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: 4 Target node size: 5 Variable importance mode: impurity OOB prediction error: 146.7542 R squared: -0.1711627 OOB RMSE: 12.114 Variable importance: [,1] Zn_M_agg30cm_AF_1km 20.67339 NMOD3avg 15.48554 rElevIndex 15.15507 Ca_M_agg30cm_AF_1km 15.09574 PET 14.91094 NCluster_1_AF_1km 14.41977 VDPMRG5 13.40194 M13NDVIA01 12.74833 EVEENV3 12.66813 Wdvi 12.36175 ECN_M_agg30cm_AF_1km 12.26274 yGapMillet 11.90068 CRFVOL_M_agg30cm_AF_1km 11.61467 PCHE3avg 11.59631 MANMCF5 11.26723 eXtreme Gradient Boosting 31 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 22, 19, 21 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 11.59086 0.02642396 0.3 2 100 11.59086 0.02642383 0.3 2 150 11.59086 0.02642383 0.3 3 50 11.71496 0.02671854 0.3 3 100 11.71496 0.02671853 0.3 3 150 11.71496 0.02671853 0.3 4 50 11.63131 0.02295557 0.3 4 100 11.63131 0.02295555 0.3 4 150 11.63131 0.02295555 0.3 5 50 11.58701 0.02667415 0.3 5 100 11.58701 0.02667413 0.3 5 150 11.58701 0.02667413 0.3 6 50 11.75902 0.01378818 0.3 6 100 11.75902 0.01378818 0.3 6 150 11.75902 0.01378818 0.3 7 50 11.64782 0.02274007 0.3 7 100 11.64782 0.02274006 0.3 7 150 11.64782 0.02274006 0.3 8 50 11.64162 0.02281182 0.3 8 100 11.64162 0.02281180 0.3 8 150 11.64162 0.02281180 0.4 2 50 11.78176 0.02407054 0.4 2 100 11.78176 0.02407054 0.4 2 150 11.78176 0.02407054 0.4 3 50 11.71637 0.02360860 0.4 3 100 11.71637 0.02360860 0.4 3 150 11.71637 0.02360860 0.4 4 50 11.65100 0.02075189 0.4 4 100 11.65100 0.02075189 0.4 4 150 11.65100 0.02075189 0.4 5 50 11.89406 0.01070799 0.4 5 100 11.89406 0.01070799 0.4 5 150 11.89406 0.01070799 0.4 6 50 11.76051 0.01907552 0.4 6 100 11.76051 0.01907552 0.4 6 150 11.76051 0.01907552 0.4 7 50 11.92905 0.01016516 0.4 7 100 11.92905 0.01016516 0.4 7 150 11.92905 0.01016516 0.4 8 50 11.63546 0.02189656 0.4 8 100 11.63546 0.02189656 0.4 8 150 11.63546 0.02189656 0.5 2 50 11.51147 0.03364025 0.5 2 100 11.51147 0.03364025 0.5 2 150 11.51147 0.03364025 0.5 3 50 11.56847 0.02835985 0.5 3 100 11.56847 0.02835985 0.5 3 150 11.56847 0.02835985 0.5 4 50 11.60023 0.02557513 0.5 4 100 11.60023 0.02557513 0.5 4 150 11.60023 0.02557513 0.5 5 50 11.60260 0.02545400 0.5 5 100 11.60260 0.02545400 0.5 5 150 11.60260 0.02545400 0.5 6 50 11.67034 0.02202219 0.5 6 100 11.67034 0.02202219 0.5 6 150 11.67034 0.02202219 0.5 7 50 11.59620 0.02521407 0.5 7 100 11.59620 0.02521407 0.5 7 150 11.59620 0.02521407 0.5 8 50 11.55700 0.02956352 0.5 8 100 11.55700 0.02956352 0.5 8 150 11.55700 0.02956352 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: 11.511 R2: 0.034 XGBoost variable importance: Feature Gain Cover Frequency 1: Zn_M_agg30cm_AF_1km 5.900966e-01 0.032107716 0.02739726 2: rElevIndex 1.656675e-01 0.080269291 0.06849315 3: af_agg_30cm_TETAs__M_1km 1.200525e-01 0.169860176 0.17808219 4: Lai_avg 3.479649e-02 0.047125842 0.04109589 5: ENTENV3 3.051312e-02 0.015535992 0.01369863 6: K_M_agg30cm_AF_1km 2.921443e-02 0.011910927 0.01369863 7: NCluster_16_AF_1km 1.879185e-02 0.031071983 0.02739726 8: af_agg_30cm_PWP__M_1km 8.193496e-03 0.043500777 0.04109589 9: C03GLC5 1.376417e-03 0.159502848 0.13698630 10: CHIRPSA 4.246854e-04 0.011910927 0.01369863 11: af_agg_30cm_AWCpF23__M_1km 3.627578e-04 0.175556706 0.15068493 12: Al_M_agg30cm_AF_1km 3.097832e-04 0.019161056 0.02739726 13: af_agg_ERZD_TAWCpF23mm__M_1km 1.412487e-04 0.026929052 0.02739726 14: CRFVOL_M_agg30cm_AF_1km 5.582707e-05 0.015535992 0.01369863 15: AAIavg_GYGA 2.097155e-06 0.009321595 0.08219178 Ensemble validation RMSE: 11.749 R2: 0 -------------------------------------- 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: 4 Target node size: 5 Variable importance mode: impurity OOB prediction error: 76.48312 R squared: 0.07346538 OOB RMSE: 8.745 Variable importance: [,1] ENTENV3 33.66248 af_agg_ERZD_TAWCpF23mm__M_1km 33.20837 VW1MOD1avg 28.40401 Water_balance 28.06704 CEC_M_agg30cm_AF_1km 26.23523 NCluster_14_AF_1km 25.61363 Lai_avg 23.67068 TMDMOD3 23.37652 M13RB1A08 20.68596 B13CHE3 20.45764 MAXENV3 19.50421 REDL00 19.07822 NIRL00 18.59236 M17NPPALTfill 18.57956 GAEZ_NPP 18.26289 eXtreme Gradient Boosting 34 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 23, 21, 24 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 8.677011 0.2865203 0.3 2 100 8.676999 0.2865226 0.3 2 150 8.676999 0.2865226 0.3 3 50 9.189562 0.2912066 0.3 3 100 9.189562 0.2912067 0.3 3 150 9.189562 0.2912067 0.3 4 50 9.219076 0.2891674 0.3 4 100 9.219077 0.2891675 0.3 4 150 9.219077 0.2891675 0.3 5 50 9.110833 0.2978842 0.3 5 100 9.110833 0.2978845 0.3 5 150 9.110833 0.2978845 0.3 6 50 8.660225 0.2794498 0.3 6 100 8.660229 0.2794498 0.3 6 150 8.660229 0.2794498 0.3 7 50 9.053801 0.2884658 0.3 7 100 9.053803 0.2884658 0.3 7 150 9.053803 0.2884658 0.3 8 50 8.908411 0.2884893 0.3 8 100 8.908413 0.2884893 0.3 8 150 8.908413 0.2884893 0.4 2 50 8.817608 0.2904161 0.4 2 100 8.817608 0.2904161 0.4 2 150 8.817608 0.2904161 0.4 3 50 9.137071 0.2874805 0.4 3 100 9.137071 0.2874805 0.4 3 150 9.137071 0.2874805 0.4 4 50 8.851937 0.2874728 0.4 4 100 8.851937 0.2874728 0.4 4 150 8.851937 0.2874728 0.4 5 50 8.812576 0.2867909 0.4 5 100 8.812576 0.2867909 0.4 5 150 8.812576 0.2867909 0.4 6 50 9.244951 0.2902271 0.4 6 100 9.244951 0.2902271 0.4 6 150 9.244951 0.2902271 0.4 7 50 9.024067 0.3194840 0.4 7 100 9.024067 0.3194840 0.4 7 150 9.024067 0.3194840 0.4 8 50 8.808467 0.2880700 0.4 8 100 8.808467 0.2880700 0.4 8 150 8.808467 0.2880700 0.5 2 50 9.015189 0.3340845 0.5 2 100 9.015189 0.3340845 0.5 2 150 9.015189 0.3340845 0.5 3 50 8.830563 0.2938456 0.5 3 100 8.830563 0.2938456 0.5 3 150 8.830563 0.2938456 0.5 4 50 8.360735 0.3274211 0.5 4 100 8.360735 0.3274211 0.5 4 150 8.360735 0.3274211 0.5 5 50 9.250129 0.2973379 0.5 5 100 9.250129 0.2973379 0.5 5 150 9.250129 0.2973379 0.5 6 50 8.936579 0.2883123 0.5 6 100 8.936579 0.2883123 0.5 6 150 8.936579 0.2883123 0.5 7 50 8.765761 0.2871082 0.5 7 100 8.765761 0.2871082 0.5 7 150 8.765761 0.2871082 0.5 8 50 9.243780 0.3148786 0.5 8 100 9.243780 0.3148786 0.5 8 150 9.243780 0.3148786 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 = 4, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 8.361 R2: 0.327 XGBoost variable importance: Feature Gain Cover Frequency 1: af_agg_ERZD_TAWCpF23mm__M_1km 4.561527e-01 0.19544697 0.15068493 2: Fe_M_agg30cm_AF_1km 2.966617e-01 0.18656302 0.13698630 3: af_ERZD__M_1km 1.050603e-01 0.03775680 0.02739726 4: af_agg_30cm_AWCpF23__M_1km 9.934138e-02 0.03164908 0.08219178 5: B13CHE3 2.421776e-02 0.06718490 0.05479452 6: C03GLC5 1.682579e-02 0.03220433 0.04109589 7: AAIavg_GYGA 1.292696e-03 0.02054414 0.02739726 8: rElev 2.522223e-04 0.03164908 0.02739726 9: af_agg_30cm_TAWCpF23mm__M_1km 1.037040e-04 0.03664631 0.05479452 10: ENAX_M_agg30cm_AF_1km 3.186185e-05 0.01665741 0.01369863 11: B14CHE3 2.709462e-05 0.01887840 0.02739726 12: B02CHE3 1.157327e-05 0.00943920 0.01369863 13: Ca_M_agg30cm_AF_1km 7.909672e-06 0.02942810 0.02739726 14: NEGMRG5 6.225754e-06 0.01610217 0.01369863 15: ASSDAC3 3.160368e-06 0.01665741 0.01369863 Ensemble validation RMSE: 8.427 R2: 0.162 -------------------------------------- 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: 20 Target node size: 5 Variable importance mode: impurity OOB prediction error: 0.0002813711 R squared: 0.8766441 OOB RMSE: 0.017 Variable importance: [,1] NMSD3avg 0.0008931829 Fe_M_agg30cm_AF_1km 0.0007574245 MAXENV3 0.0007459246 K_M_agg30cm_AF_1km 0.0006604222 fPR_MilletT2 0.0006570618 NCluster_9_AF_1km 0.0006523573 SNDPPT_M_agg30cm_AF_1km 0.0006382438 Na_M_agg30cm_AF_1km 0.0005802217 yFertilised_MilletT2 0.0005661082 Wdvi 0.0005519947 NCluster_16_AF_1km 0.0005519947 CLYPPT_M_agg30cm_AF_1km 0.0005394493 GTDHYS3 0.0005394493 POSMRG5 0.0005206313 PHIHOXagg0_30 0.0005143587 eXtreme Gradient Boosting 12 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 8, 8, 8 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 0.0003667106 1.0000000 0.3 2 100 0.0003661513 1.0000000 0.3 2 150 0.0003661513 1.0000000 0.3 3 50 0.0003667106 1.0000000 0.3 3 100 0.0003661513 1.0000000 0.3 3 150 0.0003661513 1.0000000 0.3 4 50 0.0003667106 1.0000000 0.3 4 100 0.0003661513 1.0000000 0.3 4 150 0.0003661513 1.0000000 0.3 5 50 0.0008090749 0.9997162 0.3 5 100 0.0008081065 0.9997162 0.3 5 150 0.0008081065 0.9997162 0.3 6 50 0.0004143747 0.9999937 0.3 6 100 0.0004135163 0.9999937 0.3 6 150 0.0004135163 0.9999937 0.3 7 50 0.0008822315 0.9996232 0.3 7 100 0.0008812635 0.9996232 0.3 7 150 0.0008812635 0.9996232 0.3 8 50 0.0003667106 1.0000000 0.3 8 100 0.0003661513 1.0000000 0.3 8 150 0.0003661513 1.0000000 0.4 2 50 0.0012925893 0.9988093 0.4 2 100 0.0012925893 0.9988093 0.4 2 150 0.0012925893 0.9988093 0.4 3 50 0.0003563913 1.0000000 0.4 3 100 0.0003563913 1.0000000 0.4 3 150 0.0003563913 1.0000000 0.4 4 50 0.0026998114 0.9927628 0.4 4 100 0.0026998114 0.9927628 0.4 4 150 0.0026998114 0.9927628 0.4 5 50 0.0003563913 1.0000000 0.4 5 100 0.0003563913 1.0000000 0.4 5 150 0.0003563913 1.0000000 0.4 6 50 0.0003563913 1.0000000 0.4 6 100 0.0003563913 1.0000000 0.4 6 150 0.0003563913 1.0000000 0.4 7 50 0.0018452172 0.9970644 0.4 7 100 0.0018452172 0.9970644 0.4 7 150 0.0018452172 0.9970644 0.4 8 50 0.0012925893 0.9988093 0.4 8 100 0.0012925893 0.9988093 0.4 8 150 0.0012925893 0.9988093 0.5 2 50 0.0008652890 0.9995190 0.5 2 100 0.0008652890 0.9995190 0.5 2 150 0.0008652890 0.9995190 0.5 3 50 0.0002690764 1.0000000 0.5 3 100 0.0002690764 1.0000000 0.5 3 150 0.0002690764 1.0000000 0.5 4 50 0.0002690764 1.0000000 0.5 4 100 0.0002690764 1.0000000 0.5 4 150 0.0002690764 1.0000000 0.5 5 50 0.0002690764 1.0000000 0.5 5 100 0.0002690764 1.0000000 0.5 5 150 0.0002690764 1.0000000 0.5 6 50 0.0002690764 1.0000000 0.5 6 100 0.0002690764 1.0000000 0.5 6 150 0.0002690764 1.0000000 0.5 7 50 0.0003574992 0.9999837 0.5 7 100 0.0003574992 0.9999837 0.5 7 150 0.0003574992 0.9999837 0.5 8 50 0.0008652890 0.9995190 0.5 8 100 0.0008652890 0.9995190 0.5 8 150 0.0008652890 0.9995190 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 = 0.5, 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 --------------------------------------