Assessing river water quality using Sentinel-2 products: the Po River case study

The Action Programme for the Renaturation of the Po River Area, developed by the District Basin Authority, is designed to support and accelerate the ecological transition through a comprehensive set of interventions. These include the enhancement of water resource management and the promotion of biodiversity across the riverine landscape. A key component of the programme is the morphological reconfiguration and ecological restoration of the entire fluvial corridor of the Po River - Italy’s longest watercourse - which has undergone significant anthropogenic alteration and structural modification, particularly over the past few decades. Despite such pressures, the river system retains high ecological and environmental value. Among the programme’s core activities is the continuous monitoring and assessment of the Po River’s ecological status, with a particular focus on water quality parameters as indicators of ecosystem health. Water quality was assessed using both in situ biogeochemical measurements and remote sensing data derived from Sentinel-2 (S2) satellite imagery. A total of 148 images from 2018, 145 from 2022, and 140 from 2024 were selected and downloaded. All images were resampled to a spatial resolution of 10 meters and processed using the ACOLITE v.2025 software (Vanhellemont & Ruddick, 2016), which performs atmospheric correction. In its latest version, ACOLITE also implements the RAdCor algorithm (Castagna & Vanhellemont, under review) to correct for adjacency effects. The software automatically masks water pixels and generates Remote Sensing Reflectance (Rrs) products, i.e., atmospherically corrected surface reflectances. These Rrs outputs serve as the basis for deriving thematic maps of various water quality parameters. Water pixels were identified and masked using the Water Adjusted Vegetation Index (WAVI; Villa et al., 2014). From the resulting masked Rrs images, three key water quality products were derived: Total Suspended Solids (TSS), Turbidity, and the Trophic State Index. TSS concentrations were estimated using the SPM_Novoa_2017 algorithm (Novoa et al., 2017), while Turbidity values were calculated with the TUR_Novoa_2017 algorithm (Novoa et al., 2017). The Trophic State Index was derived from chlorophyll-a (Chl-a) concentration maps, which were generated using the Chl_re_Mishra algorithm (Mishra & Mishra, 2012). These maps were then converted into trophic state classes based on the criteria defined by the Italian Phytoplankton Assessment Method (IPAM), as outlined in the updated classification standards of Decision 2018/229/EU. Specifically, we applied the classification thresholds established for type L4 water bodies—characterized as shallow, polymictic, and lacking stable stratification. The chlorophyll-a (Chl-a) concentration thresholds used to define trophic status were: high (< 4.4 μg L⁻¹), good (4.4–8 μg L⁻¹), moderate (8–14.6 μg L⁻¹), poor (14.6–26.7 μg L⁻¹), and bad (> 26.7 μg L⁻¹), as defined in Table 4.2.1/c - RQE class limits for the annual mean concentration of Chl-a (update Dec. 260/2010; from the “BV_indices_phytoplankton_2016” spreadsheet). The Total Suspended Matter (TSM) and Trophic State Index products were validated against 26 in situ sampling points collected across different dates. Validation results showed strong agreement between satellite-derived and in situ observations, with coefficients of determination (R²) exceeding 0.8 and root mean square errors (RMSE) below 10 for both TSM and trophic status. Seasonal analysis revealed that the highest concentrations of Total Suspended Solids (TSS) occurred during spring, coinciding with periods of increased precipitation, while the lowest values were recorded following extended dry periods. From a spatial perspective, turbidity levels were consistently higher in downstream areas near the river mouth. The highest trophic state values were observed in oxbow lakes and lateral zones characterized by limited water exchange and stagnation.