Variable: fNRec Ranger result Call: ranger(formulaString.lst[[j]], data = dfs, importance = "impurity", write.forest = TRUE, mtry = t.mrfX$bestTune$mtry, num.trees = 500) Type: Regression Number of trees: 500 Sample size: 31 Number of independent variables: 377 Mtry: 6 Target node size: 5 Variable importance mode: impurity OOB prediction error: 154.8799 R squared: -0.2360089 OOB RMSE: 12.445 Variable importance: [,1] VDPMRG5 20.46440 BLDFIE_M_agg30cm_AF_1km 18.51005 PET 18.02804 yGapMillet 16.61745 Zn_M_agg30cm_AF_1km 16.61639 ESMOD5avg 16.45915 Millet_intermed 16.43784 M13RB1A04 16.26094 rElev 15.48613 Fe_M_agg30cm_AF_1km 15.44207 M13NDVIA04 15.14008 N_M_agg30cm_AF_1km 14.94679 PRSCHE3 14.12243 CMCF5avg 13.99205 VBFMRG5 13.96666 eXtreme Gradient Boosting 31 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 20, 21, 21 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 10.68406 0.08136991 0.3 2 100 10.68402 0.08137418 0.3 2 150 10.68402 0.08137418 0.3 3 50 10.84064 0.07814054 0.3 3 100 10.84067 0.07813795 0.3 3 150 10.84067 0.07813795 0.3 4 50 10.68336 0.08100290 0.3 4 100 10.68339 0.08100072 0.3 4 150 10.68339 0.08100072 0.3 5 50 10.71102 0.07927737 0.3 5 100 10.71115 0.07926709 0.3 5 150 10.71115 0.07926709 0.3 6 50 10.68406 0.08143817 0.3 6 100 10.68410 0.08143515 0.3 6 150 10.68410 0.08143515 0.3 7 50 10.93803 0.06768497 0.3 7 100 10.93810 0.06767946 0.3 7 150 10.93810 0.06767946 0.3 8 50 10.85969 0.07526386 0.3 8 100 10.85974 0.07525980 0.3 8 150 10.85974 0.07525980 0.4 2 50 10.79902 0.07813664 0.4 2 100 10.79902 0.07813664 0.4 2 150 10.79902 0.07813664 0.4 3 50 10.99422 0.06162883 0.4 3 100 10.99422 0.06162881 0.4 3 150 10.99422 0.06162881 0.4 4 50 10.79538 0.08451348 0.4 4 100 10.79538 0.08451348 0.4 4 150 10.79538 0.08451348 0.4 5 50 10.87407 0.07899326 0.4 5 100 10.87407 0.07899326 0.4 5 150 10.87407 0.07899326 0.4 6 50 10.76238 0.08211735 0.4 6 100 10.76238 0.08211735 0.4 6 150 10.76238 0.08211735 0.4 7 50 10.82324 0.07908284 0.4 7 100 10.82324 0.07908284 0.4 7 150 10.82324 0.07908284 0.4 8 50 10.80150 0.07752249 0.4 8 100 10.80150 0.07752249 0.4 8 150 10.80150 0.07752249 0.5 2 50 10.59342 0.08925677 0.5 2 100 10.59342 0.08925677 0.5 2 150 10.59342 0.08925677 0.5 3 50 11.05398 0.06396339 0.5 3 100 11.05398 0.06396339 0.5 3 150 11.05398 0.06396339 0.5 4 50 10.94308 0.06660201 0.5 4 100 10.94308 0.06660201 0.5 4 150 10.94308 0.06660201 0.5 5 50 10.81045 0.08308305 0.5 5 100 10.81045 0.08308305 0.5 5 150 10.81045 0.08308305 0.5 6 50 10.96142 0.06622684 0.5 6 100 10.96142 0.06622684 0.5 6 150 10.96142 0.06622684 0.5 7 50 11.01214 0.06868563 0.5 7 100 11.01214 0.06868563 0.5 7 150 11.01214 0.06868563 0.5 8 50 10.91561 0.07507894 0.5 8 100 10.91561 0.07507894 0.5 8 150 10.91561 0.07507894 Tuning parameter 'gamma' was held constant at a value of 0 Tuning parameter 'colsample_bytree' was held constant at a value of 0.8 Tuning parameter 'min_child_weight' was held constant at a value of 1 RMSE was used to select the optimal model using the smallest value. The final values used for the model were nrounds = 50, max_depth = 2, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 10.593 R2: 0.089 XGBoost variable importance: Feature Gain Cover Frequency 1: Zn_M_agg30cm_AF_1km 5.912730e-01 0.016137428 0.01351351 2: rElevIndex 1.704223e-01 0.177511713 0.14864865 3: af_agg_30cm_TETAs__M_1km 1.501289e-01 0.223321187 0.21621622 4: Lai_avg 3.482116e-02 0.031754295 0.02702703 5: ENTENV3 3.057395e-02 0.015616866 0.01351351 6: NCluster_16_AF_1km 1.146457e-02 0.031233732 0.02702703 7: af_agg_30cm_PWP__M_1km 8.209203e-03 0.021863613 0.02702703 8: C03GLC5 1.313127e-03 0.047891723 0.04054054 9: ECN_M_agg30cm_AF_1km 5.032564e-04 0.008849558 0.01351351 10: af_agg_30cm_TAWCpF23mm__M_1km 4.991728e-04 0.011972931 0.01351351 11: af_agg_30cm_AWCpF23__M_1km 3.118711e-04 0.154086413 0.16216216 12: AAIavg_GYGA 1.810366e-04 0.034357106 0.09459459 13: B_M_agg30cm_AF_1km 1.360261e-04 0.016137428 0.01351351 14: Al_M_agg30cm_AF_1km 6.658510e-05 0.004164498 0.01351351 15: yGapMillet 5.753129e-05 0.047371161 0.04054054 Ensemble validation RMSE: 11.333 R2: 0.015 -------------------------------------- Variable: fPRec Ranger result Call: ranger(formulaString.lst[[j]], data = dfs, importance = "impurity", write.forest = TRUE, mtry = t.mrfX$bestTune$mtry, num.trees = 500) Type: Regression Number of trees: 500 Sample size: 34 Number of independent variables: 377 Mtry: 16 Target node size: 5 Variable importance mode: impurity OOB prediction error: 82.44199 R squared: 0.001278213 OOB RMSE: 9.08 Variable importance: [,1] EXMOD5avg 54.58235 Lai_avg 44.35250 Fe_M_agg30cm_AF_1km 42.68693 M17NPPALTfill 37.35817 GAEZ_NPP 36.43095 VW1MOD1avg 36.00172 M13RB3A08 35.44487 BIO12ALT 33.84246 GAEZ_ratioP_PETsea 33.36261 M17GPPALTfill 31.80711 LSTD_avgIRI_Jul2002_Sep2016_mosaicLAEA_celsius 31.57516 CEC_M_agg30cm_AF_1km 30.24007 TMOD3avg 30.15485 af_agg_ERZD_TAWCpF23mm__M_1km 29.62448 af_DRAINFAO_T__M_1kmc2 29.53430 eXtreme Gradient Boosting 34 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 22, 24, 22 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 9.220657 0.1103764 0.3 2 100 9.221077 0.1103389 0.3 2 150 9.221077 0.1103389 0.3 3 50 9.059105 0.1243584 0.3 3 100 9.059512 0.1243155 0.3 3 150 9.059512 0.1243155 0.3 4 50 9.098864 0.1230877 0.3 4 100 9.099274 0.1230450 0.3 4 150 9.099274 0.1230450 0.3 5 50 9.034869 0.1293383 0.3 5 100 9.035278 0.1292953 0.3 5 150 9.035278 0.1292953 0.3 6 50 9.017573 0.1285502 0.3 6 100 9.017992 0.1285062 0.3 6 150 9.017992 0.1285062 0.3 7 50 9.035761 0.1281337 0.3 7 100 9.036167 0.1280913 0.3 7 150 9.036167 0.1280913 0.3 8 50 9.081769 0.1252272 0.3 8 100 9.082176 0.1251844 0.3 8 150 9.082176 0.1251844 0.4 2 50 9.098057 0.1223796 0.4 2 100 9.098057 0.1223796 0.4 2 150 9.098057 0.1223796 0.4 3 50 8.982539 0.1332902 0.4 3 100 8.982539 0.1332902 0.4 3 150 8.982539 0.1332902 0.4 4 50 9.209553 0.1119668 0.4 4 100 9.209554 0.1119668 0.4 4 150 9.209554 0.1119668 0.4 5 50 9.074604 0.1227009 0.4 5 100 9.074604 0.1227009 0.4 5 150 9.074604 0.1227009 0.4 6 50 9.072445 0.1252200 0.4 6 100 9.072446 0.1252200 0.4 6 150 9.072446 0.1252200 0.4 7 50 9.182491 0.1138013 0.4 7 100 9.182491 0.1138013 0.4 7 150 9.182491 0.1138013 0.4 8 50 9.173475 0.1174265 0.4 8 100 9.173475 0.1174265 0.4 8 150 9.173475 0.1174265 0.5 2 50 9.133570 0.1213307 0.5 2 100 9.133570 0.1213307 0.5 2 150 9.133570 0.1213307 0.5 3 50 8.986683 0.1336818 0.5 3 100 8.986683 0.1336818 0.5 3 150 8.986683 0.1336818 0.5 4 50 9.008793 0.1306976 0.5 4 100 9.008793 0.1306976 0.5 4 150 9.008793 0.1306976 0.5 5 50 9.039051 0.1283927 0.5 5 100 9.039051 0.1283927 0.5 5 150 9.039051 0.1283927 0.5 6 50 9.100297 0.1232975 0.5 6 100 9.100297 0.1232975 0.5 6 150 9.100297 0.1232975 0.5 7 50 9.034269 0.1293106 0.5 7 100 9.034269 0.1293106 0.5 7 150 9.034269 0.1293106 0.5 8 50 9.205403 0.1153218 0.5 8 100 9.205403 0.1153218 0.5 8 150 9.205403 0.1153218 Tuning parameter 'gamma' was held constant at a value of 0 Tuning parameter 'colsample_bytree' was held constant at a value of 0.8 Tuning parameter 'min_child_weight' was held constant at a value of 1 RMSE was used to select the optimal model using the smallest value. The final values used for the model were nrounds = 50, max_depth = 3, eta = 0.4, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 8.983 R2: 0.133 XGBoost variable importance: Feature Gain Cover Frequency 1: af_ERZD__M_1km 6.055808e-01 0.08521303 0.06451613 2: af_agg_ERZD_TAWCpF23mm__M_1km 3.097425e-01 0.40726817 0.32258065 3: af_agg_30cm_TAWCpF23__M_1km 5.883722e-02 0.01754386 0.06451613 4: C03GLC5 9.056289e-03 0.01253133 0.01075269 5: EVEENV3 6.433834e-03 0.03258145 0.03225806 6: af_BDRICM_T__M_1km 3.575773e-03 0.03675856 0.04301075 7: yGapMillet 3.119058e-03 0.04302423 0.04301075 8: EXMOD5avg 1.400957e-03 0.02506266 0.02150538 9: fNR_MilletTrials 1.343912e-03 0.01127820 0.01075269 10: Lai_avg 6.697851e-04 0.02172097 0.02150538 11: AAIavg_GYGA 6.330687e-05 0.02464495 0.05376344 12: af_agg_30cm_TETAs__M_1km 5.659262e-05 0.04427736 0.05376344 13: K_M_agg30cm_AF_1km 5.520001e-05 0.02506266 0.02150538 14: ESMOD5avg 3.563182e-05 0.01127820 0.01075269 15: rElevIndex 9.805474e-06 0.01127820 0.01075269 Ensemble validation RMSE: 8.956 R2: 0.116 -------------------------------------- Variable: fKRec Ranger result Call: ranger(formulaString.lst[[j]], data = dfs, importance = "impurity", write.forest = TRUE, mtry = t.mrfX$bestTune$mtry, num.trees = 500) Type: Regression Number of trees: 500 Sample size: 12 Number of independent variables: 377 Mtry: 18 Target node size: 5 Variable importance mode: impurity OOB prediction error: 0.0002839891 R squared: 0.8754963 OOB RMSE: 0.017 Variable importance: [,1] Temperature 0.0009064003 SNDPPT_M_agg30cm_AF_1km 0.0008483782 NMSD3avg 0.0007919242 Na_M_agg30cm_AF_1km 0.0006915615 PHIHOXagg0_30 0.0006758798 af_agg_30cm_AWCpF23__M_1km 0.0006402600 EVEENV3 0.0006272667 CLYPPT_M_agg30cm_AF_1km 0.0006178577 P.T_M_agg30cm_AF_1km 0.0006068805 NCluster_9_AF_1km 0.0005943352 NCluster_15_AF_1km 0.0005676763 CEC_M_agg30cm_AF_1km 0.0005504265 af_agg_30cm_PWP__M_1km 0.0005300403 af_BDRICM_T__M_1km 0.0005112223 NCluster_16_AF_1km 0.0004845635 eXtreme Gradient Boosting 12 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 8, 9, 7 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 0.0003572904 1.0000000 0.3 2 100 0.0003571307 1.0000000 0.3 2 150 0.0003571307 1.0000000 0.3 3 50 0.0016920151 0.9980105 0.3 3 100 0.0016906625 0.9980105 0.3 3 150 0.0016906625 0.9980105 0.3 4 50 0.0013056118 0.9989571 0.3 4 100 0.0013042853 0.9989571 0.3 4 150 0.0013042853 0.9989571 0.3 5 50 0.0003572904 1.0000000 0.3 5 100 0.0003571307 1.0000000 0.3 5 150 0.0003571307 1.0000000 0.3 6 50 0.0003572904 1.0000000 0.3 6 100 0.0003571307 1.0000000 0.3 6 150 0.0003571307 1.0000000 0.3 7 50 0.0003572904 1.0000000 0.3 7 100 0.0003571307 1.0000000 0.3 7 150 0.0003571307 1.0000000 0.3 8 50 0.0013056118 0.9989571 0.3 8 100 0.0013042853 0.9989571 0.3 8 150 0.0013042853 0.9989571 0.4 2 50 0.0003555728 1.0000000 0.4 2 100 0.0003555728 1.0000000 0.4 2 150 0.0003555728 1.0000000 0.4 3 50 0.0003555728 1.0000000 0.4 3 100 0.0003555728 1.0000000 0.4 3 150 0.0003555728 1.0000000 0.4 4 50 0.0003555728 1.0000000 0.4 4 100 0.0003555728 1.0000000 0.4 4 150 0.0003555728 1.0000000 0.4 5 50 0.0004190462 0.9999874 0.4 5 100 0.0004190462 0.9999874 0.4 5 150 0.0004190462 0.9999874 0.4 6 50 0.0015264398 0.9984559 0.4 6 100 0.0015264398 0.9984559 0.4 6 150 0.0015264398 0.9984559 0.4 7 50 0.0003555728 1.0000000 0.4 7 100 0.0003555728 1.0000000 0.4 7 150 0.0003555728 1.0000000 0.4 8 50 0.0003555728 1.0000000 0.4 8 100 0.0003555728 1.0000000 0.4 8 150 0.0003555728 1.0000000 0.5 2 50 0.0002934923 1.0000000 0.5 2 100 0.0002934923 1.0000000 0.5 2 150 0.0002934923 1.0000000 0.5 3 50 0.0002934923 1.0000000 0.5 3 100 0.0002934923 1.0000000 0.5 3 150 0.0002934923 1.0000000 0.5 4 50 0.0004017209 0.9999762 0.5 4 100 0.0004017209 0.9999762 0.5 4 150 0.0004017209 0.9999762 0.5 5 50 0.0002934923 1.0000000 0.5 5 100 0.0002934923 1.0000000 0.5 5 150 0.0002934923 1.0000000 0.5 6 50 0.0002934923 1.0000000 0.5 6 100 0.0002934923 1.0000000 0.5 6 150 0.0002934923 1.0000000 0.5 7 50 0.0002934923 1.0000000 0.5 7 100 0.0002934923 1.0000000 0.5 7 150 0.0002934923 1.0000000 0.5 8 50 0.0002934923 1.0000000 0.5 8 100 0.0002934923 1.0000000 0.5 8 150 0.0002934923 1.0000000 Tuning parameter 'gamma' was held constant at a value of 0 Tuning parameter 'colsample_bytree' was held constant at a value of 0.8 Tuning parameter 'min_child_weight' was held constant at a value of 1 RMSE was used to select the optimal model using the smallest value. The final values used for the model were nrounds = 50, max_depth = 2, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 0 R2: 1 XGBoost variable importance: Feature Gain Cover Frequency 1: af_agg_30cm_AWCpF23__M_1km 0.96045986 0.875 0.875 2: af_agg_30cm_PWP__M_1km 0.03954014 0.125 0.125 3: NA NA NA NA 4: NA NA NA NA 5: NA NA NA NA 6: NA NA NA NA 7: NA NA NA NA 8: NA NA NA NA 9: NA NA NA NA 10: NA NA NA NA 11: NA NA NA NA 12: NA NA NA NA 13: NA NA NA NA 14: NA NA NA NA 15: NA NA NA NA Ensemble validation RMSE: 0 R2: 1 --------------------------------------