Data_Sheet_1_Dynamic Disturbance Analysis of Grasslands Using Neural Networks and Spatiotemporal Indices Fusion on the Qinghai-Tibet Plateau.PDF

Grassland is the vegetation type with the widest coverage on the Qinghai-Tibet Plateau. Under the influence of multiple factors, such as global climate change and human activities, grassland is undergoing temporal and spatially different disturbances and changes, and they have a significant impact on the grassland ecosystem of the Qinghai-Tibet Plateau. Therefore, timely and dynamic monitoring of grassland disturbances and distinguishing the reasons for the changes are essential for ecological understanding and management. The purpose of this research is to propose a knowledge-based strategy to realize grassland dynamic distribution mapping and analysis of grassland disturbance changes in the region that are suitable for the Qinghai-Tibet Plateau. The purpose of this study is to propose an analysis algorithm that uses first annual mapping and then establishes temporal disturbance rules, which is applicable to the integrated exploration of disturbance changes in highland-type grasslands. The characteristic indexes of greenness and disturbance indices in the growing period were constructed and integrated with deep neural network learning to dynamically map the grassland for many years. The overall accuracy of grassland mapping was 94.11% and that of Kappa was 0.845. The results show that the area of grassland increased by 11.18% from 2001 to 2017. Then, the grassland disturbance change analysis method is proposed in monitoring the grassland distribution range, and it is found that the area of grassland with significant disturbance change accounts for 10.86% of the total area of the Qinghai-Tibet Plateau, and the disturbance changes are specifically divided into seven types. Among them, the type of degradation after disturbance mainly occurs in Tibet, whereas the main types of vegetation greenness increase in Qinghai and Gansu. At the same time, the study finds that climate change, altitude, and human grazing activities are the main factors affecting grassland disturbance changes in the Qinghai-Tibet Plateau, and there are spatial differences.

青藏高原上，草原作为分布最广的植被类型，正受到诸如全球气候变化和人类活动等多重因素的影响。这些因素导致草原经历着时空差异性的干扰和变化，对青藏高原的草原生态系统产生了显著影响。因此，对草原干扰的及时动态监测以及区分变化原因对于生态理解和管理的至关重要。本研究旨在提出一种基于知识的策略，以实现适用于青藏高原地区的草原动态分布制图和草原干扰变化分析。研究目的在于提出一种分析算法，该算法首先进行年度映射，然后建立时间干扰规则，适用于高原型草原干扰变化的综合探索。在生长季节，构建了绿色度和干扰指数的特征指标，并将其与深度神经网络学习相结合，以动态映射多年草原。草原制图的总体准确率为94.11%，Kappa系数为0.845。结果表明，从2001年到2017年，草原面积增加了11.18%。随后，在监测草原分布范围时，提出了草原干扰变化分析方法，发现发生显著干扰变化的草原面积占青藏高原总面积的10.86%，干扰变化具体分为七种类型。其中，干扰后的退化类型主要发生在西藏，而植被绿色度增加的主要类型则见于青海和甘肃。同时，研究还发现气候变化、海拔和人类放牧活动是影响青藏高原草原干扰变化的主要因素，且存在空间差异。