Random forest code for downscaling county-level livestock statistics to 1 km resolution using environmental factors

We introduce a random forest–based framework designed to downscale county-level livestock census data to a fine spatial resolution of 1 km by integrating key environmental factors (e.g., climate, topography, and vegetation indices). This approach addresses the spatial mismatch between administrative statistics and ecological processes, thereby providing more accurate and ecologically meaningful estimates of livestock distribution. We provide code for relevance analysis to screen environmental variables as well as code for the random forest model. The county-level livestock census data for the Qinghai-Tibet Plateau from 2000 and 2020 used to build the random forest model, along with the corresponding environmental factor tables, are also provided.本研究提出一种基于随机森林（Random Forest）的降尺度框架，通过整合气候、地形、植被指数（Vegetation Indices）等关键环境因子，将县级畜牧普查数据降尺度至1千米的精细空间分辨率。该方法解决了行政统计数据与生态过程之间的空间错配问题，可提供更为精准且具备生态学意义的牲畜分布估算结果。 本数据集提供了用于环境变量筛选的相关性分析代码，以及随机森林模型代码。同时还提供了用于构建该随机森林模型的2000年与2020年青藏高原（Qinghai-Tibet Plateau）县级畜牧普查数据，以及对应的环境因子表。