Monitoring water quality in the White Clay Creek watershed using remote sensing
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
The White Clay Creek watershed in southeastern Pennsylvania and northern Delaware plays a vital role as an important drinking water source for communities in the watershed and as habitat for wildlife. As a critical link in the Delaware Estuary system it provides economic benefit to Delaware by supporting commercial, agricultural, and recreational uses. ☐ Because of its regional importance the creek water quality has been monitored since at least 1995 through physical sampling, and more recently by continuous monitoring of conductivity, turbidity, and temperature implemented at USGS monitoring stations along the creek. This study examines the viability of using satellite imagery to monitor creek water quality more frequently than a physical sampling program allows and to further expand the tools available to water resource managers for assessing the effectiveness of the resource management plan. ☐ Water quality parameters from Delaware Department of Natural Resources and Environmental Control (DNREC) water quality monitoring data for two sample locations are correlated with Landsat 8 spectral radiance data, USGS station data, and NOAA average daily precipitation values. ☐ Examination of the spectral signatures for one sample location highlighted a seasonal response for the Near Infrared band (B05), indicating overhanging vegetation was obscuring the water surface in spring and summer months. Regression analysis was limited to data similar to the spectral signature typical of unobstructed water surfaces. Even after eliminating the obstruction of overhanging vegetation, the spectral information also includes contributions from stream bank radiance due to the relative size of the pixel to the stream width. ☐ Sixty-six DNREC samples were available for the period August 2013 – January 2020. After eliminating scenes obstructed by clouds and vegetation, fourteen samples remained as the study set. Principal Components were calculated for each sample, and regression analyses were performed using either Landsat Band and Band Indices or Principal Components along with selected monitoring station variables. ☐ Results from the analyses provided a good prediction model for salinity using principal components derived from Landsat Bands 01 – 07 DN values. The R Square of the model was 0.96 with a significance factor of 0.0007. However, no other valid prediction models using spectral data were derived from this dataset. It is likely that the Landsat image resolution is too coarse for a stream of this size, and improved results would have been seen studying a wider stream. ☐ Although the frequent cloud cover in Delaware and the satellite resolution compared to the creek width present challenges, this study has shown that water quality information for a relatively small inland water body can be inferred from satellite data. Opportunities for future research include utilizing the methodology with other, larger water bodies such as the Brandywine Creek or repeating the analysis utilizing a finer resolution spectral data source such as Sentinel 2 once sufficient history is accumulated.
Description
Keywords
Stream, Water quality, White Clay Creek