A dataset mapping the temperate savanna in Northeastern China (2022)

Temperate savanna shares a wide geographical distribution. It plays a crucial role in supporting a substantial portion of rangelands and livestock, thereby contributing significantly to human livelihoods.We develop a new temperate savanna identification method for mapping the temperate savanna in Northeastern China within a minimum identifiable area of 5000 m² by integrating Unmanned Aerial Vehicle(UAV) imagery with high-resolution satellite imagery of Sentinel-1 and Sentinel-2 by using Random Forest regression and Classification and Regression Tree algorithms. The training and validation datasetswere collected from UAV imagery.The proposed method achieved with the overall accuracy of 0.82 for savanna identification.Temperate savanna in Northeastern China covering a total area of 170,000 km².This map enhances our understanding of the spatial extent and area of temperate savanna. The derived method provides a new insight to explore the extent of regional and global savannain the future.The code and generated maps are publicly accessible. To make the method to reproduce and exploit and to promote its usage, we have provided users with the python code used to produce temperate savanna map. The generated maps of fractional woody vegetation cover, fractional herbaceous vegetation cover and temperate savanna in Northeastern China are stored in TIFF format, including the fractional woody and herbaceous vegetation cover map of 20 m with the file name of FWVC20m.tif and FHVC20m.tif, and the temperate savanna map at 70.7 m (70.7 m ×70.7 m equivalent to unit area scale of 5000 m²) with the file name of tsavanna71m.tif. In temperate savanna map, we recorded temperate savanna as “1”. All image data can be viewed by geographic information system software, such as ArcGIS.The UAV imagery dataset consists of 75 TIFF files that are publicly accessible through Figshare (https://doi.org/10.6084/m9.figshare.27651468.v1).

温带稀树草原（temperate savanna）具有广泛的地理分布范围，可为大量牧场与牲畜提供支撑，对人类生计具有重要意义。本研究提出一种全新的温带稀树草原识别方法，通过整合无人机（Unmanned Aerial Vehicle, UAV）影像与哨兵1号（Sentinel-1）、哨兵2号（Sentinel-2）高分辨率卫星影像，结合随机森林回归（Random Forest regression）与分类与回归树（Classification and Regression Tree）算法，实现中国东北温带稀树草原的制图，其最小可识别面积达5000平方米。训练与验证数据集均采集自无人机影像。所提方法在稀树草原识别任务中总体精度达0.82。中国东北温带稀树草原总面积达17万平方千米。本研究生成的温带稀树草原分布图，有助于深化对温带稀树草原空间分布范围与面积的认知。该方法为未来开展区域乃至全球尺度的稀树草原范围探索提供了全新思路。本研究的代码与生成的地图均公开可获取。为便于方法复现、应用与推广，我们已向使用者提供用于生成温带稀树草原分布图的Python代码。本研究生成的中国东北木质植被覆盖度图、草本植被覆盖度图以及温带稀树草原分布图均以TIFF格式存储：其中20米分辨率的木质植被覆盖度图与草本植被覆盖度图分别命名为FWVC20m.tif与FHVC20m.tif；70.7米分辨率（70.7米×70.7米对应5000平方米的单元面积尺度）的温带稀树草原分布图命名为tsavanna71m.tif。在温带稀树草原分布图中，温带稀树草原区域的像素值标记为"1"。所有影像数据均可通过ArcGIS等地理信息系统软件查看。本研究的无人机影像数据集包含75个TIFF文件，可通过Figshare平台（https://doi.org/10.6084/m9.figshare.27651468.v1）公开获取。