# Research on the construction of growth models for dominant tree species in the Manas River Basin, Xinjiang

## TL;DR

This study builds and optimizes growth models for five main tree species in Xinjiang's Manas River Basin to improve forest management and resource allocation.

## Contribution

The study introduces deep learning to optimize multivariate nonlinear forest growth models for five dominant tree species in an arid region.

## Key findings

- Optimal DBH-height models for four tree species are S-curve models, while Mixed wood uses a logarithmic model.
- Deep learning improves model accuracy, achieving a maximum correlation coefficient of 0.960.
- Incorporating climate and topographic factors enhances model performance, with R2 reaching 0.890.

## Abstract

Research on forest growth models is not only crucial for regional ecological security and the optimal allocation of water and carbon resources but is also a key component in the study of carbon cycling in arid regions, holding scientific and practical significance for addressing climate change and promoting green sustainable development. Therefore, this study takes the Manas River Basin in Xinjiang as an example and based on the 2011 forest resource survey data from the Manas River Basin, constructs basic growth models for the diameter at breast height (DBH)-height and age-DBH relationships for five dominant tree types: Spruce, Poplar, Mixed wood, Sand jujube, and Populus euphratica. The optimal basic models for each types are selected. Secondly, climate factors (annual precipitation, Minimum of Daily Maximum Temperature, TXn) and topographic factors (Digital Elevation Model; DEM) are introduced into the optimal models to construct multivariate nonlinear forest growth models. Finally, deep learning is used to optimize the overall accuracy of the models. The results show that the optimal DBH-height models for Spruce, Poplar, Sand jujube, and Populus euphratica are S-curve models, while the optimal DBH-height model for Mixed wood is a logarithmic model. The optimal age-DBH models for Poplar and Populus euphratica are S-curve models, whereas the optimal age-DBH basic models for Spruce, Mixed wood, and Sand jujube are growth model, linear model, and logistics model, respectively. The overall accuracy of the multivariate nonlinear forest growth models is improved, with the highest R2 reaching 0.890 and the average RMSE increasing by 10.590, mainly due to the decrease in model accuracy for some tree types caused by random factors. Lastly, compared to the basic models and multivariate nonlinear forest growth models, the deep learning approach demonstrates the best performance, with the highest correlation coefficient reaching 0.960. Overall, by constructing forest growth models for five main dominant tree types in the Manas River Basin in Xinjiang, the optimal forest management strategies in the region can be determined, which helps to formulate targeted forest management strategies, effectively address the allocation of carbon and water resources, and promote healthy and sustainable forest development.