Comparison Of Coincident Hyperspectral Data From Satellite, Airborne And Fieldworks For Retrieving Water Quality Parameters In Lake Trasimeno

Nowadays, anthropogenic development and increasingly aggressive climate change phenomena are seriously affecting the stability of the world's ecosystems, especially aquatic systems. In this context, there is an increasing use of remote sensing techniques because thanks to the characteristics of satellite acquisitions, such as synoptic view and revisit time, it is possible to monitor the spatio-temporal dynamics of aquatic processes, which are usually highly variable in time and space (e.g., Tyler et al. 2016 Science of the Tot. Env., 572). In particular, the use of hyperspectral remote sensing provides measurements across several discrete narrow bands, forming a contiguous spectrum that allows the detection and identification of improved biophysical properties of the water column and bottom (e.g., Giardino et al. 2019 Surv Geophys, 40). This study focused on the analysis of hyperspectral images of Lake Trasimeno (a turbid shallow lake in central Italy) provided by PRISMA and DESIS sensors in early June 2021. In this period, a rich dataset was available for the comparison, with airborne, satellite and in-situ data characterized by different spectral and spatial resolutions: AVIRIS-NG hyperspectral data, 5 m pixels (acquired on June 4th); PRISMA and DESIS data, 30 m pixels (June 3rd and June 4th respectively); Sentinel-2 (S2) and Sentinel-3 (OLCI) multispectral data, respectively resampled at 10 m and 300 m pixels (June 4th and June 3rd); in-situ hyperspectral data from the WISPStation permanent measurement station and an ad-hoc radiometric and limnological campaign (June 4th). This dataset allowed us to assess the quality of products in terms of Remote sensing reflectance (Rrs) as well as the developing and testing of algorithms for retrieving water quality parameters. To perform the comparison in terms of Rrs, PRISMA and DESIS images were processed by using the atmospheric correction scheme implemented in the ATCOR software; moreover, because a run of AVIRIS-NG was affected by sunglint empirical methods for its removal were tested. PRISMA and DESIS spectra were comparable to AVIRIS-NG, S2, OLCI and in-situ observations and hence promising and encouraging a synergic use of imaging spectroscopy with the existing multispectral missions. Given the good match of Rrs spectra for the entire data sources and given the availability of in-situ measurements of water quality parameters - as well as radiometric measurements - the next step of the study is the generation of water quality products by testing different algorithms made available by the scientific community such as those based on bio-optical model (e.g., BOMBER, Giardino et al. 2012 Comput. Geosci., 45) having the knowledge of the Specific Inherent Optical Properties (IOPs) of the area of interest. Notably, with hyperspectral data there is the possibility to overcome the problem that band ratio algorithms are generally very sensitive to reflectance variability due to the influences of other water constituents, natural variability of IOPs and errors in atmospheric correction (Sterckx et al. 2007 Marine Geodesy, 30). This is done through the use of adaptive algorithms: the advantage of applying these algorithms to a hyperspectral image is that of not using the value of a single band in the equation, but rather, having different bands available in that spectral range, to search for the real maximum and minimum values by means of an algorithm. Once the water quality maps are generated by PRISMA, DESIS and AVIRIS-NG products, a comparison will be made with both in-situ measurements and products obtained from S2 and OLCI data which are already widely used for aquatic application.