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: 34 Number of independent variables: 377 Mtry: 4 Target node size: 5 Variable importance mode: impurity OOB prediction error: 188.1556 R squared: -0.3164793 OOB RMSE: 13.717 Variable importance: [,1] IMOD4avg 40.28973 DEMENV5 27.05193 ECN_M_agg30cm_AF_1km 25.10842 M43WVALT 24.41411 Na_M_agg30cm_AF_1km 22.59231 yGapCassava 22.57344 M43WSALT 22.51400 M13NDVIALT 21.98795 M13RB3ALT 21.63487 GAEZ_NPP 21.60191 B13CHE3 21.37508 M13RB1A01 20.84199 GAEZ_ratioP_PETsea 20.37501 PRSCHE3 20.34746 Mn_M_agg30cm_AF_1km 20.12496 eXtreme Gradient Boosting 34 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 23, 23, 22 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 14.68765 0.03506459 0.3 2 100 14.69001 0.03509001 0.3 2 150 14.69001 0.03508998 0.3 3 50 14.87305 0.03890171 0.3 3 100 14.87276 0.03887645 0.3 3 150 14.87276 0.03887645 0.3 4 50 14.84938 0.04263026 0.3 4 100 14.84936 0.04261254 0.3 4 150 14.84936 0.04261254 0.3 5 50 15.21250 0.06168841 0.3 5 100 15.21244 0.06168669 0.3 5 150 15.21244 0.06168669 0.3 6 50 14.86453 0.04545711 0.3 6 100 14.86441 0.04542761 0.3 6 150 14.86441 0.04542761 0.3 7 50 14.98456 0.03352622 0.3 7 100 14.98465 0.03351751 0.3 7 150 14.98465 0.03351749 0.3 8 50 15.04597 0.05778033 0.3 8 100 15.04602 0.05777526 0.3 8 150 15.04602 0.05777526 0.4 2 50 15.17370 0.03481639 0.4 2 100 15.17367 0.03481473 0.4 2 150 15.17367 0.03481473 0.4 3 50 15.24669 0.08317699 0.4 3 100 15.24669 0.08317699 0.4 3 150 15.24669 0.08317699 0.4 4 50 15.16607 0.05853586 0.4 4 100 15.16606 0.05853579 0.4 4 150 15.16606 0.05853579 0.4 5 50 15.17176 0.07834798 0.4 5 100 15.17175 0.07834796 0.4 5 150 15.17175 0.07834796 0.4 6 50 15.22801 0.07153820 0.4 6 100 15.22800 0.07153817 0.4 6 150 15.22800 0.07153817 0.4 7 50 15.32812 0.07814959 0.4 7 100 15.32812 0.07814959 0.4 7 150 15.32812 0.07814959 0.4 8 50 15.25138 0.04627795 0.4 8 100 15.25137 0.04627787 0.4 8 150 15.25137 0.04627787 0.5 2 50 15.34207 0.06684867 0.5 2 100 15.34207 0.06684867 0.5 2 150 15.34207 0.06684867 0.5 3 50 15.34052 0.06980297 0.5 3 100 15.34052 0.06980297 0.5 3 150 15.34052 0.06980297 0.5 4 50 15.71428 0.12316361 0.5 4 100 15.71428 0.12316361 0.5 4 150 15.71428 0.12316361 0.5 5 50 15.63974 0.07967467 0.5 5 100 15.63974 0.07967464 0.5 5 150 15.63974 0.07967464 0.5 6 50 15.27727 0.08969424 0.5 6 100 15.27727 0.08969424 0.5 6 150 15.27727 0.08969424 0.5 7 50 15.42598 0.07114112 0.5 7 100 15.42598 0.07114112 0.5 7 150 15.42598 0.07114112 0.5 8 50 15.48493 0.04971537 0.5 8 100 15.48493 0.04971537 0.5 8 150 15.48493 0.04971537 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 = 0.3, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 14.688 R2: 0.035 XGBoost variable importance: Feature Gain Cover Frequency 1: Mg_M_agg30cm_AF_1km 0.238839113 0.064147348 0.053097345 2: Cassave_actual_baseline 0.161084795 0.021594157 0.017699115 3: B13CHE3 0.145232981 0.010797078 0.008849558 4: Ca_M_agg30cm_AF_1km 0.126420079 0.053985392 0.044247788 5: BLDFIE_M_agg30cm_AF_1km 0.093822702 0.010797078 0.008849558 6: IMOD4avg 0.049766586 0.032073674 0.026548673 7: fPR_CassavaTrials 0.044353939 0.005080978 0.008849558 8: BIO12ALT 0.022942558 0.161956177 0.132743363 9: af_agg_ERZD_TAWCpF23mm__M_1km 0.017022508 0.003175611 0.008849558 10: ECN_M_agg30cm_AF_1km 0.016176389 0.020006351 0.017699115 11: AAIavg_GYGA 0.014918146 0.016830740 0.061946903 12: SSI_NCluster_AF_1km 0.011351679 0.010797078 0.008849558 13: Zn_M_agg30cm_AF_1km 0.011300836 0.013972690 0.017699115 14: CRFVOL_M_agg30cm_AF_1km 0.010135376 0.029533185 0.026548673 15: OC_M_agg30cm_AF_1km 0.008802921 0.079390283 0.070796460 Ensemble validation RMSE: 13.966 R2: 0.028 -------------------------------------- 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: 18 Target node size: 5 Variable importance mode: impurity OOB prediction error: 16.60161 R squared: 0.2216373 OOB RMSE: 4.075 Variable importance: [,1] Fcover 20.164381 PHIHOXagg0_30 16.086210 CMCF5avg 15.458823 MANMCF5 15.317406 MAXENV3 15.075972 CRFVOL_M_agg30cm_AF_1km 14.979492 Fapar 13.445480 M13RB3A08 12.732088 M13RB3ALT 10.446278 NCluster_14_AF_1km 9.942187 Riclassc2 9.740564 VW1MOD1avg 9.013491 Water_balance 8.982410 Wdvi 8.843364 NCluster_11_AF_1km 8.793340 eXtreme Gradient Boosting 34 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 23, 22, 23 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 3.593624 0.4389741 0.3 2 100 3.593989 0.4387453 0.3 2 150 3.593989 0.4387453 0.3 3 50 3.647310 0.4338695 0.3 3 100 3.646827 0.4339862 0.3 3 150 3.646827 0.4339862 0.3 4 50 3.596527 0.4362271 0.3 4 100 3.596077 0.4363185 0.3 4 150 3.596077 0.4363185 0.3 5 50 3.696588 0.4157301 0.3 5 100 3.696174 0.4158225 0.3 5 150 3.696174 0.4158225 0.3 6 50 3.649561 0.4213714 0.3 6 100 3.649099 0.4214675 0.3 6 150 3.649099 0.4214675 0.3 7 50 3.645498 0.4238297 0.3 7 100 3.645102 0.4239074 0.3 7 150 3.645102 0.4239074 0.3 8 50 3.639271 0.4225056 0.3 8 100 3.638953 0.4225713 0.3 8 150 3.638953 0.4225713 0.4 2 50 3.507004 0.4657342 0.4 2 100 3.507004 0.4657342 0.4 2 150 3.507004 0.4657342 0.4 3 50 3.577560 0.4500097 0.4 3 100 3.577559 0.4500097 0.4 3 150 3.577559 0.4500097 0.4 4 50 3.467591 0.4674221 0.4 4 100 3.467590 0.4674221 0.4 4 150 3.467590 0.4674221 0.4 5 50 3.478486 0.4703337 0.4 5 100 3.478485 0.4703337 0.4 5 150 3.478485 0.4703337 0.4 6 50 3.499017 0.4641084 0.4 6 100 3.499016 0.4641084 0.4 6 150 3.499016 0.4641084 0.4 7 50 3.413462 0.4785937 0.4 7 100 3.413461 0.4785937 0.4 7 150 3.413461 0.4785937 0.4 8 50 3.476888 0.4691880 0.4 8 100 3.476887 0.4691880 0.4 8 150 3.476887 0.4691880 0.5 2 50 3.388779 0.4899418 0.5 2 100 3.388779 0.4899418 0.5 2 150 3.388779 0.4899418 0.5 3 50 3.530002 0.4338285 0.5 3 100 3.530002 0.4338285 0.5 3 150 3.530002 0.4338285 0.5 4 50 3.445967 0.4727141 0.5 4 100 3.445967 0.4727141 0.5 4 150 3.445967 0.4727141 0.5 5 50 3.372272 0.4827002 0.5 5 100 3.372272 0.4827002 0.5 5 150 3.372272 0.4827002 0.5 6 50 3.300524 0.5150699 0.5 6 100 3.300524 0.5150699 0.5 6 150 3.300524 0.5150699 0.5 7 50 3.434904 0.4726287 0.5 7 100 3.434904 0.4726287 0.5 7 150 3.434904 0.4726287 0.5 8 50 3.384832 0.4846567 0.5 8 100 3.384832 0.4846567 0.5 8 150 3.384832 0.4846567 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 = 6, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 3.301 R2: 0.515 XGBoost variable importance: Feature Gain Cover Frequency 1: ENAX_M_agg30cm_AF_1km 3.828978e-01 0.015740741 0.0125 2: Fapar 2.757374e-01 0.059722222 0.0500 3: AAIavg_GYGA 2.499875e-01 0.221759259 0.1875 4: B02CHE3 5.069335e-02 0.098611111 0.1000 5: NCluster_7_AF_1km 1.546467e-02 0.026388889 0.0250 6: P.B_M_agg30cm_AF_1km 1.233594e-02 0.043518519 0.0375 7: GAEZ_ET 1.067380e-02 0.031481481 0.0250 8: PHIHOXagg0_30 8.140513e-04 0.029629630 0.0250 9: NEGMRG5 2.966381e-04 0.014351852 0.0125 10: IMOD4avg 2.771536e-04 0.013888889 0.0125 11: Ca_M_agg30cm_AF_1km 2.419003e-04 0.031481481 0.0250 12: K_M_agg30cm_AF_1km 1.405309e-04 0.029166667 0.0250 13: P.T_M_agg30cm_AF_1km 1.311133e-04 0.009259259 0.0125 14: af_agg_30cm_AWCpF23__M_1km 7.045275e-05 0.012962963 0.0500 15: Cassave_actual_baseline 6.926540e-05 0.014814815 0.0125 Ensemble validation RMSE: 3.506 R2: 0.428 -------------------------------------- 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: 34 Number of independent variables: 377 Mtry: 18 Target node size: 5 Variable importance mode: impurity OOB prediction error: 266.8266 R squared: 0.2629964 OOB RMSE: 16.335 Variable importance: [,1] Fapar 332.4343 Fcover 289.5037 MAXENV3 269.1298 Mn_M_agg30cm_AF_1km 240.9927 Slopeclassc3 239.9047 P_M_agg30cm_AF_1km 233.0854 EVEENV3 230.6793 M13RB3ALT 219.1532 MANMCF5 212.3941 CRFVOL_M_agg30cm_AF_1km 191.3467 ENTENV3 185.2141 BLDFIE_M_agg30cm_AF_1km 181.2104 Cassave_actual_baseline 178.6353 M13RB3A04 176.7425 CLYPPT_M_agg30cm_AF_1km 169.6373 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 16.50738 0.4551242 0.3 2 100 16.50837 0.4550985 0.3 2 150 16.50837 0.4550985 0.3 3 50 16.13134 0.4642058 0.3 3 100 16.13043 0.4642625 0.3 3 150 16.13043 0.4642625 0.3 4 50 16.44214 0.4549520 0.3 4 100 16.44056 0.4550434 0.3 4 150 16.44056 0.4550434 0.3 5 50 16.45815 0.4407725 0.3 5 100 16.45690 0.4408426 0.3 5 150 16.45690 0.4408426 0.3 6 50 15.76591 0.4977035 0.3 6 100 15.76475 0.4977700 0.3 6 150 15.76475 0.4977700 0.3 7 50 16.65711 0.4277071 0.3 7 100 16.65595 0.4277710 0.3 7 150 16.65595 0.4277710 0.3 8 50 16.58082 0.4516075 0.3 8 100 16.57959 0.4516927 0.3 8 150 16.57959 0.4516927 0.4 2 50 16.59426 0.4508633 0.4 2 100 16.59424 0.4508643 0.4 2 150 16.59424 0.4508643 0.4 3 50 16.53640 0.4496626 0.4 3 100 16.53638 0.4496640 0.4 3 150 16.53638 0.4496640 0.4 4 50 16.51759 0.4532967 0.4 4 100 16.51755 0.4532989 0.4 4 150 16.51755 0.4532989 0.4 5 50 16.52807 0.4524179 0.4 5 100 16.52805 0.4524193 0.4 5 150 16.52805 0.4524193 0.4 6 50 16.40224 0.4550093 0.4 6 100 16.40224 0.4550094 0.4 6 150 16.40224 0.4550094 0.4 7 50 15.87578 0.4708405 0.4 7 100 15.87575 0.4708422 0.4 7 150 15.87575 0.4708422 0.4 8 50 16.61075 0.4450311 0.4 8 100 16.61075 0.4450311 0.4 8 150 16.61075 0.4450311 0.5 2 50 17.26057 0.4015460 0.5 2 100 17.26057 0.4015460 0.5 2 150 17.26057 0.4015460 0.5 3 50 17.59026 0.3806589 0.5 3 100 17.59026 0.3806589 0.5 3 150 17.59026 0.3806589 0.5 4 50 17.20915 0.4229747 0.5 4 100 17.20915 0.4229747 0.5 4 150 17.20915 0.4229747 0.5 5 50 16.22205 0.4378918 0.5 5 100 16.22205 0.4378918 0.5 5 150 16.22205 0.4378918 0.5 6 50 17.35536 0.3925060 0.5 6 100 17.35536 0.3925060 0.5 6 150 17.35536 0.3925060 0.5 7 50 17.70578 0.3707339 0.5 7 100 17.70578 0.3707339 0.5 7 150 17.70578 0.3707339 0.5 8 50 17.04326 0.4205541 0.5 8 100 17.04326 0.4205541 0.5 8 150 17.04326 0.4205541 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 = 100, max_depth = 6, eta = 0.3, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 15.765 R2: 0.498 XGBoost variable importance: Feature Gain Cover Frequency 1: Fapar 0.634114783 0.014780467 0.012295082 2: af_agg_ERZD_TAWCpF23mm__M_1km 0.112195659 0.236342559 0.200819672 3: Cassave_actual_baseline 0.106121691 0.019707289 0.016393443 4: P_M_agg30cm_AF_1km 0.061623389 0.033038690 0.028688525 5: BLDFIE_M_agg30cm_AF_1km 0.028328196 0.053180698 0.049180328 6: B02CHE3 0.015333026 0.014490654 0.024590164 7: M43WNALT 0.011070554 0.018548037 0.016393443 8: ECN_M_agg30cm_AF_1km 0.010067922 0.023619765 0.024590164 9: CLYPPT_M_agg30cm_AF_1km 0.007849859 0.008259673 0.008196721 10: af_ERZD__M_1km 0.002595123 0.035067382 0.032786885 11: Wdvi 0.002488888 0.014345747 0.012295082 12: BIO1ALT 0.002042430 0.009563831 0.012295082 13: B14CHE3 0.001644379 0.014780467 0.012295082 14: af_agg_30cm_AWCpF23__M_1km 0.001438787 0.019127663 0.040983607 15: C03GLC5 0.001012484 0.160846254 0.139344262 Ensemble validation RMSE: 15.498 R2: 0.352 --------------------------------------