Soil and vegetation temperatures are crucial for various fields, including ecology, agriculture, and climate change. However, there remains a lack of entirely observation-based global datasets for these two component temperatures. To fill this gap, this study developed a multisource data Fusion-based global surface Soil and Vegetation Temperature retrieval method (FuSVeT). This novel method not only utilizes temporal and spatial information from MODIS data by adopting a temperature cycle model to capture temporal variation and using adjacent pixels to consider spatial differences and increase the number of equations solved, but also leverages ERA5-Land data to reduce unknown parameters, effectively compensating for the limitations of satellite observations. Its performances were comprehensively evaluated with simulated data, high-resolution satellite products, and in situ measurements, demonstrating competitive accuracy with root mean square errors below 2 K and Biases of under 1 K in most cases. Compared to previous retrieval method that relies solely on satellite-based temporal and spatial information, FuSVeT present enhanced accuracy, more complete spatial coverage, and improved computational efficiency, making it more applicable for global soil and vegetation temperature mapping. Using this method, we generated global 0.05° monthly mean soil and vegetation temperatures for January and July 2020. These data can capture more pronounced temperature heterogeneities within biomes than existing soil temperature products, indicating its superiority for global analyses. Importantly, FuSVeT can also be applied to satellite observations with higher spatiotemporal resolution, holding significant potential for providing accurate, long-term, global maps of surface soil and vegetation temperatures.

中文翻译：

---

基于多源数据融合的全球表层土壤和植被温度反演


土壤和植被温度对各个领域都至关重要，包括生态学、农业和气候变化。然而，对于这两个分量温度，仍然缺乏完全基于观测的全球数据集。为了填补这一空白，本研究开发了一种基于多源数据融合的全球表层土壤和植被温度检索方法 （FuSVeT）。这种新方法不仅利用了 MODIS 数据中的时间和空间信息，采用温度周期模型来捕获时间变化，并使用相邻像素来考虑空间差异并增加求解方程的数量，而且还利用 ERA5-Land 数据来减少未知参数，有效补偿了卫星观测的局限性。通过模拟数据、高分辨率卫星产品和原位测量对其性能进行了全面评估，在大多数情况下，均方根误差低于 2 K，偏差低于 1 K，表现出具有竞争力的准确性。与以往仅依赖卫星时空信息的检索方法相比，FuSVeT 具有更高的精度、更完整的空间覆盖和更高的计算效率，使其更适用于全球土壤和植被温度制图。使用这种方法，我们计算了 2020 年 1 月和 7 月的全球土壤和植被月平均温度为 0.05°。与现有的土壤温度产品相比，这些数据可以捕捉到生物群落内更明显的温度异质性，表明其在全局分析中的优越性。 重要的是，FuSVeT 还可以应用于具有更高时空分辨率的卫星观测，在提供准确、长期、全球的表层土壤和植被温度地图方面具有重要潜力。