Integration of sentinel-2 data into the SIMETAW model for assessing irrigation water requirements and evapotranspiration

Water management for crop irrigation has become a major consideration given the global scale implications of climate change and its impacts on water scarcity and security, with an increasing frequency of water restrictions in many locations during the summer months. As a consequence, decision-support tools are needed to evaluate scenarios of various agricultural practices in order to provide improved water management strategies. In this study, water requirements of orchards were assessed at the plot scale using the Simulation of Evapo-Transpiration of Applied Water (SIMETAW) model. This model was applied across a small Mediterranean watershed to evaluate and compare various water use scenarios. The standard version of SIMETAW integrates a simplified water balance model with a single uniform soil layer to calculate the crop water needs at the plot scale (Quantity of Irrigation QI). Evapotranspiration is computed using a crop coefficient specific to each crop type (Kc) and a water stress coefficient (Ks) derived from the water available in the soil (SWC). In this standard version, plots are categorized by crop type, with fixed Kc unchanged across years and between plots of the same crop type. To better capture the spatial variability of crops at the watershed scale, SIMETAW was modified to incorporate remote sensing data. Kc for each crop in the basin was estimated using the relationship between normalized difference vegetation index (NDVI) and fractional vegetation cover (FCOVER) derived from Sentinel-2 images. The modified SIMETAW model was then compared to the standard version in estimating irrigation water requirements, from the field level to the watershed scale. Model calibration was undertaken using distributed soil moisture measurements collected from 2021 to 2024. Data on water volumes used for irrigation together with farmer surveys were used to assess model performance in simulating QI both at the plot level, for thirteen distributed farms, and at the basin scale. Comparisons between the standard model version and the modified version (i.e. using Kc from remote sensing) are examined in relation to the accuracy of obtained water volumes. The model simulated soil water content (SWC) across the monitored orchards with good accuracy, providing R² values of 0.7 and 0.9 for simulation in 2022 and 2023 respectively. The simulation of the quantity of irrigation water in farms shows a strong correlation with reported data (R= 0.81). The application of the SIMETAW model to all the studied crop types offered an improved performance by incorporating the remote sensed Kc compared against the standard version, with an 17 % improvement of the water irrigation volume distributed against basin averaged irrigation volumes. Overall, the incorporation of Sentinel-2 data significantly enhanced the performance of the model by accounting for the variability in crop development across the catchment scale. By considering a typology of farms according to the irrigation practices (3 classes were defined according to observations and surveys, high, medium and low irrigation), the model was able to estimate water requirements with 10 % difference to the values provided by the manager of the water resources at regional scale (“Association Syndicale Autorisée” - ASA). Such information can prove particularly useful in helping to guide water managers in making informed decisions with regards to the dynamic management of water resources.