This article is a part of the NASA DEVELOP’s Spring 2017 Article Session. For more articles like these, click here
This project analyzed the efficacy of NASA Earth observations in providing an additional tool for the Virginia Department of Environmental Quality to use in their analysis of water quality using turbidity as the parameter.
The Chesapeake Bay watershed is a diverse ecosystem consisting of freshwater, brackish, and saltwater spanning Virginia, Maryland, Delaware, West Virginia, Pennsylvania, New York, and the District of Columbia .
On Dec. 29, 2010, the EPA established the Chesapeake Bay Total Maximum Daily Load (TMDL) that serves as a calculation for the maximum amount of point and non-point source pollutants permissible throughout the bay. Point source pollution is pollution that comes from a specific area like an industrial or sewage treatment plant. Non-point source pollution is caused by rainfall picking up pollutants such as excess fertilizer, sediment, or grease from urban runoff and carrying it into the bay .
Monitoring efforts within the bay are imperative because changes in water quality can have dramatic effects on the ecosystem. Submerged Aquatic Vegetation (SAV) are keystone species within the Chesapeake Bay that provide food for and support 29 species of waterfowl, 348 species of finfish and various species of shellfish, including the economically-important blue crab . The bay also serves as a resting ground for more than 1 million waterfowl as they migrate . Healthy beds of SAV add oxygen to the water through photosynthesis, remove suspended sediments and stabilize the near-shore regions where they grow by preventing erosion .
SAV are important ecosystem health indicators because they are quick to react to changes in water quality. According to Lefcheck et al. (2017), the bay has experienced a nearly 30 percent eelgrass decline since the 1970s due to increasing temperatures and declining water clarity . While extreme and gradual dieback events occurred over this period, it is thought that eelgrass responds to multiple drivers and in some cases was able to recover to pre-dieback levels . This variation in SAV abundance has ecological and economic implications. Declines in SAV have been associated with increased nutrients and sediments throughout the bay from anthropogenic sources, ecosystem service and economic loss, and overall Chesapeake Bay health and its need for improvement .
This project addresses NASA’s Water Resources National Application Area by developing a model to estimate turbidity from NASA Earth observation data within the Chesapeake Bay. The Virginia Department of Environmental Quality approached DEVELOP to determine the effectiveness of NASA Earth observations in monitoring SAV populations in the Virginia portion of the watershed (Rappahannock, York, James and the southern bank of the Potomac River) from 2003 to 2016. The DEQ aims to address the Environmental Protection Agency’s requirements of the Total Maximum Daily Load for Chesapeake Bay health, and aims to improve their data collection and monitoring abilities using Earth observations.
The team utilized Landsat scenes, which are available at a 30 m spatial resolution compared to the 250 m resolution of the MODerate resolution Imaging Spectroradiometer (MODIS) sensor, and thus serve better in the analysis of the near-shore waters such as the rivers and estuaries of the Chesapeake Bay . Previous studies found that Landsat data was capable of tracking populations of SAV, but that several difficulties arise when using satellite imagery such as atmospheric interference , variation in water depth and water column light attenuation .
This project utilized the surface reflectance products from Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) data collections in Google Earth Engine API. Landsat data were acquired from 2003 to 2016 to correspond with continuous monitoring water quality data from VIMS data. The data collected from Landsat 5 occurred between 2003 and 2011, and data collection from Landsat 8 occurred between 2013 through 2016.
Visual inspection of the scenes from Landsat 5 and Landsat 8 through the study period allowed the team to remove images that had obvious cloud cover throughout the study area. We filtered our images using the cfmask to identify pixels that were both cloud free and over water. Approximately 38 percent of the Landsat images from the study period were usable after this inspection. The useable dates obtained from these satellite overpasses were compared to the continuous monitoring station turbidity data from VIMS, leaving us with 64 data points annually, approximately eight points per day for Landsat 8 (based on the data available on the days with useable Landsat images). We then extracted the single band and NDTI values at each pixel corresponding with a continuous monitoring station.
Continuous monitoring station turbidity data were spatially and temporally co-located with Landsat satellite data. Turbidity measurements, in Nephelometric Turbidity Units (NTU), were filtered from VIMS data at the 15-minute increment corresponding to the satellite overpass, at 11:30 a.m. EDT (10:30 a.m. EST) for Landsat 5 and 11:45 a.m. EDT (10:45 a.m. EST) for Landsat 8. Turbidity observations flagged by VIMS as suspect were removed from the final in situ dataset before being converted into a fusion table for use in Google Earth Engine API. The fusion table was matched up with its corresponding Landsat NDTI and extracted to a csv spreadsheet to be used in the regression model.
When the data were annualized using the SAV growing season of June through May, excluding November through February, the correlations improved, likely due to the differences in substrate type and changes in turbidity because of the growing seasons. Furthermore, due to the difference in dynamic range between Landsat 5 and Landsat 8, it was determined that further study should be pursued regarding the use of Landsat 8 for continued monitoring efforts. While a linear regression was not entirely effective, alternative regression model types and turbidity data transformations, such as using band ratios, the annual median instead of the annual mean and spatially or temporally aggregating data throughout the bay, should be used to determine the relationship between Landsat derived and in situ turbidity as an indicator of SAV health within the Chesapeake Bay. Additionally, this study will be further expanded upon to incorporate the use of the software program, ACOLITE, to account for challenges related to coastal remote sensing including marine reflectance and atmospheric interference.
The results of this project will be utilized at the state and federal level to determine how remotely-sensed data can be incorporated into the current monitoring efforts of SAV. Satellite-derived data could supplement their current monitoring efforts by providing a greater spatial and temporal resolution of water quality conditions than current in situ sampling. Further, these products can be used to assess areas of the bay and its tributaries that were not sampled in person. By combining these methods, state and federal program managers will have a more holistic suite of information available for their decision-making.
 K. Beard et al., “Chesapeake Bay water resources and ecological forecasting: Mapping submerged aquatic vegetation in the Chesapeake Bay watershed using remote sensing and GIS algorithms,” NASA DEVELOP National Program, Hampton, VA, Tech. Rep., Apr. 2010.
 T. Beard and A. Kennedy, “Chesapeake Bay water resources and ecological forecasting: Mapping submerged aquatic vegetation in the Chesapeake Bay watershed using remote sensing and GIS algorithms,” NASA DEVELOP National Program, Hampton, VA, Tech. Rep., 2009.
 A. Dogliotti et al., “A single algorithm to retrieve turbidity from remotely-sensed data in all coastal and estuarine waters,” Remote Sens Env, vol. 156, pp. 157-168, Jan. 2015.
 W.M. Kemp et al., “Habitat requirements for submerged aquatic vegetation in Chesapeake Bay: Water quality, light regime, and physical-chemical factors,” Estuaries, vol. 27, no. 3, pp. 363-377, Jun. 2004.
 J. Lefcheck et al., “Multiple stressors threaten the imperiled coastal foundation species eelgrass (Zostera marina) in Chesapeake Bay, USA,” Global Change Biol, Feb. 2017.
 N. Pahlevan et al., “Landsat 8 remote sensing reflectance (R rs) products: Evaluations, intercomparisons, and enhancements,” Remote Sens Env, vol. 190, pp. 289-301, Mar. 2017.
 R. Samp et al., “Chesapeake Bay water resources: The integration of NASA Earth observation data into the Chesapeake Bay watershed model,” NASA DEVELOP National Program, Hampton, VA, Tech. Rep., Aug. 2009.
Danielle Quick is a recent graduate from Fayetteville State University who worked with the NASA DEVELOP National Program at NASA Langley Research Center as an independent research consultant on the Chesapeake Bay Water Resources project.
Amanda Clayton is a master’s graduate from the University of Illinois at Springfield working as a project coordination fellow with the NASA DEVELOP National Program at NASA Langley Research Center.
Cole Cowher is a student from Christopher Newport University who worked with the NASA DEVELOP National Program at NASA Langley Research Center as an independent research consultant on the Chesapeake Bay Water Resources project.
Collin Henson is a graduate student at the College of William and Mary who worked with the NASA DEVELOP National Program at NASA Langley Research Center as an independent research consultant on the Chesapeake Bay Water Resources project.
Gregory Hoobchaak is a recent graduate from Christopher Newport University who worked with the NASA DEVELOP National Program at NASA Langley Research Center as an independent research consultant on the Chesapeake Bay Water Resources project.