An innovative data driven approach improves drought impact analysis using earth observation data

Drought is a devastating natural hazard that can be of diverse magnitude, duration and intensity. It leads to economic and social losses and ecological imbalances. Ascribing to climate change, drought has occurred more frequently with high intensity worldwide in recent decades, such as the striking droughts in the summer of year 2022. In water resource aspect, one direct consequence of drought is the decrease of water amount in the rivers, which could further develop into water shortage for irrigation and drinking water supply, and cargo shipping disruption. Therefore, in order to make management decisions that help mitigate the drought damage, it is important to monitor river water anomalies and identify the vulnerable shrinking sections along the river network. Traditional river gauging stations only provide us limited observations of particular spots. A proper utilization of spatially distributed remote sensing data is necessary and effective. In this work, we develop a novel framework to monitor river water shrinking anomaly by including image processing and machine learning approaches, based on earth observation data. The Rhine, a major cargo-route river, is selected as the pilot case, because it had huge water decrease and caused shipping disruption during the 2022 summer’s drought in Germany. The Modified Normalized Difference Water Index (MNDWI) is calculated from the green and Shortwave-Infrared bands of Sentinel-2 satellite images.  MNDWI images of a specific non-drought period is defined as the reference datasets representing normal conditions. Afterwards, a new water shrinking index is introduced to quantify the river water anomaly during drought periods.  Specifically, a python based algorithm which includes image processing and machine learning clustering methods is developed to scan along the MNDWI images to compute the water shrinking index with adjustable river section size. With the index datasets, river sections are further grouped into categories with drought vulnerable levels. By parameterizing the section size, the algorithm is able to quantify and identify the vulnerable shrinking river sections with varying scales. It provides classified references of drought affected hotspots for the regional water management plans in case of drought mitigation. Such a scalable framework can offer a timely distributed monitoring of the drought impacts on the water resource along the river network. As a long term benefit, numerical connections can be identified between the river shrinking status and the economic losses of cargo shipping disruption due to drought.  This is of great value to facilitate the drought impact analysis and forecasts.