Neural Network-Based Soil Moisture Estimation Using Satellite SAR Data

Soil moisture measurement using satellite radar images for various kinds of crops is a subject widely described in the literature. Typical solutions presented concern the crops grown in areas with flat surfaces (e.g. wheat). However, some crop cultivation with uneven ground surfaces poses a totally different problem if satellite radar imaging is to be applied. The shape of the farmland surface can significantly affect the satellite radar cross-section e.g. rows usually formed in parallel ridges separated from each other by some distance. To investigate how the shape of the cultivation surface affects the RF-signal reflections, simulations were carried out. For verification purposes, real radar images (SAR) were obtained from the SENTINEL-1 satellite along with ground truth information about soil moisture from sensors on site.In this article, the authors present some preliminary results of research on the estimation of soil moisture by using neural networks based on satellite SAR radar data.