Comparison among allometric models for tree biomass estimation using non-destructive trees’ data

The non-destructive method—using allometric models—of forest biomass estimation is one of the most effective and efficient approaches, however, there is an ambiguity to choose a particular type of model for predicting biomass. The study aimed to compare 11 different allometric models generally employing in Nepal using empirical data of 4,942 individual standing trees of 36 species from 134 sample plots considering 20 environmental variables from five community forests belonging in two ecological regions. Akaike Information Criterion (AIC), coefficient of determination (R²), and root mean square error (RMSE) of each model was estimated and compared. The biomass density predicted by different models differed significantly (p < 0.05) despite employing the same data set. A variation on the estimated biomass density differed from 4% (M4 and M11) to 364% (M5 and M10) among the models considered for the study. Results show that no models have relatively higher R² (less than 0.61), lower AIC (more than 813) and RMSE (more than 16). This indicates that hardly any existing general models in Nepal estimates actual biomass of the forest ecosystems. The results suggest that a universal model may not be applicable for the given condition, indicating an imminent need for the development of allometric models for biomass prediction based on species type, ecological regions, degree of disturbances, considering diversity variables, and ground and crown cover scenario of the forests. Such models require regular calibration for validity, reliability, credibility, and integrity of forest data management. The results would be a reference for policymaker and forest conservationists to select appropriate allometric equation and, thereby accurate prediction of biomass and timber volume, which in turn, will have good value for REDD + (carbon payment from forestry sector) and other economic conversions of the forests' biomass.