Variable: yControl000 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: 360 Number of independent variables: 372 Mtry: 4 Target node size: 5 Variable importance mode: impurity OOB prediction error: 616362 R squared: 0.4257458 OOB RMSE: 785.087 Variable importance: [,1] MAXENV3 4024506 Lai_avg 3884946 M13NDVIA01 3512816 ENTENV3 3376220 rElev 3295017 M13NDVIA08 2881614 M13RB3ALT 2864659 M13RB1A08 2743240 Maize_actualbaseline 2682629 M13NDVIA04 2672524 M13RB3A08 2639354 Maize_intermed 2587389 N_M_agg30cm_AF_1km 2582334 EVEENV3 2553521 AAIavg_GYGA 2539035 eXtreme Gradient Boosting 360 samples 372 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 240, 240, 240 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 890.6756 0.2929716 0.3 2 100 905.0505 0.2894869 0.3 2 150 907.9662 0.2889389 0.3 3 50 925.7790 0.2710491 0.3 3 100 931.6370 0.2689832 0.3 3 150 932.0850 0.2687249 0.3 4 50 927.7459 0.2708398 0.3 4 100 928.4078 0.2707268 0.3 4 150 928.4175 0.2707233 0.3 5 50 921.9250 0.2787804 0.3 5 100 922.1340 0.2787058 0.3 5 150 922.1349 0.2787053 0.3 6 50 935.3812 0.2647928 0.3 6 100 935.4434 0.2647785 0.3 6 150 935.4434 0.2647785 0.3 7 50 941.4233 0.2620351 0.3 7 100 941.4812 0.2620090 0.3 7 150 941.4812 0.2620090 0.3 8 50 928.9436 0.2706108 0.3 8 100 928.9845 0.2706023 0.3 8 150 928.9845 0.2706023 0.4 2 50 900.8730 0.2920113 0.4 2 100 915.4616 0.2846317 0.4 2 150 917.5647 0.2836040 0.4 3 50 931.0302 0.2668460 0.4 3 100 932.0495 0.2667012 0.4 3 150 932.0751 0.2666950 0.4 4 50 922.0035 0.2806311 0.4 4 100 922.1337 0.2805645 0.4 4 150 922.1345 0.2805639 0.4 5 50 944.3330 0.2610873 0.4 5 100 944.3496 0.2610846 0.4 5 150 944.3496 0.2610846 0.4 6 50 924.4067 0.2758163 0.4 6 100 924.4098 0.2758161 0.4 6 150 924.4098 0.2758161 0.4 7 50 946.4454 0.2555406 0.4 7 100 946.4480 0.2555396 0.4 7 150 946.4480 0.2555396 0.4 8 50 941.9094 0.2581769 0.4 8 100 941.9110 0.2581765 0.4 8 150 941.9110 0.2581765 0.5 2 50 927.0215 0.2707289 0.5 2 100 933.6247 0.2688035 0.5 2 150 933.9309 0.2687845 0.5 3 50 935.4171 0.2672072 0.5 3 100 935.6342 0.2671917 0.5 3 150 935.6353 0.2671913 0.5 4 50 927.6850 0.2734849 0.5 4 100 927.7148 0.2734662 0.5 4 150 927.7148 0.2734662 0.5 5 50 931.9737 0.2721313 0.5 5 100 931.9762 0.2721297 0.5 5 150 931.9762 0.2721297 0.5 6 50 944.0865 0.2589016 0.5 6 100 944.0866 0.2589016 0.5 6 150 944.0866 0.2589016 0.5 7 50 939.9412 0.2679181 0.5 7 100 939.9414 0.2679179 0.5 7 150 939.9414 0.2679179 0.5 8 50 922.8905 0.2754338 0.5 8 100 922.8905 0.2754338 0.5 8 150 922.8905 0.2754338 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: 890.676 R2: 0.293 XGBoost variable importance: Feature Gain Cover Frequency 1: Lai_avg 0.14087274 0.0200172371 0.01459854 2: M13NDVIA01 0.10167382 0.0300258556 0.02189781 3: M13RB3A08 0.09698693 0.0206566766 0.02189781 4: NMSD3avg 0.09064032 0.0211571075 0.03649635 5: M13NDVIA08 0.08899982 0.0123161611 0.02919708 6: Maize_actualbaseline 0.06989111 0.0639995552 0.05109489 7: Water_balance 0.04889821 0.0110928855 0.01459854 8: ENAX_M_agg30cm_AF_1km 0.03214313 0.0593566682 0.04379562 9: ENTENV3 0.02635230 0.0677805888 0.05109489 10: MAXENV3 0.02548919 0.0100086185 0.00729927 11: M17NPPALTfill 0.02177826 0.0092579721 0.00729927 12: Ca_M_agg30cm_AF_1km 0.02177591 0.0403124913 0.03649635 13: RANENV3 0.02061926 0.0199616336 0.01459854 14: M13RB1ALT 0.01907658 0.0105924546 0.01459854 15: SW1L14 0.01880487 0.0007506464 0.00729927 Ensemble validation RMSE: 816.555 R2: 0.377 -------------------------------------- Variable: ymx000 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: 399 Number of independent variables: 372 Mtry: 10 Target node size: 5 Variable importance mode: impurity OOB prediction error: 1199457 R squared: 0.4906418 OOB RMSE: 1095.197 Variable importance: [,1] NIRL14 15906577 M13NDVIA01 15612368 CLYPPT_M_agg30cm_AF_1km 15146359 NCluster_8_AF_1km 14873422 N_M_agg30cm_AF_1km 14112729 Lai_avg 13640417 RANENV3 10386896 C08GLC5 10207892 ENTENV3 9722738 Thermoc35 9655308 C03GLC5 9522928 B_M_agg30cm_AF_1km 8132568 M13RB3A01 7952407 Mg_M_agg30cm_AF_1km 7775708 M13NDVIA04 7294035 eXtreme Gradient Boosting 399 samples 372 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 266, 266, 266 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 1363.435 0.2460801 0.3 2 100 1368.353 0.2520173 0.3 2 150 1371.949 0.2522273 0.3 3 50 1361.103 0.2544230 0.3 3 100 1365.651 0.2541435 0.3 3 150 1366.032 0.2540372 0.3 4 50 1349.556 0.2605524 0.3 4 100 1351.071 0.2604697 0.3 4 150 1351.089 0.2604641 0.3 5 50 1352.832 0.2652000 0.3 5 100 1353.085 0.2651289 0.3 5 150 1353.085 0.2651288 0.3 6 50 1334.888 0.2777568 0.3 6 100 1334.997 0.2777230 0.3 6 150 1334.997 0.2777230 0.3 7 50 1385.508 0.2410361 0.3 7 100 1385.549 0.2410310 0.3 7 150 1385.548 0.2410311 0.3 8 50 1350.250 0.2675560 0.3 8 100 1350.279 0.2675481 0.3 8 150 1350.279 0.2675482 0.4 2 50 1350.722 0.2588561 0.4 2 100 1360.527 0.2585579 0.4 2 150 1362.195 0.2581547 0.4 3 50 1377.239 0.2396309 0.4 3 100 1378.523 0.2397561 0.4 3 150 1378.553 0.2397512 0.4 4 50 1339.001 0.2743649 0.4 4 100 1339.232 0.2742802 0.4 4 150 1339.233 0.2742802 0.4 5 50 1368.799 0.2523812 0.4 5 100 1368.816 0.2523817 0.4 5 150 1368.816 0.2523817 0.4 6 50 1376.967 0.2460310 0.4 6 100 1376.971 0.2460307 0.4 6 150 1376.971 0.2460308 0.4 7 50 1347.225 0.2636880 0.4 7 100 1347.227 0.2636874 0.4 7 150 1347.227 0.2636874 0.4 8 50 1371.734 0.2548546 0.4 8 100 1371.735 0.2548544 0.4 8 150 1371.735 0.2548544 0.5 2 50 1423.292 0.2174065 0.5 2 100 1427.295 0.2186991 0.5 2 150 1427.721 0.2187845 0.5 3 50 1361.579 0.2587500 0.5 3 100 1361.977 0.2586851 0.5 3 150 1361.982 0.2586830 0.5 4 50 1364.470 0.2565245 0.5 4 100 1364.494 0.2565232 0.5 4 150 1364.494 0.2565232 0.5 5 50 1363.884 0.2578308 0.5 5 100 1363.885 0.2578305 0.5 5 150 1363.885 0.2578305 0.5 6 50 1370.382 0.2571387 0.5 6 100 1370.383 0.2571386 0.5 6 150 1370.383 0.2571386 0.5 7 50 1366.600 0.2545819 0.5 7 100 1366.600 0.2545818 0.5 7 150 1366.600 0.2545819 0.5 8 50 1413.633 0.2310335 0.5 8 100 1413.633 0.2310335 0.5 8 150 1413.633 0.2310335 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 = 0.3, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 1334.888 R2: 0.278 XGBoost variable importance: Feature Gain Cover Frequency 1: CLYPPT_M_agg30cm_AF_1km 0.32669203 0.030403000 0.015015015 2: C03GLC5 0.09419703 0.025280850 0.019519520 3: Maize_actualbaseline 0.09258799 0.017467401 0.010510511 4: NMSD3avg 0.05571992 0.023527165 0.013513514 5: AfSIS_WRBc93 0.05087765 0.004019586 0.003003003 6: RANENV3 0.04363041 0.034700398 0.021021021 7: M13NDVIA01 0.04297794 0.009410866 0.007507508 8: REDL00 0.03542382 0.004592572 0.004504505 9: PCHE3avg 0.02222511 0.004288716 0.009009009 10: M43BNALT 0.01931313 0.038945705 0.021021021 11: F05USG5 0.01827677 0.010114076 0.006006006 12: C08GLC5 0.01719630 0.017510809 0.009009009 13: Al_M_agg30cm_AF_1km 0.01422661 0.009150418 0.016516517 14: af_agg_30cm_PWP__M_1km 0.01082610 0.008360391 0.048048048 15: Temperature 0.01051362 0.003116698 0.003003003 Ensemble validation RMSE: 1178.109 R2: 0.41 -------------------------------------- Variable: fRyld 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: 306 Number of independent variables: 372 Mtry: 6 Target node size: 5 Variable importance mode: impurity OOB prediction error: 888153.1 R squared: 0.5989279 OOB RMSE: 942.419 Variable importance: [,1] B_M_agg30cm_AF_1km 9964630 M13NDVIA01 9669228 NIRL14 9344032 NCluster_8_AF_1km 8947555 M13NDVIALT 8569833 M43BVALT 8214062 N_M_agg30cm_AF_1km 8007791 REDL00 7983891 LSTD_avgIRI_Jul2002_Sep2016_mosaicLAEA_celsius 7349924 M43WVALT 7042124 RANENV3 6722998 M13RB3ALT 6176832 CLYPPT_M_agg30cm_AF_1km 6121735 M13RB3A01 6055216 M13NDVIA04 5871113 eXtreme Gradient Boosting 306 samples 372 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 203, 204, 205 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 1139.637 0.4138348 0.3 2 100 1144.164 0.4164919 0.3 2 150 1143.908 0.4174713 0.3 3 50 1134.201 0.4275494 0.3 3 100 1135.811 0.4271053 0.3 3 150 1135.830 0.4271205 0.3 4 50 1096.898 0.4609990 0.3 4 100 1096.934 0.4612409 0.3 4 150 1096.934 0.4612433 0.3 5 50 1147.299 0.4173021 0.3 5 100 1147.317 0.4173246 0.3 5 150 1147.317 0.4173247 0.3 6 50 1120.727 0.4393178 0.3 6 100 1120.737 0.4393226 0.3 6 150 1120.737 0.4393227 0.3 7 50 1095.228 0.4628411 0.3 7 100 1095.247 0.4628306 0.3 7 150 1095.247 0.4628306 0.3 8 50 1132.965 0.4302561 0.3 8 100 1132.949 0.4302701 0.3 8 150 1132.949 0.4302702 0.4 2 50 1097.921 0.4565654 0.4 2 100 1101.165 0.4564955 0.4 2 150 1101.311 0.4565049 0.4 3 50 1118.237 0.4436005 0.4 3 100 1118.408 0.4438275 0.4 3 150 1118.408 0.4438303 0.4 4 50 1132.628 0.4275806 0.4 4 100 1132.627 0.4276002 0.4 4 150 1132.627 0.4276002 0.4 5 50 1124.920 0.4344439 0.4 5 100 1124.920 0.4344472 0.4 5 150 1124.920 0.4344472 0.4 6 50 1134.780 0.4273705 0.4 6 100 1134.782 0.4273699 0.4 6 150 1134.782 0.4273699 0.4 7 50 1122.418 0.4358682 0.4 7 100 1122.419 0.4358681 0.4 7 150 1122.419 0.4358681 0.4 8 50 1124.618 0.4345916 0.4 8 100 1124.618 0.4345918 0.4 8 150 1124.618 0.4345917 0.5 2 50 1089.358 0.4667046 0.5 2 100 1089.497 0.4672916 0.5 2 150 1089.545 0.4672703 0.5 3 50 1168.187 0.4013777 0.5 3 100 1168.168 0.4014782 0.5 3 150 1168.168 0.4014785 0.5 4 50 1121.680 0.4374078 0.5 4 100 1121.677 0.4374152 0.5 4 150 1121.677 0.4374152 0.5 5 50 1120.864 0.4406614 0.5 5 100 1120.864 0.4406615 0.5 5 150 1120.864 0.4406615 0.5 6 50 1132.492 0.4298027 0.5 6 100 1132.492 0.4298027 0.5 6 150 1132.492 0.4298027 0.5 7 50 1130.476 0.4282632 0.5 7 100 1130.476 0.4282632 0.5 7 150 1130.476 0.4282632 0.5 8 50 1160.728 0.4082474 0.5 8 100 1160.728 0.4082473 0.5 8 150 1160.728 0.4082472 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: 1089.358 R2: 0.467 XGBoost variable importance: Feature Gain Cover Frequency 1: B_M_agg30cm_AF_1km 0.16996274 0.020746073 0.021739130 2: CLYPPT_M_agg30cm_AF_1km 0.16960024 0.019502618 0.014492754 3: C03GLC5 0.12456099 0.009914921 0.007246377 4: REDL00 0.10421065 0.017048429 0.014492754 5: Mg_M_agg30cm_AF_1km 0.05727727 0.039790576 0.028985507 6: EXBX_M_agg30cm_AF_1km 0.04416803 0.011714660 0.014492754 7: ENAX_M_agg30cm_AF_1km 0.03058490 0.029450262 0.021739130 8: PHIHOXagg0_30 0.02914944 0.002912304 0.007246377 9: M13NDVIA08 0.02906945 0.010013089 0.007246377 10: NMOD3avg 0.02381205 0.010013089 0.007246377 11: VBFMRG5 0.02252953 0.008409686 0.014492754 12: rElevIndex 0.02116215 0.021989529 0.021739130 13: LSTD_avgIRI_Jul2002_Sep2016_mosaicLAEA_celsius 0.02042432 0.029842932 0.021739130 14: SW2L14 0.01619882 0.029155759 0.021739130 15: ECN_M_agg30cm_AF_1km 0.01501337 0.009653141 0.014492754 Ensemble validation RMSE: 1000.8 R2: 0.553 --------------------------------------