Simulating and predicting soil water content by combining soil water balance calculations, weather forecasts and soil sensors with inverse modelling for optimal irrigation advice: A case study in Flanders, 2022

Abstract

In Flanders, a cumulative precipitation deficit of no less than 330 mm was calculated during the growing season of 2022 (April - September) (Soil Service of Belgium). This high precipitation deficit reflects the importance and need of additional water supply to meet the water demand and the yield potential of the crop. However, this additional water must be administered as efficiently as possible to avoid water waste, while maximizing yields. For decades, the Soil Service of Belgium already offers paid irrigation advice based on simulations with a soil water balance model calibrated with manual soil samples, and weather data, while considering weather predictions separately. With the rise of affordable, autonomous sensors and IoT (Internet-of-Things) technology, it is possible to monitor the soil moisture in a field online and in real time. The use of these sensors offers opportunities such as data accessibility, model calibration, and optimization of irrigation advice.Soil moisture model simulations and forecasts alone may be less accurate than in situ soil moisture measurements. However, soil moisture forecasts make it possible to anticipate drought or precipitation forecasts, which makes it easier to plan irrigation in advance. Sensor data alone fall short in this respect, as sensors only provide data on the previous and current soil moisture content, but do not provide information on future soil moisture development. Both approaches can be combined by calibrating the model with sensor data via inverse modelling. In this study, DREAM is used as inverse modelling approach to estimate model parameters, including soil and crop growth parameters, as well as their uncertainty. These parameter distributions result in soil moisture simulations, and, when inserting weather forecasts, predictions, along with their uncertainty. The uncertainty of the calibrated model simulations can be used to determine the probability of the soil moisture dropping below the critical water stress threshold.When this combined approach is compared to the irrigation advice based on a model alone, the soil moisture is simulated and predicted more accurately, resulting in a more efficient water application, while the crop experiences less stress. In the dry growing season of 2022, for example, a celery trial in Flanders (Research Station for Vegetable Production) saved about 45 mm (21%) of water without sacrificing crop quality and yield. In addition to irrigation yield responses, the approach is also validated in light of parameter estimation, and soil moisture simulations, comparing simulated and measured soil moisture content.