Holistic Approach for Estimating Water Quality Ecosystem Services of Danube Floodplains: Field Measures, Remote Sensing, and Machine Learning

Abstract

Human pressure has caused river ecosystems to be severely damaged. To improve river ecosystems, “working with nature”, i.e., nature-based Solutions (NbS), should be supported. The purpose of this paper is to evaluate the effects of a specific NbS, i.e., floodplain restoration, which provides, among others, the ecosystem service of nutrient retention. For these, an in-depth time series analysis of different nutrients’ concentrations and water physiochemical parameters was performed to obtain Water Quality Indices (WQI), which were calculated along the river. To estimate water quality from remote sensing data and to generate water quality maps along the river, Sentinel-2 water products were validated against in situ data, and linear regression (LR), random forest (RF), and support vector regression (SVR) were trained with atmospherically corrected data for chlorophyll-a and TSM. The results show different outcomes in diverse floodplains in terms of improvement of the water quality downstream of the floodplains. RF demonstrated higher performance to model Chl-a, and LR demonstrated higher performance to model TSM. Based on this, we provide an insightful discussion about the benefits of NbS. These methodologies contribute to the evaluation of already existing NbS on the Danube River based on a quantitative analysis of the effects of floodplain ecosystems to water quality.