Estimation of soil moisture from Rongowai GNSS-R using machine learning

The Rongowai mission comprises of a next-generation receiver for Global Navigation Satellite System Reflectometry (GNSS-R), mounted on a domestic Air New Zealand Q300 aircraft. During routine operations, the receiver autonomously records GNSS signals which are reflected from New Zealand’s varied landscapes and change depending on surface properties. We report early progress towards development of algorithms for estimating soil moisture from Rongowai data, through a data assimilation and machine learning approach using a Random Forest algorithm. Initial results show reasonable accuracy (r
²
= 0.75; RMSE 6.6%), with good scope for improvement as additional data are collected through the planned 8-year mission.