Calibration of the SWAT model using remote sensing based ET data of an intensively used and irrigated agricultural lowland area of Lombardy, Italy

Global changes are impacting water availability, which touch a wide range of human activities, especially agriculture. For this reason, hydrological models have been developed in recent years, which are an important support in the management of water resources.

The aim of this study is to setup and calibrate a hydrological model using remote sensing-based evapotranspiration (ET) data, in an area free of natural streams where irrigation channels are the only watercourses, expect for Ticino River. From a hydrological point of view, the study area is quite complex. Rainwater infiltrates into the permeable soils characterizing the area, while the rest of the precipitation leave the soil system through evapotranspiration. In fact, we noticed after periods of rain or irrigation a variation of the discharges of local springs located at the base of the fluvial terrace escarpments of the Ticino River. Moreover, being in a flat area the surface runoff component is almost nil, except for ponding that occur after precipitation or during the period in which the rice fields are flooded. During the spring-summer period, actually, large quantities of water are distributed through a complex network of channels to irrigate the rice and maize fields. So, water distributed for irrigation use is not only important for the agriculture, but also contributes to the recharge of the water table, which then feeds springs, forming a unique cascade system of water reuse that was already created in the15^th century. However, calibrating a spatially distributed hydrological model of an intensively irrigated and flat agricultural area is a difficult challenge. In this study the Soil Water Assessment Tool (SWAT) was applied, a physically based model used worldwide for soil and water management studies. The SUFI-2 program for model calibration and uncertainty analysis was utilized and Kling-Gupta Efficiency (KGE) was applied as objective function. In the calibration process we used ET data derived from MODIS sensor with a spatial resolution of 1 km².

The results show that despite the complexity of the area a calibration of the model with ET’s MODIS data yield a KGE of 0.59. The results indeed highlight that the model simulates well the hydrological dynamics of the area. Although there are some differences between observed and simulated data, due to a strong control of the hydrological dynamics by human activities, as well as the difference in model input data and satellite data used for calibration. Model validation through on-site measured soil water content, with 12 TEROS sensors installed on three different land uses, confirm the feasibility of using satellite data for SWAT model calibration in a complex area. Moreover, with these sensors we assessed the differences between the different crops and get information about the irrigation activities that modify the hydrological cycle of the area.

Finally, the calibrated and validated SWAT model allows for a further hydrological analysis of a system altered by human activities in terms of future scenarios. Particularly, we evaluate vertical soil water dynamics and assess the impact of land use change and land management (e.g., irrigation).