Ecosystem health assessment based on deep learning in a mountain-basin system in Central Asia's arid regions, China
Xu Bi,Yongyong Fu,Ping Wang,Yushuo Zhang,Zihan Yang,Fen Hou,Bo Li
DOI: https://doi.org/10.2139/ssrn.4750918
IF: 6.9
2024-05-27
Ecological Indicators
Abstract:Arid pastoral ecosystems are particularly vulnerable to degradation and desertification, exacerbated by factors such as overgrazing, land use changes, and climate change impacts. Therefore, research on ecosystem health is imperative for ecosystem management and restoration in arid pastoral areas. Although ecosystem health studies have been conducted in various regions, research on ecological health assessment in arid pastoral areas of Central Asia remains limited, especially within a mountain-basin system (MBS). Taking the Chinese pastoral region of Fuyun County in Altay Prefecture, Xinjiang, as our study area, we improved an evaluation framework in terms of Vigor-Organization-Resilience-Service (VORS) based on remote sensing and GIS technology, and innovatively employed a deep learning method to assess the ecological system health status and spatiotemporal patterns for the years 2000, 2010, and 2020. Our results indicate the following: (1) The evaluation method based on deep learning can achieve a more comprehensive, efficient and objective evaluation of ecosystem health. (2) The ecosystem health index (EHI) ranged from 0.0759 to 0.5735 in 2000, 0.0798 to 0.5952 in 2010, and 0.0415 to 0.5657 in 2020. (3) The distribution pattern of EHI exhibited spatial heterogeneity, with the indexes decreasing from north to south. The highest proportion of high ecosystem health level (EHL) was found in summer pastures, followed by those in north of spring/autumn pastures and along the Irtysh and Ulungur rivers. On the other hand, the EHLs were relatively low in spring/autumn and winter pasture areas. The EHL exhibited minimal alteration between 2000 and 2010. From 2010 to 2020, areas experiencing an increase in EHL were primarily concentrated along the Ulungur River, with another part of the region focused on the interface between winter pastures and spring/autumn pastures. Based on the results, rotational grazing and grazing prohibition should be put into place for the winter pastures in the central and southern regions of the study area, as well as for the spring and autumn pastures. Meanwhile, it is recommended to plant premium, high-yield artificial forage grass alongside the Irtysh and Ulungur rivers. These findings will assist ecosystem managers in implementing effective measures to enhance ecosystem health in arid pastoral areas. Moreover, the presented evaluation framework and method enable a more objective, comprehensive, and efficient assessment of complex ecosystems, and they can be applied to ecosystem health assessments in similar regions.
environmental sciences