Evaluation of Earth Observation Systems for Estimating Environmental Determinants of Microbial Contamination in Recreational Waters

Public health risks related to the microbial contamination of recreational waters are increased by global environmental change. Intensification of agriculture, urban sprawl, and climate change are some of the changes which can lead to favorable conditions for the emergence of waterborne diseases. Earth observation (EO) images have several advantages for the characterization and monitoring of environmental determinants that could be associated with the risk of microbial contamination of recreational waters in vast territories like Canada. There are a large number of EO systems characterized by different spatial, temporal, spectral, and radiometric resolutions. Also, they have different levels of accessibility. In this study, we compared several EO systems for the estimation of environmental determinants to assess their usefulness and their added value in monitoring programs of recreational waters. Satellite images from EO systems WorldView-2, GeoEye-1, SPOT-5/HRG, Landsat-5/TM, Envisat/MERIS, Terra/MODIS, NOAA/AVHRR, and Radarsat-2 were acquired in 2010 and 2011 in southern Quebec, Canada. A supervised classification of these images with a maximum likelihood algorithm was used to estimate five key environmental determinants (agricultural land, impervious surfaces, water, forest, and wetlands) within the area of influence of 78 beaches. Logistic regression models were developed to establish the relationship between fecal contamination of beaches and environmental determinants derived from satellite images. The power prediction of these models and criteria such as accuracy of classified images, the ability of the sensor to detect environmental determinants in the area of influence of beaches, the correlation between the estimated environmental determinants in the area of influence by the sensor with those estimated by very high spatial resolution reference sensors (WorldView-2 and GeoEye-1), and general criteria of accessibility (cost of the images, imaging swath, sat- llite revisit interval, hours of work, and expertise and material required to process the images) were used to evaluate the EO systems. The logistic regression model establishing the relationship between environmental determinants from Landsat-5/TM images and the level of fecal contamination of beaches is the one which performs best. These images are also those that best meet all of the evaluation criteria. This study showed that environmental determinants like agricultural lands and impervious surfaces present in the area of influence of beaches are those which contribute the most to the microbial contamination of beaches. Our study demonstrated the utility and the added value that EO images could bring to programs monitoring the microbial contamination of recreational waters.