Pakistan Desertification Distribution Dataset, 1990-2020

Desertification is one of the most serious ecological and environmental problems in arid and semi-arid regions worldwide. Pakistan, located in South Asia, has long faced challenges such as desertification and its resulting consequences, including declining agricultural yields, water scarcity, food security crises, and loss of biodiversity. The lack of baseline and change data on desertification in this region directly impacts the sustainable development of the China-Pakistan Economic Corridor. This study selected five desertification indicators: fractional vegetation cover (FVC), normalized difference vegetation index (NDVI), modified soil-adjusted vegetation index (MSAVI), surface albedo, and topsoil granularity index (TGSI). Through comparative evaluation, the optimal machine learning model (gradient boosting decision tree) was obtained. Based on Google Earth Engine and using Landsat remote sensing images, a dataset of Pakistan's desertification distribution with a 30 m spatial resolution was obtained for every five years from 1990 to 2020. The quality and accuracy of this dataset were verified using high-resolution Google Earth images, resulting in an overall accuracy of 98.41% and a Kappa coefficient of 0.97. This dataset objectively reflects the spatial distribution of different degrees of desertification in Pakistan and can provide detailed and reliable data support for delineating and restoring key areas for desertification control, which is of great significance for decision-making regarding the green and sustainable development of the China-Pakistan Economic Corridor.

荒漠化是全球干旱与半干旱地区最严峻的生态环境问题之一。地处南亚的巴基斯坦长期面临荒漠化及其衍生挑战，包括农业减产、水资源短缺、粮食安全危机与生物多样性丧失。该区域荒漠化基准数据与动态变化数据的缺失，直接制约了中巴经济走廊的可持续发展。
本研究选取5项荒漠化表征指标：植被覆盖度（fractional vegetation cover, FVC）、归一化差分植被指数（normalized difference vegetation index, NDVI）、修正土壤调整植被指数（modified soil-adjusted vegetation index, MSAVI）、地表反照率以及表层土壤粒度指数（topsoil granularity index, TGSI）。通过对比评估，最终获得最优机器学习模型：梯度提升决策树（gradient boosting decision tree）。本研究基于谷歌地球引擎（Google Earth Engine）平台，结合Landsat遥感影像，生成了1990年至2020年期间每5年一期、空间分辨率为30米的巴基斯坦荒漠化分布数据集。
本研究采用高分辨率谷歌地球影像对该数据集的质量与精度进行验证，最终整体准确率达98.41%，Kappa系数为0.97。
该数据集客观反映了巴基斯坦不同等级荒漠化的空间分布特征，可为荒漠化防治重点区域的划定与生态修复提供详实可靠的数据支撑，对中巴经济走廊绿色可持续发展的决策制定具有重要意义。