Satellite imagery for lake water quality assessments

By Patrick Brezonik

The use of satellite imagery to measure water quality conditions in lakes and other surface waters has made tremendous advances in the last 20 years – from a technique typically viewed by limnologists as more a curiosity than a viable measurement approach to a sophisticated technique now producing critical and detailed data on major water quality characteristics, such as chlorophyll, suspended solids, colored dissolved organic matter and water clarity, at spatial scales impossible or impractical to achieve by conventional land-based sampling approaches.

Major technological advances in two areas are largely responsible for these advancements. First, satellites, and especially their sensors, have become more sophisticated, with greater radiometric sensitivity, more and better bands to measure spectral reflectance associated with specific water quality parameters, and in some cases better spatial resolution. The addition of the European Sentinel satellites to the continuing program of U.S. Landsat satellites has resulted in much higher frequency of temporal coverage, thus increasing the likely of overpasses during clear weather conditions. Second, computing capacities have increased many fold over the past two decades, enabling the ready handling of giga- and even peta-scale data sets. The development of Google Earth Engine (GEE) enabled ready display and analysis of all available imagery for a given spatial area, thus facilitating multi-temporal analyses of trends over years or even decades. The recent publication1 of a large data set linking reflectance data from clear imagery with concurrently gathered ground-based measurements should foster further model development and deepen integration of satellite remote sensing into inland water science.

Nonetheless, it is fair to say that scientific advancements by remote sensing scientists in understanding how to process and use satellite imagery also have been critical to the development and maturation of the field. Of special importance has been the fairly recent development of reliable techniques to correct raw imagery, which essentially measures what the satellite sensor (at the top of the atmosphere) detects, rather than what was reflected from the water surface. Variable atmospheric conditions can dramatically affect what the satellite sensor “sees,” correction for these variable atmospheric effects is essential so that separate ground-based calibration data are not needed for every individual satellite image. One of our colleagues, Ben Page recently developed and published a robust atmospheric correction routine called MAIN2 that works very well on both Landsat and Sentinel imagery.

Also of great importance in enhancing the usefulness of satellite imagery for water quality assessments are numerous studies conducted by scientists over the past two decades to develop procedures or “algorithms” to extract water quality data from satellite sensor signals. Our group at the University of Minnesota has contributed to these efforts in several ways. For example, we developed reliable procedures to retrieve water clarity data, expressed as the common limnological variable Secchi depth, from satellite imagery nearly 20 years ago3 and have since used variants of the technique to make census-level water-clarity measurements on all Minnesota lakes larger than ~ 20 acres at approximately 5-year intervals – a total of 9 censuses since 1975.4,5

figure 1

Figure 1. Map of colored dissolved organic matter (CDOM) for Minnesota lakes including statistical summaries by ecoregion. Landsat 8 data were used for the map.

More recently, we developed retrieval methods for colored dissolved organic matter (CDOM), a common constituent in northern Minnesota lakes,6,7 and we have applied these techniques to retrieve CDOM on all Minnesota lakes > 20 acres. Statistical distributions of the lake CDOM levels vary greatly among Minnesota’s seven aquatic ecoregions, with the highest levels found in the Northern Lakes and Forests Ecoregion of northeastern Minnesota (Figure 1).

Most recently, we began using satellite imagery to measure chlorophyll in lakes. A map of the statewide distribution of chlorophyll in Minnesota’s lakes is available on our website and results for individual lakes and statistical summaries of chlorophyll concentrations by county, ecoregion, and other spatial divisions are readily available on our lake “browser”. An indication of the type of detailed information that will be forthcoming in future years, as we take full advantage of the existing satellites and retrieval methods, is shown in a time sequence of chlorophyll concentrations in small lakes near Lake Minnetonka during summer 2018, which is on the chlorophyll page of the water.rs web site. We invite interested persons to visit the web site and lake browser, both of which recently underwent major updating and expansion, for more information on recent CDOM, chlorophyll, suspended solids, and water clarity studies.

Finally, it is pertinent to note that the studies described above mostly were conducted using manual procedures—that is, each satellite image was downloaded and analyzed separately. Under the auspices of funding from the LCCMR and Minnesota’s Environmental and Natural Resources Trust Fund, we have been developing procedures whereby imagery will be downloaded and processed automatically using resources of the Minnesota Supercomputing Center. A beta version of this new technology will be available in early 2020, and we are excited at the prospects of the huge amount of new information this technology will produce.

 

 

References

1. Ross, M. R., et al. 2019. Water Res. Research, doi.org/10.1029/2019WR024883.

2. Page, B., et al. 2019. Rem. Sens. Environ. 231: 15 September, 111284.

3. Kloiber, S. M., et al. 2002. Rem. Sens. Environ. 82: 38.

3. Olmanson, L. G., et al. 2008. Rem. Sens. Environ.112: 4086.

4. Olmanson, L. G., et al. 2014. J. Amer. Water Resour. Assoc. 50: 748.

5. Brezonik, P. L., et al. 2005. Lake Res. Manage. 21: 373.

6. Olmanson, L. G., et al. 2016. Rem. Sens. Environ. 185: 119-128.