Effect of soil moisture uncertainty on irrigation water use and farm profits

Feeding the planet sustainably requires a substantial increase in agricultural water productivity. Water managers and policymakers often view digital technologies and big data as key solutions for helping farmers to grow more food while reducing pressure on limited freshwater resources. Soil moisture probes, for example, could be used to improve the timing and efficiency of farmers’ irrigation management decisions. However, current adoption rates are low with most farmers, relying instead on the visual appearance of the crop or the feel of the soil to schedule irrigation decisions. These methods have potentially large uncertainties, which may lead farmers to schedule their irrigation sub-optimally. Despite the possible impact on water use and profits, little research to date has evaluated the effects of imperfect soil moisture information, and hence the value proposition to farmers and policy makers of investing in better information. 

 

In this study we investigate the effect of soil moisture uncertainty on irrigation water use and farm profits. We focus our analysis on a case study of irrigated maize production in Nebraska, USA. Nebraska has the second largest number of irrigated acres by state in the United States, with almost all that water being pumped up from the High Plains Aquifer (HPA). The HPA has seen large decreases in groundwater storage over recent decades, resulting in mounting pressure for more efficient irrigation practices. Using a crop-water model (AquaCrop-OS) in combination with a particle swarm optimisation algorithm, we define an optimal irrigation schedule - represented by a set of soil moisture thresholds - that maximise average profits over a 30-year historic weather period. Under this perfect-information strategy, we assess the impact on profits and water use of adding random errors to the water-flux and soil-texture inputs to the model. These random errors result in a divergence between the true water content and the farmer’s perception - potentially leading to irrigation being triggered too early or too late when compared with perfect information. 

 

Our results show that increasing levels of uncertainty lead to decreasing water-use efficiency and profits. However, we also find that the effect of increasing water-flux and soil-texture error is not linear, and that there is diminishing returns to further reductions in uncertainty below a standard error of 15%. In contrast, reductions in water-use efficiency and profits due to sub-optimal selection of irrigation management strategies are much larger. This implies that improving the quality of irrigation scheduling could have more impact on agricultural water productivity than solely improving the accuracy of soil-water information. Our findings highlight the need for further research to evaluate different methods of irrigation scheduling by using models and optimisation techniques to develop irrigation strategies that incorporate information uncertainty.