Jump to Main Content
Long-Term Agroecosystem Research in the Central Mississippi River Basin: Hyperspectral Remote Sensing of Reservoir Water Quality
- Kenneth A. Sudduth, Gab-Sue Jang, Robert N. Lerch, E. John Sadler
- Journal of environmental quality 2015 v.44 no.1 pp. 71-83
- agroecosystems, ammonia, calibration, chlorophyll, hyperspectral imagery, lakes, least squares, long term experiments, model validation, models, monitoring, nitrogen, nutrients, phosphorus, reflectance, remote sensing, suspended sediment, turbidity, water quality, water reservoirs, watersheds, Mississippi River, Missouri
- In situ methods for estimating water quality parameters would facilitate efforts in spatial and temporal monitoring, and optical reflectance sensing has shown potential in this regard, particularly for chlorophyll, suspended sediment, and turbidity. The objective of this research was to develop and evaluate relationships between hyperspectral remote sensing and lake water quality parameters—chlorophyll, turbidity, and N and P species. Proximal hyperspectral water reflectance data were obtained on seven sampling dates for multiple arms of Mark Twain Lake, a large man-made reservoir in northeastern Missouri. Aerial hyperspectral data were also obtained on two dates. Water samples were collected and analyzed in the laboratory for chlorophyll, nutrients, and turbidity. Previously reported reflectance indices and full-spectrum (i.e., partial least squares regression) methods were used to develop relationships between spectral and water quality data. With the exception of dissolved NH₃, all measured water quality parameters were strongly related (R² ≥ 0.7) to proximal reflectance across all measurement dates. Aerial hyperspectral sensing was somewhat less accurate than proximal sensing for the two measurement dates where both were obtained. Although full-spectrum calibrations were more accurate for chlorophyll and turbidity than results from previously reported models, those previous models performed better for an independent test set. Because extrapolation of estimation models to dates other than those used to calibrate the model greatly increased estimation error for some parameters, collection of calibration samples at each sensing date would be required for the most accurate remote sensing estimates of water quality.