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: 82 Number of independent variables: 372 Mtry: 8 Target node size: 5 Variable importance mode: impurity OOB prediction error: 309793 R squared: 0.3636049 OOB RMSE: 556.591 Variable importance: [,1] Cu_M_agg30cm_AF_1km 686086.0 NCluster_19_AF_1km 486172.1 CEC_M_agg30cm_AF_1km 463830.0 rElev 463413.4 Ca_M_agg30cm_AF_1km 442183.1 PHIHOXagg0_30 419847.6 P_M_agg30cm_AF_1km 413499.8 BIO1ALT 400254.3 Wdvi 394861.5 M13RB3A01 387023.6 NIRL00 379952.8 SW1L14 378171.1 MAXENV3 374141.5 M13NDVIA08 362360.0 VBFMRG5 361955.0 eXtreme Gradient Boosting 82 samples 372 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 54, 55, 55 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 692.9612 0.3961765 0.3 2 100 693.3338 0.3970040 0.3 2 150 693.3387 0.3970119 0.3 3 50 700.2206 0.4237388 0.3 3 100 700.2878 0.4238076 0.3 3 150 700.2878 0.4238076 0.3 4 50 754.3855 0.3922310 0.3 4 100 754.4765 0.3922809 0.3 4 150 754.4765 0.3922809 0.3 5 50 703.5150 0.4200306 0.3 5 100 703.6074 0.4200664 0.3 5 150 703.6074 0.4200664 0.3 6 50 701.5145 0.4253967 0.3 6 100 701.5987 0.4254390 0.3 6 150 701.5987 0.4254390 0.3 7 50 755.7084 0.3877445 0.3 7 100 755.7975 0.3878006 0.3 7 150 755.7975 0.3878006 0.3 8 50 708.5913 0.3942082 0.3 8 100 708.6896 0.3942508 0.3 8 150 708.6896 0.3942508 0.4 2 50 719.2970 0.3754583 0.4 2 100 719.4295 0.3754803 0.4 2 150 719.4293 0.3754813 0.4 3 50 674.5514 0.3833701 0.4 3 100 674.5548 0.3833730 0.4 3 150 674.5548 0.3833730 0.4 4 50 737.4646 0.3617401 0.4 4 100 737.4685 0.3617422 0.4 4 150 737.4685 0.3617422 0.4 5 50 671.8696 0.4088449 0.4 5 100 671.8745 0.4088462 0.4 5 150 671.8745 0.4088462 0.4 6 50 692.4480 0.4155466 0.4 6 100 692.4526 0.4155486 0.4 6 150 692.4526 0.4155486 0.4 7 50 709.8659 0.3832795 0.4 7 100 709.8710 0.3832812 0.4 7 150 709.8710 0.3832812 0.4 8 50 692.1297 0.4118416 0.4 8 100 692.1331 0.4118440 0.4 8 150 692.1331 0.4118440 0.5 2 50 710.2926 0.3632619 0.5 2 100 710.2912 0.3632834 0.5 2 150 710.2912 0.3632834 0.5 3 50 688.7063 0.3944040 0.5 3 100 688.7063 0.3944043 0.5 3 150 688.7063 0.3944043 0.5 4 50 738.2523 0.3259021 0.5 4 100 738.2524 0.3259022 0.5 4 150 738.2524 0.3259022 0.5 5 50 692.0849 0.4001369 0.5 5 100 692.0849 0.4001369 0.5 5 150 692.0849 0.4001369 0.5 6 50 723.5429 0.3398193 0.5 6 100 723.5430 0.3398193 0.5 6 150 723.5430 0.3398193 0.5 7 50 707.4393 0.4231886 0.5 7 100 707.4393 0.4231887 0.5 7 150 707.4393 0.4231887 0.5 8 50 760.9814 0.3231512 0.5 8 100 760.9814 0.3231512 0.5 8 150 760.9814 0.3231512 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 = 5, eta = 0.4, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 671.87 R2: 0.409 XGBoost variable importance: Feature Gain Cover Frequency 1: CEC_M_agg30cm_AF_1km 0.21596809 0.009312473 0.012433393 2: P_M_agg30cm_AF_1km 0.20477901 0.007343550 0.003552398 3: Wdvi 0.10076683 0.010855683 0.005328597 4: C01GLC5 0.07428467 0.014899957 0.021314387 5: MAXENV3 0.07389658 0.054916986 0.024866785 6: Cu_M_agg30cm_AF_1km 0.05853736 0.137718178 0.062166963 7: rElev 0.04201545 0.038899532 0.019538188 8: af_agg_ERZD_TAWCpF23mm__M_1km 0.03537520 0.021338868 0.044404973 9: B13CHE3 0.03365636 0.064974457 0.072824156 10: af_agg_30cm_PWP__M_1km 0.02494519 0.013303533 0.060390764 11: EVEENV3 0.02336347 0.030119200 0.017761989 12: NCluster_16_AF_1km 0.02205663 0.005800341 0.005328597 13: AAIavg_GYGA 0.01844630 0.012718178 0.069271758 14: af_BDRICM_T__M_1km 0.01390318 0.004469987 0.017761989 15: NCluster_12_AF_1km 0.01162762 0.007875692 0.003552398 Ensemble validation RMSE: 571.713 R2: 0.35 -------------------------------------- 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: 57 Number of independent variables: 372 Mtry: 14 Target node size: 5 Variable importance mode: impurity OOB prediction error: 691181.6 R squared: 0.4537381 OOB RMSE: 831.373 Variable importance: [,1] NCluster_16_AF_1km 1655513.3 M02MOD4 1644495.4 SW2L00 1639956.2 SW1L00 1503081.6 Cu_M_agg30cm_AF_1km 1409044.2 Sorghum_actual_baseline 1335637.3 NCluster_19_AF_1km 1234162.8 MMOD4avg 1172564.3 SLTPPT_M_agg30cm_AF_1km 1152953.7 BIO1ALT 982916.7 NIRL00 975587.2 SNDPPT_M_agg30cm_AF_1km 879703.1 Al_M_agg30cm_AF_1km 871700.4 PET 845345.1 M43WNALT 836437.2 eXtreme Gradient Boosting 57 samples 372 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 39, 37, 38 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 1010.4518 0.3083235 0.3 2 100 1010.6906 0.3083030 0.3 2 150 1010.6907 0.3083030 0.3 3 50 1007.3315 0.3112695 0.3 3 100 1007.4525 0.3112477 0.3 3 150 1007.4525 0.3112477 0.3 4 50 1016.1066 0.3069928 0.3 4 100 1016.2266 0.3069755 0.3 4 150 1016.2266 0.3069755 0.3 5 50 1009.6257 0.3124437 0.3 5 100 1009.7463 0.3124246 0.3 5 150 1009.7463 0.3124246 0.3 6 50 1009.6779 0.3086865 0.3 6 100 1009.7980 0.3086649 0.3 6 150 1009.7980 0.3086649 0.3 7 50 1010.9959 0.3100165 0.3 7 100 1011.1130 0.3099998 0.3 7 150 1011.1130 0.3099998 0.3 8 50 1009.1339 0.3102384 0.3 8 100 1009.2561 0.3102180 0.3 8 150 1009.2561 0.3102180 0.4 2 50 1006.6828 0.3143324 0.4 2 100 1006.6927 0.3143339 0.4 2 150 1006.6927 0.3143339 0.4 3 50 999.5643 0.3206135 0.4 3 100 999.5698 0.3206127 0.4 3 150 999.5698 0.3206127 0.4 4 50 1002.9243 0.3197612 0.4 4 100 1002.9299 0.3197605 0.4 4 150 1002.9299 0.3197605 0.4 5 50 997.4751 0.3222763 0.4 5 100 997.4807 0.3222756 0.4 5 150 997.4807 0.3222756 0.4 6 50 1005.2571 0.3151016 0.4 6 100 1005.2621 0.3151013 0.4 6 150 1005.2621 0.3151013 0.4 7 50 1004.4462 0.3158235 0.4 7 100 1004.4515 0.3158230 0.4 7 150 1004.4515 0.3158230 0.4 8 50 1001.3371 0.3205746 0.4 8 100 1001.3429 0.3205736 0.4 8 150 1001.3429 0.3205736 0.5 2 50 1005.6042 0.3145960 0.5 2 100 1005.6048 0.3145960 0.5 2 150 1005.6048 0.3145960 0.5 3 50 1000.4715 0.3191110 0.5 3 100 1000.4716 0.3191110 0.5 3 150 1000.4716 0.3191110 0.5 4 50 1014.1195 0.3161379 0.5 4 100 1014.1197 0.3161379 0.5 4 150 1014.1197 0.3161379 0.5 5 50 997.1513 0.3251231 0.5 5 100 997.1514 0.3251230 0.5 5 150 997.1514 0.3251230 0.5 6 50 1006.6196 0.3180741 0.5 6 100 1006.6198 0.3180741 0.5 6 150 1006.6198 0.3180741 0.5 7 50 998.3193 0.3220035 0.5 7 100 998.3194 0.3220034 0.5 7 150 998.3194 0.3220034 0.5 8 50 997.5511 0.3239727 0.5 8 100 997.5513 0.3239727 0.5 8 150 997.5513 0.3239727 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 = 5, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 997.151 R2: 0.325 XGBoost variable importance: Feature Gain Cover Frequency 1: DEMENV5 0.4339909869 0.017488076 0.017142857 2: NCluster_16_AF_1km 0.1995737746 0.007551669 0.005714286 3: af_agg_ERZD_TAWCpF23mm__M_1km 0.1183315403 0.054981452 0.062857143 4: Al_M_agg30cm_AF_1km 0.0837225735 0.017090620 0.022857143 5: af_BDRICM_T__M_1km 0.0602012685 0.191043985 0.148571429 6: C02GLC5 0.0487320771 0.023184950 0.022857143 7: Mg_M_agg30cm_AF_1km 0.0216607655 0.067965024 0.051428571 8: af_ERZD__M_1km 0.0114391929 0.014043455 0.011428571 9: af_agg_30cm_PWP__M_1km 0.0076333720 0.008876524 0.017142857 10: NCluster_1_AF_1km 0.0053649924 0.005431902 0.005714286 11: C03GLC5 0.0032607054 0.127848437 0.102857143 12: af_agg_30cm_AWCpF23__M_1km 0.0029488161 0.011128776 0.045714286 13: M13RB3A08 0.0014483153 0.014573397 0.011428571 14: P.B_M_agg30cm_AF_1km 0.0011456728 0.005696873 0.005714286 15: SW2L00 0.0002289663 0.006226815 0.005714286 Ensemble validation RMSE: 864.45 R2: 0.422 -------------------------------------- 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: 47 Number of independent variables: 372 Mtry: 4 Target node size: 5 Variable importance mode: impurity OOB prediction error: 573687.5 R squared: 0.4309717 OOB RMSE: 757.422 Variable importance: [,1] NMSD3avg 665620.5 SLTPPT_M_agg30cm_AF_1km 607417.3 SW1L00 570370.9 Cu_M_agg30cm_AF_1km 544178.5 NCluster_19_AF_1km 465036.4 af_BDRICM_T__M_1km 461648.9 SW2L00 445394.4 af_agg_30cm_TAWCpF23mm__M_1km 444866.6 Na_M_agg30cm_AF_1km 403850.1 ECN_M_agg30cm_AF_1km 402456.4 Wdvi 400071.0 NCluster_16_AF_1km 398873.1 PHIHOXagg0_30 384457.9 SW2L14 383396.4 MMOD4avg 376939.0 eXtreme Gradient Boosting 47 samples 372 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 31, 32, 31 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 886.4052 0.5148098 0.3 2 100 886.4918 0.5147548 0.3 2 150 886.4918 0.5147548 0.3 3 50 867.7303 0.5074357 0.3 3 100 867.7952 0.5074540 0.3 3 150 867.7952 0.5074540 0.3 4 50 909.8932 0.4867289 0.3 4 100 909.9549 0.4867577 0.3 4 150 909.9549 0.4867577 0.3 5 50 899.1388 0.5065488 0.3 5 100 899.1980 0.5065729 0.3 5 150 899.1980 0.5065729 0.3 6 50 888.9790 0.5207271 0.3 6 100 889.0394 0.5207471 0.3 6 150 889.0394 0.5207470 0.3 7 50 896.3443 0.4988973 0.3 7 100 896.4049 0.4989217 0.3 7 150 896.4049 0.4989217 0.3 8 50 894.2249 0.5071663 0.3 8 100 894.2861 0.5071927 0.3 8 150 894.2861 0.5071927 0.4 2 50 901.7725 0.5075222 0.4 2 100 901.7782 0.5075186 0.4 2 150 901.7782 0.5075186 0.4 3 50 817.9720 0.5173963 0.4 3 100 817.9745 0.5173976 0.4 3 150 817.9745 0.5173976 0.4 4 50 890.9691 0.5134837 0.4 4 100 890.9714 0.5134848 0.4 4 150 890.9714 0.5134848 0.4 5 50 900.1856 0.5089054 0.4 5 100 900.1882 0.5089065 0.4 5 150 900.1882 0.5089065 0.4 6 50 895.3888 0.5132266 0.4 6 100 895.3913 0.5132277 0.4 6 150 895.3913 0.5132277 0.4 7 50 874.2806 0.5209343 0.4 7 100 874.2830 0.5209356 0.4 7 150 874.2830 0.5209356 0.4 8 50 892.3911 0.5153423 0.4 8 100 892.3936 0.5153434 0.4 8 150 892.3936 0.5153434 0.5 2 50 842.9587 0.5072316 0.5 2 100 842.9589 0.5072316 0.5 2 150 842.9589 0.5072316 0.5 3 50 845.8091 0.5053401 0.5 3 100 845.8092 0.5053401 0.5 3 150 845.8092 0.5053401 0.5 4 50 899.7948 0.5065796 0.5 4 100 899.7949 0.5065796 0.5 4 150 899.7949 0.5065796 0.5 5 50 820.5056 0.5017547 0.5 5 100 820.5057 0.5017547 0.5 5 150 820.5057 0.5017547 0.5 6 50 907.0180 0.5031658 0.5 6 100 907.0181 0.5031658 0.5 6 150 907.0181 0.5031658 0.5 7 50 921.0108 0.4799885 0.5 7 100 921.0109 0.4799885 0.5 7 150 921.0109 0.4799885 0.5 8 50 859.7595 0.5009871 0.5 8 100 859.7596 0.5009871 0.5 8 150 859.7596 0.5009871 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: 817.972 R2: 0.517 XGBoost variable importance: Feature Gain Cover Frequency 1: af_BDRICM_T__M_1km 0.5440868097 0.153732882 0.129213483 2: NIRL00 0.2717371668 0.013841850 0.011235955 3: C02GLC5 0.0902811305 0.164776911 0.140449438 4: af_agg_ERZD_TAWCpF23mm__M_1km 0.0380720865 0.046973936 0.050561798 5: Mg_M_agg30cm_AF_1km 0.0169131513 0.027683699 0.022471910 6: M13NDVIALT 0.0130069195 0.010896775 0.011235955 7: P.T_M_agg30cm_AF_1km 0.0074745567 0.006626417 0.005617978 8: C08GLC5 0.0061323894 0.079369754 0.067415730 9: af_ERZD__M_1km 0.0045331850 0.015019879 0.016853933 10: af_agg_30cm_AWCpF23__M_1km 0.0031400990 0.009276984 0.033707865 11: N_M_agg30cm_AF_1km 0.0015192388 0.012369312 0.011235955 12: M13NDVIA08 0.0011702591 0.006479164 0.011235955 13: AAIavg_GYGA 0.0010368630 0.051391548 0.106741573 14: EVEENV3 0.0005536113 0.033868355 0.033707865 15: BARL10 0.0002356564 0.005890149 0.005617978 Ensemble validation RMSE: 733.895 R2: 0.455 --------------------------------------