Precision agriculture aims to optimize field management to increase agronomic yield, reduce environmental impact, and potentially foster soil carbon sequestration. In 2015, the Copernicus mission, with Sentinel-1 and -2, opened a new era by providing freely available high spatial and temporal resolution satellite data. Since then, many studies have been conducted to understand, monitor and improve agricultural systems. This paper presents results from the SolumScire project, focusing on the prediction of the spatial distribution of soil zones and topsoil properties, such as pH, soil organic matter (SOM) and clay content in agricultural fields through random forest algorithms. For this purpose, samples from 120 fields were investigated. The zoning and soil property prediction has an accuracy greater than 90%. This is supported by a high agreement of the derived zones with farmer’s observations. The trained models revealed a prediction accuracy of 94%, 89% and 96% for pH, SOM and clay content, respectively. The obtained models for soil properties can support precision field management, the improvement of soil sampling and fertilization strategies, and eventually the management of soil properties such as SOM.

中文翻译：

---

通过遥感了解田野：土壤分区和性质图

精准农业旨在优化田间管理，以增加农艺产量，减少环境影响并潜在地促进土壤碳固存。2015年，哥白尼任务与Sentinel-1和-2一起，通过提供免费可用的高时空分辨率卫星数据，开启了一个新时代。从那时起，进行了许多研究以了解，监测和改善农业系统。本文介绍了SolumScire项目的结果，重点是通过随机森林算法预测土壤区域和表土属性（例如pH值，土壤有机质（SOM）和农田中粘土含量）的空间分布。为此，对120个油田的样品进行了调查。分区和土壤属性预测的精度大于90％。这在派生区域与农民的观察结果高度吻合的情况下得到了支持。经过训练的模型显示出对pH，SOM和黏土含量的预测准确度分别为94％，89％和96％。所获得的土壤特性模型可以支持精确的田间管理，土壤采样和施肥策略的改进，并最终支持土壤特性（如SOM）的管理。