Assessing potential impacts of COVID-19 on water quality using combined Landsat-8 and Sentinel-2 data products

Abstract:

The COVID-19 virus has changed the course of daily life throughout the world. The fight against the pandemic requires global-scale assessments. However, most of the available data is generated locally. Geographic information systems (GIS) in general and remote sensing (RS) in particular can play a crucial role in this fight. Satellite imagery is one of the sources we can rely on to study the impact of the virus in a wide range of areas including, but not limited to, water quality.

This work aims to investigate the effect of COVID-19 on water quality by applying anomaly detection methods on remote sensing products from Landsat-8 and Sentinel-2 missions. Taking advantage of both Landsat-8 and Sentinel-2 data enables us to achieve higher temporal coverage. Here we take advantage of a collection of level 2 products such as chlorophyll a (Chl_a) and total suspended solids (TSS). These products serve as inputs to our anomaly detection algorithms, such as MAD-Z score, isolation forest (IF) and convolutional auto-encoders (CAE).

We evaluate the performance of each approach by cross-validation with high-frequency in-situ measurements. This study spans over multiple sites in the proximity of human settlements in Italy, Peru, and the U.S. Our analysis for each site includes all historical data for each mission, namly from 2013 for the Landsat-8 and from 2015 for the Sentinel-2 mission. The lessons learned at the well-monitored sites with operational sewage and wastewater treatments allow us to quantify potential anomalies surrounding other coastal cities at global scales.