Variable: YA 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: 19 Number of independent variables: 377 Mtry: 18 Target node size: 5 Variable importance mode: impurity OOB prediction error: 2555795 R squared: -0.1188475 OOB RMSE: 1598.685 Variable importance: [,1] yGapRice 2421946.1 TMSD3avg 1300438.8 VBFMRG5 1280826.3 NCluster_M_AF_1km 699335.8 NCluster_14_AF_1km 585995.0 af_BDRICM_T__M_1km 568832.0 M17NPPALTfill 548645.7 EACKCL_M_agg30cm_AF_1km 473624.1 fPR_RiceTrials 469469.0 NCluster_4_AF_1km 462269.2 ENTENV3 447831.2 GPMIMERGALT 447774.7 af_agg_ERZD_TAWCpF23mm__M_1km 445105.5 Al_M_agg30cm_AF_1km 441727.6 AfSIS_WRBc109 417179.0 eXtreme Gradient Boosting 19 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 14, 12, 12 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 1085.7496 0.6880073 0.3 2 100 1085.7331 0.6879573 0.3 2 150 1085.7334 0.6879572 0.3 3 50 1086.3620 0.5103513 0.3 3 100 1086.3729 0.5103515 0.3 3 150 1086.3730 0.5103516 0.3 4 50 1228.4008 0.3322422 0.3 4 100 1228.4336 0.3322553 0.3 4 150 1228.4331 0.3322560 0.3 5 50 1011.4216 0.6953931 0.3 5 100 1011.4744 0.6952818 0.3 5 150 1011.4745 0.6952818 0.3 6 50 1114.6293 0.5755556 0.3 6 100 1114.6921 0.5755404 0.3 6 150 1114.6914 0.5755412 0.3 7 50 1233.5266 0.3454147 0.3 7 100 1233.5348 0.3454137 0.3 7 150 1233.5336 0.3454168 0.3 8 50 1075.1924 0.5867874 0.3 8 100 1075.2195 0.5867828 0.3 8 150 1075.2202 0.5867823 0.4 2 50 1158.8744 0.5188652 0.4 2 100 1158.8745 0.5188653 0.4 2 150 1158.8745 0.5188653 0.4 3 50 982.4759 0.6896219 0.4 3 100 982.4752 0.6896245 0.4 3 150 982.4752 0.6896245 0.4 4 50 1218.2603 0.4776524 0.4 4 100 1218.2612 0.4776501 0.4 4 150 1218.2612 0.4776501 0.4 5 50 1182.6548 0.4141813 0.4 5 100 1182.6607 0.4141803 0.4 5 150 1182.6607 0.4141803 0.4 6 50 1083.4191 0.5654359 0.4 6 100 1083.4234 0.5654356 0.4 6 150 1083.4234 0.5654356 0.4 7 50 1100.5640 0.4376570 0.4 7 100 1100.5644 0.4376577 0.4 7 150 1100.5644 0.4376577 0.4 8 50 1171.7738 0.5512170 0.4 8 100 1171.7771 0.5512168 0.4 8 150 1171.7771 0.5512168 0.5 2 50 953.0407 0.5949815 0.5 2 100 953.0407 0.5949815 0.5 2 150 953.0407 0.5949815 0.5 3 50 1327.3666 0.4615555 0.5 3 100 1327.3668 0.4615555 0.5 3 150 1327.3668 0.4615555 0.5 4 50 1225.6023 0.5015551 0.5 4 100 1225.6024 0.5015551 0.5 4 150 1225.6024 0.5015551 0.5 5 50 1197.8765 0.4771426 0.5 5 100 1197.8766 0.4771426 0.5 5 150 1197.8766 0.4771426 0.5 6 50 1106.1126 0.6521782 0.5 6 100 1106.1128 0.6521781 0.5 6 150 1106.1128 0.6521781 0.5 7 50 1131.7094 0.5292643 0.5 7 100 1131.7094 0.5292643 0.5 7 150 1131.7094 0.5292643 0.5 8 50 1190.4189 0.3635468 0.5 8 100 1190.4190 0.3635468 0.5 8 150 1190.4190 0.3635468 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 = 2, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 953.041 R2: 0.595 XGBoost variable importance: Feature Gain Cover Frequency 1: yGapRice 8.515225e-01 0.149149149 0.117647059 2: NCluster_M_AF_1km 6.262316e-02 0.018518519 0.014705882 3: AAIavg_GYGA 4.086164e-02 0.005005005 0.014705882 4: Wdvi 2.231933e-02 0.012012012 0.014705882 5: af_agg_30cm_PWP__M_1km 1.360163e-02 0.101601602 0.102941176 6: Ca_M_agg30cm_AF_1km 2.323345e-03 0.008008008 0.007352941 7: NCluster_6_AF_1km 1.993480e-03 0.023523524 0.022058824 8: C04GLC5 1.300865e-03 0.066066066 0.058823529 9: P_M_agg30cm_AF_1km 1.056166e-03 0.011011011 0.014705882 10: EVEENV3 8.198309e-04 0.005005005 0.007352941 11: TMSD3avg 4.047212e-04 0.013513514 0.014705882 12: af_agg_30cm_TETAs__M_1km 3.527643e-04 0.007507508 0.014705882 13: L10USG5 2.632308e-04 0.004504505 0.007352941 14: OC_M_agg30cm_AF_1km 1.234785e-04 0.006006006 0.007352941 15: NCluster_18_AF_1km 9.915007e-05 0.038038038 0.029411765 Ensemble validation RMSE: 1180.26 R2: 0.366 -------------------------------------- Variable: YW 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: 19 Number of independent variables: 377 Mtry: 6 Target node size: 5 Variable importance mode: impurity OOB prediction error: 802732.4 R squared: -0.144322 OOB RMSE: 895.953 Variable importance: [,1] CRFVOL_M_agg30cm_AF_1km 260697.8 BARL10 163344.1 P.B_M_agg30cm_AF_1km 152409.3 af_agg_30cm_TETAs__M_1km 152293.3 BLDFIE_M_agg30cm_AF_1km 141221.1 B02CHE3 133708.0 VDPMRG5 132433.7 NEGMRG5 130214.4 M02MOD4 123161.7 Na_M_agg30cm_AF_1km 121496.7 GTDHYS3 120086.4 Temperature 118402.2 TMOD3avg 112957.4 M13NDVIA04 109745.2 RANENV3 103100.5 eXtreme Gradient Boosting 19 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 12, 13, 13 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 1001.3352 0.15740825 0.3 2 100 1001.3272 0.15747760 0.3 2 150 1001.3269 0.15747758 0.3 3 50 973.4849 0.10239804 0.3 3 100 973.4736 0.10248721 0.3 3 150 973.4729 0.10248539 0.3 4 50 958.6235 0.07497265 0.3 4 100 958.6263 0.07496169 0.3 4 150 958.6234 0.07496081 0.3 5 50 984.1137 0.03791158 0.3 5 100 984.1303 0.03789726 0.3 5 150 984.1291 0.03789596 0.3 6 50 960.0574 0.18623945 0.3 6 100 960.0669 0.18639481 0.3 6 150 960.0668 0.18639478 0.3 7 50 964.6116 0.12313806 0.3 7 100 964.6267 0.12316491 0.3 7 150 964.6252 0.12316400 0.3 8 50 1000.9639 0.05615739 0.3 8 100 1000.9461 0.05610209 0.3 8 150 1000.9449 0.05610196 0.4 2 50 955.0438 0.09795298 0.4 2 100 955.0534 0.09792837 0.4 2 150 955.0552 0.09792779 0.4 3 50 1036.3741 0.22079519 0.4 3 100 1036.3806 0.22077889 0.4 3 150 1036.3806 0.22077889 0.4 4 50 925.1082 0.19084927 0.4 4 100 925.1093 0.19085126 0.4 4 150 925.1091 0.19085127 0.4 5 50 925.7736 0.36551027 0.4 5 100 925.7745 0.36550932 0.4 5 150 925.7743 0.36550936 0.4 6 50 1240.4128 0.28549423 0.4 6 100 1240.4121 0.28549496 0.4 6 150 1240.4116 0.28549428 0.4 7 50 951.1623 0.12196660 0.4 7 100 951.1631 0.12196658 0.4 7 150 951.1629 0.12196662 0.4 8 50 906.4906 0.21761988 0.4 8 100 906.4925 0.21761429 0.4 8 150 906.4925 0.21761429 0.5 2 50 953.0126 0.10014881 0.5 2 100 953.0132 0.10014805 0.5 2 150 953.0133 0.10014812 0.5 3 50 950.3554 0.03742890 0.5 3 100 950.3554 0.03742890 0.5 3 150 950.3554 0.03742896 0.5 4 50 916.7357 0.04146457 0.5 4 100 916.7342 0.04146421 0.5 4 150 916.7326 0.04146383 0.5 5 50 987.4272 0.01526266 0.5 5 100 987.4272 0.01526266 0.5 5 150 987.4272 0.01526266 0.5 6 50 938.6747 0.09320550 0.5 6 100 938.6744 0.09320528 0.5 6 150 938.6740 0.09320505 0.5 7 50 836.1290 0.15154626 0.5 7 100 836.1290 0.15154626 0.5 7 150 836.1290 0.15154626 0.5 8 50 958.6024 0.10982185 0.5 8 100 958.6025 0.10982190 0.5 8 150 958.6025 0.10982190 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 = 7, eta = 0.5, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 836.129 R2: 0.152 XGBoost variable importance: Feature Gain Cover Frequency 1: CRFVOL_M_agg30cm_AF_1km 0.509441406 0.010835472 0.007936508 2: Na_M_agg30cm_AF_1km 0.142589539 0.003706872 0.003968254 3: af_BDRICM_T__M_1km 0.116139823 0.011120616 0.015873016 4: MAXENV3 0.093230331 0.005702880 0.007936508 5: M02MOD4 0.044548406 0.005417736 0.003968254 6: SW1L00 0.018229453 0.005417736 0.003968254 7: Fe_M_agg30cm_AF_1km 0.011930057 0.008269176 0.007936508 8: yGapRice 0.009759043 0.005417736 0.003968254 9: Fapar 0.008616251 0.009409752 0.007936508 10: af_agg_30cm_AWCpF23__M_1km 0.007344295 0.019389792 0.051587302 11: af_agg_30cm_TETAs__M_1km 0.007243871 0.186769319 0.138888889 12: GPMIMERGALT 0.006468798 0.019674936 0.019841270 13: EACKCL_M_agg30cm_AF_1km 0.006076593 0.003992016 0.003968254 14: af_agg_30cm_PWP__M_1km 0.005320133 0.027658968 0.051587302 15: Mn_M_agg30cm_AF_1km 0.003995559 0.005417736 0.003968254 Ensemble validation RMSE: 857.655 R2: 0 -------------------------------------- Variable: YW_SD 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: 19 Number of independent variables: 377 Mtry: 4 Target node size: 5 Variable importance mode: impurity OOB prediction error: 720667.8 R squared: -0.1637277 OOB RMSE: 848.922 Variable importance: [,1] RANENV3 149534.99 M13RB3A04 124568.65 CRFVOL_M_agg30cm_AF_1km 121475.46 NEGMRG5 111372.06 Temperature 103898.91 NCluster_14_AF_1km 96295.77 af_agg_30cm_TETAs__M_1km 93274.16 K_M_agg30cm_AF_1km 93000.94 MAXENV3 91761.66 rElevIndex 86418.33 GAEZ_ET 85488.74 SNDPPT_M_agg30cm_AF_1km 84553.20 P.B_M_agg30cm_AF_1km 84040.80 SW2L14 82082.90 yFertilised_RiceTrials 78784.84 eXtreme Gradient Boosting 19 samples 377 predictors No pre-processing Resampling: Cross-Validated (3 fold, repeated 1 times) Summary of sample sizes: 13, 13, 12 Resampling results across tuning parameters: eta max_depth nrounds RMSE Rsquared 0.3 2 50 861.9559 0.27224039 0.3 2 100 862.0165 0.27220380 0.3 2 150 862.0168 0.27220630 0.3 3 50 979.3389 0.27932113 0.3 3 100 979.3426 0.27926583 0.3 3 150 979.3426 0.27926583 0.3 4 50 906.9419 0.21681416 0.3 4 100 906.9537 0.21681575 0.3 4 150 906.9537 0.21681575 0.3 5 50 889.5386 0.10315218 0.3 5 100 889.5968 0.10326509 0.3 5 150 889.5953 0.10326173 0.3 6 50 849.2789 0.30221201 0.3 6 100 849.3352 0.30225606 0.3 6 150 849.3338 0.30225343 0.3 7 50 833.3239 0.06671046 0.3 7 100 833.3623 0.06675944 0.3 7 150 833.3620 0.06675971 0.3 8 50 935.9967 0.27129844 0.3 8 100 936.0460 0.27120070 0.3 8 150 936.0455 0.27120032 0.4 2 50 948.3261 0.21943044 0.4 2 100 948.3266 0.21943235 0.4 2 150 948.3266 0.21943235 0.4 3 50 946.1550 0.25577948 0.4 3 100 946.1571 0.25578070 0.4 3 150 946.1571 0.25578070 0.4 4 50 925.5634 0.09189713 0.4 4 100 925.5659 0.09190039 0.4 4 150 925.5659 0.09190039 0.4 5 50 895.7256 0.28418038 0.4 5 100 895.7273 0.28417731 0.4 5 150 895.7273 0.28417731 0.4 6 50 951.8043 0.17519022 0.4 6 100 951.8095 0.17519283 0.4 6 150 951.8101 0.17519426 0.4 7 50 961.6389 0.13401010 0.4 7 100 961.6401 0.13401226 0.4 7 150 961.6407 0.13401522 0.4 8 50 981.8319 0.35839408 0.4 8 100 981.8329 0.35839842 0.4 8 150 981.8336 0.35839933 0.5 2 50 902.4299 0.12366750 0.5 2 100 902.4301 0.12366768 0.5 2 150 902.4300 0.12366749 0.5 3 50 950.6992 0.14039769 0.5 3 100 950.6991 0.14039740 0.5 3 150 950.6991 0.14039763 0.5 4 50 860.1578 0.35906094 0.5 4 100 860.1585 0.35906089 0.5 4 150 860.1589 0.35906090 0.5 5 50 962.2639 0.13171674 0.5 5 100 962.2639 0.13171724 0.5 5 150 962.2639 0.13171736 0.5 6 50 893.4256 0.17219525 0.5 6 100 893.4261 0.17219691 0.5 6 150 893.4267 0.17219861 0.5 7 50 932.8830 0.15142969 0.5 7 100 932.8834 0.15142948 0.5 7 150 932.8834 0.15142948 0.5 8 50 868.6636 0.05943421 0.5 8 100 868.6637 0.05943440 0.5 8 150 868.6637 0.05943440 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 = 7, eta = 0.3, gamma = 0, colsample_bytree = 0.8 and min_child_weight = 1. RMSE: 833.324 R2: 0.067 XGBoost variable importance: Feature Gain Cover Frequency 1: Na_M_agg30cm_AF_1km 0.262431782 0.009874327 0.007194245 2: CRFVOL_M_agg30cm_AF_1km 0.235399716 0.037103531 0.025179856 3: BICUSG5 0.106511329 0.019449431 0.014388489 4: NEGMRG5 0.075446132 0.010173549 0.007194245 5: B02CHE3 0.066601967 0.029622980 0.021582734 6: RANENV3 0.059280378 0.034410533 0.025179856 7: yGapRice 0.032158377 0.009275883 0.010791367 8: M02MOD4 0.030163548 0.015858767 0.010791367 9: NCluster_8_AF_1km 0.020212494 0.003889886 0.003597122 10: PHIHOXagg0_30 0.015337285 0.004488330 0.003597122 11: af_agg_30cm_PWP__M_1km 0.014566523 0.015858767 0.039568345 12: Zn_M_agg30cm_AF_1km 0.013020884 0.003291442 0.003597122 13: rElevIndex 0.011315104 0.008078995 0.007194245 14: yFertilised_RiceTrials 0.010274854 0.008378217 0.007194245 15: af_BDRICM_T__M_1km 0.007895117 0.011669659 0.017985612 Ensemble validation RMSE: 825.144 R2: 0.027 --------------------------------------