Plant species richness across grasslands in Germany

This geospatial dataset provides an estimation of plant species richness in grasslands across Germany for 2021. Each pixel covers an area of 10 by 10 m and represents the estimated number of herbaceous species per 16m2. It is derived using Sentinel-2 time series from April to October 2021, and a multilayer perceptron model recalibrted following Muro et al. (2022). Two raster files are provided. In the first version the model has been applied to all pixels of Germany, weather grasslands or not, so that users can mask it to their own grassland map. In the second version, the model has been applied only to pixels that are within the feature space of the training dataset, limiting the analysis to (most) grasslands in Germany. Values range between 1 and 70 species per 16m2. The estimated accuracy for these areas has a relative root squared error of 0.28 and a root mean squared error of ±8 species

本地理空间数据集针对2021年德国全境草原的植物物种丰富度开展估算。每个像素覆盖10×10米的区域，代表每16平方米内的草本植物物种估计数量。该数据集基于2021年4月至10月的Sentinel-2时间序列影像，并参照Muro等人（2022）的研究重新校准的多层感知器（Multilayer Perceptron）模型生成。本次共提供两份栅格文件：第一版将模型应用于德国全境所有像素，无论该像素是否为草原，以便用户可基于自身的草原地图进行掩膜处理；第二版仅将模型应用于落在训练数据集特征空间内的像素，将分析范围限定在德国境内的（绝大多数）草原区域。每16平方米的物种估算值介于1至70种之间。上述区域的估算精度为相对均方根误差0.28，均方根误差为±8种。