Linking soil moisture sensors and crop models for irrigation management

A number of challenges must be faced when using soil moisture sensors, such as accounting for soil heterogeneity in measurements or dealing with sensor faults. As a consequence, it is difficult to obtain reliable estimations of the water status in the root zone and using sensor data for irrigation planning is not straightforward. In this work, a method is proposed to interpret soil water content measurements that is based on the use of a model to correct and complement sensor data, in particular in the case of a non-uniform water distribution. This approach relies on the decomposition of the sensor's signal in two factors, one space dependent, accounting for heterogeneity in hydraulic properties at small scales and assumed to represent porosity, and the other factor accounting for the time variation of soil water status. With practical applications in mind, a simple model and an efficient calibration procedure are developed, in particular considering the online application of the method to complement sensor data in real time. The capabilities of the model are illustrated with data from experiments on the growth of lettuce in greenhouses with reclaimed wastewater irrigation. Requiring only a short calibration period, the model is successfully validated, with a mean relative calibration error of 5.85% and a mean relative validation error of 6.8% for the soil water content data, and is proven to be a valuable tool to correct for sensor malfunctions. Moreover, the proposed method is shown to allow the meaningful estimation of the water status of the soil crop system, in particular when measurements of sensors positioned close to each other showed important differences.