Smart irrigation using novel cosmic ray neutron sensors and land-surface modelling approaches

Innovative soil moisture (SM) monitoring and modelling methods are key to reduce irrigation water use in the face of expected water scarcity and increase of droughts related to climate change. A promising irrigation monitoring method is Cosmic-Ray Neutron Sensing (CRNS), which is based on the negative correlation between fast neutrons originating from Cosmic-Ray neutron intensities and SM content. The CRNS key advantage lies in its relatively large sensing volume of several hectares, which allows to use a single CRNS instead of a network of point-scale sensors. Additionally, land surface models such as the Community Land Model (CLM5) that simulate the exchange of water, energy, carbon and nitrogen at the land–atmosphere interface can be a valuable tool to study the efficiency of irrigation and effects on crop growth. In this study, novel CRNS and the newly developed CLM5-FruitTree were tested in two small (~1.2 ha) irrigated apple orchards located in the Pinios Hydrologic Observatory (Greece). In 2020, a climate station (Atmos21) and a network of 12 SoilNet nodes, each with two SM sensors at 5, 20 and 50 cm depth, were installed in each field, as well as water meters to measure irrigation timing and amounts. In addition, a CRNS was installed in each field to test the possibility of monitoring irrigation and informing irrigation models. We found that the CRNS was very sensitive to the weekly irrigation events. However, the magnitude of the SM fluctuations caused by the irrigation was underestimated by the CRNS resulting in an RMSE of up to 0.058 cm³ cm^-3. This can be attributed to the fact that the CRNS has a large footprint, and the neutron counts were therefore also influenced by the surroundings of the irrigated field. Therefore, to compensate for this influence, an additional SoilNet node was installed outside one of the two irrigated fields in 2022. By combining these data with neutron transport simulations of the study area, a correction of CRNS-derived SM was developed to better capture both timing and magnitude of SM changes (RMSE reduced to 0.031 cm³ cm^-3). In parallel, CLM5-FruitTree was able to reproduce the observed SM response to irrigation when the local irrigation schedule was considered (i.e., defining starting date, timing, and target soil moisture for irrigation). Interestingly, the simulated irrigation in 2021 and 2022 used 10 to 60 % less water than the amount applied by the farmer. This suggests a great water saving potential through a reduction in irrigation amounts or through improvements in irrigation efficiency by reducing losses through evaporation or deep percolation. However, existing model weaknesses in the representation of soil properties and water fluxes need to be further addressed for this modelling approach. Nevertheless, the results of this study are a further step towards the use of novel CRNS and modelling tools as a decision support system in irrigation for more efficient use of water resources.