Modeling Spatial Variation in Stand Volume of <i>Acacia mangium </i>Plantations Using Geographically Weighted Regression

Abstract

Stand volume can be estimated from other stand variables by using multiple linear regression (MLR) or other ordinary regression models. MLR, however, only produces global parameter estimates that cannot reveal spatial variations in stand variables. In this study, we used a geographical weighted regression (GWR) method to investigate local spatial variations in the relationship between stand volume, stand age, and basal area of Acacia mangium plantations, and to examine whether a GWR model could provide better prediction accuracy than an MLR model. Stand data and geographical coordinates were obtained from 247 plantation sample plots. We analyzed the data using MLR and GWR methods by formulating a linear model that relates stand volume to stand age and basal area. Performance of the GWR model was compared with the MLR model in terms of their parameter estimates and goodness-of-fit statistics. We found that the GWR model was not only able to reveal local spatial variations in the relationship between stand volume, stand age, and basal area, but it also produced better prediction accuracy than the MLR model. The GWR model reduced AIC by 2%, increased R 2 adj up to 3%, and reduced RMSE by 14%, compared with those of the MLR model. The GWR model, therefore, could be useful for modeling spatial variations in stand attributes that cannot be revealed by ordinary regression models.

Evaluation of MLR and GWR Regression Models of Current Annual Increment Predicted by Growing Stock, Topographical and Landsat 8 Spectral Data in the Northeast Bosnian Mixed Forest

“…These findings are based on GWR showing distinctive performance related to MLR, Also, other research found that GWR provided statistically significant estimates and was important tool for modeling, detecting, and mapping the spatial variability in stand attributes and variables that cannot be revealed by ordinary regression models. (SHRİESTA, 2006;PROPASTİN, 2008;TİRYANA, 2010;BENÍTEZ, 2016, ČNÍTEZ, 2016, ČABARAVDİĆ ET ALL., 2016. TİRYANA (2010) registered that comparing with MLR, GWR produced better prediction accuracy, revealed local spatial variations in the relationship between structural predictors, reduced AIC, increased adjusted coefficient of determination up and reduced RMSE, what is in accordance with our results.…”