Providing optimum models for predicting the water loss after harvesting and extracting trees aboveground biomass from the Hyrcanian mixed-beech forests

Abstract

The ecological and biological attributes of trees stand as well as the water cycle in forests are substantially related to variations of water storage capacity in forest ecosystems. This study aimed to figure out a protocol for monitoring the water storage capacity variations in the Hyrcanian mixed-beech stands after harvesting and extracting trees form the forest. A total of 174 trees were felled and weighed, and destructive sampling following lines of exploitation was carried out for measuring the water content in aboveground biomass of trees. Curve estimation regression analyses including the tree biophysical variables (breast height diameter DBH, total height H, basic wood density $$\rho$$ ) were used for examining the prediction accuracy. Nonlinear models were log-transformed, and systematic bias was corrected by correction factor depending on standard error of estimate when back-transforming to the originally dependent value. The findings showed that the power-law models were the best functional form for predicting the dependent variables. Using only DBH in the simple power model explained 76% of total variance (Adj.R 2 = 0.76) with a low Akaike information criterion (AIC) and normalized root-mean-square error RMSE%, indicating a high accuracy of prediction ( $${\text{AIC}} \approx - 238$$ ; RMSE% = 11.7). Adding H and $$\rho$$ in the linear log-transformed power models with different interaction terms increased the certainty of prediction with the highest accuracy ( $${\text{Adj}}.R^{2} = 0.86,\;{\text{AIC}} = - 329,\;{\text{RMSE}}\% = 9$$ ). Considering diverse conditions for natural forest sites, the optimum models including the biophysical variables may have associated parameters in the other forests having different stand types and compositions.