Global Pasture Watch - Annual short vegetation height maps at 30-m spatial resolution (2000—2022)
收藏资源简介:
Sub-dataset: Median vegetation height, 2000—2001 Description Global annual maps of median vegetation height for 2000—2022 produced within the scope of the Global Pasture Wath initiative, integrating multi-sensor remote sensing data (LIDAR ICESat-2, Landsat, MODIS and DTM) and based on spatiotemporal Machine Learning (Gradient boosting trees). While the primary focus is on improving monitoring in short vegetation ecosystems, the dataset provides wall-to-wall coverage to all terrestrial ecosystems and is organized in 70 global mosaics in COG (Cloud Optimized GeoTIFF) format, WGS84 Coordinate Systems (EPSG:4326) and pixel size equal to 0.00025 (STAC and GEE) and 0.00075 degrees (Zenodo), including: Median vegetation height: Mean predicted value: 23 annual maps with values range from 0–20 (embedded scale factor 0.1) Percentile 5th value (upper prediction boundary): 23 annual maps with values range from 0–20 (embedded scale factor 0.1) Percentile 95th value (lower prediction boundary): 23 annual maps with values range from 0–20 (embedded scale factor 0.1) Trend analysis (2000—2022): Linear regression intercept (filtered p-value < 0.05, Mann-Kendall test): 1 long-term map with values range from -0.92–0.7 (embedded scale factor 0.0001) All raster files are in 16-bit integer format and use -3200 as no-data value (pixels covering ocean waters, lakes and water bodie), following an specific naming convention: Project name: Global Pasture Watch (gpw) Class name: Short vegetation height (short.veg.height) Procedure combination: Ensemble Gradient Boosting Trees (egbt), trend/intercept derived per-pixel via linear regression (trend), Mann-Kendall test (mk). Variable type: mean predicted value (m), 5th (p.05) and 95th percentiles (p.95) Spatial resolution: 30m Begin of time reference: date of first Landsat composite used by the modeling (20220101) End of time reference: date of last Landsat composite used by the modeling (20221231) Spatial extent: global (go) Coordinate system: World Geodetic System 1984, used in GPS (epsg.4326) Version: v1 Related resources Maps of median vegetation height:2000-2001 2002-2003 2004-2005 2006-2007 2008-2009 2010-2011 2012-2013 2014-2015 2016-2017 2018-2019 2020-2021 2022 Trend analysis map:2000—2022 Global reference samples and machine learning models:Parquet and joblib python files
子数据集:2000—2001年植被中位高度 数据集说明 本数据集为「全球牧场监测(Global Pasture Watch)」计划产出的2000—2022年全球年度植被中位高度图,整合了多传感器遥感数据(激光雷达ICESat-2、Landsat、MODIS与数字地形模型DTM),并基于时空机器学习(梯度提升树)构建。尽管其核心目标为优化短植被生态系统的监测工作,本数据集仍实现了所有陆地生态系统的全覆盖,以70幅全球镶嵌图的形式组织,采用云优化地理TIFF(Cloud Optimized GeoTIFF,COG)格式,坐标系为WGS84(EPSG:4326),像素分辨率在STAC与GEE平台为0.00025度,在Zenodo平台为0.00075度,包含以下内容: 一、植被中位高度相关指标: 1. 预测均值:共23幅年度地图,数值范围为0–20(内置缩放因子0.1) 2. 第5百分位值(预测上边界):共23幅年度地图,数值范围为0–20(内置缩放因子0.1) 3. 第95百分位值(预测下边界):共23幅年度地图,数值范围为0–20(内置缩放因子0.1) 二、2000—2022年趋势分析: 线性回归截距(经Mann-Kendall检验,过滤p值<0.05):共1幅长期地图,数值范围为-0.92–0.7(内置缩放因子0.0001) 所有栅格文件均采用16位整数格式,以-3200作为无数据值(对应海洋、湖泊及其他水体覆盖的像素),并遵循特定命名规范: 项目名称:Global Pasture Watch(缩写gpw) 类别名称:短植被高度(short.veg.height) 处理组合:集成梯度提升树(egbt)、逐像素线性回归衍生趋势/截距(trend)、Mann-Kendall检验(mk) 变量类型:预测均值(m)、第5百分位(p.05)与第95百分位(p.95) 空间分辨率:30米 时间参考起始:建模所用首幅Landsat合成影像日期(20220101) 时间参考结束:建模所用末幅Landsat合成影像日期(20221231) 空间范围:全球(go) 坐标系:全球大地测量系统1984,用于GPS定位(EPSG:4326) 版本:v1 相关资源 植被中位高度图:2000-2001、2002-2003、2004-2005、2006-2007、2008-2009、2010-2011、2012-2013、2014-2015、2016-2017、2018-2019、2020-2021、2022 趋势分析图:2000—2022 全球参考样本与机器学习模型:Parquet与Python joblib格式文件



