Home

InversePELMO Manual (pdf, 1 MB, not barrier-free) - BVL

1. 29 Figure 19 InversePELMO Experimental concentrations in the percolate 30 Figure 20 InversePELMO Experimental soil Concentrations serrrnvrnnvnnvennvvrnnvnvverrvvrnnvnnn 31 Figure 21 InversePELMO define fitting parameters for the hydrology in soil 32 Figure 22 InversePELMO Parameter for fitting pesticide fate rrrrrnnnnnnrrrrnnrrnrrrrnnnnnnrnnnn 33 Figure 23 InversePELMO start optimisation of the pesticide fate rrrrrrrrnrnnrrnnrnnnnnrnnnnn 34 Figure 24 InversePELMO Analyse the results of the optimisation for soil hydrology HOT COALS ae eisen Ba ea IN pe 35 Figure 25 InversePELMO View the results of the optimisation soil concentrations 36 Figure 26 InversePELMO View the results of the optimisation cumulative flux 37 Figure 27 InversePELMO Evaluate the results of the optimisation pesticide fate 38 Figure 28 Parameters used in the optimisation of the percolate hypothetical test data set 1 EE OSE EE NT EE EN ne 43 Figure 29 Results of the optimisation percolate hypothetical test data set 1 43 Figure 30 Parameters used in the optimisation of the substance flux test data set 1 44 Figure 31 Results of the optimisation substance flux hypothetical test data set 1 45 Figure 32 Daily precipitation at the lysimeter station from August 2008 to December 2009 46 aes
2. 0031 0092 0122 0153 0184 0212 0396 0457 0487 0518 08 10 11 12 01 02 08 10 11 12 01 01 01 01 02 02 02 02 02 02 69 143 120 Lede 221 264 331 451 564 710 790 7133 0434 62815 20235 35835 87768 81468 48953 618272 86 PEST output file Pest water rec PEST RUN RECORD CASE PEST WATER PEST run mode Parameter estimation mode Case dimensions Number of parameters Number of adjustable parameters Number of parameter groups Number of observations 1 Number of prior estimates oon a Model command line s run_water Jacobian command line na Model interface files Templates SCENARIO TPL for model input files BORSTEL SZE Parameter values written using single precision protocol Decimal point always included Instruction files PEST INS for reading model output files PEST_WATER PLM PEST to model message file na Derivatives calculation Param Increment Increment Increment Forward or Multiplier Method group type low bound central central central kcO relative 1 0000E 02 none switch 2 000 parabolic kel relative 1 0000E 02 none switch 2 000 parabolic kc2 relative 1 0000E 02 none switch 2 000 parabolic moid relative 1 0000E 02 none switch 2 000 parabolic Parameter definitions ame Trans Change Initial Lower Upper formation limit value bound bound kco none relative 1 00000 0 500000 10 0000 kc1 none rel
3. Figure 2 Flow chart File handling of a flux optimisation with InversePELMO 13 4 3 InversePELMO Main form After successful installation the main form of inverse PELMO appears as shown in Figure 3 inversePELMO Path to PELMO DAFOCUS aktuelles_FOCUSPELMO Version 1 0 13 Sep 2011 Table of Projects Sickerwasser und Boden Sickerwasser Test UBA project Exit Figure 3 InversePELMO Main form 4 3 1 Status Information Two fields on top of the form give information about the path to the FOCUS PELMO installation and about the current version of the software If the user clicks at the path a form will be loaded to change the current setting as shown in Figure 4 14 Enter Path to PELMO DAFOCUS aktuelles_ FOCUSPELMO D EJ inverse Modellierung amp Projects Source Done Figure 4 InversePELMO Path to FOCUS PELMO d Daten After a mouse click at the information field on current the software version the form shown in Figure 5 will appear a Release Info nversePELMO Tool fo perform inverse modelling studies with PELO Version 1 0 13 Sep 2011 developed by Michael Klein Fraunhofer Institut f r Molekularbiologie und Angewandte Okologie D 57392 Schmallenberg Germany Phone 49 2972 302 317 Fax 49 2972 302 319 E mail michael klein ime fraunhofer de Umwelt Bundes Amt F r Mensch und Umwelt check for update ok Figure 5 I
4. DAY 1000E 09 1600E 10 9000E 11 1300E 10 0000 ooo0o0 0O DISP COEFF 1 DISP LENGTH 0 INPUT 1 PRZM 2 PELMO 0 INPUT 1 CALCULATED 0 INPUT 1 CALCULATED INIT SOIL WATE CONT G CM 3 CM IAL R ENT CM DRAINAGE PARAMETER DAY 30 0000 5 7000 30 0000 4 9000 15 0000 4 9000 15 0000 5 0000 20 0000 4 8000 OUTPUT FILE PARAMETERS OUTPUT WATR PEST TIME STEP DAY DAY 5000 0 6000 0 5800 0 6200 0 6000 0 LAYER FREQ 2000 2000 2000 2000 2000 2 3000 2 3000 2 2000 0 0000 0 0000 0 0000 0 0000 0 0000 0 0000 19 00 0 0000 0 0000 0 0000 19 00 0 0000 0 0000 0 0000 19 00 0 0000 0 0000 0 0000 0 0000 100 0 0 7000 0 0000 14 79 0 9000 eg 0 0000 0 0000 0 1000E 19 MET B1 MET Cl MET D1 BR CO2 DAY DAY DAY DAY 0 1000E 09 0 1000E 09 0 1000E 09 0 3304E 01 0 1600E 10 0 1600E 10 0 1600E 10 0 5286E 02 0 9000E 11 0 9000E 11 0 9000E 11 0 2974E 02 0 1300E 10 0 1300E 10 0 1300E 10 0 4295E 02 0 0000 0 0000 0 0000 0 0000 110 0 5 44 0 2 1 0 free drainage ORGANIC DISPERSION SAND CLAY CARBON COEFFICIENT CM 2 DAY 68 3000 7 2000 1 5000 0000 67 0000 6 7000 1 0000 0000 96 2000 0 9000 0 2000 0000 99 8000 0 0000 0 0000 0000 100 0000 0 0000 0 0000 0000 82 CONC DAY 1 CALCULATED HYDRAULIC PROPERTIES HORIZON FIELD CAPACITY WILTING POINT CM3 CM3 CM3 CM3 L 0 2778 0 0592 2 0 2794 0 0586 3 0 2094 0 0336 4 0 2004 0
5. Figure 33 Actual ET at the lysimeter station from from August 2008 to December 2009 Figure 34 Parameters used in the optimisation of the percolate test data set 2 Figure 35 Results of the optimisation percolate test data Set 2 rrnrrnnnnnnnnnrnnnnnrrrrnnnnnnenn Figure 36 Parameters used in the optimisation of the substance flux test data set 2 Figure 37 Results of the optimisation substance flux test data set 2 Figure 38 Results of the standard simulation with optimised parameters FOCUS Hamburg 1 Summary A software called InversePELMO was developed that can be used to perform inverse modelling studies with PELMO using the results of higher tier outdoor studies e g lysimeter experiments as input This is done in order to obtain key parameters for leaching models such as Kfoc Freundlich sorption constant related to organic carbon and DT50 degradation time to 50 Aim of such a study is on one hand to get a deeper look into the processes that led to a certain lysimeter result On the other hand inverse modelling studies can be used to improve the standard modelling on tier 1 by considering additional information from higher tier studies The results of InversePELMO can be used to make e Predictions about the most likely behaviour if the lysimeter study had been conducted over a longer time period e Translations of the lysimeter results to a different situation with respect to the
6. 4 4 3 Step 3 Check initial simulation After clicking at the button it is checked whether the initial simulation runs without problems After PELMO terminates the user has to confirm that the simulation didn t quit with an error condition Only after confirmation the arrow will move to the next button see Figure 10 Optimisation testtest Optimisation sequence ae Enter experimental data KE Create PELMO input files oo Import PELMO input files V Check initial simulation view opt or view optimisatior PEMO simulation control Start simulation day dd mm Start simulation day dd mm bi gt bi End simulation day dd mm 31 gt fiz gt Number of years BM Study begin dd mm yy gt u Pesticide input file Pesticide A Maizepm StS Scenario input file IH MAIZEsze Climate input file s HMBGNORM CLI Figure 10 InversePELMO experimental data percolate 4 4 4 Step 4 Enter experimental data percolate In step 4 the experimental percolate has to be entered in a specific form see Figure 11 22 Experimental Percolate 3 D Percolate Lim Weighting factor 60 31 10 28 70 03 0 0 0 2 3 4 5 6 Z 8 0 Figure 11 InversePELMO Experimental Percolate For each sampling during the study the date and the amount of percolate in L m is needed The user should not enter any cumulative numbers here because they will be automatically calculated b
7. 98 Objective function gt Sum of squared weighted residuals ie phi 10 19 Correlation Coefficient gt Correlation coefficient 0 9995 Analysis of residuals gt All residuals Number of residuals with non zero weight 10 Mean value of non zero weighted residuals 8 8320E 02 Maximum weighted residual observation 08 2 556 Minimum weighted residual observation 09 1 620 Standard variance of weighted residuals 1 274 Standard error of weighted residuals 1 129 Note the above variance was obtained by dividing the objective function by the number of system degrees of freedom ie number of observations with non zero weight plus number of prior information articles with non zero weight minus the number of adjustable parameters If the degrees of freedom is negative the divisor becomes the number of observations with non zero weight plus the number of prior information items with non zero weight Parameter covariance matrix gt koc kdeg koc 2 698 1 5992E 03 kdeg 1 5992E 03 9 5288E 07 Parameter correlation coefficient matrix gt koc kdeg koc 1 000 0 9975 kdeg 0 9913 1 000 ormalized eigenvectors of parameter covariance matrix gt Vector 1 Vector 2 koc 5 9283E 04 1 000 kdeg 1 000 5 9283E 04 Eigenvalues gt 4 8263E 09 2 698 Parameter Estimated 95 percent confidence limits value lower limit upper limit koc 95 1700 91 3826 98 9574 DT50 22
8. CROP INFORMATION MAXIMUM INTERCEPT MAXIMUM MAXIMUM MAXIMUM USLE COVER MANAGEMENT CROP POTENTIAL ROOT DEPTH COVER WEIGHT AMC RUNOFF CURVE NUMBERS C FACTOR 1 000 1 000 1 000 15 00 MONTH MAR JUNE SEP DEC 0 4600 12 IRRIGATION PERENNIAL TILLAGE FLAG 0 NO DAY HOURS 11 59 16 78 12 33 7 221 CROP 0 NO FLAG 0 NO SURFACE CONDITION AFTER NUMBER FALLOW 100 CM CM 5 KG M 2 1 YES 1 YES CROP RESIDUE FALLOW CROP RESIDUE 72 S 72 0 0000 100 0 90 00 0 0000 0 0 86 70 86 1 0000 1 0000 1 0000 94 84 94 CROP ROTATION INFORMATION CROP TILLAGE EMERGENCE MATURATION HARVEST NUMBER DATE DATE DATE DATE Winter Rape 2 SEP 1 5 MAY 28 JULY 2 Winter Rape 2 SEP 2 5 MAY 28 JULY 3 Winter Rape 2 SEP 3 5 MAY 28 JULY 4 PARAMETERS OF ACTIVE SUBSTANCE Parent KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK PESTICIDE APPLICATION INFORMATION PESTICIDE INCORPORATION APPLICATION APPLIED DEPTH DATE KG HA CM 21 AUG 1 200 0 0000 PLANT PESTICIDE PARAMETERS MODEL UTILIZED 1 SOIL 2 LINEAR 3 EXPONENTIAL 1 VOLATILIZATION PARAMETERS ACTIVE SUBSTANCE TEMPERATURE deg C 20 00 HENRY CONSTANT Pa m3 mole or J mole 0 2000E 04 CALCULATED USING VAPOUR PRESSURE Pa 0 1000E 03 MOLECULAR MASS g mole 100 0 WATER SOLUBILITY mg 1 500 0 TEMPERATURE deg C 30 00 HENRY CONSTANT Pa m3 mole or J mole 0 4000E 04 CALCULATED USING VAPOU
9. Figure 38 Results of the standard simulation with optimised parameters FOCUS Hamburg 55 Table 15 Pesticide in the percolate at 1 m soil depth FOCUS Hamburg Year Pesticide Flux Percolate Pesticide Conc g ha Um g L 1 3 47E 10 273 300 0 000 2 2 50E 07 78 8200 0 000 3 0 0215800 265 900 0 008 4 0 1968000 252 600 0 078 5 1 3630000 431 100 0 316 6 4 5400000 470 500 0 965 7 1 6590000 140 900 1 177 8 0 8473000 138 600 0 611 9 0 4894000 234 600 0 209 10 0 4048000 281 600 0 144 11 0 4784000 226 100 0 212 12 1 7700000 459 900 0 385 13 2 6330000 432 900 0 608 14 1 1830000 185 600 0 637 15 1 5250000 372 000 0 410 16 1 4950000 308 300 0 485 17 0 6419000 176 500 0 364 18 0 6015000 234 800 0 256 19 0 6459000 266 800 0 242 20 1 1640000 314 600 0 370 21 0 4984000 156 000 0 319 22 0 1471000 78 8200 0 187 23 0 1690000 265 900 0 064 24 0 2439000 252 600 0 097 25 1 3740000 431 100 0 319 26 4 5400000 470 500 0 965 Total 22 5106 5428 12 0 415 Perc 8 13 3 4803000 571 500 0 610 annual applications of 1 2 kg ha on 21 August 2011 optimised parameter setting 56 6 Discussion and Conclusions The example simulations with all four test data set demonstrate that the link between InversePELMO and PELMO works sufficiently and results in adequate descriptions of what processes may have been occurred in lysimeter studies Furthermore higher tier data on mobility and degradation can be obtained which can be used for refined
10. Number of prior estimates CONN N Model command line s run_pesticide Jacobian command line na Model interface files Templates PESTICIDE TPL for model input files SUBSTANCE_2 PSM Parameter values written using single precision protocol Decimal point always included Instruction files PEST INS for reading model output files PEST_FLUX PLM PEST to model message file na Derivatives calculation Param Increment Increment Increment Forward or Multiplier Method group type low bound central central central koc relative 1 0000E 02 none switch 2 000 parabolic kdeg relative 1 0000E 02 none switch 2 000 parabolic Parameter definitions Name Trans Change Initial formation limit value koc none relative 50 0000 kdeg none relative 1 099000E 02 Name Group Scale Offset koc koc 1 00000 0 00000 kdeg kdeg 1 00000 0 00000 Prior information No prior information supplied Observations Observation name Observation Weight Group Lower Upper bound bound 1 00000 1000 00 6 900000E 04 0 693150 Model command number 1 1 94 ol 0 00000 000 no_name o2 0 00000 000 no_name 03 0 00000 000 no_name 04 0 00000 000 no_name 05 0 00000 000 no_name 06 0 00000 000 no name o7 0 00000 000 no_name 08 12 2710 000 no name 09 54 1210 000 no name o10 104 309 000 no_name Control settings Initial lambda 5 0000 Lambda adjustment factor 2 0000 Sufficient new ol
11. environmental conditions e g different climate e Translations of the lysimeter result to a different situation with respect to the application pattern of the substance e g change of the rate e Use of optimised parameter settings for a refined standard tier 1 simulation Test simulations are performed in order to check the suitability of InversePELMO for the above mentioned aims The example simulations with two different test data sets demonstrate that the link between InversePELMO and PELMO works sufficiently and leads to adequate descriptions of the processes occurred in lysimeter studies InversePELMO guides the user to a sequential process including first the calibration of soil hydrology followed by the optimisation of the substance fluxes The user can stop the inverse modelling study at any times exit InversePELMO and continue at a later time where he left All inverse modelling studies are saved as projects and can be re evaluated at a later time The quality of the optimised parameters is reported according to the recommendations of FOCUS degradation kinetics e g chi test t test 2 Background Within the registration of pesticides for all active compounds and metabolites concentrations in the environment soil groundwater and surface water have to be calculated This is done using different computer models such as Exposit UBA 2011 PELMO Klein 1995 Jene 1998 FOCUS 2009 EVA Holdt et al 2011 ESCAPE Klein 2008 co
12. 02 kdeg OPTIMISATION ITERATION NO 3 Model calls so far 8 Starting phi for this iteration 1 55900E 05 Lambda 0 62500 sees gt Phi 9 59417E 05 6 154 times starting phi Lambda 0 31250 gt Phi 60826E 06 10 316 times starting phi Lambda 1 2500 gt Phi 80902E 05 1 160 times starting phi Lambda 2 50000 gt Phi 33430E 05 0 856 of starting phi Lambda 5 0000 x Haste gt Phi 51729E 05 0 973 of starting phi No more lambdas phi rising Lowest phi this iteration 1 33430E 05 Current parameter values Previous parameter values koc 6 01900 koc 5 04200 kdeg 2 063000E 02 kdeg 1 671000E 02 Maximum relative change 0 2346 kdeg OPTIMISATION ITERATION NO 4 Model calls so far 17 Starting phi for this iteration Lambda 25000 Phi 1 31355E 05 Lambda 1 2500 Phi 1 30446E 05 No more lambdas Lowest phi this iteration Current parameter values 1 33430E 05 0 984 of starting phi 0 978 of starting phi relative phi reduction between lambdas less than 0 0300 1 30446E 05 Previous parameter values koc 6 10300 koc 6 01900 kdeg 2 147000E 02 kdeg 2 063000E 02 Maximum relative change 4 0717E 02 kdeg OPTIMISATION ITERATION NO 5 Model calls so far 23 Starting phi for this iteration 1 30446E 05 Lambda 0 62500 Phi 1 19313E 05 0 915 of starting phi Lambda 0 31250 gt Phi 91098 0 698 of starting phi Lambda
13. 0301 5 0 2000 0 0300 Total number of layers in the top meter 41 PLOT FILE INFORMATION NUMBER OF PLOTTING VARIABLES 1 TIMSER NAME MODE DEPTH CM ARGUMENT CONSTANT SUBSTANCE LEAC TSER 100 al 0 1000E 10 PESTIC 83 8 2 Example data set 2 8 2 1 Optimisation of soil hydrology InversePELMO control file sampling water txt 01 08 01 10 31 8 4 47 9 31 10 1 124 6 30 11 1 177 38 31 12 1 222 4 31 1 2 25545 28 2 2 346 9 31 8 2 495 25 31 10 2 582 9 30 1 2 67549 31 12 2 767 15 84 PEST control file Pest water pst pef control data restart 4 10 4 0 A 1 single point 5x 0 2 0 03 0 03 10 3 0 300001 0 1 30 0 01 3 3 GAL 3 1 1 1 group definitions and derivative data KcO relative 0 01 0 0 switch 2 0 parabolic Kel relative 0 01 0 0 switch 2 0 parabolic Kc2 relative 0 01 0 0 switch 2 0 parabolic OIO relative 0 01 0 0 switch 2 0 parabolic parameter data c0 none relative 1 0 5 10 KcO 1 0000 0 00000 cl none relative 1 0 5 10 Kel 1 0000 0 00000 c2 none relative 1 0 5 10 Kc2 1 0000 0 00000 OIO none relative 0 2 0 05 0 5 MOIO 1 0000 0 00000 observation data ol 47 9 1 02 124 6 03 171 235 1 04 222 4 1 05 255 5 1 06 346 9 T 07 495 25 L o8 582 9 1 09 675 9 a 010 767 15 1 model command line run_water model input output scenario tpl BORSTEL SZE pest ins pest_water plm prior information PELMO output file PEST water pim 85
14. 1 06E 01 0 5 1 30E 02 5 10 2 95E 02 10 15 5 31E 02 15 20 7 56E 02 0 5 1 03E 03 5 10 2 32E 03 10 15 4 45E 03 15 20 7 51E 03 20 25 1 14E 02 25 30 1 57E 02 30 35 1 47E 02 35 40 1 79E 02 40 45 2 08E 02 45 50 2 30E 02 2 43E 02 C Percolate concentrations Soil concentrations Figure 20 InversePELMO Experimental soil concentrations Soil concentrations have to be given in ug kg together with the sampling date and the sampling depth For both tables it is not necessary to type all information the user can also paste them in as a table e g from MS Excel or MS Word If the form has been filled correctly e g no negative figures it can be closed using the Done button and the arrow on the optimisation form will jump to the next button to enter all information about the input parameters used in the optimisation see Figure 21 32 x Optimisation testtest Optimisation sequence a 3 Enter experimental data MEN Create PELMO input files Import PELMO input files Start optimisation Start optimisation V Check initial simulation View optimisation View optimisation PEMO simulation control Start simulation day dd mm Start simulation day dd mm bi gt bi End simulation day dd mm 31 fiz gt Number of years b Study begin dd mm yy bi jor or Pesticide input file Pesticide A Maizepsm ss Scenario input file HMAZEse i i OCO Climate input file
15. 486 38523 31 3 3 490 07456 30 4 3 490 07456 70 PEST control file Pest pesticide pst pot control data restart 2 24 2 0 1 1 single point 5 0 2 0 0 3 0 03 10 3 0 3 0 0 001 OL 30 0 01 3 3 0 01 3 1 1 1 group definitions and derivative data KOC relative 0 01 0 0 switch 2 0 parabolic KDEG relative 0 01 0 0 switch 2 0 parabolic parameter data KOC none relative 5 1 1000 KOC 1 0000 0 00000 KDEG none relative 1 09861228866811E 02 6 93147180559945E 04 0 693147180559945 KDEG 1 0000 0 00000 observation data ol 0 1 o2 0 Al 03 0 1 04 0 1 05 0 1 06 0 1 o7 0 T o8 1 46124 1 09 25 03332 1 010 112 65269 oll 215 17544 012 273 90689 013 330 16014 ol4 330 16014 015 364 9787 1 016 418 91302 017 418 91302 018 418 91302 1 019 434 77234 020 477 43926 0o21 485 54307 022 486 38523 023 490 07456 024 490 07456 I model command line run_pesticide model input output pesticide tpl Pesticide_B_example_1 psm pest ins pest_flux plm prior information PELMO output file PEST flux pim 0031 31 05 0 0 0061 30 06 01 0 0092 31 07 0 0 0123 31 08 01 0 0153 30 09 01 0 0184 31 10 0 0 0000019122 0214 30 11 01 0 00010610627 0245 31 12 0 0 51728890627 0276 31 01 02 17 42258290627 0304 28 02 02 101 64738290627 0335 31 03 02 210 59816290627 0365 30 04 02 274 92216290627 0396 31 05 02 333 87266290627 0426 30 06 02 333 87266290627 0457 31 07 02 367
16. 80666290627 0488 31 08 02 421 59516290627 0518 30 09 02 421 59516290627 0549 31 10 02 421 59516290627 0579 30 11 02 435 93256290627 0610 31 12 02 476 53836290627 0641 31 01 03 483 26630590627 0669 28 02 03 483 92267590627 0700 31 03 03 486 58290790627 0730 30 04 03 486 58290790627 72 PEST output file Pest_pesticide rec PEST RUN RECORD CASE PEST_PESTICIDE PEST run mode Parameter estimation mode Case dimensions Number of parameters Number of adjustable parameters Number of parameter groups Number of observations 5 2 Number of prior estimates OPDNDDNDMN Model command line s run_pesticide Jacobian command line na Model interface files Templates PESTICIDE TPL for model input files PESTICIDE_B_EXAMPLE_1 PSM Parameter values written using single precision protocol Decimal point always included Instruction files PEST INS for reading model output files PEST_FLUX PLM PEST to model message file na Derivatives calculation Param Increment Increment Increment Forward or Multiplier Method group type low bound central central central koc relative 1 0000E 02 none switch 2 000 parabolic kdeg relative 1 0000E 02 none switch 2 000 parabolic Parameter definitions Name Trans Change Initial formation limit value koc none relative 5 00000 kdeg none relative 1 099000E 02 Name Group Scale Offset koc koc 1 00000 0 00000 kdeg kdeg 1 00000 0 00000
17. Concentrations for Pesticides Dependent on FOCUS Degradation Kinetics FKZ 360 03 037 Umweltbundesamt Dessau Ro lau Watermark 2003 PEST Model Independent Parameter Estimation Watermark Numerical Computing http www pesthomepage org Downloads php UBA 2011 Exposit 3 0 beta Available at http www bvi bund de DE 04 Pflanzenschutzmittel 03 Antragsteller 04 Zulassungs verfahren 07 Naturhaushalt psm naturhaush node htmli 59 8 Documentation of Model Output InversePELMO 8 1 Example data set 1 8 1 1 Optimisation of soil hydrology InversePELMO control file sampling water txt 01 05 01 24 31 5 0 30 6 0 31 7 0 31 8 0 30 9 51 60 31 10 32 96 30 11 51 18 3a 12 L 117 6 31 1 2 179 96 28 2 2 242 59 31 3 2 289 34 30 4 2 315 27 31 5 2 342 38 30 6 2 342 38 31 7 2 361 81 31 8 2 402 18 30 9 2 402 18 31 10 2 402 18 30 11 2 419 16 31 12 2 513 14 31 1 3 569 81 28 2 3 579 49 31 3 3 649 1 30 4 3 649 1 60 PEST control file Pest water pst pef control data restart 5 24 SG 1 single point 5 0 2 0 0 3 0 03 10 3 0 3 0 0 001 0 1 30 0 01 3 3 0 01 3 1 1 1 group definitions and derivative data Kc0 relative 0 01 0 0 switch 2 0 parabolic Ke relative 0 01 0 0 switch 2 0 parabolic Kc2 relative 0 01 0 0 switch 2 0 parabolic ANETD relative 0 01 0 0 switch 2 0 parabolic OIO relative 0 01 0 0 switch 2 0 parabolic parameter data c0 none relative 0
18. Prior information No prior information supplied Observations Observation name Observation Weight Group Lower Upper bound bound 1 00000 1000 00 6 900000E 04 0 693150 Model command number 1 1 73 ol 0 00000 000 no_name o2 0 00000 000 no_name 03 0 00000 000 no_name 04 0 00000 000 no_name 05 0 00000 000 no_name 06 0 00000 000 no_name o7 0 00000 000 no_name 08 1 46124 000 no_name 09 23 0333 000 no_name 0o10 112 653 000 no_name oll 215 115 000 no_name 012 273 907 000 no_name o13 330 160 000 no_name 014 330 160 000 no_name 015 364 979 000 no_name 016 418 913 000 no_name 017 418 913 000 no_name o18 418 913 000 no_name 019 434 772 000 no_name 020 477 439 000 no_name o21 485 543 000 no name 022 486 385 000 no_name 023 490 075 000 no_name 024 490 075 000 no name Control settings Initial lambda 5 0000 Lambda adjustment factor 2 0000 Sufficient new old phi ratio per optimisation iteration 0 30000 Limiting relative phi reduction between lambdas 3 00000E 02 Maximum trial lambdas per iteration 10 Maximum factor parameter change factor limited changes na Maximum relative parameter change relative limited changes 3 0000 Fraction of initial parameter values used in computing change limit for near zero parameters 1 00000E 03 Allow bending of parameter upgrade vector no Allow parameters to stick to their bounds no Relative phi reduction below which to begin use of central derivatives
19. TOTAL HORIZONS IN CORE TOTAL COMPARTMENTS IN CORE DPFLAG FLAG THETA FLAG PARTITION COEFFICIENT FLAG BULK DENSITY FLAG SOIL HYDRAULICS MODULE SOIL HORIZON INFORMATION BIODEG FACTOR HORIZON PH THICKNESS DENSITY CM winter TRANSFORMATION RATE TO MET Al DAY 0 1000E 09 0 5000E 10 0 3000E 10 0 3000E 10 0 3000E 10 MET Bl DAY 0 1000E 09 0 5000E 10 0 3000E 10 0 3000E 10 0 3000E 10 100 0 5 20 0O DISP COEFF 1 DISP LENGTH 1 0 INPUT 1 PRZM 2 PELMO 0 0 INPUT 1 CALCULATED 0 INPUT 1 CALCULATED DRAINAGE PARAMETER DAY 1 0 100 0 100 0 0 0000 100 0 TES EQ SITES MET C1 0 0 0 0 0 DAY 1000E 09 5000E 10 3000E 10 3000E 10 3000E 10 free drainage FIELD CAPACITY WATER CONTENT CM CM 7000 7000 0000 7000 19 55 1000E 19 MET D1 DAY 0 1000E 09 0 5000E 10 0 3000E 10 0 3000E 10 0 3000E 10 OO BR CO2 DAY 0 3086E 01 0 1543E 01 0 9258E 02 0 9258E 02 0 9258E 02 0 3000 4 0 3000 5 0 3000 30 0000 6 4000 30 0000 5 6000 15 0000 5 6000 15 0000 5 7000 10 0000 5 5000 OUTPUT FILE PARAMETERS OUTPUT WATR PEST CONC TIME STEP DAY DAY DAY INITIAL SOIL BULK WATER CONTENT G CM 3 CM CM 5000 0 28 6000 0 28 5600 0 28 6200 0 28 6000 0 28 LAYER FREQ Hm 2 3000 2 3000 2 3000 2 3000 0 0000 0 2920 0 2770 0 2290
20. appear see section 4 4 Open project After a click at this button InversePELMO will load all details of the project The user may have a look at the current parameter setting improve the optimisation or produce tabular or graphical output see also the section on new project More information about project can be found in section 4 4 Copy project A click at that button will first open a form where the name for the new project can be entered see Figure 6 Then all information of the selected project is copied into a second folder The option can be useful when a certain modification of an existing inverse modelling study should be performed without loosing information of the current status and without going through whole sequence of an inverse modelling study Delete project After a click at this button the selected project will be removed from the system To avoid accidental deleting the user has to confirm the command in an additional message box Exit A click at that button will terminate InversePELMO 18 4 4 InversePELMO Optimisation This form is used for new projects see Figure 7 as well as existing projects see Figure 8 However if a new inverse modelling study should be performed a certain sequence has to be followed A red arrow is used to guide the user through this process For the same reason most of the buttons are disabled at the beginning Optimisation user project Optimisation sequence Cr
21. end Optimisation of the hydrology in soil Fitting parameters evapotranspiration min depth for evaporation initial soil water Optimisation of chemicals fate software PEST oder R Tool parameters in optimisation KOC DT50 Freundlich 1 n Re assessment of Kfoc and DT50 Quality check based on information provided in standard PELMO output files Figure 1 General flowchart of inverse modelling studies As shown by the previous considerations inverse modelling studies for the calibration of lysimeter results are not totally uncomplicated but require detailed knowledge about the leaching model with its input and output file structure Only part of this will be the creation of input data for weather and soil properties users must be also able to manipulate pesticide input files in a dos environment So even if users are familiar with the normal shell and are 10 able to create input files for standard simulations it will be not sufficient to go through the complex inverse modelling procedure unless special supporting software is available Background is the additional optimisation tool in the procedure which is part of the package and which needs special input files created by the user These additional input files are read in by this optimisation tool and used to create PELMO input files automatically by the within the optimisation sequence Also the Also the post processing of PELMO output files during the sequence
22. example is based on real lysimeter data 5 1 Example data set 1 Leaching of Parent over a two years 5 1 1 Environmental data For the soil data the standard Borstel soil was used with exactly the same description as given in the PELMO 3 0 soil data base The soil profile information is summarised in Table 1 Table 1 Borstel soil profile in the lysimeter hypothetical test data set 1 Horizon cm 0 30 30 57 57 73 73 90 90 110 Soil density g cm 1 5 1 6 1 58 1 62 1 6 Sand 68 3 67 0 96 2 98 8 100 Silt 24 5 26 3 2 9 0 2 0 Clay 7 2 6 7 0 9 0 0 OC 1 5 1 0 0 2 0 0 initial soil water content m m 0 05 0 05 0 05 0 05 0 05 Biodegradation factor 1 0 16 0 09 0 13 0 pH value 5 7 4 9 4 9 5 0 4 8 The Hamon equation was used to estimate potential evapotranspiration The parameter linked to that process are summarised in the following table Table 2 Further input parameters influencing evapotranspiration hypothetical test data set 1 Parameter Value Minimum depth for evaporation cm 15 KcO no crop 1 0 Kc1 mid season 1 3 Kc2 late season 0 5 The crop considered for the simulation was maize with standard crop parameter setting 40 For the climate during the lysimeter study the standard PELMO 3 0 climate files Hamburg normal and wet weather are used The monthly and annual precipitation and temperature data is given in Table 3 Begin of the study 1 the 1 May Table 3 Climate data during
23. kdeg 4 159000E 02 Maximum relative change 0 1763 OPTIMISATION ITERATION NO Model calls so far 3 for this iteration Starting phi Lambda 9 76563E 03 gt Phi 36 266 0 161 No more lambdas phi is less than 0 Lowest phi this iteration 36 266 Current parameter values koc 89 3100 kdeg 3 423000E 02 Maximum relative change 0 1770 OPTIMISATION ITERATION NO Model calls so far Starting phi for this iteration Lambda 4 88281E 03 gt Phi 12 915 0 356 Lambda 2 44141E 03 gt Phi 11 162 0 308 Lambda 1 22070E 03 gt Phi 10 545 0 291 No more lambdas phi is less than 0 Lowest phi this iteration 10 545 Current parameter values koc 94 2900 kdeg 3 138000E 02 Maximum relative change 8 3260E 02 OPTIMISATION ITERATION NO Model calls so far Starting phi for this iteration Lambda 6 10352E 04 gt Phi 10 203 0 968 Lambda 3 05176E 04 gt Phi 10 203 0 968 No more lambdas Lowest phi this iteration 10 203 96 Previous parameter values koc 53 3100 kdeg 6 387000E 02 koc 8 24 1095 6 of starting phi 3000 of starting phi Previous parameter values koc 66 4100 kdeg 5 008000E 02 koc 9 27 225 40 of starting phi 3000 of starting phi Previous parameter values koc 78 1200 kdeg 4 159000E 02 kdeg 10 30 36 266 of starting phi of starting phi 3000 of starting phi Previous parameter values koc 89 3100
24. modelling studies With test simulations it was demonstrated that the results of inverse modelling studies with InversePELMO can also be used to transfer the lysimeter study to hypothetical situations such as e extension of the study e modification of the application pattern e modification of crop data e other climate scenarios 57 7 References FOCUS 2000 FOCUS groundwater scenarios in the EU review of active substances Report of the FOCUS Groundwater Scenarios Workgroup EC Document Reference Sanco 321 2000 rev 2 202pp FOCUS 2006 Guidance Document on Estimating Persistence and Degradation Kinetics from Environmental Fate Studies on Pesticides in EU Registration Report of the FOCUS Work Group on Degradation Kinetics EC Document Reference Sanco 10058 2005 version 2 0 434 pp FOCUS 2009 Assessing Potential for Movement of Active Substances and their Metabolites to Ground Water in the EU Report of the FOCUS Ground Water Work Group EC Document Reference Sanco 13144 2010 version 1 604 pp FOCUS 2009 Assessing Potential for Movement of Active Substances and their 32 Metabolites to Ground Water in the EU Report of the FOCUS Ground Water Work Group 33 EC Document Reference Sanco 2009 version 1 594 pp FOCUS 2009 Assessing Potential for Movement of Active Substances and their Metabolites to Ground Water in the EU Bericht der FOCUS Groundwater Work Group EC Document Reference Sanco 2009 ve
25. optimisation PEST will call PELMO several times After PEST terminated the user has to confirm that the optimisation didn t quit with an obvious error conditions Optimisation testtest Optimisation sequence Mg 3 Enter experimental data KREM Create PELMO input files oo Import PELMO input files V Check initial simulation PEMO simulation control Start simulation day dd mm Start simulation day dd mm for gt for End simulation day dd mm 31 gt h2 gt Number of years E Study begin dd mm yy for gt for gt for gt Pesticide input file Pesticide A Maizepm Scenario input file HMAZEse Climate input file s HMBGNORM CLI Figure 14 InversePELMO start optimisation for the hydrology in soil Only after confirmation the arrow will move to the next button see Figure 15 4 4 7 Step 7 View the optimisation percolate In step 7 the user can evaluate the results of the percolate optimisation 26 Optimisation testtest m Optimisation sequence IH Enter experimental data Create PELMO input files o Lie wae Import PELMO input files Start optimisation Start optimisation V Check initial simulation View optimisation view optimisation PEMO simulation control Start simulation day dd mm Start simulation day dd mm for 01 End simulation day dd mm s number EEE E Study begin dd mm yy jor bi gt or Pesticide input fi
26. s HMBGNORM CLI gt About a Figure 21 InversePELMO define fitting parameters for the hydrology in soil 4 4 9 Step 9 Enter fitting parameter pesticide fate In step 9 the parameters used in the optimisation have to be characterised in a specific form see Figure 22 33 Substance considered FOCUS DUMMY A v Parameter Initial value Min value Max value M Freundlich 1 n vw DT50 d 3 58 1 000 2 Figure 22 InversePELMO Parameter for fitting pesticide fate Three PELMO input parameters dominating pesticide fate in soil can be used to do the fitting e KOC Kfoc linear or non linear sorption factor e Freundlich 1 n Kc factor mid season linear correction factor for daily potential evapotranpiration when the crop is growing e DT50 Time to reach 50 degradation in soil If parameters shall be considered for the optimisation their initial values and their range have to be specified If a parameter is not checked the respective input field is invisible As the DT50 in soil is not an input parameter in PELMO it will be converted into the respective rate constant which is the actual input parameter internally InversePELMO is able to analyse the fate of pesticides as well as of transformation products If metabolites have been defined in the PSM file previously the user can select the compound using the list box on top of the form If the form has been filled correctly e g no negative figu
27. study substance flux hypothetical test data set 1 Month Leachate L m Concentration L L Remark August 69 14 0 September 10 63 0 October 40 94 0 November 90 33 0 December 50 59 0 January 42 58 0 February 67 16 0 March 77 79 0 002 inverse modelling April 0 9 study May 10 8 0 006 June 0 0 July 31 93 0 017 August 0 0 September 13 82 0 028 October 99 12 0 087 November 145 7 0 316 Decenmiber 80 14 0 597 January 53 46 0 75 February 67 16 0 889 March 77 79 0 991 April 9 2 May 10 8 0 956 June 0 0 prediction July 31 93 0 89 August 0 0 September 13 82 0 618 October 99 12 0 424 November 145 7 0 248 December 80 14 0 18 optimised parameter setting 54 Based on the results of the previously performed inverse modelling study the PELMO calculation showed that the lysimeter study did not cover the peak maximum instead the maximum peak is estimated to occur in March of the following year The calculation furthermore showed that also in the next winter concentration above 0 1 ug L can be expected 5 2 6 Translation into standard conditions The most interesting question usually is what concentrations can be considered if the lysimeter study had been performed under the official standard conditions The results of the respective simulation FOCUS Hamburg 26 years of annual applications are presented in Table 15 Average Pesticide concentration in leachate 1m depth 1g L 1 5 0 5 0 5 10 15 20 Period
28. the study hypothetical test data set 1 Month Montly Annual Monthly Annual Precipitation Precipitation Temperature Temperature mm mm C C January 62 8 0 25 February 75 9 4 80 March 69 8 6 09 April 66 4 10 25 May 80 4 10 45 June 30 4 16 29 July 142 6 15 23 August 110 9 15 27 September 25 8 15 78 October 51 7 11 17 November 55 5 4 37 December 99 8 0 75 January 67 6 1 81 February 18 3 0 69 March 93 7 4 95 April 13 0 789 7 5 39 8 3 May 21 0 12 54 June 94 5 15 55 July 73 7 15 57 August 72 9 15 74 September 153 2 12 20 October 52 9 10 45 November 33 5 6 57 December 83 2 0 23 January 62 8 0 25 February 75 9 4 80 March 69 8 6 09 April 66 4 859 8 10 25 9 1 5 1 2 Pesticide data For the pesticide input data the example compound FOCUS B was considered a fast leaching substance with KOC 17 L kg of and DT50 of 20 d Q10 2 58 The application pattern was a single application of 1 kg ha to the soil surface on 1 May An overview on all pesticide data is given in Table 4 41 Table 4 Pesticide input parameters used for the test simulations Parameter Unit Value Molar mass g mol 1 300 Solubility in water mg L 1 90 Molar enthalpy of dissolution kJ mol 1 27 Vapour pressure at 20 C mPa 0 1 Molar enthalpy of vaporisation kJ mol 1 95 Diffusion coefficient in water m2 d 1 4 3 10 5 Gas diffusion coefficient m2 d 1 0 43 Reference temperature for degradation vaporisation and di
29. tools provided in the Control Panel under Add Remove Programs After successful installation the main form of InversePELMO will appear when calling the file InversePELMO exe Figure 3 If you call InversePELMO for the first time please make sure that the path to PELMO which is given on top of the main form is correct You can modify the path after clicking at the input field on the form 4 2 File handling between InversePELMO and PELMO PELMO is the standard model for doing leaching simulations for registration purposes in Germany Holdt et al 2011 and in Europe FOCUS 2009 However PELMO with its normal shell is not designed to perform inverse modelling studies because these studies require several model runs including automatic modification of input files based on the comparison with experimental results A scheme that shows the file handling is presented in Figure 2 for an optimisation of pesticide properties based on cumulative fluxes in the leachate All pesticide and application parameters are gathered in text files with extension psm The scenario input data can be found in files with extension sze Before starting the inverse modelling calculation a first simulation with initial conditions for either the soil hydrology or pesticide properties should be prepared using the normal shell which can be called directly from InversePELMO The optimisation itself is done automatically by InversePELMO As shown in
30. value 0 00000 0 00000 0 00000 0 00000 0 00000 1 912200E 06 1 061063E 04 0 517289 17 4226 101 598 210 274 333 333 367 595 421 421 435 476 483 483 486 583 421 486 647 922 873 873 807 Residual 0 00000 0 00000 0 00000 0 00000 0 00000 1 912200E 06 1 061063E 04 0 943951 7 61074 11 0053 4 57728 1 02527 F3x71252 3 71252 2 82796 2 68214 2 68214 2 68214 1 16022 0 900897 2 27676 2 46255 3 49165 3 49165 Weight 000 000 000 000 000 Group no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name no name See file PEST PESTICIDE RES for more details of residuals in graph ready format See file PEST PESTICIDE SEO for composite observation sensitivities Objective function Sum of squared weighted residuals Correlation Coefficient Correlation coefficient ie phi 296 8 0 9999 68 Analysis of residuals gt All residuals Number of residuals with non zero weight 24 Mean value of non zero weighted residuals 0 6786 Maximum weighted residual observation o010 Lis Oo Minimum weighted residual observation o013 3 713 Standard variance of weighted residuals 13 49 Standard error of weighted residuals 36 13 Note the above variance was obtained by dividing th
31. values Previous parameter values koc 5 04200 koc 5 00600 kdeg 1 671000E 02 kdeg 1 560000E 02 Maximum relative change 7 1154E 02 kdeg OPTIMISATION ITERATION NO 3 Model calls so far 8 Starting phi for this iteration 1 55900E 05 Lambda 0 62500 gt Phi 9 59417E 05 6 154 times starting phi Lambda 0 31250 gt Phi 60826E 06 10 316 times starting phi Lambda 152500 gt Phi 80902E 05 1 160 times starting phi Lambda 2 5000 gt Phi 33430E 05 0 856 of starting phi Lambda 5 0000 gt Phi 51729E 05 0 973 of starting phi No more lambdas phi rising Lowest phi this iteration 1 33430E 05 Current parameter values Previous parameter values koc 6 01900 koc 5 04200 kdeg 2 063000E 02 kdeg 1 671000E 02 Maximum relative change 0 2346 kdeg OPTIMISATION ITERATION NO 4 Model calls so far 17 Starting phi for this iteration 1 33430E 05 Lambda 2 5000 gt Phi 1 31355E 05 0 984 of starting phi Lambda 1 2300 gt See gt Phi 1 30446E 05 0 978 of starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 1 30446E 05 Current parameter values Previous parameter values koc 6 10300 koc 6 01900 kdeg 2 147000E 02 kdeg 2 063000E 02 Maximum relative change 4 0717E 02 kdeg OPTIMISATION ITERATION NO 5 Model calls so far 23 Starting phi for this iteratio
32. z e DEG Number of years i Study begin dd mm yy Pesticide input file Scenario input file Climate input file s se aize psm H MAIZE sze HMBGNORM CLI Figure 18 InversePELMO InversePELMO experimental data pesticide fate 4 4 8 Step 8 Enter experimental data pesticide fate For the optimisation of the pesticide fate the experimental results has to be entered ina specific form see Figure 19 30 Experimental Residues o 3 D Cone ug L Weighting factor 7 2 3 4 5 6 8 N o Percolate concentrations Soil concentrations Figure 19 InversePELMO Experimental concentrations in the percolate Dependent on what input data is available see the radio button on the form in Figure 19 either percolate concentrations or soil concentrations have to be entered If percolate concentrations should be used for the optimisation the concentration in ug L are needed for each sampling date during the study If appropriate the individual sampling could be characterised by weighting factors in the final column The user cannot enter any date her because the sampling dates are taken from the previous percolate optimisation If the user wants to use soil concentrations instead the respective table is loaded when the radio button is used Figure 20 31 Experimental Residues Upper depth cm Lower depth cm Conc ug kg Weighting factor 15 20
33. 0 10000 Relative phi reduction indicating convergence 0 10000E 01 Number of phi values required within this range 3 Maximum number of consecutive failures to lower phi 3 Minimal relative parameter change indicating convergence 0 10000E 01 Number of consecutive iterations with minimal param change 3 Maximum number of optimisation iterations 30 Attempt automatic user intervention no OPTIMISATION RECORD INITIAL CONDITIONS Sum of squared weighted residuals ie phi 6 10098E 05 Current parameter values koc 5 00000 kdeg 1 099000E 02 OPTIMISATION ITERATION NO 1 Model calls so far 5 1 Starting phi for this iteration 6 10098E 05 Lambda 55 0000 gt Phi 1 65784E 05 0 272 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 1 65784E 05 Current parameter values Previous parameter values koc 5 00600 koc 5 00000 kdeg 1 560000E 02 kdeg 1 099000E 02 Maximum relative change 0 4195 kdeg 74 OPTIMISATION ITERATION NO 5 2 Model calls so far 4 Starting phi for this iteration 1 65784E 05 Lambda 225000 gt Phi 1 56340E 05 0 943 of starting phi Lambda 1 25000 gt Phi 1 55900E 05 0 940 of starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 1 55900E 05 Relative phi reduction between optimisation iterations less than 0 1000 Switch to central derivatives calculation Current parameter
34. 0 15625 gt Phi 2 56993E 05 1 970 times starting phi No more lambdas phi rising 65 Lowest phi this iteration 91098 Current parameter values Previous parameter values koc 7 81500 koc 6 10300 kdeg 2 668000E 02 kdeg 2 147000E 02 Maximum relative change 0 2805 kogt OPTIMISATION ITERATION NO 6 Model calls so far x 30 Starting phi for this iteration 91098 Lambda 0 15625 gt Phi 1 92513E 05 2 113 times starting phi Lambda 7 81250E 02 gt Phi 4 90305E 05 5 382 times starting phi Lambda 0 31250 gt Phi 47871 0 525 of starting phi Lambda 0 62500 gt Phi 72670 0 798 of starting phi No more lambdas phi rising Lowest phi this iteration 47871 Current parameter values Previous parameter values koc 10 2400 koc 7 81500 kdeg 3 122000E 02 kdeg 2 668000E 02 Maximum relative change 0 3103 koc OPTIMISATION ITERATION NO 7 Model calls so far 38 Starting phi for this iteration 47871 Lambda 0 31250 gt Phi 15984 0 334 of starting phi Lambda 0 15625 gt Phi LYLL2 0 399 of starting phi Lambda 0 62500 gt Phi 29252 0 611 of starting phi No more lambdas phi rising Lowest phi this iteration 15984 Current parameter values Previous parameter values koc 12 1100 koc 10 2400 kdeg 3 225000E 02 kdeg 3 122000E 02 Maximum relative change 0 1826 koc OPTIMISATION ITERATION
35. 0 1630 0 1630 WILTING POINT WATER DISPERSION ORGANIC CONTENT LENGTH CARBON CM CM CM 3 0 0640 5 0000 5000 0 0470 5 0000 0000 0 0400 5 0000 2000 0 0220 5 0000 0000 0 0220 5 0000 0000 Total number of layers in the top meter PLOT FILE INFORMATION NUMBER OF PLOTTING VARIABLES TIMSER NAME MODE LEAC TSER DEPTH CM ARGUMENT 100 102 21 21 CONSTANT 0 1000E 10 SUBSTANCE PESTIC
36. 00 kdeg 2 103000E 02 kdeg 1 099000E 02 Maximum relative change 0 9136 kdeg OPTIMISATION ITERATION NO 2 Model calls so far 4 Starting phi for this iteration 8 43319E 07 Lambda 2 5000 gt sene gt Phi 1 14493E 07 0 136 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 1 14493E 07 Current parameter values Previous parameter values koc 50 1100 koc 50 0400 kdeg 3 124000E 02 kdeg 2 103000E 02 Maximum relative change 0 4855 kdeg 95 OPTIMISATION ITERATION NO 3 Model calls so far i 7 Starting phi for this iteration 1 14493E 07 Lambda 1 2500 _ 0 e gt Phi 1 49660E 06 0 131 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 1 49660E 06 Current parameter values Previous parameter values koc 50 2400 koc 50 1100 kdeg 4 157000E 02 kdeg 3 124000E 02 Maximum relative change 0 3307 kdeg OPTIMISATION ITERATION NO 4 Model calls so far i 10 Starting phi for this iteration 1 49660E 06 Lambda 0 62500 ees gt Phi 1 81613E 05 0 121 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 1 81613E 05 Current parameter values Previous parameter values koc 50 5300 koc 50 2400 kdeg 5 155000E 02 kdeg 4 157000E 02 Maximum relative change 0 2401 kdeg OPTIMISATION ITERATION NO 5 Model calls so far 13 Starting phi for this iteration 1 816
37. 085 1 91 PEST control file Pest pesticide pst pct control data restart 25 TEO 2 0 1 1 single point 5 0 2 0 0 3 0 03 10 3 0 3 0 0 001 0 30 0 01 3 3 0 01 3 1 1 1 group definitions and derivative data KOC relative 0 01 0 0 switch 2 0 parabolic KDEG relative 0 01 0 0 switch 2 0 parabolic parameter data KOC none relative 50 1 1000 KOC 1 0000 0 00000 KDEG none relative 1 09861228866811E 02 6 93147180559945E 04 0 693147180559945 KDEG 1 0000 0 00000 observation data ol o2 o3 o4 o5 o6 o7 0 08 12 271 ab 09 54 121 1 010 104 3085 1 model command line run_pesticide model input output pesticide tpl Substance 2 psm pest ins pest flux plm prior information ooooo0o0 HFHrHrHreHhrrr 92 PELMO output file PEST flux pim 0031 0092 0122 0153 0184 0212 0396 0457 0487 0518 31 31 30 31 31 28 3 31 30 31 08 10 TI 12 01 02 08 10 11 12 01 01 01 01 02 02 02 02 02 02 2 NER 5 GU aS 4360315E 16 517069544315E 11 25404336295443E 07 03420123362954E 05 8029550123363E 03 742968255012336 9 71547825501234 55 7414252550123 103 614612255012 93 PEST output file Pest_pesticide rec PEST RUN RECORD CASE PEST_PESTICIDE PEST run mode Parameter estimation mode Case dimensions Number of parameters Number of adjustable parameters Number of parameter groups Number of observations alt
38. 13E 05 Lambda 0 31250 gt Phi 19869 0 109 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 19869 Current parameter values Previous parameter values koc 51 2500 koc 50 5300 kdeg 5 993000E 02 kdeg 5 155000E 02 Maximum relative change 0 1626 kdeg OPTIMISATION ITERATION NO 7 6 Model calls so far 16 Starting phi for this iteration 19869 Lambda 0 15625 gt Phi 3407 5 0 172 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 3407 5 Current parameter values Previous parameter values koc 53 3100 koc 51 2500 kdeg 6 387000E 02 kdeg 5 993000E 02 Maximum relative change 6 5743E 02 kdeg OPTIMISATION ITERATION NO 7 Model calls so far 19 Starting phi for this iteration 3407 5 Lambda 7 81250E 02 gt Phi 1394 7 0 409 of starting phi Lambda 3 90625E 02 gt Phi 1095 6 0 322 of starting phi Lambda 1 95313E 02 gt Phi 2830 2 0 831 of starting phi No more lambdas phi rising Lowest phi this iteration 1095 6 Current parameter values koc 66 4100 kdeg 5 008000E 02 Maximum relative change 0 2457 OPTIMISATION ITERATION NO Model calls so far s for this iteration Starting phi Lambda 1 95313E 02 gt Phi 225 40 C 0 206 No more lambdas Lowest phi this iteration phi is less than 0 225 40 Current parameter values koc 78 1200
39. 2200E 06 000 no_name o7 0 00000 1 061063E 04 1 061063E 04 000 no_name 08 1 46124 0 517289 0 943951 000 no_name 09 25 0333 17 4226 7 61074 000 no_name o10 112 653 101 647 11 0053 000 no_name oll 215 175 210 598 4 57728 000 no_name 012 273 907 274 922 1 01527 000 no name olg 330 160 333 873 3 71252 000 no name ol4 330 160 333 813 3 71252 000 no name ols 364 979 367 807 2 82796 000 no name o16 418 913 421 595 2 68214 000 no name o17 418 913 421 595 2 68214 000 no name 018 418 913 421 595 2 68214 000 no_name 019 434 772 435 933 1 16022 000 no name 020 477 439 476 538 0 900897 000 no name 021 485 543 483 266 2 27676 000 no name 022 486 385 483 923 2 46255 000 no name 023 490 075 486 583 3 49165 000 no name 024 490 075 486 583 3 49165 000 no name See file PEST PESTICIDE RES for more details of residuals in graph ready format See file PEST PESTICIDE SEO for composite observation sensitivities Objective function gt Sum of squared weighted residuals ie phi 296 8 Correlation Coefficient Correlation coefficient 0 9999 Analysis of residuals 78 All residuals Number of residuals with non zero weight 24 Mean value of non zero weighted residuals 0 6786 Maximum weighted residual observation o010 11 01 Minimum weighted residual observation o013 3 713 Standard variance of weighted residuals 13 49 Standard error of weighted residuals 34673 Note the above variance was obtained by divi
40. 24 501524 136824 136824 136824 461424 514924 392164 281764 351524 351524 62 PEST output file Pest water rec PEST RUN RECORD CASE PEST PESTICIDE PEST run mode Parameter estimation mode Case dimensions Number of parameters Number of adjustable parameters Number of parameter groups Number of observations 2 Number of prior estimates OPNN DN Model command line s run_pesticide Jacobian command line na Model interface files Templates PESTICIDE TPL for model input files PESTICIDE_B_EXAMPLE_1 PSM Parameter values written using single precision protocol Decimal point always included Instruction files PEST INS for reading model output files PEST_FLUX PLM PEST to model message file na Derivatives calculation Param Increment Increment Increment Forward group type low bound central koc relative 1 0000E 02 none switch kdeg relative 1 0000E 02 none switch Parameter definitions Name Trans Change Initial formation limit value koc none relative 5 00000 kdeg none relative 1 099000E 02 Name Group Scale Offset koc koc 1 00000 0 00000 kdeg kdeg 1 00000 0 00000 Prior information No prior information supplied Observations or Multiplier Method central central 2 000 parabolic 2 000 parabolic Lower Upper bound bound 1 00000 1000 00 6 900000E 04 0 693150 Model command number 1 63 Observation name Observa
41. 46 20 93 24 23 Minimum error for which the Chi Test passes according to FOCUS 4 75 99 PELMO output file ECHO PLM KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK PESTICIDE LEACHING MODEL 8 i PELMO 4 00 Dec 2010 x KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK DEVELOPED BY U S ENVIRONMENTAL PROTECTION AGENCY OFFICE OF REASEARCH AND DEVELOPMENT ATHENS ENVIRONMENTAL RESEARCH LABORATORY ATHENS GA 30613 404 546 3138 AND ANDERSON NICHOLS 2666 EAST BAYSHORE RD PALO ALTO CA 94303 AND FRAUNHOFER INSTITUTE POSTFACH 1260 D 57377 SCHMALLENBERG Tel 49 2972 302 317 AND SLFA Neustadt DEPARTMENT ECOLOGY D 67435 NEUSTADT WSTR Tel 49 6321 671 422 PELMO 4 00 Dec 2010 kkkkkkkxkkkx kx x xx x xHYDROLOGY DATAS KKK KKK KKK KKK YEAR 1 Grafschaft 1 YEAR 2 Grafschaft 2 HYDROLOGY AND SEDIMENT RELATED PARAMETERS Temperature data are used to calculate pan evaporation LATTITUDE OF THE LOCATION 53 50 PAN COEFFICIENT FOR EVAPORATION NO CROP PAN COEFFICIENT FOR EVAPORATION MID SEASON PAN COEFFICIENT FOR EVAPORATION LATE SEASON FLAG FOR ET 0 EVAP 1 TEMP 2 EVAP TEMP 3 HAUDE DEPTH TO WHICH ET IS COMPUTED YEAR ROUND CM MONTHLY DAYLIGHT HOURS MONTH DAY HOURS MONTH DAY HOURS JAN 7 728 FEB 9 314 APR 14 04 MAY 15 98 JULY 16 33 AUG 14 68 OCT 9 890 NOV 8 016 SNOW MELT COEFFICIENT CM DEG C DAY INITIAL CROP NUMBER INITIAL CROP CONDITION NO CALCULATION OF RUNOFF EVENTS
42. 5 0 1 10 KcO 1 0000 ei none relative 0 5 0 1 10 Kel 1 0000 c2 none relative 0 5 0 1 10 Kc2 1 0000 ANETD none relative 30 5 100 ANETD 1 0000 O10 none relative 0 2 0 01 0 5 MOTO observation data ol 0 1 o2 0 1 03 0 1 04 0 1 05 5 6 1 06 32 96 1 07 51 18 T o8 117 6 1 09 179 96 di 010 242 59 oll 289 34 012 315 27 013 342 38 ol4 342 38 015 361 81 016 402 18 017 402 18 018 402 18 019 419 16 020 513 14 o21 569 81 022 579 49 023 649 1 024 649 1 model command line run_water model input output scenario tpl BORSTEL_Example_1 SZE pest ins pest_water plm prior information 1 0000 0 00000 0 00000 0 00000 0 00000 0 00000 PELMO output file PEST water pim 61 0031 3105 01 0061 0092 0123 0153 0184 0214 0245 0276 0304 0335 0365 0396 0426 0457 0488 0518 0549 0579 0610 0641 0669 0700 0730 30 31 3 30 3 30 31 3 28 3 30 31 30 31 31 30 31 30 31 31 28 3 30 06 07 08 09 10 11 12 01 02 03 04 05 06 07 08 09 10 11 12 01 02 03 04 01 01 01 01 01 01 01 02 02 02 02 02 02 02 02 02 02 02 02 03 03 03 03 22 4 5 13 13 13 32 oly 117 180 243 290 314 344 344 362 402 402 402 417 io 568 578 648 648 205E 14 7984E 14 8470000000001 8470000000001 8470000000001 6090000000001 1211000000001 7021 24083 13033 329424 046424 161524 1615
43. 6 15 6 5 November 159 0 5 8 December 96 483 1 126 1 4 8 0 PET potential evpotranspiration 48 The crop considered for the simulation was winter rape two seasons The standard crop and crop rotation information was used Begin of the study 2 was the 1 August with an application of 1 2 kg ha on 21 August 5 2 2 Pesticide data An overview on all pesticide data is given in Table 10 Table 10 Pesticide input parameters used for the test simulations Parameter Unit Value Molar mass g mol 1 100 Solubility in water mg L 1 90 Molar enthalpy of dissolution kJ mol 1 27 Vapour pressure at 20 C mPa 0 1 Molar enthalpy of vaporisation kJ mol 1 95 Diffusion coefficient in water m2 d 1 4 3 10 5 Gas diffusion coefficient m2 d 1 0 43 Reference temperature for degradation vaporisation and C 20 dissolution Reference soil moisture for degradation at 10 kPa field capacity Q10 factor increase of degradation rate with an increase of 2 58 temperature of 10 C Arrhenius activation energy kJ mol 1 65 4 B exponent of degradation moisture relationship according to 0 7 Walker Exponent of the FREUNDLICH Isotherm 0 9 Non equilibrium sorption TSCF transpiration stream concentration factor 5 2 3 Lysimeter results test data set 2 not considered 0 0 The main results of the study 2 are summarised in Table 11 The maximum concentration in the leachate was detecte
44. 6 85 77 1 000 times starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration Current parameter values 14 7900 koc kdeg 296 3 304000E 02 Maximum relative change Optimisation complete Total model calls 79 0 000 85 Previous parameter values 14 7900 koc kdeg koc 3 successive iterations 3 304000E 02 relative parameter change less than 1 0000E 02 over The model has been run one final time using best parameters Thus all model input files contain best parameter values and model output files contain model results based on these parameters OPTIMISATION RESULTS Parameters gt Parameter Estimated 95 percent confidence limits value lower limit upper limit koc 14 7900 14 6360 14 9440 kdeg 3 304000E 02 3 298439E 02 3 309561E 02 Note confidence limits provide only an indication of parameter uncertainty They rely on a linearity assumption which may not extend as far in parameter space as the confidence limits themselves see PEST manual See file PEST_PESTICIDE SEN for parameter sensitivities Observations gt Observation Measured Calculated Residual Weight Group value value ol 0 00000 0 00000 0 00000 000 no name o2 0 00000 0 00000 0 00000 000 no_name o3 0 00000 0 00000 0 00000 000 no_name o4 0 00000 0 00000 0 00000 000 no_name o5 0 00000 0 00000 0 00000 000 no_name o6 0 00000 1 912200E 06 1 91
45. ATION DATE 15 BRP 2 15 SEP 15 SEP LY SEP G 20 00 0 3333E 03 0 1000E 03 300 0 90 00 30 00 0 3333E 03 0 1000E 03 4303 0 1000 0 1368E 06 0 1322E 06 0 5000 MOISTURE DURING STUDY ABSOLUTE RELATIVE 1 YES MOISTURE EXPONENT HARVEST 3 SENESCENCE DATE 21 AUG 1 21 AUG 1 21 AUG 1 21 AUG 1 REL IN NEO DOMAIN MET Al MET Bl MET C1 MET D1 BR CO2 DAY 0 1000E 09 0 1000E 09 0 1000E 09 0 1000E 09 0 3304E 01 SORPTION PARAMETERS DEPTH DEPENDEND SORPTION AND TRANSFORMAT PARAMETERS TO CALCULATE KD VALUES WIT KOC CM 3 G FREUNDLICH SORPTION EXPONENT 1 n PEARL FACTOR DESCRIBING NON EQ S PEARL DESORPTION RATE C 0 0000 20 00 20 00 20 00 20 00 CIZDI 81 MIN CONC FOR FREUNDLICH SORPTION amp G L 5 TH KOC TES EQ SITES ON PARAMETERS HORIZON RUNE 5 Pest KOC KD CM 3 G CM 3 G 14 79 0 2219 14 79 0 1479 14 79 0 2958E 01 0 0000 0 0000 0 0000 0 0000 Borstel Mais Mais Mais GENERAL SOIL INFORMATION CORE PART BUL DEPTH CM TOTAL HORIZONS IN CORE TOTAL COMPARTMENTS IN CORE DPFLAG FLAG THETA FLAG TION COEFFICIENT FLAG DENSITY FLAG SOIL HYDRAULICS MODULE SOIL HORIZON INFORMATION BIODEG FACTOR HORIZON CM BUL PH THICKNESS DEN K SITY FR EXP 0 9000 0 9000 0 9000 0 9000 0 9000 TRANSFORMATION RATE TO MET Al
46. D PALO ALTO CA FRAUNHOFER INSTI POSTFACH 1260 13 94303 TUTE D 57377 SCHMALLENBERG Tel 49 2972 30 SLFA Neustadt DEPARTMENT ECOLOGY WSTR Tel 49 6321 671 422 D 67435 NEUSTADT Dec 2 317 2010 kkkkkkkkkkx kx kx x x xHYDROLOGY DATAS X kkkkkkKKK HYDROLOGY AND YEAR 1 Hamburg 1978 normal YEAR 2 Hamburg 1961 na YEAR 3 Hamburg 1978 normal SEDIMENT RELATED PARAMETERS Temperature data are used to calculate pan evaporation LATTITUDE OF THE LOCATION 50 PAN COEFFICIENT FOR EVAPORATION PAN COEFFICIENT FOR EVAPORATION PAN COEFFICIENT FOR EVAPORATION 00 NO CROP MID SEASON LATE SEASON FLAG FOR ET 0 EVAP 1 TEMP 2 EVAP TEMP 3 HAUDE DEPTH TO WHICH ET IS COMPUTED YEAR ROUND CM MONTHLY DAYLIGHT HOURS MONTH DAY HOURS MONTH DAY HOURS JAN 8 312 FEB 9 681 APR 12 76 MAY 15 40 JULY 15593 AUG 14 31 OCT 10 18 NOV 8 561 SNOW MELT COEFFICIENT CM DEG C DAY INITIAL CROP NUMBER INITIAL CROP CONDITION NO CALCULATION OF RUNOFF EVENTS CROP INFORMATION MAXIMUM INTERCEPT MAXIMUM MAXIMUM MAXIMUM USLE COVER MANAGEMENT CROP POTENTIAL ROOT DEPTH COVER WEIGHT AMC RUNOFF CURVE NUMBERS C FACTOR 0 9805 1 340 4 072 30 00 MONTH MAR JUNE SEP DEC 0 4600 IRRIGATION PERENNIAL TILLAGE FLAG 0 NO DAY 11 64 16 12 28 7 874 CROP 0 NO HOURS FLAG 0 NO SURFACE CONDITION AFTER NUMBER CM CM 3 FALLOW CROP RES
47. Figure 2 InversePELMO calls PEST which then reads the control file pest pesticide pst with all information about the parameters considered for the optimisation including their initial values and their allowed ranges Also the experimental data e g cumulative fluxes can be found in pest pesticide pst 12 According to the information in pesticide tpl PEST exe is able to create pesticide input files pesticide psm for PELMO including the correct position for the input parameters used in the optimisation After this file has been written PEST calls PELMO for a simulation To make the interface between PELMO and PEST more stable a second program is always executed after PELMO in the example presented in Figure 2 PELMO results pesticide exe which gathers the important simulation results e g calculated cumulative pesticide fluxes and writes them into the file pest plm After both programs PELMO and PELMO results pesticide exe are finished PEST gets control again and will read the important simulation results listed in pest plm instructions for PEST to read pest plm is given in pest ins According to the simulation results a new iteration is initiated with new DT50 and Kfoc data for the optimisation until the optimisation is finalised pest pesticide pst PEST control file pesticide tpl PELMO Input file description pest ins PELMO Output file description Pest plm Pesticide psm Sze files Cli files pestflux plm
48. IDUE FALLOW CROP I 72 67 72 3 0 0000 100 0 90 00 II 86 83 86 1 0 III 94 92 94 CROP ROTATION INFORMATION CROP TILLAGE HARVEST NUMBER DATE DATE Sillage Maize 20 OCT s 1 Sillage Maize 20 OCT 54 2 Sillage Maize 20 06T4 3 Sillage Maize 20 OCT 4 PARAMETERS OF ACTIVE SUBSTANCE KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK PESTICIDE APPLICATION INFORMATION KG M 2 80 RESIDUE 0 0000 000 1 0000 EMERGENCE DATE 5 MAY 5 MAY 5 MAY 5 MAY FOCUS DUMMY B PESTICIDE INCORPORATION APPLICATION APPLIED DEPTH DATE KG HA CM 1 MAY 1 000 0 0000 PLANT PESTICIDE PARAMETERS MODEL UTILIZED 1 SOIL 2 LINEAR 3 EXPONENTIAL VOLATILIZATION PARAMETERS ACTIVE SUBSTANCE TEMPERATURE deg C HENRY CONSTANT Pa m3 mole o CALCULATED USING VAPOUR PRESSURE MOLECULAR MASS WATER SOLUBIL Pa g mole TY mg 1 r J mole TEMPERATURE di HENRY CONSTANT CALCULATED US eg C Pa m3 mole or NG VAPOUR PRESSURE MOLECULAR MASS WATER SOLUBIL Pa g mole mg 1 TY Q10 Factor for Henry s consta DIFFUSION COEFF AIR cm2 d nt J mole DEPTH OF SURFACE LAYER FOR VOLATILIZATION CM HENRY CONSTANT AT HENRY CONSTANT AT 20 0 deg C 30 0 deg C PLANT UPTAKE OF ACTIVE SUBSTANCE PLANT UPTAKE FACTOR TRANSFORMATION PARAMETERS TRANSFORM TRANSFORM TEMP Q10 TRANSFORM TO in EQ Domaine OF STUDY VALUE 1 YES 1 0000 1 YES MATUR
49. InversePELMO A tool to perform inverse modelling studies with PELMO 4 0 FKZ 360 03 050 Fraunhofer Institut f r Molekularbiologie und Angewandte kologie 57392 Schmallenberg Head of the Institute Prof Dr R Fischer Developed by Dr M Klein Schmallenberg 30 November 2011 Content 1 2 4 1 4 2 4 3 4 4 5 1 5 2 8 1 8 2 Summary Background Description of InversePELMO Working with InversePELMO Installing InversePELMO File handling between InversePELMO and PELMO InversePELMO Main form InversePELMO Optimisation Results of test simulations Example data set 1 Leaching of Parent over a two years Test data set 2 Leaching of Parent over a 17 months Discussion and Conclusions References Documentation of Model Output InversePELMO Example data set 1 Example data set 2 11 11 11 13 18 39 39 46 56 57 59 59 83 List of Tables Table 1 Borstel soil profile in the lysimeter hypothetical test data set 1 39 Table 2 Further input parameters influencing evapotranspiration hypothetical test data set EE EEE EE EE 39 Table 3 Climate data during the study hypothetical test data set 1 40 Table 4 Pesticide input parameters used for the test simulations rrrnrrnnrrnrrrrrrrrrrrnnnrn 41 Table 5 Percolate and percolate concentrations in the lysimeter hypothetical test data set 1 EN KE EEE EE EE
50. NE NR ee 42 Table 6 Optimised parameter for the percolate hypothetical test data set 1 44 Table 7 Optimised parameter for the substance fluxes hypothetical test data set 1 45 Table 8 Borstel soil profile in the lysimeter test data Set 2 uunsssssrssssenennnnnnnnnnnnnen nenn 46 Table 9 Climate data during the study test data Set 2 rrnrrrnnnnnnvrvnnnnnnnvrnnnrnnnrnnnrnnnrrnnnnnnr 47 Table 10 Pesticide input parameters used for the test Simulations en 48 Table 11 Percolate and percolate concentrations in the lysimeter test data set 2 49 Table 12 Optimised parameter for the percolate test data set 2 rrrrrrrrrrnnnnrnrnrrrrrrrrrnnnnn 50 Table 13 Optimised parameter for the substance fluxes hypothetical test data set 2 52 Table 14 Extension of the lysimeter study substance flux hypothetical test data set 1 53 Table 15 Pesticide in the percolate at 1 m soil depth FOCUS Hamburg 55 List of Figures Figure 1 General flowchart of inverse modelling studies rrrrrrrrrnrvnvvrnrvnrnrrnvvenrvnrnrrnvanrrnne 9 Figure 2 Flow chart File handling of a flux optimisation with InversePELMO 12 Figure 3 InversePELMO Main form 4444444444440RnHnnnnannnnnnnnnnnnnnnnnnnnnnnnnnnonnnnnnnnnnnn nn 13 Figure 4 InversePELMO Path to FOCUS PELMO ernnnnnvnnnnnnnrrnnnnnnrrrrnnnnrnnrr
51. NO 8 Model calls so far 45 Starting phi for this iteration 15984 Lambda 0 31250 gt Phi 11485 0 719 of starting phi Lambda 0 15625 gt Phi 12288 0 769 of starting phi Lambda 0 62500 gt Phi 99057 0 620 of starting phi Lambda 1 2500 gt Phi 6987 3 0 437 of starting phi Lambda 2 5000 ees gt Phi 2734 8 0 171 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 2734 8 Current parameter values koc 14 4200 kdeg 3 334000E 02 Previous parameter values koc 12 1100 kdeg 3 225000E 02 Maximum relative change 0 1908 OPTIMISATION ITERATION NO Model calls so far for this iteration Starting phi Lambda 2 5000 gt Phi 478 67 6 051 28 phi is less than 0 478 67 No more lambdas Lowest phi this iteration Current parameter values koc 14 5200 kdeg 3 300000E 02 Maximum relative change 1 0198E 02 OPTIMISATION ITERATION NO Model calls so far Starting phi for this iteration Lambda 142500 r gt Phi 309 72 0 647 Lambda 0 62500 gt Phi 300 94 No more lambdas Lowest phi this iteration 300 94 Current parameter values koc 14 7800 kdeg 3 305000E 02 Maximum relative change OPTIMISATION ITERATION NO Model calls so far for this iteration Starting phi Lambda 0 31250 gt Phi 296 85 Lambda 0 15625 gt Phi 296 85 No more lambda
52. R PRESSURE Pa 0 4000E 03 MOLECULAR MASS g mole 100 0 WATER SOLUBILITY mg 1 1000 Q10 Factor for Henry s constant 2 000 DIFFUSION COEFF AIR cm2 d 4303 DEPTH OF SURFACE LAYER FOR VOLATILIZATION CM 0 1000 HENRY CONSTANT AT 20 0 deg C 0 8206E 08 HENRY CONSTANT AT 30 0 deg C 0 1587E 07 PLANT UPTAKE OF ACTIVE SUBSTANCE PLANT UPTAKE FACTOR 0 0000 TRANSFORMATION PARAMETERS TRANSFORM TRANSFORM TO MET Al TRANSFORM TEMP 010 MOISTURE DURING STUDY in EQ Domaine OF STUDY VALUE ABSOLUTE RELATIVE DAY C 0 1000E 09 20 00 2 580 0 0000 100 0 2 3 4 1 YES MOISTURE EXPONENT 0 7000 HARVEST SENESCENCE DATE 28 JUNE 2 28 JUNE 3 28 JUNE 4 REL IN NEQ DOMAIN 0 0000 MET Bl MET C1 MET D1 BR CO2 1000E 09 1000E 09 1000E 09 3086E 01 SORPTION PARAMETERS PARAMETERS TO CALCULATE KD VALUES WIT DEPTH DEPENDEND SORPTION AND TRANSFORMAT KOC CM 3 G FREUNDLICH SORPTION EXPONENT 1 n PEARL FACTOR DESCRIBING NON EQ S PEARL DESORPTION RATE 20 00 20 00 20 00 20 00 1 D 101 580 580 NNNN 580 MIN CONC FOR FREUNDLICH SORPTION amp G L 0 0000 0 0000 19 00 0 0000 TH KOC ON PARAMETERS HORIZON KOC CM 3 G 1 9517 2 95 17 0 3 95 17 0 4 0 0000 0 5 0 0000 0 Subs Ver 3 Hamburg oil KD seed rape CM 3 G 1 428 9357 1903 0000 0000 GENERAL SOIL INFORMATION CORE DEPTH CM
53. See file PEST_WATER SEO for composite observation sensitivities Objective function gt Sum of squared weighted residuals ie phi 4779 Correlation Coefficient gt Correlation coefficient 0 9958 Analysis of residuals gt All residuals Number of residuals with non zero weight 10 Mean value of non zero weighted residuals 3 8712E 03 Maximum weighted residual observation o7 43 37 Minimum weighted residual observation 09 34 59 Standard variance of weighted residuals 796 6 Standard error of weighted residuals 28 22 Note the above variance was obtained by dividing the objective function by the number of system degrees of freedom ie number of observations with non zero weight plus number of prior information articles with non zero weight minus the number of adjustable parameters If the degrees of freedom is negative the divisor becomes the number of observations with non zero weight plus the number of prior information items with non zero weight Covariance and other statistical matricies cannot be determined Normal matrix nearly singular cannot be inverted Minimum error for which the Chi Test passes according to FOCUS 4 75 90 8 2 2 Optimisation of pesticide fate InversePELMO control file sampling pesticide txt 01 08 01 0 10 5 31 8 1 0 31 0 1 0 30 1 1 0 31 2 1 0 31 2 0 28 2 2 0 31 8 2 0 31 0 2 12 271 30 1 2 54 121 1 31 2 2 104 3
54. Their initial values and their possible range are shown in Figure 28 43 Parameter Initial value Min value Max value Kc factor no crop Kc factor mid season Kc factor late season Minimum depth for evaporation cm Initial soil water content m m Cancel o a Done Figure 28 Parameters used in the optimisation of the percolate hypothetical test data set 1 After the optimisation the results summarised in Figure 29 were obtained Zumulative percolate Lim 600 500 400 300 200 100 ays 0 100 200 300 400 500 600 700 Figure 29 Results of the optimisation percolate hypothetical test data set 1 44 The minimum error for which the Chi Test passes according to FOCUS was found to be 1 28 which supports the excellent agreement shown in the figure Table 6 Optimised parameter for the percolate hypothetical test data set 1 Parameter Estimated value KCO 0 98 KC1 1 34 KC2 4 07 ANETD 30 MOIO 0 2 The results summarised in Table 6 shows that the inverse modelling tool did not find the same parameter setting as used when producing the hypothetical test data However more important than identical parameters is the correct description of the percolate by the leaching model because many parameter combinations lead to similar results For the optimisation of the substance fluxes the parameters DT50 and KOC were considered in the fitting Their initial values and their po
55. ative 1 00000 0 500000 10 0000 ko none relative 1 00000 0 500000 10 0000 moid none relative 0 200000 5 000000E 02 0 500000 ame Group Scale Offset Model command number kco kc0 1 00000 0 00000 1 kel kel 1 00000 0 00000 T kc2 kc2 1 00000 0 00000 I moid moid 1 00000 0 00000 1 Pri Obs Obs ol 02 03 o4 o5 87 or information No prior information supplied ervations ervation name Observation Weight Group 47 9000 000 no name 124 600 000 no_name 177 350 000 no_name 222 400 000 no_name 255 500 000 no_name 346 900 000 no_name 495 250 000 no_name 582 900 000 no_name 675 900 000 no_name 0 767 150 000 no_name Control settings Initial lambda 5 0000 Lambda adjustment factor 2 0000 Sufficient new old phi ratio per optimisation iteration 0 30000 Limiting relative phi reduction between lambdas 3 00000E 02 Maximum trial lambdas per iteration 10 Maximum factor parameter change factor limited changes na Maximum relative parameter change relative limited changes 3 0000 Fraction of initial parameter values used in computing change limit for near zero parameters 1 00000E 03 Allow bending of parameter upgrade vector no Allow parameters to stick to their bounds no Relative phi reduction below which to begin use of central derivatives 0 10000 Relative phi reduction indicating convergence 0 10000E 01 Number of phi values required within this range 3 Maximum number of consecutive failures to lower phi 3 Minimal re
56. ative phi reduction between optimisation iterations less than 0 1000 Switch to central derivatives calculation Current parameter values Previous parameter values kco 1 00000 kc0 1 00000 kc1 1 00000 kc1 1 00000 kc2 1 00000 kc2 1 00000 moid 0 281140 moiO0 0 285800 Maximum relative change 1 6305E 02 moi0 OPTIMISATION ITERATION NO 3 Model calls so far 12 Starting phi for this iteration 4779 6 Parameter kc0 has no effect on observations Parameter kcl has no effect on observations Parameter kc2 has no effect on observations Lambda 1 2500 gt Phi 4779 4 1 000 of starting phi Lambda 0 62500 gt Phi 4779 4 1 000 of starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 4779 4 Current parameter values Previous parameter values kcO 1 00000 kco 1 00000 kel 1 00000 kel 1 00000 kc2 1 00000 kc2 1 00000 moid 0 280990 moid 0 281140 Maximum relative change 5 3354E 04 moi0 OPTIMISATION ITERATION NO 4 Model calls so far 22 Starting phi for this iteration 4779 4 Parameter kc0 has no effect on observations Parameter kc1 has no effect on observations Parameter kc2 has no effect on observations Lambda 0 62500 gt Phi 4779 4 1 000 of starting phi Lambda 0 31250 gt Phi 4779 4 1 000 of starting phi No more lambdas Lowest phi this iteration Current parameter values kco kc1 kc2 moid Maxim
57. bdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 300 94 Current parameter values Previous parameter values koc 14 7800 koc 14 5200 kdeg 3 305000E 02 kdeg 3 300000E 02 Maximum relative change 1 7906E 02 koc OPTIMISATION ITERATION NO TA Model calls so far 65 Starting phi for this iteration 300 94 Lambda 0 31250 gt Phi 296 85 0 986 of starting phi Lambda 0 15625 gt Phi 296 85 0 986 of starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 296 85 Current parameter values Previous parameter values koc 14 7900 koc 14 7800 kdeg 3 304000E 02 kdeg 3 305000E 02 Maximum relative change 6 7659E 04 koc OPTIMISATION ITERATION NO 12 Model calls so far E 71 Starting phi for this iteration 296 85 Lambda 0 15625 gt Phi 296 85 1 000 times starting phi Lambda 7 81250E 02 gt Phi 296 85 1 000 times starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 296 85 Current parameter values Previous parameter values koc 14 7900 koc 14 7900 kdeg 3 304000E 02 kdeg 3 304000E 02 Maximum relative change 0 000 tkoc OPTIMISATION ITERATION NO B 13 Model calls so far ED Starting phi for this iteration 296 85 Lambda 7 81250E 02 gt Phi 296 85 1 000 times starting phi Lambda 3 90625E 02 gt Phi 29
58. d at the end of the study December 0 55 ug L The total percolate collected was 767 L m2 49 Table 11 Percolate and percolate concentrations in the lysimeter test data set 2 Month Percolate L m Concentration ug L August 47 9 0 00 October 76 7 0 00 November 52 75 0 00 December 45 05 0 00 January 33 1 0 00 February 91 4 0 00 August 148 35 0 00 October 87 65 0 14 November 93 0 45 December 91 25 0 55 5 2 4 Optimisation test data set 2 For the optimisation of the percolate all possible parameters were considered in the fitting Their initial values and their possible range are shown in Figure 28 Initial value Min value Max value Kc factor mid season Kc factor late season Minimum depth for evaporation cm Initial soil water content m m Cancel o a Done J Figure 34 Parameters used in the optimisation of the percolate test data set 2 After the optimisation the results summarised in Figure 35 were obtained 50 a Pulang percolate Lim 700 600 500 400 300 200 100 days 0 100 200 300 400 500 Figure 35 Results of the optimisation percolate test data set 2 The minimum error for which the Chi Test passes according to FOCUS was found to be 4 13 which supports the excellent agreement shown in the figure Table 12 Optimised parameter for the percolate test data set 2 Parameter Estimated value KCO 0 5 KC1 2 304 KC2 1 00 MOIO 0 269 For
59. d phi ratio per optimisation iteration 0 30000 Limiting relative phi reduction between lambdas 3 00000E 02 Maximum trial lambdas per iteration go GA Maximum factor parameter change factor limited changes na Maximum relative parameter change relative limited changes 3 0000 Fraction of initial parameter values used in computing change limit for near zero parameters 1 00000E 03 Allow bending of parameter upgrade vector no Allow parameters to stick to their bounds no Relative phi reduction below which to begin use of central derivatives 0 10000 Relative phi reduction indicating convergence 0 10000E 01 Number of phi values required within this range 3 Maximum number of consecutive failures to lower phi i 3 Minimal relative parameter change indicating convergence 0 10000E 01 Number of consecutive iterations with minimal param change 8 3 Maximum number of optimisation iterations t 30 Attempt automatic user intervention no OPTIMISATION RECORD INITIAL CONDITIONS Sum of squared weighted residuals ie phi 6 20080E 08 Current parameter values koc 50 0000 kdeg 1 099000E 02 OPTIMISATION ITERATION NO 1 Model calls so far 1 Starting phi for this iteration 6 20080E 08 Lambda 30000 sees gt Phi 8 43319E 07 0 136 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 8 43319E 07 Current parameter values Previous parameter values koc 50 0400 koc 50 00
60. ding the objective function by the number of system degrees of freedom ie number of observations with non zero weight plus number of prior information articles with non zero weight minus the number of adjustable parameters If the degrees of freedom is negative the divisor becomes the number of observations with non zero weight plus the number of prior information items with non zero weight Parameter covariance matrix gt koc kdeg koc 5 5159E 03 3 6736E 07 kdeg 3 6736E 07 7 1906E 10 Parameter correlation coefficient matrix gt koc kdeg koc 1 000 0 1845 kdeg 0 1845 1 000 ormalized eigenvectors of parameter covariance matrix gt Vector 1 Vector 2 koc 6 6602E 05 1 000 kdeg 1 000 6 6602E 05 Eigenvalues gt 6 9459E 10 5 5159E 03 Parameter Estimated 95 percent confidence limits value lower limit upper limit koc 14 7900 14 6360 14 9440 DT50 20 98 20 94 21 01 Minimum error for which the Chi Test passes according to FOCUS 1 23 79 PELMO output file ECHO PLM KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK Dec 20 DEVELOPED BY AND AND AND PELMO 4 00 U S PESTICIDE LEACHING MODEL PELMO 4 00 10 KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK ENVIRONMENTAL PROTECTION AGENCY OFFICE OF REASEARCH AND DEVELOPMENT ATHENS ENVIRONMENTAL RESEARCH LABORATORY ATHENS GA 306 404 546 3138 ANDERSON NICHOLS 2666 EAST BAYSHORE R
61. e factor late season v Minimum depth for evaporation cm M Initial soil water content nm m oja Figure 13 InversePELMO Parameter for fitting the percolate In total five PELMO input parameters influencing the hydrology in soil can be used to do the fitting Kc factor no crop linear correction factor for daily potential evapotranpiration when there is no crop in the field Kc factor mid season linear correction factor for daily potential evapotranpiration when the crop is growing Kc factor late season linear correction factor for daily potential evapotranpiration when the crop is at senescence Minimum depth for evaporation cm minimum depth for evaporation which is used independent whether or not a crop is present initial soil water content soil water content at the beginning of the simulation If parameters shall be considered for the optimisation their initial values and their range have to be specified If a parameter is not checked the respective input fields is invisible If the form has been filled correctly e g no negative figures no minimum values higher than maximum values it can be closed using the Done button and the arrow on the optimisation form will jump to the next button which is the button for starting the optimisation see Figure 13 25 4 4 6 Step 6 Optimisation of the hydrology in soil percolate In step 6 of the sequence the optimisation is performed During the
62. e objective function by the number of system degrees of freedom ie number of observations with non zero weight plus number of prior information articles with non zero weight minus the number of adjustable parameters If the degrees of freedom is negative the divisor becomes the number of observations with non zero weight plus the number of prior information items with non zero weight Parameter covariance matrix gt koc kdeg koc 5 5159E 03 3 6736E 07 kdeg 3 6736E 07 7 1906E 10 Parameter correlation coefficient matrix gt koc kdeg koc 1 000 0 1845 kdeg 0 1845 1 000 ormalized eigenvectors of parameter covariance matrix gt Vector 1 Vector 2 koc 6 6602E 05 1 000 kdeg 1 000 6 6602E 05 Eigenvalues gt 6 9459E 10 9 9 L99E 03 Parameter Estimated 95 percent confidence limits value lower limit upper limit koc 14 7900 14 6360 14 9440 DT50 20 98 20 94 21 01 Minimum error for which the Chi Test passes according to FOCUS 1 23 69 8 1 2 Optimisation of pesticide fate InversePELMO control file sampling pesticide txt 01 05 01 0 24 22 31 5 1 0 l 30 6 1 0 1 31 7 1 0 a 31 8 1 0 L 30 9 1 0 a 31 10 1 0 1 30 11 1 0 1 31 T2 1 1 461241 31 1 2 25 03332 28 2 2 112 65269 31 3 2 215 17544 30 4 2 273 90689 31 5 2 330 16014 30 6 2 330 16014 31 7 2 364 9787 31 8 2 418 91302 30 9 2 418 91302 31 10 2 418 91302 30 11 2 434 77234 31 12 2 477 43926 31 1 3 485 54307 28 2 3
63. eate PELMO input files oo Import PELMO input files Figure 7 InversePELMO Optimisation form new project However these limitations do not held for inverse modelling studies previously performed see Figure 8 If the user did not go through all steps of the study so far last time some buttons may be still disabled when the project is called again The red arrow will then indicate where to continue 19 Optimisation Test Optimisation sequence 5 i Enter experimental data Enter experimental data GJEN Create PELMO input files Define fitting parameters Define fitting parameters 22 LA ii Import PELMO input files Start optimisation Start optimisation NG SS Check initial simulation View optimisation View optimisation PEMO simulation control Start simulation day dd mm Start simulation day dd mm 91 Jol 1 ape End simulation day dd mm 12 Number of years E Study begin dd mm yy a or bi Pesticide input file Pesticide_A_Maize psm Scenario input file H_MAIZE sze Climate input file s HMBGNORM CLI Figure 8 InversePELMO Optimisation form existing project In the following detailed information about the sequence of doing inverse modelling studies with InversePELMO is given assuming that a new project has been created 4 4 1 Step 1 Create PELMO input files In the first step the necessary input files for the inverse modelling study has to be created The click at t
64. escription as given in the PELMO 3 0 soil data base The soil profile information is summarised in Table 1 Table 8 Borstel soil profile in the lysimeter test data set 2 Horizon cm 0 30 30 57 57 73 73 90 90 110 Soil density g cm 1 5 1 6 1 58 1 62 1 6 Sand 68 3 67 0 96 2 98 8 100 Silt 24 5 26 3 29 0 2 0 Clay 7 2 6 7 0 9 0 0 OC 1 5 1 0 0 2 0 0 initial soil water content m8 m 0 05 0 05 0 05 0 05 0 05 Biodegradation factor 1 0 5 0 3 0 3 0 3 pH value 5 7 4 9 4 9 5 0 4 8 The Climatic conditions during the lysimeter study are summarised in Table 9 whereas the following two figures show the daily pattern Precipitation cm d il ML HIMAL ll VL SE Figure 32 Daily precipitation at the lysimeter station from August 2008 to December 2009 47 Actual evapotranspiration cm d 2 1 5 Day 500 l Nm PN b l LG Figure 33 Actual ET at the lysimeter station from from August 2008 to December 2009 Table 9 Climate data during the study test data set 2 Month Montly Annual Monthly PET Annual PET Monthly Annual Precipitatio Precipitation mm C Temperature Temperature n mm mm C C August 87 105 13 9 September 50 2 8 8 October 60 20 7 9 November 60 0 4 2 December 56 1 0 January 50 6 2 6 February 83 5 1 March 89 20 2 2 April 25 3 10 7 May 72 96 11 4 June 65 3 12 3 July 135 832 108 369 15 8 7 0 August 38 108 16 4 September 64 2 12 6 October 12
65. he respective button see Figure 7 will open the normal FOUCS PELMO shell for the editing If all scenario climate and pesticide input files are available and FOCUS PELMO closed the red arrow will jump to the next button 4 4 2 Step 2 Import PELMO input files In step 2 all PELMO input files will be copied into the project folder and some additional information about the duration of the lysimeter study has to be provided This information has 20 to be entered in the frame PELMO simulation control The start and end date of the simulation is related to the PELMO simulation whereas the study begin is related to the start of the lysimeter study which may be different when a warming up period is considered for PELMO in order to optimise the soil hydrology Optimisation user EIER Optimisation sequence inne Create PELMO input files heck initial simulatio view optimisatior view optimisation r PEMO simulation control Start simulation day dd mm Start simulation day dd mm g End simulation day damm Number of years Study begin dd mm yy x x Pesticide input file a ne Scenario input file E Climate input file s Select climate input files Figure 9 InversePELMO Import PELMO input files During this step also all necessary executables are copied into the project folder That should make the documentation of the various programmes used for the analysis more transparent 21
66. i 296 3 304000E 02 Maximum relative change Optimisation complete Total model calls The model has been run one final time using best parameters Thus all model input files contain best parameter values 79 0 000 Previous parameter values 14 7900 koc kdeg koc 3 successive iterations relative phi reduction between lambdas less than 0 0300 85 3 304000E 02 relative parameter change less than 1 0000E 02 over and model output files contain model results based on these parameters Parameters Parameter koc kdeg OPTIMISATION RESULTS Estimated value 14 7900 3 304000E 02 lower limit 14 6360 3 298439E 02 14 9440 95 percent confidence limits upper limit 3 309561E 02 Note confidence limits provide only an indication of parameter uncertainty They rely on a linearity assumption which may not extend as far in parameter space as the confidence limits themselves see PEST manual See file PEST_PESTICIDE SEN for parameter sensitivities Observations Observation ol o2 o3 o4 o5 o6 07 08 09 O 000000000 vooauPwMN Ho 00006 NNNN wWNHO 024 Measured value 0 00000 0 0 0 0 0 0 00000 00000 00000 00000 00000 00000 1 46124 25 0333 112 215 2735 330 330 364 418 418 418 434 477 485 653 17 5 907 160 160 979 913 913 913 772 439 543 486 490 490 385 075 075 Calculated
67. ific PELMO output file containing information used by PEST perkolat plm flux plm soil concentrationXX plm XX number of soil layer e g soil concentration01 plm e MS DOS batch files Run_water bat and Run pesticide bat created by PEST used to start the PELMO simulation and to prepare the simulation results for PEST e PEST output files e g pest soil concentrations rec with detailed information on the optimisation Additional files with the same name but different file extensions are created by PEST such as pest soil concentrations jac with other information e g about parameter sensitivity The files in the project folder should be not removed manually by the user because it may lead InversePELMO to crash when scrolling through the project list After a double click at one of the items in the list box or when using the button Open project InversePELMO will load the project 17 3 Study name Enter a name for the new project Type here new project o a Figure 6 InversePELMO New project 4 3 3 Command buttons Five buttons are placed on the main form allowing direct access to important functions of InversePELMO New project This button calls the dialogue shown in Figure 6 to name the new project It is not possible to use the name of an existing project If existing projects should be overwritten please delete the old project first After a click at Done the optimisation form will
68. kdeg 3 423000E 02 kdeg 11 35 10 545 of starting phi relative phi reduction between lambdas less than 0 0300 Relative phi reduction between optimisation iterations less than 0 1000 Switch to central derivatives calculation Current parameter values Previous parameter values koc 95 1900 koc 94 2900 kdeg 3 085000E 02 kdeg 3 138000E 02 Maximum relative change 1 6890E 02 kdeg OPTIMISATION ITERATION NO 12 Model calls so far Starting phi Lambda Phi Lambda Phi No more lambdas 97 39 for this iteration 10 203 3 05176E 04 gt 20 130 0 999 of starting phi 1 52588E 04 gt 10 190 0 999 of starting phi relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 10 190 Current parameter values Previous parameter values koc 95 1700 koc 95 1900 kdeg 3 086000E 02 kdeg 3 085000E 02 Maximum relative change 3 2415E 04 kdeg OPTIMISATION ITERATION NO 13 Model calls so far 45 Starting phi for this iteration 10 190 Lambda 1 52588E 04 gt Phi 10 195 1 000 times starting phi Lambda 7 62939E 05 gt Phi 10 195 1 000 times starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 10 195 Current parameter values Previous parameter values koc 95 1600 koc 95 1700 kdeg 3 086000E 02 kdeg 3 086000E 02 Maximum relative change 1 0508E 04 koc Optimisation complete the 3 lo
69. lative parameter change indicating convergence 0 10000E 01 Number of consecutive iterations with minimal param change 3 Maximum number of optimisation iterations 30 Attempt automatic user intervention no OPTIMISATION RECORD INITIAL CONDITIONS Sum of squared weighted residuals ie phi 56927 Current parameter values kco 1 00000 kc1 1 00000 kc2 1 00000 moid 0 200000 OPTIMISATION ITERATION NO f 1 Model calls so far B 1 Starting phi for this iteration 56927 Parameter kc0 has no effect on observations Parameter kc1 has no effect on observations Parameter kc2 has no effect on observations Lambda 5 0000 gt Phi 4984 2 0 088 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 4984 2 Current parameter values Previous parameter values kco 1 00000 kcO 1 00000 kel 1 00000 kel 1 00000 kc2 1 00000 kc2 1 00000 moi0 0 285800 moiO0 0 200000 Maximum relative change 0 4290 moiO 88 OPTIMISATION ITERATION NO 5 2 Model calls so far i 6 Starting phi for this iteration 4984 2 Parameter kc0 has no effect on observations Parameter kcl has no effect on observations Parameter kc2 has no effect on observations Lambda 25000 esse gt Phi 4779 6 0 959 of starting phi Lambda 1 2500 gt Phi 4779 6 0 959 of starting phi No more lambdas relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 4779 6 Rel
70. le Pesticide A Maizepsm ss Scenario input file HMAZEse 2 28 iti SCS S Climate input file s HMBGNORM CLI Figure 15 InversePELMO Analyse the results of the optimisation for soil hydrology percolate After clicking at the respective button arrow on Figure 15 a form is loaded showing the experimental and optimised cumulative percolate graphically see Figure 16 27 Evaluation Cumulative percolate Lim Figure 16 InversePELMO View the results of the optimisation percolate The circles represent the experimental data the curve stands for the PELMO optimisation Detailed output describing the optimisation procedure is available via the button View output file Figure 17 28 Evaluation Sum of squared weighted residuals ie phi 3 58ZZE 03 Correlation Coefficient Correlation coefficient Analysis of residuals All residuals Number of residuals with non zero weight Mean walue of non zero weighted residuals Maximum weighted residual observation olZ Minimum weighted residual observation ol Standard variance of weighted residuals Standard error of weighted residuals 12 58343E 02 6550E 02 0000E 03 980Z2E 04 9950E 0Z Note the above variance was obtained by dividing the objective function by the number of system degrees of freedom ie number of observations with non zero weight plus number of prior information articles with non zero weigh
71. mum number of optimisation iterations Attempt automatic user intervention OPTIMISATION RECORD INITIAL CONDITIONS Sum of squared weighted residuals ie phi 6 10098E 05 Current parameter values koc 5 00000 kdeg 1 099000E 02 OPTIMISATION ITERATION NO 1 Model calls so far 1 Starting phi for this iteration 5 0000 1 65784E 05 0 2 Lambda Phi No more lambdas phi is less than Lowest phi this iteration 1 657 Current parameter values koc 5 00600 kdeg 1 560000E 02 Maximum relative change 0 4195 6 10098E 05 72 of starting phi 0 3000 of starting phi 84E 05 5 0000 2 0000 0 30000 3 00000E 02 10 na 3 0000 1 00000E 03 no no 0 10000 0 10000E 01 3 3 0 10000E 01 3 30 no Previous parameter values koc kdeg kdeg 5 00000 1 099000E 02 OPTIMISATION ITERATION NO Model calls so far Starting phi for this iteration Lambda 2 5000 Phi 1 56340E 05 Lambda 1 2300 Phi 1 55900E 05 No more lambdas relative 64 2 4 1 65784E 05 0 943 of starting phi 0 940 of starting phi phi reduction between lambdas less than 0 0300 Lowest phi this iteration 1 55900E 05 Relative phi reduction between optimisation iterations less than 0 1000 Switch to central derivatives calculation Current parameter values Previous parameter values koc 5 04200 koc 5 00600 kdeg 1 671000E 02 kdeg 1 560000E 02 Maximum relative change 7 1154E
72. n 1 30446E 05 Lambda 0 62500 Phi 1 19313E 05 0 915 of starting phi Lambda 0 31250 gt Phi 91098 0 698 of starting phi Lambda 0 15625 gt Phi 2 56993E 05 1 970 times starting phi No more lambdas phi rising 75 Lowest phi this iteration 91098 Current parameter values Previous parameter values koc 7 81500 koc 6 10300 kdeg 2 668000E 02 kdeg 2 147000E 02 Maximum relative change 0 2805 koc OPTIMISATION ITERATION NO 6 Model calls so far 30 Starting phi for this iteration 91098 Lambda 0 15625 gt Phi 1 92513E 05 2 113 times starting phi Lambda 7 81250E 02 gt Phi 4 90305E 05 5 382 times starting phi Lambda 0 31250 gt Phi 47871 0 525 of starting phi Lambda 0 62500 gt Phi 72670 0 798 of starting phi No more lambdas phi rising Lowest phi this iteration 47871 Current parameter values Previous parameter values koc 10 2400 koc 7 81500 kdeg 3 122000E 02 kdeg 2 668000E 02 Maximum relative change 0 3103 koc OPTIMISATION ITERATION NO 7 Model calls so far 38 Starting phi for this iteration 47871 Lambda 0 31250 gt Phi 15984 0 334 of starting phi Lambda 0 15625 gt Phi 19112 0 399 of starting phi Lambda 0 62500 gt Phi 29252 0 611 of starting phi No more lambdas phi rising Lowest phi this iteration 15984 Current paramete
73. nnnnnnnrennnennnnene 14 Figure 5 InversePELMO Release information 44 us4444400nnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnannn 15 Figure 6 InversePELMO New project 444444444Hnnannnnnnnnnnnnnnnnnnnnnnnnnnnnannnnnnnnnnnnnnnnnnn 17 Figure 7 InversePELMO Optimisation form new project 4msss4444snnnensnnnnnnnnnnnnnnn 18 Figure 8 InversePELMO Optimisation form existing project 444n44 nennen 19 Figure 9 InversePELMO Import PELMO input files sur44444s00r nn nnnnnnnnnnnnnnnnnnnnnnn 20 Figure 10 InversePELMO experimental data percolate ss 44444nn nn nnnnnnnnnnnennnn 21 Figure 11 InversePELMO Experimental Percolate aronnnrrnronnnonornrrnnnnnnnrrnnnnnannnnnernnnnenne 22 Figure 12 InversePELMO define fitting parameters for the hydrology in soil 23 Figure 13 InversePELMO Parameter for fitting the percolate 4u0444444nn nennen 24 Figure 14 InversePELMO start optimisation for the hydrology in soil en 25 Figure 15 InversePELMO Analyse the results of the optimisation for soil hydrology peter 26 Figure 16 InversePELMO View the results of the optimisation percolate 27 Figure 17 InversePELMO Evaluate the results of the optimisation percolate 28 Figure 18 InversePELMO InversePELMO experimental data pesticide fate
74. nsidering the results of different studies on degradation and sorption of these compounds in soil Normally laboratory studies are performed to get the input parameters for the models considering the recommendations of FOCUS e g FOCUS 1997 FOCUS 2000 FOCUS 2006 Alternatively the necessary input parameters can be also obtained based on outdoor studies e g field dissipation studies Recently the new FOCUS groundwater group suggested a third methodology for the input parameter setting FOCUS 2009 The idea is to analyse outdoor studies especially lysimeter studies using the inverse modelling technique which allows the estimation of sorption and degradation parameters within a single step For this procedure an optimisation tool in this project the program PEST Watermark 2005 has to be combined with a leaching model in this project PELMO 4 Aim of inverse modelling studies is to find the KOC and DT50 values that could describe the outdoor study best considering all data recorded during the experiments e g rainfall temperatures percolate and substance fluxes However the whole procedure is rather complicated and detailed knowledge on leaching models is needed to successfully perform inverse modelling studies The software presented here was developed in order to make available a user friendly tool to perform or check inverse modelling studies performed with PELMO and PEST 3 Description of InversePELMO Inverse modelling st
75. nversePELMO Release information On this form the button check for update can be used to look for new releases of InversePELMO 16 4 3 2 Project list On the left side the form shows a list of all inverse modelling studies projects available Information for each project is gathered in a respective directory of the same name All project directories are located in the folder projects Usually a project folder has following content e Upto 5 executables the optimisation tool PEST pest exe the simulation model PELMO pelmo400 exe and its post processors pelmo results water exe gathering information on simulated percolate belmo results pesticide exe gathering information on cumulative substance fluxes and pelmo_results_soil_profile exe PELMO gathering information on soil concentrations e Standard PELMO input files the normal pesticide and scenario input files with extension psm and sze respectively Furthermore at least one climate data file extension cli and the normal control file of PELMO called pelmo inp e Standard PEST control files pest_pesticde pst and pest water pst scenario tpl information how to create PELMO scenario input files pesticide tpl information how to create PELMO pesticide input files pest ins information how to extract simulation results out of PELMO e Standard PELMO output files echo plm wasser plm chem plm plot plm e Spec
76. of the inverse modelling study cannot be simply done based on the standard user shell The optimisation software needs exact information about the location of individual simulation results e g percolate at a specific day of the simulation in order to compare predictions with experimental data The new shell InversePELMO is able to provide both programs the optimisation tool and the simulation model with the necessary input files in the correct format InversePELMO has also a built in module to perform standard statistical tests to check the quality of the optimisation such as the determination of the FOCUS error at which the chi error passes described in FOCUS 2006 The optimisation tool selected for InversePELMO was PEST version 7 2 Watermark 2003 because it met all criteria necessary to combine it with PELMO in a DOS environment It was also tested that PEST works under all relevant windows systems Windows XP Windows VISTA and Windows 7 32 bit as well as 64 bit version 11 4 Working with InversePELMO 4 1 Installing InversePELMO To install InversePELMO following steps have to be done 1 Install PELMO 4 if it has not been done so far Call InversePELMO setup zip Select a directory and start unzipping the files into a temp folder After unzipping close InversePELMO setup zip O W DR call setup exe in the folder where the files were unzipped InversePELMO may be un installed using first the standard MS Windows un install
77. ptimisation testtest Optimisation sequence v3 Enter experimental data AEA Create PELMO input files Import PELMO input files Start optimisation Start optimisation V Check initial simulation View optimisation View optimisation PEMO simulation control Start simulation day dd mm Start simulation day dd mm for for End simulation day dd mm 31 gt fiz Number of years 8 Study begin dd mm yy for gt for for gt Pesticide input file Pesticide A Maizepm ss sS Scenario input file IH MAIZEsze Climate input file s HMBGNORM CLI Figure 24 InversePELMO Analyse the results of the optimisation for soil hydrology percolate After clicking at the respective button arrow on Figure 24 a form is loaded showing the experimental and optimised results graphically either soil Figure 25 or percolate concentrations Figure 26 If soil concentrations are presented different diagrams are available for each soil layer After a click with the left mouse button the next soil layer will be shown 36 Evaluation Concentration in soil mg kg Soil depth 10 cm to 15 cm Figure 25 InversePELMO View the results of the optimisation soil concentrations JF Evaluation Cumulative Flux m m Figure 26 InversePELMO View the results of the optimisation cumulative flux The circles always represent the experimental data the curve stands for the PELMO optimisation De
78. r values Previous parameter values koc 12 1100 koc 10 2400 kdeg 3 225000E 02 kdeg 3 122000E 02 Maximum relative change 0 1826 koc OPTIMISATION ITERATION NO 8 Model calls so far 45 Starting phi for this iteration 15984 Lambda 0 31250 gt Phi 11485 0 719 of starting phi Lambda 0 15625 gt Phi 12288 0 769 of starting phi Lambda 0 62500 gt Phi 9905 7 0 620 of starting phi Lambda 1 2500 gt Phi 6987 3 0 437 of starting phi Lambda 25000 stans gt Phi 2734 8 0 171 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 2734 8 Current parameter values Previous parameter values koc 14 4200 koc 12 1100 kdeg 3 334000E 02 kdeg 3 225000E 02 Maximum relative change 0 1908 koc 76 OPTIMISATION ITERATION NO 9 Model calls so far 54 Starting phi for this iteration 2734 8 Lambda 2 5000 gt Phi 478 67 0 175 of starting phi No more lambdas phi is less than 0 3000 of starting phi Lowest phi this iteration 478 67 Current parameter values Previous parameter values koc 14 5200 koc 14 4200 kdeg 3 300000E 02 kdeg 3 334000E 02 Maximum relative change 1 0198E 02 kdeg OPTIMISATION ITERATION NO 10 Model calls so far 59 Starting phi for this iteration 478 67 Lambda 125000 gt Phi 309 72 0 647 of starting phi Lambda 0 62500 gt Phi 300 94 0 629 of starting phi No more lam
79. res no minimum values higher than maximum values it can be closed using the Done button and the arrow on the optimisation form will jump to the next button which is the button for starting the optimisation see Figure 23 34 4 4 10 Step 10 Optimisation of the hydrology in soil pesticide fate In step 10 of the sequence the pesticide fate is optimised During the optimisation PEST will call PELMO several times After PEST terminated the user has to confirm that the optimisation didn t quit with an obvious error conditions Optimisation testtest Optimisation sequence 23 Y Enter experimental data Enter experimental data EN Create PELMO input files eo Import PELMO input files Start optimisation Start optimisation V Check initial simulation View optimisation View optimisation PEMO simulation control Start simulation day dd mm Start simulation day dd mm bi gt oi gt End simulation day dd mm 31 gt fiz gt Number of years 8 Study begin dd mm yy for gt for gt for gt Pesticide input file Pesticide A Maizepm 277 Scenario input file HMAZEse Climate input file s HMBGNORM CLI Figure 23 InversePELMO start optimisation of the pesticide fate Only after confirmation the arrow will move to the next button see Figure 24 4 4 11 Step 11 View the optimisation pesticide fate In step 11 the user can evaluate the results of the fate optimisation 35 O
80. rsion 1 Holdt G Gallien P Nehls A Bonath l Osterwald A K nig W Gottesb ren B Jene B Resseler H Sur R and Zillgens B 2011 Recommendations for Simulations to Predict Environmental Con centrations of Active Substances of Plant Protection Products and their Metabolites in Groundwater PECgw in the National Assess ment for Authorisation in Germany Part 1 Tier 1 and Tier 2 available at http www bvl bund de SharedDocs Downloads 04 Pflanzenschutzmittel zul umwelt pelmo pdf jsessionid E7F54E78DA37DBDDCDO6E4901C7EBB08 1 cid094 blob publicationFile amp v 3 Holdt G Gro mann D H llriegl Rosta A Christina Pickl 2011 EVA 2 1 Exposure via Air Assessment of the Short Range Transport and Deposition of Pesticides for Aquatic and Terrestrial Ecosystems spray drift and volatilisation considered Umweltbundesamt Dessau Ro lau Available at http www bvi bund de DE 04 Pflanzenschutzmittel 03 Antragsteller 04 Zulassungs verfahren 07 Naturhaushalt psm naturhaush_node html Jene B 1998 PELMO 3 0 User manual extension SLFA Neustadt 58 Klein 2011 Erarbeitung eines Tools zur routinem igen Durchf hrung von Simulationsrechnungen zur inversen Modellierung FKZ 360 03 050 Umweltbundesamt laufendes Vorhaben Klein M 1995 PELMO Pesticide Leaching Model version 2 01 Benutzerhandbuch Fraunhofer Institut Schmallenberg Klein M 2008 Calculation of PECsoil including Plateau
81. s Lowest phi this iteration Current parameter values koc 14 7900 kdeg 3 304000E 02 Maximum relative change OPTIMISATION ITERATION NO Model calls so far Starting phi for this iteration Lambda 0 15625 gt Phi 296 85 Lambda 7 81250E 02 gt Phi 296 85 No more lambdas 1 7906E 02 6 7659E 04 66 koc 9 54 2734 8 of starting phi 3000 of starting phi Previous parameter values koc 14 4200 kdeg 3 334000E 02 kdeg 10 99 478 67 0 629 of starting phi relative phi reduction between lambdas less than 0 0300 Previous parameter values koc 14 5200 kdeg 3 300000E 02 koc tL 65 300 94 0 986 of starting phi relative phi reduction between lambdas less than 0 0300 296 85 Previous parameter values koc 14 7800 kdeg 3 305000E 02 koc 12 71 296 85 1 000 times starting phi relative phi reduction between lambdas less than 0 0300 Lowest phi this iteration 296 85 Current parameter values Previous parameter values koc 14 7900 koc 14 7900 kdeg 3 304000E 02 kdeg 3 304000E 02 Maximum relative change 0 000 koc OPTIMISATION ITERATION NO 13 Model calls so far LD Starting phi for this iteration 296 85 Lambda 7 81250E 02 gt Phi 296 85 1 000 times starting phi Lambda Phi 3 90625E 02 296 85 No more lambdas Lowest phi this iteration Current parameter values 14 7900 koc kdeg 67 1 000 times starting ph
82. ssible range are shown in Figure 30 Substance considered FOCUS DUMMY B X Parameter Initial value Min value Max value M Freundlich 1 n vw DT50 d 50 I 1 000 Psa Figure 30 Parameters used in the optimisation of the substance flux test data set 1 After the optimisation the results summarised in Figure 31 were obtained 45 Cumulative Flux m m 500 400 300 200 100 days 0 100 200 300 400 500 600 700 Figure 31 Results of the optimisation substance flux hypothetical test data set 1 The agreement is excellent as also expressed by the small FOCUS chi error of 2 2 Table 7 Optimised parameter for the substance fluxes hypothetical test data set 1 Parameter Estimated 95 confidence limits Original parameters value lower limit value koc 13 570 13 29 13 570 17 DT50 12 82 12 77 12 82 20 Nevertheless Table 7 shows that PEST did not find back the original parameter setting but suggested different sorption and degradation data Obviously there are different combinations that lead to the same substance fluxes PEST suggested a slightly lower koc and compensated the higher mobility by a shorter half life But according to Figure 31 the alternative parameter combination leads to the same leaching behaviour in the study 46 5 2 Test data set 2 Leaching of Parent over a 17 months 5 2 1 Environmental data For the soil data the standard Borstel soil was used with exactly the same d
83. ssolution C 20 Reference soil moisture for degradation at 10 kPa field capacity Q10 factor increase of degradation rate with an increase of 2 58 temperature of 10 C Arrhenius activation energy kJ mol 1 65 4 B exponent of degradation moisture relationship according to 0 7 Walker Exponent of the FREUNDLICH Isotherm 0 9 Non equilibrium sorption not considered TSCF transpiration stream concentration factor 0 5 5 1 3 Lysimeter results hypothetical test data set 1 The main results of the hypothetical study 1 are summarised in Table 5 The maximum concentration in the leachate was detected at the end of the first winter February 3 37 ug L The total percolate collected was 649 1 L m 42 Table 5 Percolate and percolate concentrations in the lysimeter hypothetical test data set 1 Month Percolate L m Concentration ug L May 0 0 June 0 0 July 0 0 August 0 0 September 5 6 0 October 27 36 0 November 18 22 0 December 66 42 0 022 January 62 36 0 378 February 62 63 1 399 March 46 75 2 193 April 25 93 2 265 May 27 11 2 075 June 0 1 95 July 19 43 1 792 August 40 37 1 336 September 0 0 October 0 0 November 16 98 0 934 December 93 98 0 454 January 56 67 0 143 February 9 68 0 087 March 69 61 0 053 April 0 0 034 5 1 4 Optimisation hypothetical test data set 1 For the optimisation of the percolate all possible parameters were considered in the fitting
84. t minus the number of adjustable parameters If the degrees of freedom is negative the divisor becomes the number of observations with non zero weight plus the number of prior information items with non zero weight Covariance and other statistical matricies cannot be determined Jacobian and or Normal Matrix not yet calculated or normal matrix singular Minimum error for which the Chi Test passes according to FOCUS 0 01 View diagramm Copy Done Figure 17 InversePELMO Evaluate the results of the optimisation percolate The standard output file of PEST extension rec is used here but with additional information included about the error at which the chi test passes The methodology is according to FOCUS degradation kinetics FOCUS 2006 The copy button puts either graphic or the text into the clipboard whereas the Print button can be used to print out the optimisation results After the form was closed Done the arrow will jump to the next position the begin of the pesticide optimisation see Figure 18 29 Optimisation testtest m Optimisation sequence Enter experimental data Enter experimental data MEN Create PELMO input files ine ng parameters Start optimisation Import PELMO input files V Check initial simulation Der optimisatior PEMO simulation control Start simulation day dd mm Start simulation day dd mm 01 01 End simulation day dd mm
85. tailed output describing the optimisation procedure is available via the button View output file Figure 27 38 i Evaluation 1 1032E 05 1 3354E 08 ara t ati ici ma i Parameter correlation coefficient matrix koc kdeg 1 000 0 9571 0 9571 1 000 5 9488E 03 Parameter Estimeted 95 percent confidence limits value lower limit upper limit koc 14 7600 14 5531 14 9669 DT50 11 54 11 50 11 59 Minimum error for which the Chi Test passes according to FOCUS 1 44 View diagramm Figure 27 InversePELMO Evaluate the results of the optimisation pesticide fate The standard output file of PEST extension rec is used here but with additional information included about the error at which the chi test passes The methodology is according to FOCUS degradation kinetics FOCUS 2006 As the DT50 in soil is not an input parameter in PELMO it has to be converted into the respective rate constant Consequently PEST will not give information about the optimisation for DT50 values This information was therefore also added to the original PEST output file The copy button puts either graphic or the text into the clipboard whereas the Print button can be used to print out the optimisation results 39 5 Results of test simulations In this chapter results of two different studies are used to check the inverse modelling abilities of InversePELMO and PEST Whereas the first case is a hypothetic example the second
86. the optimisation of the substance fluxes the parameters DT50 and KOC were considered in the fitting Their initial values and their possible range are shown in Figure 36 51 Substance considered Parent Parameter Initial value Min value Max value M Freundlich 1 n vw DT50 d i 00 i fi 000 a Figure 36 Parameters used in the optimisation of the substance flux test data set 2 After the optimisation the results summarised in Figure 37 were obtained 52 Eyaluation Cumulative Flux m m Figure 37 Results of the optimisation substance flux test data set 2 The minimum error for which the Chi Test passes according to FOCUS was found to be 4 75 which supports the excellent agreement shown in the figure Table 13 Optimised parameter for the substance fluxes hypothetical test data set 2 Parameter Estimated 95 confidence limits value lower limit value koc 95 2 91 4 99 0 DT50 22 46 20 93 24 23 53 5 2 5 Hypothetical extension of the study test data set 2 An interesting problem in connection with lysimeter studies is the question what would have been if the study had been extended A prediction can be made based on the results of the inverse modelling studies as shown in the following example which uses the result of test data set 2 and assuming the same weather conditions as in the second year The results are presented in Table 14 Table 14 Extension of the lysimeter
87. tion Weight Group ol 0 00000 000 no_name o2 0 00000 000 no_name 03 0 00000 000 no_name 04 0 00000 000 no name 05 0 00000 000 no_name 06 0 00000 000 no_name o7 0 00000 000 no_name 08 1 46124 000 no_name 09 23 0333 000 no_name o10 112 653 000 no_name oll 215 175 000 no name 012 213 307 000 no_name 013 330 160 000 no name ol4 330 160 000 no name o15 364 979 000 no_name 016 418 913 000 no_name 017 418 913 000 no_name 018 418 913 000 no_name 019 434 772 000 no_name 020 477 439 000 no_name o21 485 543 000 no name 022 486 385 000 no name 023 490 075 000 no_name 024 490 075 000 no_name Control settings Initial lambda Lambda adjustment factor Sufficient new old phi ratio per optimisation iteration Limiting relative phi reduction between lambdas Maximum trial lambdas per iteration Maximum factor parameter change factor limited changes Maximum relative parameter change relative limited changes Fraction of initial parameter values used in computing change limit for near zero parameters Allow bending of parameter upgrade vector Allow parameters to stick to their bounds Relative phi reduction below which to begin use of central derivatives Relative phi reduction indicating convergence Number of phi values required within this range Maximum number of consecutive failures to lower phi Minimal relative parameter change indicating convergence Number of consecutive iterations with minimal param change Maxi
88. udies are performed in order to obtain key parameters for leaching models such as Kfoc Freundlich sorption constant related to organic carbon and DT50 degradation time to 50 from higher tier studies e g lysimeter studies instead directly from standard laboratory studies on sorption and degradation Aim of such a study is on one hand to get a deeper look into the processes that led to a certain lysimeter result On the other hand inverse modelling studies can be used to improve the standard modelling on tier 1 by considering additional information from higher tier studies Furher questions that can be answered based on inverse modelling studies are Prediction about the most likely behaviour if the lysimeter study had been conducted over a longer time period Translation of the lysimeter results to a different situation with respect to the environmental conditions e g different climate Translation of the lysimeter result to a different situation with respect to the application pattern of the substance e g change of the rate Use of the optimised parameter setting for a refined standard tier 1 simulation Generally two steps have to be conducted when performing inverse modelling studies First the hydrology in soil is optimised followed by the optimisation of pesticide fate as shown in Figure 1 Collection of available information from lysimeter studies cumulative fluxes water substance soil residues at study
89. um relative change Optimisation complete Total model calls 4779 4 1 00000 1 00000 1 00000 0 281000 3 5588E 05 the 3 lowest phi of eachother of 1 32 relative phi reduction between lambdas less than 0 0300 Previous parameter values kcO 1 00000 kel 1 00000 kc2 1 00000 moi0 0 280990 moiO s are within a relative distance 000E 02 89 The model has been run one final time using best parameters Thus all model input files contain best parameter values and model output files contain model results based on these parameters OPTIMISATION RESULTS Covariance matrix and parameter confidence intervals cannot be determined Normal matrix nearly singular cannot be inverted Parameters gt Parameter Estimated value kc0 1 00000 kel 1 00000 kc2 1 00000 moid 0 281000 See file PEST_WATER SEN for parameter sensitivities Observations gt Observation Measured Calculated Residual Weight Group value value ol 47 9000 69 1430 21 2430 000 no_name 02 124 600 120713 3 88670 000 no name 03 177 350 171 043 6 30660 000 no_name o4 222 400 221 628 0 771850 000 no_name 05 255 500 264 202 8 70235 000 no_name 06 346 900 331 338 15 5417 000 no_name 07 495 250 451 878 43 3723 000 no_name 08 582 900 564 815 18 0853 000 no_name 09 675 900 710 490 34 5895 000 no_name o10 767 150 790 618 23 4683 000 no_name See file PEST_WATER RES for more details of residuals in graph ready format
90. west phi s are within a relative distance of eachother of 1 000E 02 Total model calls 51 The model has been run one final time using best parameters Thus all model input files contain best parameter values and model output files contain model results based on these parameters OPTIMISATION RESULTS Parameters gt Parameter Estimated 95 percent confidence limits value lower limit upper limit koc 95 1700 91 3826 98 9574 kdeg 3 086000E 02 2 860899E 02 3 311101E 02 Note confidence limits provide only an indication of parameter uncertainty They rely on a linearity assumption which may not extend as far in parameter space as the confidence limits themselves see PEST manual See file PEST_PESTICIDE SEN for parameter sensitivities Observations gt Observation Measured Calculated Residual Weight Group value value ol 0 00000 0 00000 0 00000 000 no_name o2 0 00000 5 436032E 16 5 436032E 16 000 no_name 03 0 00000 7 517070E 11 7 517070E 11 000 no_name o4 0 00000 1 254043E 07 1 254043E 07 000 no_name 05 0 00000 1 034201E 05 1 034201E 05 000 no_name 06 0 00000 2 802955E 03 2 802955E 03 000 no_name 07 0 00000 0 742968 0 742968 000 no_name 08 12 270 9 71548 2455552 000 no_name 09 54 1210 55 7414 1 62043 000 no_name 010 104 309 103 615 0 693888 000 no_name See file PEST PESTICIDE RES for more details of residuals in graph ready format See file PEST_PESTICIDE SEO for composite observation sensitivities
91. y InversePELMO If appropriate the individual sampling can be characterised by weighting factors in the final column If the form has been filled correctly e g no negative figures it can be closed using the Done button and the arrow on the optimisation form will jump to the next button to enter all information about the input parameters used in the optimisation see Figure 12 It is not necessary to type all information the user can also paste them in as a table e g from MS Excel or MS Word 93 Optimisation testtest m Optimisation sequence a 3 Enter experimental data MEN Create PELMO input files x 5 Import PELMO input files Start optimisatior Start optimisation Check initial simulation view optimisation view optimisatior m PEMO simulation control Start simulation day dd mm Start simulation day dd mm 01 01 End simulation day dd mm Bi h2 x number EE p 3 Study begin dd mm yy for gt for gt or Pesticide input file Pesticide A Maizepsm ss Scenario input file HMAZEse 2 28 iti SCS S Climate input file s HMBGNORM CLI Figure 12 InversePELMO define fitting parameters for the hydrology in soil 4 4 5 Step 5 Enter fitting parameter percolate In step 5 the parameters used in the optimisation have to be characterised in a specific form see Figure 13 24 Initial value Min value Max value M Ke factor mid season l K

InversePELMO Manual (pdf, 1 MB, not barrier-free) - BVL

Contents

Download Pdf Manuals

Related Search

Related Contents