Characterization of spatio-temporal patterns of grassland utilization intensity in the Selinco watershed of the Qinghai-Tibetan Plateau from 2001 to 2019 based on multisource remote sensing and artificial intelligence algorithms. Issue 1 (31st December 2022)

• Record Type: Journal Article
• Authors: Ma, Changhui; Xie, Yaowen; Duan, Si-Bo; Qin, Wenhua; Guo, Zecheng; Xi, Guilin; Zhang, Xueyuan; Bie, Qiang; Duan, Hanming; He, Lei
• Is Part Of: GIScience & remote sensing. Volume 59:Issue 1(2022)
• Page Start: 2217
• Page End: 2246
• Publication Date: 2022-12-31
• Subjects: Grassland utilization intensity -- spatial quantification method -- artificial intelligence algorithms -- spatio-temporal patterns -- Qinghai-Tibetan Plateau -- nature reserve; Geodesy -- Periodicals; Cartography -- Periodicals; Aerial photogrammetry -- Periodicals; Remote sensing -- Periodicals; 526.05
• DOI: 10.1080/15481603.2022.2153447
• Languages: English
• ISSNs: 1548-1603

Abstract

ABSTRACT: Due to the limitations of spatial quantification methods, the spatio-temporal patterns of grassland utilization intensity (GUI) in the Selinco watershed (SLCW), the core region of ecological security on the Qinghai-Tibetan Plateau, is unclear under multiple utilization modes. This paper quantified GUI by constructing the association between the potential and actual Enhanced Vegetation Index (EVI) of grasslands in terms of interannual variability. To obtain an accurate spatio-temporal dataset of potential EVI, the following two components were considered on. Firstly, the temporal lag effects of each raw climate factor were investigated to determine the optimal climate variables affecting vegetation productivity. Secondly, four machine learning (ML) algorithms, including an artificial neural network, random forest, support vector machine, and gradient boosting regression tree combined with the Bayesian model average, were used to construct grassland potential EVI models involving EVI, grassland type, and environmental factors (topography, soil, raw climate, and bioclimatic). Meanwhile, to maximize the performance of ML models, variable selection, variable transformation, and hyperparameter optimization were systematically implemented, where the hyperparameter optimization algorithms employ the grid search algorithm, Bayesian optimization, genetic algorithm, and particle swarm optimization. Then, the spatio-temporal dataset of GUI in the SLCW from 2001 to 2019 was established by using the above quantification method based on multisource remote sensing and artificial intelligence algorithms. The analysis of spatio-temporal variation in GUI showed that the implementation of ecological restoration projects leads to a significant and rapid decline in the overall GUI of the SLCW after 2010 (declining by 4.8%), which is more obvious in the non-nature reserve (declining by 9.3%). In the Qiangtang Nature Reserve within the SLCW, although the GUI shows a declining trend after 2010 because of the implementation of ecological restoration projects, it shows an insignificant increase from 2001 to 2019 due to the recovery increase of wildlife populations in recent decades. Besides, by exploring the effects of elevation and slope on the GUI, it is found that grasslands on higher slopes at lower elevations are at a greater risk of degradation due to more intensive grassland utilization. Graphical abstract: uf0001