A procedure for regional lake water clarity assessment using Landsat multispectral data

Steven M. Kloiber, Patrick L. Brezonik, Leif Olmanson, Marvin E Bauer

Research output: Contribution to journalArticlepeer-review

312 Scopus citations


Although previous investigations have demonstrated reliable empirical relationships between satellite data and nearly contemporaneous ground observations, satellite imagery has not been incorporated into routine lake monitoring programs. This paper focuses on key issues involved in applying satellite imagery to the regional assessment of lake clarity. Ten Landsat Thematic Mapper (TM) images and four Multispectral Scanner (MSS) images of the Twin Cities Metropolitan Area (TCMA) spanning 25 years (1973-1998) were analyzed. Based on this analysis, recommendations are made for a Landsat-based procedure for water clarity assessment. Closeness of fit (r2 values) of regression models between satellite brightness data and measured Secchi disk transparency (SDT) decreased with increasing size of the time window between image collection and ground observation of SDT. Use of SDT data collected within ±1 day of the image date is recommended, but where SDT data are limited, windows up to ±7 days yield reasonable results, especially in late summer when water clarity is relatively constant. Average brightness data from at least nine pixels in the deep open area of a lake should be used to predict lake clarity, but the accuracy of predicted SDT did not improve much as the number of pixels in the area of interest (AOI) increased above this value. A three-coefficient regression model using the TM1/TM3 ratio and TM1 was a consistent and reliable predictor of SDT (r2 values of .7-.8). A similar relationship involving the MSS1/MSS2 ratio and MSS1 was a reasonable predictor of SDT for MSS data. Efforts to produce a standard prediction equation for SDT applicable to images collected on different dates were not successful, but a simple regression procedure to account for differences in atmospheric conditions among image collection dates substantially decreased the range of coefficients in the regression model.

Original languageEnglish (US)
Pages (from-to)38-47
Number of pages10
JournalRemote Sensing of Environment
Issue number1
StatePublished - Sep 2002

Bibliographical note

Funding Information:
This work was funded by the (Twin Cities) Metropolitan Council through the Twin Cities Water Quality Initiative grant program, with additional support from the University of Minnesota Water Resource Center and Agricultural Experiment Station. We gratefully acknowledge the assistance of Erin Day who helped process the image data.


Dive into the research topics of 'A procedure for regional lake water clarity assessment using Landsat multispectral data'. Together they form a unique fingerprint.

Cite this