内蒙古自治区1991-2020年30米分辨率草原年度数据集

草原是一种分布广泛的土地利用类型，既是畜牧业生产的重要资源，也是生态系统的关键组成部分。过去几十年间，受气候变化和人类活动影响，草原显著退化和萎缩，其面积持续发生变化。本研究基于谷歌地球引擎（GEE）平台，采用Res-UNet++模型分析Landsat影像中草原的物候和光谱特征，实现了草原的长期年度监测；同时结合LandTrendr算法对草原时序数据进行校正，生成了1991-2020年内蒙古草原分布图。结果表明：1991-2020年间草原空间总体精度超过96%，Kappa系数达0.92以上，监测精度较高；其中2019年草原监测结果与第三次全国国土调查（TNLS）数据高度吻合，决定系数（R²）达0.97，体现了方法的准确性和可靠性。Grasslands are a widely distributed land use type, serving as both an important resource for livestock production and a critical component of terrestrial ecosystems. Over the past decades, grasslands have undergone significant degradation and shrinkage due to climate change and human activities, with their areas experiencing continuous changes. This study is based on the Google Earth Engine (GEE) platform, adopting the Res-UNet++ model to analyze the phenological and spectral characteristics of grasslands in Landsat imagery to achieve long-term annual monitoring of grasslands; meanwhile, it combines the LandTrendr algorithm to correct the grassland time-series data, generating distribution maps of Inner Mongolia grasslands spanning the period from 1991 to 2020. The results show that the overall spatial accuracy of grassland monitoring from 1991 to 2020 exceeds 96%, with the Kappa coefficient reaching above 0.92, indicating high monitoring precision; specifically, the 2019 grassland monitoring results are highly consistent with the data from the Third National Land Survey (TNLS), with the coefficient of determination (R²) reaching 0.97, which demonstrates the accuracy and reliability of the proposed method.

搜集汇总

背景与挑战

背景概述

该数据集提供了内蒙古自治区1991年至2020年每年度的30米分辨率草原分布栅格数据，基于Google Earth Engine平台和Landsat影像，采用Res-UNet++模型和LandTrendr算法生成，整体监测精度高（空间精度>96%，Kappa>0.92），并与第三次全国国土调查数据高度一致（R²=0.97），适用于长期草原变化分析和生态研究。