Table 5.

Best fit models for estimating percentage of seeds damaged by Leptoglossus occidentalis, based on the sample of 192 Pinus strobiformis trees.

Variable selection method	Machine learning algorithm	Independent variables	RMSE	MAE	R2
ROC	avNNet	SMRPB, DD5, WINP, Mnppm, KCEC, GSP, Arbutus xalapensis, Juniperus deppeana	0.199	0.162	0.122
ROC	brnn	SMRPB, DD5, WINP, Mnppm, KCEC, GSP, Arbutus xalapensis, Juniperus deppeana	0.201	0.160	0.107
KWE	rf	MMAX, FDAY, SMRPB, SPRP, dieback, shootdeath, Slope, Znppm	0.203	0.163	0.140
KWE	nnet	MMAX, FDAY, SMRPB, SPRP, dieback, shootdeath, Slope, Znppm	0.203	0.164	0.076
ROC	lm	SMRPB, DD5, WINP, Mnppm, KCEC, GSP, Arbutus xalapensis, Juniperus deppeana	0.204	0.165	0.111
KWE	mlpWeightDecay	MMAX, FDAY, SMRPB, SPRP, dieback, shoot death, Slope, Znppm	0.219	0.174	0.082

ROC, Receiver Operating Characteristic; KW_E, Kruskal Wallis selected with elevation; avNNet, Model Averaged Neural Network; brnn, Bayesian Regularized Neural Networks; rf, Random Forest; nnet, Neural Network; lm, linear regression; mlpWeightDecay, Multi-Layer Perceptron; RMSE, Root-mean-squared error; MAE, Mean Absolute Error; R², Determination coefficient; SMRPB, Summer precipitation balance: (jul+aug+sep)/(apr+may+jun); DD5, Degree-days >5 degrees C (based on mean monthly temperature); WINP, Winter precipitation: (nov+dec+jan+feb); Mn_ppm, manganese concentration in the soil (ppm); K_CEC, proportion of potassium in CEC (cation exchange capacity); GSP, Growing season precipitation (April to September); Arbutus xalapensis, frequency of occurrence of Arbutus xalapensis in the neighborhood; Juniperus deppeana, frequency of occurrence of Juniperus deppeana in the neighborhood; MMAX, Mean maximum temperature in the warmest month (degrees C); FDAY, Julian date of the first freezing date of autumn; SPRP, Spring precipitation: (apr+may), dieback, Cronartium ribicola damage measurement; shoot death, Cronartium ribicola damage measurement; Slope, slope position (%); Zn_ppm, zinc concentration in the soil (ppm).