Satellite-assisted Lake Water Quality Monitoring

By Dr. Thomas M. Lillesand and Dr. Jonathan W. Chipman, University of Wisconsin-Madison

Using Satellite Data to Observe Regional Trends in Lake Transparency

The state of Wisconsin (USA) includes more than 15,000 freshwater inland lakes. Costs and logistics limit traditional detailed water quality sampling to as few as 50 lakes annually. The Satellite Lake Observatory Initiative (SLOI) is aimed at monitoring spatial and temporal trends in lake water quality on a continuous, statewide, operational basis. Field-measured Secchi disk data for a limited number of lakes are being used to empirically "calibrate" Landsat and MODIS images in order to estimate the transparency of all lakes included in the satellite scenes.

Lakes are important economic, environmental and recreational resources, and Wisconsin is blessed with one of the world's highest densities of lakes. Protecting and monitoring water quality is a daunting task in such an area. It is impractical to monitor directly more than a small fraction of the state's lakes.

Regional Land Use and Climatic Context

The above situation also typifies the neighboring states of Minnesota and Michigan, which together with Wisconsin form the "Upper Great Lakes States" of the US. Land use is changing rapidly throughout this region as "baby boomers" develop second homes for recreation and/or retirement. At the same time, population is increasing generally and conversion of agricultural and other rural land into suburban uses is rapidly urbanizing many watersheds in which lakes reside. In turn, lakes are often subject to "flashier" hydrologic events, increased nutrient loadings and toxic runoff constituents. Add to this the impact of increased climatic variability in the region, and the need to better understand spatial and temporal water quality trends in support of a range of environmental planning and management activities becomes acute.

The Satellite Lake Observatory Initiative

The Satellite Lake Observatory Initiative (SLOI) is an inter-agency program (Table 1) aimed at integrating satellite remote sensing into the regional lake water quality process. SLOI is designed to exploit the complementarity and synergism among multiple satellite systems as well as multiple sources of water quality ground reference data. The initial suite of sensing systems being emphasized includes the Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) systems as well as the Moderate Resolution Imaging Spectrometer (MODIS) onboard the Terra satellite (Table 2).

Table 1. Participants in the Satellite Lake Observatory Initiative (SLOI)

Table 2. Comparison between Landsat TM and ETM+ vs. MODIS data for regional lake water quality assessment

Collecting Ground Reference Data

Because we correlate the satellite data to ground reference data, our efforts require field sampling of as many lakes as possible, as frequently as possible. This ensures that we have reference data available that corresponds to the nearly biweekly (16 day) coverage from Landsat or the near-daily imagery from MODIS. The Self-help program volunteers are critical in this process. There are now over 700 such volunteers who sample more than 600 lakes distributed across the state. All volunteers measure water clarity using the Secchi disk method, using a rope graduated in meters to lower a 20 cm diameter white metal disk until it is no longer visible. A subset also collects temperature data and measures phosphorus and chlorophyll concentrations. The data are quality checked by DNR, and eventually incorporated with the DNR's own sampling data into the EPA's national STORET surface water quality database.

Landsat TM and ETM+ Processing

The way we "calibrate" each TM or ETM+ image with respect to the field measured data is patterned after an approach developed by Dr. Marvin Bauer and his colleagues at the University of Minnesota. This amounts to modeling Secchi disk depth as a function of observed digital numbers from a combination of multispectral bands on the TM or ETM+ sensor. This is done using a multiple regression model of the following form:

In(Secchi) = b₀ + (b₁ TM1) + (b₂ TM1/TM3 )

where the dependent variable is a natural log transformation of Secchi disk depth (in meters); the independent variables TM1 and TM3 are the observed digital numbers from the blue and red bands (respectively) of the Landsat TM or ETM+; and b₀, b₁, and b₂ are regression model coefficients. Figure 1 summarizes the status of the application of this technique on a statewide basis.

Understanding the Temporal Dimension of Lake Clarity via Landsat Data

Another issue that SLOI is considering is the temporal stability or dynamics of lake reflectance. Using multiple dates of imagery, statistics have been generated for many lakes showing relative differences in reflectance (among lakes on a single date) as well as different trends in reflectance (among lakes on multiple dates). As shown in Figure 2, although some lakes remain quite stable over time in their spectral reflectance, others may change quite dramatically from year to year.

Figure 1. Status of statewide water clarity mapping via Landsat TM and ETM+ data.

(Click to enlarge)

Figure 2. Two-dimensional spectral reflectance trajectories for ten lakes, based on Landsat-TM bands 2 and 3, for four dates of imagery from 1984 to 1993. The lake reflectance values have been normalized, being represented as number of standard deviations above or below the overall mean for all lakes.

(Click to enlarge)

The Role of MODIS Data

A major factor limiting the usefulness of Landsat TM or ETM+ data for lake water clarity assessment in our area is cloud cover. MODIS, with its near-daily repeat coverage is being assessed as an alternative and complementary data source in this light. This sensor also features a much larger coverage area, at the expense of its spatial resolution (Table 2). Figure 3 illustrates the use of MODIS data to estimate the clarity of large lakes on a regional basis.

Figure 3. Predicted and observed lake water clarity, for 40 large lakes imaged by MODIS on 8 Sept. and 17 Sept. 2000.

(Click to enlarge)

Historically, More Promise than Practice

From a research perspective, satellite remote sensing of lake water quality dates back to the 1970s, when several investigators demonstrated the potential of using first-generation Landsat Multispectral Scanner (MSS) data for this purpose. This was followed by early work with Landsat TM and SPOT data in the 1980s. However, there has not been widespread adoption of these techniques by either resource managers or policy makers. In our view, there are three primary reasons why this has been the case:

In short, there was no clear aggregated market of end users requiring such information to meet their day-to-day needs.

Times are Changing

At present we are in the midst of a sweeping transformation from being a spatially literate society to one that is spatially dependant and enabled. Individuals and organizations abound that can exploit such tools as GIS, GPS, the Internet, and decision support and visualization systems. Add to this the impending international explosion of governmental and commercial sources of satellite data to become available over a range of spectral, spatial, and radiometric resolutions.

The above capabilities provide a technological "push" to the future of satellite remote sensing in general. In the case of the application of this technology to lake monitoring and management there are many "pulls" at work as well. These include such concerns as the impact of urban sprawl on water quality and quantity, the over-exploitation of groundwater supplies and the impacts of climatic variability, to name only a few. There is now clear recognition that many of the problems confronting local planners and resource managers have regional origins.

Future Outlook

With multi-stage remote sensing we now have the tools to offer literally a global to local perspective on such issues as those above. In the not too distant future we believe satellite-based regional lake water quality assessments over time and space will become commonplace. Beyond this, high-resolution satellite data will be a key tool for the lake manager and land use planning consultant, hydrologic engineer, local zoning administrator and the various commercial firms who provide lake management goods and services (e.g., weed control, sediment disposal, erosion control, algae control, etc.). The future for application of satellite remote sensing to lake management is indeed a bright one.

Biography of the Authors

Dr. Lillesand is Director of the Environmental Remote Sensing Center (ERSC) at the University of Wisconsin-Madison (UW). Previously, he taught remote sensing at the State University of New York College of Environmental Science and Forestry and at the University of Minnesota. He is a Fellow and Past-president of the American Society for Photogrammetry and Remote Sensing. He (along with Dr. Ralph W. Kiefer) is a co-author of the book Remote Sensing and Image Interpretation.

Dr. Chipman is a Research Associate at ERSC and the UW Center for Limnology. He obtained his MS and Ph.D. from the UW in Environmental Monitoring. In addition to SLOI, he has participated in a diverse range of research projects, including large-area land cover mapping, estimation of land cover and forest characteristics using microwave systems, development of commercial remote sensing applications, and topographic mapping by means of radar interferometry.

Affiliation

Environmental Remote Sensing Center (ERSC), University of Wisconsin-Madison, 1225 W. Dayton Street, Madison, WI 53706, Phone: 608-262-1585, FAX: 608-262-5964, E-mail: tmlilles@facstaff.wisc.edu and jchipman@facstaff.wisc.edu