SoilMoistureMapper: a GNSS-R approach for soil moisture retrieval on UAV

volkan yusuf senyurek, Mehedi Farhad, Ali Cafer Gurbuz, Mehmet Kurum, Robert Moorhead

Keywords: GNSS-R, Soil moisture, Random forest, UAV

TL;DR: A practical and low-cost soil moisture mapping techniques based on Global Navigation Satellite System (GNSS) Reflectometry (GNSS-R) observations via an unmanned-aerial vehicle (UAV).

Abstract: Measuring of distribution of the soil moisture (SM) content is an essential requirement in precision agriculture. This paper demonstrates practical and low-cost soil moisture mapping techniques based on Global Navigation Satellite System (GNSS) Reflectometry (GNSS-R) observations via a small-size unmanned-aerial vehicle (UAV). An SM estimation model is developed using a random forest (RF) machine-learning (ML) algorithm combining GNSS-R signals with ancillary vegetation indices from a multispectral camera. The ML model is trained and tested using in-situ data from eight SM probes located in a 2.48ha farm. The study results showed that SM maps of the field can be obtained with about 13 mins flight with 5m $\times$ 5m spatial resolution. The developed ML model reached RMSE of 0.032$m^{3}m^{-3}$ and R-value of 0.93 in 10-fold cross-validation.