Reducing the Uncertainty of Radiata Pine Site Index Maps Using an Spatial Ensemble of Machine Learning Models

Site Index has been widely used as an age normalised metric in order to account for variation forest height at a range of spatial scales. Although previous research modelling methods describe the regional Index, little examined gains that can be achieved through use regression kriging or ensemble methods. In this study, extensive set environmental surfaces were covariates predict measurements covering Pinus radiata D. Don plantations Chile. Using dataset, objectives (i) compare predictive precision geostatistical, parametric, and non-parametric models, (ii) determine whether significant attained kriging, (iii) evaluate model utilises predictions from five most precise using prediction with lowest error given pixel, (iv) produce map across study area. The models all geostatistical they included ordinary four based on partial least squares random forests. A was constructed these those developed (RMSE = 1.851 m, RMSE% 6.38%) it had relatively bias. Climatic edaphic variables strongest determinants and, particular, are related soil water balance well represented within models. These results highlight utility predicting approaches, construct may more than constituent