Machine learning prediction of tree species diversity using forest structure and environmental factors: a case study from the Hyrcanian forest, Iran

Biomass patterns in Srivilliputhur Wildlife Sanctuary: exploring factors and gradients with machine learning approach

Forest biomass plays a crucial role in the global carbon cycle as a significant contributor derived from both soil and trees. This study focuses on investigating tree carbon stock (TCS) and estimating aboveground biomass (AGB) based on elevation within the Srivilliputhur Wildlife Sanctuary forest, while also exploring the various factors that influence their contribution. Utilizing a non-destructive approach for carbon estimation, we found that the total tree biomass in this region ranged from 220.9 Mg/ha (in Z6) to 720.6 Mg/ha (Z2), while tree carbon stock ranged from 103.8 to 338.7 Mg/ha. While Kruskal-Wallis tests did not reveal a significant relationship (p = 0.09) between TCS and elevation, linear regression showed a weak correlation (R2 = 0.002, p < 0.05) with elevation. To delve deeper into the factors influencing TCS and biomass distribution, we employed a random forest (RF) machine learning algorithm, demonstrating that stand structural attributes, such as basal area (BA), diameter at breast height (DBH), and density, held a more prominent role than climatic variables, including temperature, precipitation, and slope. Generalized linear models (GLM) were also utilized, confirming that BA, mean DBH, and elevation significantly influenced AGB (p ≤ 0.001), with species richness, precipitation, and temperature having lower significance (p ≤ 0.01) comparatively. Overall, the RF model exhibited superior performance (R2 = 0.92, RMSE = 0.12) in terms of root mean square error (RMSE) compared to GLM (R2 = 0.88, RMSE = 0.35). These findings shed light on the intricate dynamics of biomass distribution and the importance of both stand structural and climatic factors in shaping forest ecosystems.