Soils are the largest terrestrial carbon pool and a valuable good that provides important ecosystem services. Since soils are threatened by degradation and in order to fight climate change the knowledge of the status quo especially of its soil organic carbon (SOC) content is required. A promising tool to map and monitor our soils are spaceborne imaging spectrometers which are able to produce up-to-date, inexpensive and spatially explicit maps. Especially the recent launch of new imaging spectroscopy sensors with a high signal-to-noise ratio opens up new possibilities. One of those is the combination of multitemporal spaceborne imaging spectroscopy data into SOC composite maps with a higher spatial coverage. This study explores different multitemporal combination workflows in order to support finding a best practice. To our knowledge for the first time, a spatially more complete SOC composite map was generated using four PRISMA images recorded over the same study site in northern Germany. Two different workflows of computation were compared: workflow one, creates a synthetical bare soil composite using averaged spectra as a basis for model development. Workflow two uses compositing after individual SOC modeling for each image. Within these workflows, different approaches were tested to estimate the SOC content, amongst them are a range of SOC spectral features and a two-step local PLSR which replaces the wet-chemistry SOC analyses for model calibration and validation by laboratory spectra and a large scale soil spectral library. Results show that the best method to produce a multitemporal composite SOC map based on imaging spectroscopy data was workflow two: the SOC maps composite, using the SOC spectral feature approach (R2 = 0.83, RPD = 2.42). While workflow two and the traditional PLSR approach was more robust for all input dates (R2 = 0.79, RPD = 2.21). Best results of the single images reached R2 values of 0.76-0.91 and RPD values ranging between 2.06-3.42. Three of the tested SOC spectral features provided accuracies comparable to the modeling approaches. These results are promising regarding the improvement of the spatial coverage and the refinement of the mapping and monitoring of SOC and other soil parameters. Further investigations in this direction are needed as they are precursors of what will be feasible by upcoming operational imaging spectroscopy missions with increased availability.

中文翻译：

---

使用多时相 PRISMA 成像光谱数据估算土壤有机碳


土壤是最大的陆地碳库，也是提供重要生态系统服务的宝贵商品。由于土壤受到退化的威胁，并且为了应对气候变化，需要了解现状，尤其是土壤有机碳（SOC）含量。星载成像光谱仪是一种很有前途的土壤测绘和监测工具，它能够生成最新、廉价且空间清晰的地图。特别是最近推出的具有高信噪比的新型成像光谱传感器开辟了新的可能性。其中之一是将多时相星载成像光谱数据组合成具有更高空间覆盖范围的 SOC 合成图。本研究探索了不同的多时相组合工作流程，以支持寻找最佳实践。据我们所知，首次使用在德国北部同一研究地点记录的四张 PRISMA 图像生成了空间上更完整的 SOC 合成图。比较了两种不同的计算工作流程：工作流程一，使用平均光谱作为模型开发的基础创建合成裸土复合材料。工作流程二在对每个图像进行单独 SOC 建模后使用合成。在这些工作流程中，测试了不同的方法来估计 SOC 含量，其中包括一系列 SOC 光谱特征和两步局部 PLSR，该两步局部 PLSR 取代了湿化学 SOC 分析，通过实验室光谱和大规模数据进行模型校准和验证土壤光谱库。结果表明，基于成像光谱数据生成多时相复合 SOC 图的最佳方法是工作流程二：使用 SOC 光谱特征方法合成 SOC 图（R2 = 0.83，RPD = 2.42）。 而工作流程二和传统 PLSR 方法对于所有输入日期都更加稳健（R2 = 0.79，RPD = 2.21）。单幅图像的最佳结果达到R2值0.76-0.91，RPD值范围2.06-3.42。三个测试的 SOC 光谱特征提供了与建模方法相当的精度。这些结果对于改善空间覆盖范围以及精细化 SOC 和其他土壤参数的绘图和监测来说是有希望的。需要在这个方向上进行进一步的研究，因为它们是即将到来的可操作性成像光谱任务的可行性的先驱。