional compilation flags The AMD Core Math Library ACML is mandatory and should be at least version 3 6 0 ACML is available from htt
33. ction feature might cause a slight per formance downgrade even when not actually used by the user One can hence leave it out of the program altogether by disabling it in the makefile if the performance is crucial The flag COMPARISONSEEDING is a debugging feature is off by default and when enabled it makes the simulator artificially seed the random number generator in order to produce comparable simu lation results To compile SPATE HPC run the command make from the base directory of SPATE HPC in a command line prompt If the compilation process succeeds an executable file named spate hpc on Unix or spate hpc exe on Windows will be created in the base directory This is the program file used to run the simulator and it may be freely moved to another directory if the user so desires Ifthe makefile has been altered after a partial or full compilation it is necessary to clear out old object files before recompiling the program This is done by first running make clean Then make can be run again 2 20 Running SPATE HPC SPATE HPC is a command line program i e it must be run from a command prompt SPATE HPC requires one command line argument the name of the XML parameter file On a single processor system SPATE HPC can be run with the following command spate hpc input file xml Here the current working directory is the one where the executable spate hpc resides and the XML input file is named input file xml
34. d EIN eB A modelTerm element corresponds to a term in a model function The elements homogeneousPoisson inhomogeneousPoisson homogeneousClus tered and inhomogeneousClustered 2 The elements homogeneousRegular and inhomogeneousRegular A species element contains parameters specific to a certain species A forestType element defines a particular forest type A managementTemplate element defines a set of management rules An assortment element defines a specific assortment A compartment element contains compartment specific parameters The inspection element enables more detailed inspection of the simulation process 41 42 44 45 46 47 48 49 50 52 53 54 55 Chapter 1 Introduction SPATE HPC Spatio Temporal Stand Simulator with High Performance Computing is a tree pop ulation dynamics simulator that simulates each tree individually Forest growth and silvicultural treatments can be simulated in order to obtain statistics on wood raw material volumes SPATE HPC is a highly scalable parallelized computer program that can be run both on small home computers as well as on large supercomputers SPATE GUI is an input file editor that can be used to produce simulation parameter files for SPATE HPC SPATE GUI is a stand alone program that can be run on home computers This user manual describes how the simulator works how to use it in o
35. d r is the competition distance The fundamental algorithm for calculating the correlated random numbers is to construct a variance covariance matrix whose cells are given by formula 3 4 The matrix is then de composed into a lower triangular matrix with the Cholesky decomposition L LI M Finally correlated random numbers are obtained by multiplying the lower triangular matrix by a vector of uncorrelated normally distributed random numbers Lez z N 0 1 Because Xe must be a positive definite matrix in order to perform a Cholesky decomposition the values of 9 and 04 must be chosen so that 2 is positive definite in all situations Due to technical limitations the full variance covariance matrix and the lower triangular ma trix for all trees in the forest cannot be calculated Instead the correlated random numbers are ap proximated by performing multiple smaller local Cholesky decompositions The local Cholesky decomposition algorithm performs one decomposition per tree where the variance covariance matrix is constructed of only the immediate neighboring trees of the target tree The trees included are the ones that are located within the competition distance multiplied by the ocal correlation factor The default value of the local correlation factor is 1 but it can be increased by the user Increasing it to a higher value will enlarge the local variance covariance matrix and thus likely decrease the approximation error but a
36. d reproduction phases by assigning diameter factors Initial stand transformations One may provide target distributions for the diameter and height marks that are applied in the initial stand generation phase Reproduction transformations One may provide target distributions for the diameter and height marks that are applied in the reproduction phase 3 9 Management methods Management methods silvicultural treatments are applied to compartments in order to thin or harvest the stand by removing trees according to a specific procedure The management methods implemented in SPATE HPC are dimension cutting energy wood cutting low thinning selective thinning and clearcutting 17 Management methods are assigned to compartments via management templates A manage ment template may contain one of dimension cutting energy wood cutting low thinning selective thinning or none of them In addition a management template may contain a clear cut A com partment can associated with none one or several management templates Furthermore the same management template can be assigned to several different compartments 3 9 1 Dimension and energy wood cutting Dimension and energy wood cutting operate in the same way but their results are diametrically opposite Dimension cutting is only given one parameter diameter limit and when dimension cutting is carried out all trees with a diameter wider than the diameter limit are cut down Con versely
37. e also the buttons Edit and Clear which are used to modify and empty the model respectively On the Mortality page there is an additional field called Mortality threshold The mortality threshold is used to control the regular mortality whereas the mortality models are used for the irregular mortality The reproduction model is described in section 3 5 4 the growth model is described in sec tion 3 5 2 and the mortality model is described in section 3 5 3 4 4 Editing a model Models are edited with the model editor dialog which is brought up by clicking the Edit button within a mark model panel The user is first presented the main part of the model which corre sponds to equation 3 2 A screenshot of the window for editing the main part of a model is shown in figure 4 4 The terms of the model are listed vertically where each term comprises a smoothing function coefficient represented by a Beta label and a set of variables Here the Beta coefficients cor respond to and the variables to x in equation 3 2 New model terms are added by clicking the Add model term button and existing terms are removed by clicking the X button at the right end of the term s row Furthermore new variables are added to a term by clicking the Add variable button and existing variables are removed by selecting the Delete variable option from the variable s drop down menu Naturally the same vari
38. e log file 20 3 10 Assortments Assortments are categories of wood raw material that are obtained by cutting the stem of a felled tree into logs for example saw log and pulpwood If assortments are defined SPATE HPC will compile assortment specific volume statistics after a simulation has completed An assortment may have a set of size criteria minimum diameter and log length that dictate how the assortment shall be calculated Additionally there may be different size criteria for different species SPATE HPC calculates assortment volumes of a felled tree according to the following algo rithm It starts at the height of the stump and progresses toward the top of the tree The program tries to cut a log according to the first assortment s length criterion If the diameter at both ends of the log is at least as wide as the minimum diameter criterion the log is cut off the stem and the program continues by trying to cut another log of the same assortment If the size criteria are not fulfilled the log will not be cut and the program switches to the next assortment and tries to cut logs according to that one s criteria The process continues until there is no stem left or there are no assortments whose size criteria can be fulfilled If an assortment does not have size criteria it can be used for the top of the tree When the program switches to an assortment without size criteria the rest of the stem all the way to the top of the tree will b
39. e the trunk of pre existent ones Finally the new trees are assigned marks according to a model of formula 3 2 and the mark values are transformed to target distributions analogously to the corresponding process in the initial stand Section 3 6 contains more information about mark transformations Regeneration model The regeneration model which is used to determine the expected number of new trees in the reproduction phase is a linear model of the form n max 0 azo ba cz2 3 7 14 where a b etc are constants and zo 11 etc are variables The model is hence restricted to produce only non negative results The available variables which cannot be combined or expo nentiated are the following Intercept The constant 1 Volume The volume of all trees in the compartment per hectare m ha Basal area The basal area of all trees in the compartment per hectare m ha Trees The total number of trees in the compartment per hectare 1 ha Inverse volume The inverse of Volume Inverse basal area The inverse of Basal area Inverse trees The inverse of Trees Epsilon A normally distributed random number N 0 1 The value is limited to the standard normal limiting factor 3 6 Mark transformation When new tree marks diameter and height are generated for the initial stand and in the repro duction phase they will be normally distributed with an infinite support if epsilon or xi are used in the
40. e thinning can be applied Thinning is carried out if the compartment s basal area is larger than the lower basal area and the unit of the lower basal area parameter is square meters per hectare m ha Analogously to the lower basal area the lower compartment volume is the smallest volume the compartment may have before thinning can be applied The unit of the lower compartment volume is cubic meters per hectare m ha If both the lower basal area and the lower compartment volume criteria are specified thinning will be carried out whenever either of them is fulfilled The thinning interval is the minimum number of inactive years between two subsequent thin nings In practice the actual interval between two thinnings might be longer than the number of years given in case the other criteria lower basal area or minimum volume have not been satis fied earlier The thinning interval is only enforced between two subsequent thinnings and is thus not affected by the stand s initial age The first thinning might very well take place the first year of the simulation 3 9 4 Clear cut Clearcutting cuts almost every tree in the compartment Clearcutting is not affected by the criteria described in section 3 9 3 Instead the frequency of clear cuts is specified with the years between clear cuts parameter The first clear cut takes place when the stand has reached an age that is a multiple of the years between clear cuts parameter For example if the age
41. e used for that assortment 3 11 Output files SPATE HPC produces four types of output files during and after a simulation compartment statis tics files statistics table files tree data files and log files There is one compartment statistics file for each compartment which is named simulationname_statistics_compartment_xxxx dat where simulationname is replaced by the name of the simulation and xxxx by the correspond ing compartment s numerical ID A compartment statistics file contains statistics pertaining only to that compartment In addition to the compartment statistics files the simulator also produces a statistics table file The statistics table file is named simulationname statistics table dat and contains a table of the final characteristics of all compartments Compartment statistics files are described further in section 5 6 and statistics table files in section 5 7 Tree data files are files containing the characteristics of all simulated trees in a table format Tree data files are optional and can be produced for each replication at the end of the simulation and additionally for each replication at the end of each iteration step The files produced at the end of the simulation are named simulationname replication xxxx dat where xxxx is the replication number and the files produced after each iteration step are named simulation name replication xxxx iteration yyyy dat where yyyy is the ite
42. ed Ee ER RE ee oe ee ae 20 3 10 Assortments 2 2 ei Ge ee Ue o e 21 SL Output files 32 4 4 xe bt eed rep He ar A A 21 4 The input file editor 23 Al General page eis da LS e o RO do eo adu as Bee UR ce e 23 4 2 Regeneration page ls 25 4 3 Reproduction Growth and Mortality pages o 25 4 4 Editing a model io ae Sal far Ge a eun Ped 26 4 5 Initial forest page ve 6 id xe har BG be a a a a he a 29 4 6 Foresttype page as te oes a ee GES ae Ee wc ee e Ode a 29 47 SPECIES Pages iso some es arta a ao p eh Leo AR a 32 4 8 Management os ou eux oe p n NW eru due p RR ERO es 32 4 9 Assortments page 36 4 10 Compartments page 37 4 11 Notes page i ie Es ee ee ea E A 38 5 File formats 39 L Parameter file 3 konnte Barta ae OR E RUN Rd 39 SAT Simulation 2 3s e eae ge BGS al bee v e ke a 40 5 1 2 General parameters es 40 311 3 Model ferms uos aD i RG ae ae co we a 46 5 1 4 Peintprocesses zoe e o Uo ome 47 5 1 5 Species parameters een 47 5 1 6 Forest type parameters es 48 5 1 7 Management parameters e 51 5 1 8 Assortment parameters es 53 5 1 9 Compartment parameters 0 02000 0000 54 5 1 10 Inspection parameters ees 54 2 2 Polygon file 22e uno er tbe dS arte St d 56 5 2 1 Cut outregions oo ce ea es 56 5 3 Empirical data coa e e E e e REESE 57 5 4 Empirical cumulative distribution e
43. egeneration Reproduction Growth Mortality Initial forest Forest types Species Management Assortments E Simulation area Area Mestmap txt y Load area Compartments testmap txt testmap txt Mestmap txt testmap txt pine dominated Mestmap txt testmap txt Management testmap txt M Select estmap txt ctiveThinning Mestmap txt Mestmap txt Mestmap txt testmap txt testmap txt testmap txt testmap txt testmap txt Number of initial trees factor testmap txt Er j Mestmap txt Merge compartments Age of initial trees factor o2 When merging compartments the resulting merged compartment will be assigned the parameter values of the compartment with Polygons the lowest ID Figure 4 11 The compartments page showing the forest area A compartment consisting of three polygons is selected Selecting a compartment from the list or clicking on the corresponding polygon in the area view will show its settings in the right part of the window A compartment has the following configurable settings 37 Include in simulation Whether the compartment should be included in the simulation Compartment id Each compartment should be given a unique ID number If not explicitly changed the default is the polygon ID Forest type The forest type of the compartment Management Here
44. er lt transformation mark string gt container Figure 5 10 A forestType element defines a particular forest type A transformation element see figure 5 10 contains the parameter of a target distribution of a transformation It should contain one of the following child elements weibullDistribution normalDistribution empiricalData or empiricalCumulativeDistribution The weibullDistribution 50 element which designates a theoretical Weibull distribution should contain the elements scale and shape and may contain the element origin The normalDistribution element which designates a theoretical normal distribution should contain the elements mean and standardDeviation The elements empiricalData and empiricalCumulativeDistribution should contain a filename that refers to a file containing empirical data or an empirical cumulative distribution respectively The file format of these files are described in sections 5 3 and 5 4 5 1 7 Management parameters The managementParameters may contain a set of managementTemplate elements see figure 5 11 Each managementTemplate contains a set of rules for carrying out management and can be referred to by a compartment A managementTemplate element contains the following child elements thinningIntervalInYears If specified the value of this element dictates the minimum number of years between two consecutive thinning instances This rule only applies to non clear cut methods lowerBasal
45. er DN 1 R Diameter sum The sum of all competitors diameters gt T di Diameter difference The sum of the differences between the target tree s diameter and the competitors diame ters Y di di Height difference The sum of the differences between the target tree s height and the competitors heights DT hi Height diameter ratio The quotient of the target trees height and diameter h d Basal area of competitors The competitors total basal area relative to a hectare 10000B A Relative basal area The target tree s basal area relative to the competitors basal area 7 d 200 B Diameter quotient The sum of all larger competitors diameters divided by the target tree s diameter gt gt A di d 11 Basal area to mean diameter ratio The competitors basal area divided by the mean diameter of the competitors B A d N Tau The target tree s initial heterogeneity T Xi A correlated random number Epsilon A normally distributed random number It is drawn anew for each calculation N 0 1 The variables referred to are defined as N Number of competitors r Competition distance m Ri Distance between target tree and neighbor i m d Diameter of target tree cm h Height of target tree m dj Diameter of competitor cm h Height of competitor i m A Competition area m zr B Basal area of competitors m B 7 37 d 200 The variables that c
46. er curve 32 thinning interval 20 35 thinning level 19 35 thinning rate 18 tree data file 13 21 volume 16 32 Weibull thinning 18 wxWidgets 6 Xerces 3 XML 39 year 25 years between clear cuts 20 62
47. erforms a simulation as an iterative process consisting of a sequence of time steps The length of a time step is chosen by the user but is at least one year long A time step is further decomposed into a series of individual phases which each governs a certain aspect of the stand development The phases are growth mortality reproduction and management In the growth phase pre existent trees grow in the mortality phase trees die in the reproduction phase new trees are generated and in the management phase silvicultural treatments are carried out The changes introduced in the growth mortality and reproduction phases do not take affect until after all three phases have been performed When the growth mortality and reproduction phases have been performed and the changes applied the management phase is carried out SPATE HPC can run several replications which are simulations with identical initial condi tions both with differently drawn random numbers The results of the different replications are combined for the final statistics in order to produce figures within a Monte Carlo confidence inter val Figure 3 1 shows a diagram of the simulation workflow 3 1 Tree In SPATE HPC each tree is simulated individually A single tree is also the smallest entity within a simulation A tree can be viewed as collection of attributes or characteristics which are the following X and Y coordinates Two real numbers that represent the tree s coordinates in
48. est types page showing the forest type spruce dominated which con tains 80 spruce and has diameter transformations defined for both the initial stand and the reproduction phase ees The Species page Here the species Pine is selected and its settings are shown in the panel to the tight o o o 0 020000000 Management templates are defined on the Management page The selected tem plate LowThinning is set to carry out low thinning removing 50m ha every 13 year if the compartment volume exceeds 200m ha o o o Assortments are defined on the Assortments page Here the assortment sawlog is selected and its size criteria for different species are shown in the panel to the The compartments page showing the forest area A compartment consisting of three polygons is selected o o e e 111 2 26 27 28 28 30 31 33 34 36 5 1 5 2 5 3 5 4 5 5 5 6 5 7 5 8 5 9 5 10 5 11 5 12 5 13 5 14 XML elements are illustrated with these types of boxes The root element simulation and its direct children The generalParameters element contains assorted simulation parameters Parameters governing growth mortality reproduction and regeneration initialCompartment contains parameters for generating or loading trees for the Initial Stands 21 euo opus Gap df aha Seapets cA RR UN
49. he small volume coefficient which is species dependent and has the default value 0 6 A species also has a competition distance model Different species may have different compe tition distance models and the computed value from competition distance model might be different for different trees However all trees have the same effective competition distance which is the maximum value obtained by calculating the competition distance of all trees in the forest The competition distance model takes the same form as model 3 2 and use variables and secondary level variables described in section 3 4 1 However the model calculation cannot take neighboring trees into account which makes only the following variables useful in this model Intercept Diameter Diameter log Height diameter ratio Tau and Epsilon All the secondary level variables can be used though 3 8 Forest types A forest type defines a particular composition of tree species and a compartment is of a specific forest type A forest type has the following attributes Name A unique name that is used to associate a forest type with a compartment List of species Each species is assigned a number that corresponds to its relative amount in the initial stand If desired the species can be assigned a separate relative amount for the reproduction phase Furthermore the species mean diameter can be offset from the global one in the initial stand an
50. iable 10 7 Species 1 Add variable Figure 4 5 The smoothing function editor window showing the secondary level model 5 0 4 0 7 Species 28 model is shown in figure 4 5 Analogously to the main model window the terms of the secondary level model are listed vertically where each term comprises a coefficient and a set of variables The coefficients correspond to 3 and the variables to w in equation 3 3 The procedures for adding and removing variables and adding and removing terms are identical to the ones used in the main model window When a new model is created from scratch or when new model terms are added to an existing model the model terms smoothing functions will be uninitialized This is indicated by the font of the Beta label a normal font means the function is uninitialized whereas a bold font means the function has been initialized In order to be able to save the model all smoothing functions must be properly initialized 4 5 Initial forest page The settings for creating an initial stand are located on the Initial forest page see figure 4 6 The user may choose to load the initial stand from a tree data file or generate it Selecting a tree data file is done by browsing for it using the Load initial forest data from file input field The tree data file should be in the file format described in section 5 5 If the trees are to be generated one of the options Homogeneous Poisson Inh
51. ide of the window there is a list of the assortments that are currently defined Selecting an assortment in the list brings up its settings on the right side of the window A new assortment can be created by clicking New and an existing assortment removed by clicking Delete An assortment must be given a unique name using the Name input field and a set of size criteria Species specific size criteria are defined by clicking the Add button The size criteria that can be given are Minimum diameter and Log length If these fields are left empty no size criteria are enforced and the assortment can be used for the top of the tree One can specify the species for which the criteria apply by selecting it in the list If no species is selected the size criteria will apply to all species 36 4 10 Compartments page Compartments are set up on the Compartments page see figure 4 11 There must be at least one compartment defined in order to run the simulation Compartments are described in section 3 3 First of all a map must be loaded This is done by clicking the Load area button and selecting a polygon file that conforms to the specification in section 5 2 Upon successfully loading a polygon file the area will be presented in the main part of the window where polygons are enclosed with black lines The compartments are also listed in the left part of the window IIl SPATE GUI v 0 0912 svn726 DER General R
52. imulation project move or copy polygon and other external files to that directory and then use SPATE GUI to create the input parameter file in that directory 4 1 General page On the General page are settings controlling general aspects of the simulation Here are also the Save settings and Load settings buttons located which are used to save and load the XML input file respectively A screenshot of the General page is shown in figure 4 1 The settings that are configurable on the General page are the following Output directory The directory where output files should be created If a relative path is given SPATE HPC will create the output directory relatively to the directory from which SPATE HPC is run Simulation name A name describing the simulation Output files will be prefixed with this name Save tree files Whether to save tree data files Enabling this option saves tree files for each repli cation at the end of a simulation Save tree files for each iteration Whether to save tree data files after each iteration step as well This feature is likely to consume large amount of hard disk space and slow down the simu lation process noticeably Size classes Size classes are used to classify trees according to DBH Age classes Age classes are used to classify trees according to age 23 E SPATE GUI v 0 0912 svn726 General Regeneration Reproduction Growth Mortality Initial forest Fores
53. ion re generation or competition distance An element with one of the aforementioned values enables calculation inspection of the corresponding model inspectThinning There may be none or multiple inspectThinning elements which each may con tain one of the values weibull iteration or weibull thinning probability 54 lt inspectionParameters gt container Figure 5 14 The inspection element enables more detailed inspection of the simulation process iterationInterval If given the element should have a first and a last attribute Inspection is then enabled only for the inclusive iteration interval given by the attributes 55 5 2 Polygon file SPATE HPC uses polygon files to identify the polygonal areas that constitute compartments There must exist at least one polygon file in a simulation A polygon file is a plain text file con taining the coordinates of one or several polygons The coordinate system is a two dimensional Cartesian coordinate system where the units are meters A polygon has a unique ID number that is used to associate it with a compartment and a list of vertices points The vertices must be given in order so that a coherent polygonal shape can be outlined by drawing consecutive straight lines between the vertices The last vertex must coincide with the first one This way the last line will end exactly where the first line started thus enclosing the polygon The minimum number of vertices
54. ith arrows There are three color codes for the elements light green blue and maroon see figure 5 2 A light green element represents a container element It does not contain a value itself but is merely a container for other elements A blue element contains a value The type of the value is written on the last line within the box Maroon elements are mutually exclusive when a set of child elements have a maroon indicator color only one of the siblings may be chosen Maroon elements are either container or value elements Elements may also have attributes This is indicated by a name value type string within the element box See the blue element in figure 5 2 The vertical structure of the XML file is illustrated with arrows An arrow from one element to another denotes that the latter element is a child of the former There is also a cardinality label accompanying the arrowhead which states how many instances of the child element that may occur See http www w3 org XML for the XML specification 39 5 1 1 Simulation The root element of the XML file is named simulation see figure 5 2 Its child elements are note inspectionParameters generalParameters speciesParameters forest TypeParameters man agementParameters assortmentParameters and compartmentParameters The element note is ig nored by the simulator It can be used by the user to attach explanatory notes to the input file The other elements contain parame
55. l terms A modelTerm element is a child of a mark or a competitionDistanceModel element A modelTerm element corresponds to a term in a model function which contains variables and a coefficient There may be any number of sibling modelTerms inside a mark or competitionDistanceModel container whereupon each modelTerm element represents a separate term see figure 5 6 lt modelTerm gt container many lt termCoefficient gt container many lt smoothingFunctionTerm coefficient double gt container many Figure 5 6 A model Term element corresponds to a term in a model function model Term elements correspond to the terms in formula 3 2 A modelTerm must contain a termCoefficient element which represents the term s coefficient and a set of model Variable elements which represent variables A modelVariable element has a name attribute whose value should be one of the variable names listed in section 3 4 1 Additionally a model Variable may have a order attribute whose value is the order to which the variable is to be exponentiated Moreover variables can be multiplied by each other by having several sibling modelVariable elements A termCoefficient element corresponds to a secondary level model as in formula 3 3 A termCoefficient element has a set of smoothingFunctionTerm elements that represent secondary level terms and each term contains a set of smoothingFunction Variable elements that represent secondary
56. level variables A smoothingFunctionTerm has a coefficient attribute that represents the term s coefficient A smoothingFunction Variable has a name attribute whose value should be one of the secondary level variables listed in section 3 4 1 Secondary level variables can also be exponentiated by pro viding the smoothingFunction Variable with an order attribute and they can also be multiplied by each other by providing several sibling smoothingFunction Variable elements 46 5 1 4 Point processes The elements homogeneousPoisson inhomogeneousPoisson homogeneousClustered and inho mogeneousClustered and their child elements are shown in figure 5 7 The elements homogeneous Regular and inhomogeneousRegular and their child elements are shown in figure 5 8 0 1 0 1 lt estimatedPoissonRegression Coefficients gt container lt estimatedPoissonRegression CoefficientsForMotherPoints gt container Figure 5 7 The elements homogeneousPoisson inhomogeneousPoisson homogeneousClustered and inhomogeneousClustered 5 1 5 Species parameters The speciesParameters element contains a set of species elements The content of a species ele ment defines characteristics that are specific to a certain species see figure 5 9 A species element may contain the following elements name The name element is obligatory It should contain a unique name identifying the species secondaryModel Value If given it should contain a numerical val
57. lso prolong the simulation time 3 5 Simulation phases This section describes how the simulation phases initial stand growth mortality and reproduction are performed 3 5 1 Initial stand The initial stand can either be generated or loaded from a file The file may also be incomplete i e it may lack certain tree characteristics whereupon the missing characteristics are generated with a model If no tree data file is given the first step is to generate tree positions for the new trees The number of new trees to create is drawn as a Poisson random number with a mean value provided by the user The trees are then placed by a point process When the trees have been placed they are assigned tree marks diameter and height The tree marks are calculated according to models given by the user which follow the form of formula 3 2 After a mark has been calculated for all trees in the stand it can optionally be transformed to adhere to a given distribution The marks are calculated and transformed in order first the diameter is calculated and transformed then the height is calculated and transformed The height 13 calculation can hence be made dependent upon the transformed diameter Section 3 6 contains more information about mark transformations 3 5 2 Growth A tree grows by multiplying a mark s value of the previous time step by a growth factor The growth factor is the value of the exponential function with y from formula 3 2 as
58. models To allow generated marks to adhere to an arbitrary distribution and to prevent negative mark values marks can be transformed to adhere to a user defined distribution The transformation process begins by building a cumulative distribution of the calculated un transformed mark values This is done when mark values for all trees in the entire forest have been calculated The second step is to build a cumulative distribution according to the target distribution specified by the user Different compartments may have different target distributions Finally the mark values are transformed to adhere to the target distribution through cubic spline interpolation When mark values are calculated and transformed these steps are performed separately for the diameter and the height First the diameter values are calculated and transformed and then the height values are calculated and transformed This means that if the height model uses the diameter as a model variable the variable s value will be the transformed one The following are the options the user may choose for the target distribution Normal distribution A normal distribution with a mean y and standard deviation o given by the user The support is truncated to u u where 6 min u 3c i e the support is no more than six standard deviations wide but not allowed to encompass negative values Weibull distribution A Weibull distribution with the shape scale and origin lo
59. mpetition distance 9 model 17 32 competition index 9 competitor 9 compiling 3 6 correlated random numbers 9 13 covariance function 13 CppUnit 3 CSV 58 59 cut out region 56 density coefficient 19 35 diameter limit 18 35 dimension cutting 18 editor 23 empirical cumulative distribution 16 empirical data 16 energy wood cutting 18 environmental effects 4 13 forest type 17 29 GCC 3 6 growth 8 14 25 GSL 3 initial stand 13 29 input file 39 inspection 4 iteration length 25 linear tail 35 load balancing 5 25 local correlation factor 13 log file 21 log length 21 36 logistic regression 14 19 low thinning 18 19 lower basal area 20 35 lower compartment volume 20 35 makefile 3 6 management 8 method 17 32 template 18 32 minimum diameter 21 36 model 10 26 secondary level 26 value 16 mortality 8 14 25 threshold 14 MPI 3 5 MPICH 3 output directory 23 output file 21 PGL3 point process 13 Poisson random number 13 14 19 polygon 9 37 56 61 regeneration 14 25 removal per hectare 19 35 removal tolerance 19 35 replication 25 reproduction 8 14 25 secondary model value 32 35 secondary level model 10 selective thinning 18 19 size class 23 35 small volume coefficient 17 smoothing function 11 species 16 32 SPRNG 3 standard normal limiting factor 15 statistics 21 stump 16 32 tap
60. ng trees are referred to as competitors The competition distance is global for the entire forest but may vary between iterations 3 3 Compartments The entire forest may be subdivided into one or several compartments and a compartment com prises one or several polygons see figure 3 2 Different compartments may have different forest types and may be assigned different management methods The forest is spatially continuous across compartment boundaries i e trees in different compartments near compartment borders interact with each other as if no compartment boundaries existed The management methods however are applied separate to each compartment When a simulation has finished statistics are compiled individually for each compartment Each compartment is associated with a unique user defined numerical ID that is used to identify the compartment Figure 3 2 A sample forest area with 38 compartments consisting of 57 polygons 3 4 Simulation model Conventionally the models used in the growth mortality and reproduction phases as well as in the initial stand generation follow the pattern y xB For Fy 3 1 where x is a row vector of competition indices 3 is a column vector of coefficients 7 is the initial heterogeneity is a correlated random number o and y are coefficients and e is a normally distributed random number corresponding to the error term The models defined by the user however need not strictl
61. ntain a unique name identifying the forest type species There should be one or several species elements one for each species that occurs in the forest type The mandatory name attribute is used to refer to the species in question The element s content is the relative species ratio which is used when generating trees for the initial stand The optional attribute reproductionA mount is the relative species ratio used in 48 4euigjuoo lt SJa 9weJe qure d duns Jeujgjuoo une jepow gt Jeuigjuoo Jeurejuo2 epojyeouejsiquonneduioo s lt SJ0 J9wWeJegenunogiede gt Jeurejuoa lt seneds gt ueul Figure 5 9 A species element contains parameters specific to a certain species 49 the reproduction phase The optional attributes initialTreeDiameterFactor and new TreeDi ameterFactor are factors that the trees diameters are multiplied by in the initial stand and reproduction phase respectively initialStandTransformations This element may contain a transformation element for each mark that is to be transformed in the initial stand phase reproductionTransformations This element may contain a transformation element for each mark that is to be transformed in the reproduction phase 1 many lt forestType gt container lt initialStandTransforma lt reproductionTransform tions gt ations gt container container lt transformation lt transformation mark string gt mark string gt container contain
62. omogeneous Poisson Homogeneous clustered Inhomogeneous clustered Homogeneous regular or In homogeneous regular should be selected The selected option corresponds to the point process used to place the trees Moreover one should choose an average age of the trees in the Aver age initial age input field and an average number of trees per hectare in the Expected number of initial points per hectare input field The average initial age and trees per hectare parameters are common to the whole simulation but they can be readjusted for different compartments by specifying compartment specific factors The diameter and height marks of initial trees are generated according to the models speci fied on this page This is done by setting the Number of marks input field and adding models according the description in section 4 4 The mechanisms for generating the initial stand are described in section 3 5 1 4 6 Forest type page Forest types are described in section 3 8 A forest type is a composition of different tree species and there must be at least one forest type Forest types are edited on the Forest types page see figure 4 7 There must be at least one species defined in order for the forest types page to be usable On the left side of the window there is a list of the forest types that are currently defined Selecting a forest type in the list brings up its settings on the right side of the window
63. onParameters container element may con tain a set of variable elements which each has a name attribute and a value The value of the name attribute must be one of the variable names listed in section 3 5 4 and there must be no two vari able elements with the same name The value of the variable element corresponds to the variable s coefficient see figure 5 4 43 lt mortalityParameters gt 0 1 container lt reproductionParameters gt 0 1 container mark name string gt 0 2 container mark name string container lt environmentalEff ects gt container lt environment alEffects gt container lt modelTerm gt container lt modelTerm gt container lt growthParameters gt 0 1 lt regenerationParameters gt container mark name string gt container lt environmentalEff ects gt container lt modelTerm gt container Figure 5 4 Parameters governing growth mortality reproduction and regeneration 44 1 lt initialCompartment gt container 0 1 0 1 lt generate gt container 1 1 1 lt pointProcess gt lt initializationParameters gt container container 0 2 lt mark name string gt container many 0 1 lt modelTerm gt lt environmentalEffects gt container container Figure 5 5 initialCompartment contains parameters for generating or loading trees for the initial stand Initial compartment The initialCompartment element
64. p developer amd com Xerces C is mandatory should be at least version 2 8 Xerces C is available from http xerces apache org xerces c The Scalable Parallel Pseudo Random Number Generators Library SPRNG is mandatory and should be at least version 4 0 for C SPRNG is available from http sprng cs fsu edu The GNU Scientific Library GSL is mandatory and should be at least version 1 11 GSL is available from http www gnu org software gsl Message Passing Interface MPI is used to run SPATE HPC in parallel SPATE HPC can be compiled without MPI support but will in that case only be able to run on one processor and not in parallel MPICH is available from http www mcs anl gov research projects mpi CppUnit is used for unit tests and is not normally needed If unit tests are explicitly enabled at compile time however the CppUnit library must be present CppUnit is available from http sourceforge net projects cppunit SPATE HPC is compiled with the Make program In order to run Make copy the file named makefile sample and name it makefile The compilation process is configured by editing the makefile It is most likely necessary to edit the makefile to reflect the system set up in order for SPATE HPC to compile successfully The following is a list of makefile variables that can be changed CC The name of the compiler executable It should be set to mpicxx if the simulator is to be compiled with MPI suppor
65. ration number The format of the tree data files are described in section 5 5 Enabling iteration wise tree data files will likely consume large amounts of hard disk space and slow down the simulation process noticeably Log files are written continuously as the simulation is running and contains information about each simulation step as it is carried out Log files may hence prove a valuable resource in under standing the simulation process or tracking down error conditions There is one log file written by each processor and the corresponding log file contains only information related to the that pro cessor If SPATE HPC is compiled without MPI support the single log file is named output txt whereas if SPATE HPC is compiled with MPI support the log files are named xxxx output txt 21 where xxxx is the processor number 22 Chapter 4 The input file editor This chapter describes the input file editor SPATE GUI SPATE GUI can be used to create and modify the input parameter file which is in the XML file format Other input files such as polygon files tree data files and files containing empirical distributions cannot be edited with SPATE GUI and must be created using other programs Although the input parameter file can reference external files located in any location it is recommended for the sake of coherency that all input files be put in the same directory One can thus begin by creating a new directory for the s
66. rder to run simulations how to create input files and how to interpret output files In chapter 2 it is explained how to compile and run SPATE HPC and SPATE GUI The growth models simulation mechanisms and implemented features of SPATE HPC are described in chapter 3 How to use SPATE GUI to create and edit input parameter files is shown in chapter 4 And finally the different file formats of input and output files are detailed in chapter 5 1 4 Programs input and output files The simulator program SPATE HPC can be run either on the user s own workstation another computer or a supercomputer The simulator program is a command line application i e it has no graphical user interface It is run by providing it a set of input files and it produces a set of output files after the simulation has completed The main input parameter file is in the XML extensible markup language file format and serves as the main input file Other input files such as the polygon file that define the forest area and empirical data files for diameter and height distributions are referenced by the XML input parameter file When a simulation has finished statistics files on the stand s development and the final stand are compiled During the simulation the program also writes log files that contain information about the different simulation steps performed Additionally tree data files containing the characteristics of all individual trees can be produced To facilitate
67. ree marks in the initial stand and the reproduction phase are to be transformed to adhere to an arbitrary distribution this is specified in the Initial stand transformations and Reproduction transformations panels There can be one transformation for each mark diameter and height and they are specified by clicking the Add button The user may select one of the options Weibull distribution Normal distribution Empirical data or Empirical cumulative distribution If Empirical data is selected one should provide a file of the format described in section 5 3 and if Empirical cumulative distribution is selected one should provide a file of the format described in section 5 4 The process of transformation marks is explained in section 3 6 4 7 Species page Tree species are edited on the Species page see figure 4 8 There must be at least one species defined and species must be defined before defining forest types Species are described in sec tion 3 7 On the left side of the window there is a list of the species that are currently defined Selecting a species in the list brings up its settings on the right side of the window A new species can be created by clicking New and an existing species removed by clicking Delete The following is a list of the settings that are configurable for a species Name The species must have a unique name which is used to refer to the species
68. ry height and the diameter at 20 of the height is calculated using the model falx I b z box baa bsx baa bisz biz baya 3 8 where dj is the diameter at height l d 7 is the diameter at 20 height h is the tree s height and x 1 l h Since d gt is not known it is estimated as 3 di d h gt STR 3 9 eR estan oe where d 3 is the DBH Finally the diameter at an arbitrary height l is estimated as dy fid 1 l h 3 10 The coefficients b1 b2 034 are species specific and are provided by the user When the volume of a tree is calculated the calculation does not start from the ground but from the heigh position of an imagined stump The stump height A in meters is calculated using the model hs max a di 3 b h 100 hms 3 11 Jouko Laasasenaho Taper Curve and Volume Functions for Pine Spruce and Birch Finnish Forest Research Institute Helsinki 1982 16 where a and b are species specific coefficients given by the user and Ams gt 0 is the lowest possible stump The diameter is in centimeters and the height in meters Taper curve functions are sidestepped and a simple cylinder model is used instead when the tree is shorter than four meters or has a diameter less than three centimeters The cylinder model approximates the tree to a cylinder with the same height as the tree and with the tree s DBH as the cylinder s diameter In addition the cylinder volume is multiplied by t
69. ry out it is often beneficial to allow a slight imbalance among the processors and not perform a rebalance after every iteration step This allowed imbalanced is controlled with the allowed imbalance percentage parameter The imbalance percentage is the ratio between the difference in number of trees and the minimum number of trees of any processor When the imbalance percentage exceeds the allowed imbalance percentage a rebalance is carried out 2 3 Compiling SPATE GUI from source code The input file editor SPATE GUI can be compiled from source code with the GNU Compiler Collection GCC C compiler The program uses the widget toolkit wx Widgets for the graphical components which makes it possible to compile it natively on Microsoft Windows using Cygwin Linux and Mac OS X The only third party library dependency of SPATE GUI is wxWidgets which should be at least of version 2 8 wxWidgets is available from http www wxwidgets org Akin to SPATE HPC SPATE GUI is compiled with the Make program There is a sample makefile called makefile sample which should be copied an named makefile The makefile should then be edited to reflect the system on which SPATE GUI is to be compiled The following is a list of makefile variables that can be changed CXX The name of the compiler executable The default is g which is the GCC C com piler WX CONFIG The name of the wxWidgets compilation utility The default value wx config
70. should be appropriate on most systems TARGET OS This variable should be set to one of windows linux or mac depending on the operating system for which SPATE GUI is to be compiled EXTRA LIBS This variable can be used to include additional libraries LDFLAGS This variable is used to provide additional flags to the linker CXXFLAGS This variable is used to provide additional flags to the C compiler To compile SPATE GUI run the command make from the base directory of SPATE GUI in a command line prompt If the compilation process succeeds an executable file named spate gui on Linux or spate gui exe on Windows will be created in the base directory This is the program file used to run the program and it may be freely moved to another directory if the user so desires On Mac OS X a proper application bundle named SpateGUI is created Ifthe makefile has been altered after a partial or full compilation it is necessary to clear out old object files before recompiling the program This is done by first running make clean Then make can be run again 2 4 Running SPATE GUI SPATE GUI is graphical user interface program and can be started on most operating systems by double clicking on the executable file On Windows this file is named spate gui exe and on Linux it is named spate gui On Mac OS X the application bundle SpateGUI is used to run the program Chapter 3 The simulator SPATE HPC p
71. t on MPICH or set to CC if it is to be compiled with PGI If the simulator is to be compiled without MPI support with GCC it should be set to g USE MPI A flag controlling whether to enable MPI support It should be set to either true or false xx DIR Specifies the directory where library xx is installed Replace xx by ACML XER CES SPRNG GSL or CPPUNIT xx LIB Specifies the names of the library files of library xx to be linked into the resulting program file Replace xx by ACML XERCES SPRNG GSL or CPPUNIT INCLUDE_TESTS A flag that controls whether to build unit tests It should be set to either true or false The default value is false since unit tests are not needed in order to use the simulator EXTRA_LIBS This variable can be used to include additional libraries CFLAGS This variable can be used to provide additional flags to the C compiler In addition there are three C preprocessor flags that can be enabled or disabled in the make file with the corresponding CFLAGS line CALCULATION INSPECTION INCLUDE ENVIRONMENTAL EFFECTS and COMPARISONSEEDING The flags CALCULATION INSPECTION and INCLUDE ENVIRONMENTAL EFFECTS are both on by default and enable the inspection of model calculations and the generation of correlated random numbers for environmental effects respectively The calculation inspe
72. t types Species Management Assortments Compartments Notes Control Save settings Load settings General parameters m Output m Simulation Output directory joutput t Browse Number of years The directory name may contain a t wildcard EUM which will be replaced by the current date and time Number of replications Simulation name simulation Iteration length years IV Save tree files eet Save tree files for each iteration Standard normal limiting factor Correlation inclusion factor IV Iteration load balancing Allow imbalance percentage IV Use calculation inspection Inspection parameters SPATE HPC must be compiled with calculation inspection explicitly enabled in order for these parameters to take effect Use interval fo E o E Initial Stand Growth Mortality Reproduction Regeneration Competition Distance Figure 4 1 The general simulation settings 24 Height classes Height classes are used to classify trees according to height Number of years The number of years to simulate ahead of time Number of replications The number of replications to simulate Iteration length The length of an iteration step in years Rng seed A preferably large number that is used to seed the random number generator Standard normal limiting factor Some Gaussian random numbers are limited to the standard de viation multiplied by this factor Correlation inclusion factor
73. te values are allowed and there must be no two sibling mark elements with the same name see figure 5 4 A mark element contains a set of modelTerm elements The modelTerm elements are described in section 5 1 3 Furthermore if environmental effects are used in the model a mark element should have a environmentalEffects child element The environmentalEffects element may have two coefficient child elements whose values correspond to 0 and 62 in formula 3 4 the first coefficient element corresponds to 0 and the second one to 05 If either one or both coefficients are left out their default value is zero In the mortalityParameters element a mark element may contain an additional element called regularMortalityThreshold The value of regularMortality Threshold corresponds to the regular mortality threshold described in section 3 5 3 If regularMortality Threshold is not given the default value is 1 0 If one of the growthParameters mortalityParameters or reproductionParameters container ele ments is missing or if it exists but contains no mark elements the corresponding phase will not be carried out in the simulation In other words it is possible to disable a particular simulation phase by not including the corresponding parameters element in the XML input file Regeneration parameters The regenerationParameters element contains the model for calculating regeneration i e the num ber of new trees in a reproduction phase The regenerati
74. ted The section contains the following tables Age frequency table The age frequency of the final stand 58 Iteration statistics The basal area and volume development in the different simulation phases and for each iteration step Stem statistics The total number of trees in the different simulation phases of different age class and for each iteration step Assortment volumes The growing stock volumes of different assortments and species for each iteration step Removal assortment volumes The removal volumes of different assortments and species for each iteration step Stand summary The stand summary section contains mean values of tree characteristics and the final species composition 5 7 Statistics table file The statistics table file contains a table in the comma separated values CSV format with the final characteristics of each compartment Each row in the table corresponds to a compartment polygon and the columns contain specific polygon or compartment characteristics The columns are the following in order Polygon id The polygon s ID number Compartment id The corresponding compartment s ID number Polygon area The polygon s area in hectares Compartment area The compartment s area in hectares Forest type The name of the forest type Basal area growth The total basal area growth per hectare Final basal area The final basal area per hectare Volume growth The total volume growth per hectare Dead tree volume
75. ters which are described in the following sections 5 1 2 General parameters The generalParameters element contains assorted simulation parameters see figure 5 3 Its child elements containing values are the following name Name of the simulation numberOfyears The number of years that should be simulated The value can be zero whereupon only the initial stand is generated iterationLength The length of one iteration step in number of years numberOfReplications The number of replications identical simulations to perform rngSeed A number preferably very large to seed the random number generator with saveTreeFiles The value of the element which can be either 0 or 1 indicates whether tree data files should be produced after the simulation has completed Additionally if the itera tion attribute with the value 1 is given tree data files are produced at every intermediate iteration step iterationLoadBalancing The value of the element which can be either 0 or 1 indicates whether the simulator should rebalance trees between processors between iterations The element takes the attribute imbalancePct whose value is an integer greater than or equal to Zero and represents the maximum allowed imbalance percentage outputDirectory If given the value of this element is the name of the directory that should be used for all output files The name may contain the character t which is replaced by the
76. the template from a compartment Lower basal area If specified the management is carried out when the compartment s basal area per hectare exceeds this parameter value Thinning interval Two consecutive managements cannot be carried out closer to each other than this number of years Lower compartment volume If specified the management is carried out when the compartment s total volume per hectare exceeds this parameter value Removal per hectare If low or selective thinning is used this is the target volume to remove Secondary model value The value used for the variable Management in the secondary level model see section 3 4 1 Diameter limit If dimension or energy wood cutting is used trees with larger or smaller diameter than this value respectively are cut Thinning level The parameter 6o in formula 3 12 when selective thinning is used Density coefficient The parameter 8 in formula 3 12 when selective thinning is used Class width The width of size classes The default is 4cm Removal tolerance The percentage the resulting removal volume in low and selective thinning may deviate from the target volume Linear tail ratio If specified the tail of the Weibull curve in low or selective thinning is linearized according to this ratio see section 3 9 2 Use clear cuts Whether to carry out clear cuts Years between clear cuts The number of years between to consecutive clear cuts Trees left per hectare The number
77. ue that is subsequently assigned the Species variable in the secondary level model taperCurveParameters This is a container element which if present must contain all of the fol lowing elements coefficient1 coefficient2 coefficient3 coefficient5 coefficient8 coef ficient13 coefficient21 and coefficient34 The child elements should contain numerical values for the corresponding taper curve coefficients competitionDistanceModel This is a container element that contains a model for the competition distance Its child elements are modelTerm elements 47 lt estimatedElevationSurface ModelCoefficients gt container Figure 5 8 The elements homogeneousRegular and inhomogeneousRegular stumpHeightParameters This element contains parameters used for calculating the height of the stump It should contain each one of the following elements minimumHeight heightCoef ficient and diameterCoefficient smallVolumeCoefficient When a tree is too small the program forgoes taper curves and uses a cylindrical model to calculate the volume The cylindrical volume is multiplied by this weight coefficient If not given its default value is 0 6 5 1 6 Forest type parameters The forestTypeParameters element contains a set of forestType elements The content of a forest Type element defines a particular forest type see figure 5 10 A forestType element may contain the following elements name The name element is obligatory It should co
78. value is 596 It is possible that the program concludes that the target volume cannot be achieved In this case the program will reject a resulting volume larger than the target volume but will accept a smaller resulting volume Weibull curve with linear tail The tail of the Weibull curve can be linearized so that it does not approach zero asymptotically like it normally would but instead transition into a linear function that reaches zero prematurely The purpose of this functionality is to reduce the Weibull curve s tail in order to increase the thinning rate of the largest size classes If the linear tail ratio parameter is specified the linear tail will start when the ratio between the gradients of two consecutive size classes both with negative slopes is 19 greater than the linear tail ratio parameter Formally if a ax 1 gt a where aj is the gradient at the centre of size class k and a is the linear tail ratio the linear tail will start at the centre of class k and have the gradient az while the rest of the Weibull curve is discarded 3 9 3 Whenis thinning carried out There are three parameters to control when the thinning methods are to be applied lower basal area lower compartment volume and thinning interval in years These parameters apply to all management methods save for the thinning interval in years which does not apply to clearcutting The lower basal area is the smallest basal area the compartment may have befor
79. when energy wood cutting is carried out trees with a thinner diameter than the diameter limit are cut down 3 9 2 Weibull thinning Low and selective thinning are collectively referred to as Weibull thinning In Weibull thinning trees are categorized according to diameter into size classes The default width of a size class is 4cm but this can be changed by the user Furthermore the diameter distribution is approximated by a Weibull distribution The curve that represent the approximate Weibull distribution is then shifted in either direction to the right in low thinning and to the left in selective thinning EX Simulated data Estimated distribution AS Shifted curve N e a e Estimated distribution EX Simulated data Shifted curve Frequency Frequency SA Diameter Diameter a Low thinning b Selective thinning Figure 3 3 The simulated tree distribution is represented by the histogram and the estimated distribution by the solid curve The dashed curve is the shifted curve This is readily illustrated by drawing a histogram of the number of trees in the size classes and the Weibull curves scaled to conform to the height of the histogram bars There is now an overlapping area between the two Weibull curves where the histogram bars are likely to protrude above the shifted Weibull curve This area is to the left in low thinning and to the right in selective thinning The objective
80. y take this form but may follow a more generic pattern y Bo abe R lt Bey p ees 3 2 Here 0 are coefficients z are variables and p are exponents The variables in equation 3 2 can be competition indices as well as 7 and e A model in SPATE HPC is a set of terms where each term has a coefficient and a set of variables There can be one or several variables in a term and they can be exponentiated Furthermore the coefficients can be expanded into a secondary level model Bo Boo wQo wor gt B wig wi re tee 3 3 Here Gi are coefficients w are secondary level variables and q are exponents Analogously to the top level model a secondary level model also comprises a set of terms where each term has a 10 coefficient and a set of variables Secondary level variables can also be multiplied by each other and exponentiated Secondary level models are also referred to as smoothing functions 3 4 Variables The following is a list of variables that can be used as elements of x Intercept The constant 1 Diameter The diameter at breast height DBH d Diameter log The natural logarithm of diameter In d Local density The tree density within the competition radius Trees per hectare 10000 N A Average distance The average distance to competitors M TUN Inverse distance The sum of the inverse distance to all larger competitors A competitor is larger if it has a larger diamet

Download Pdf Manuals

Related Search

Related Contents

Manual - Microtronics Pakistan  Powered-Lift RC Aircraft Operator`s Manual V1.5  Panasonic WV-SW558 Installation Guide  Energizer ENULTIMATEAAA8  EnVision™ FLEX, High pH, (Dako Autostainer/Autostainer Plus)  ロータリコントロールバルブ  Primer User Manual r09  新たに路線測量機能をプラス!  

Copyright © All rights reserved. Failed to retrieve file