MODIS data is well suited for the assessment of large lakes and offers better spectral sensitivity and temporal (daily) coverage than Landsat which has a 16 day overpass interval. However due to the relatively low spatial resolution 250, 500 and 1000 m the number of lakes that can be assessed using MODIS is considerable lower than for Landsat's 30 m resolution. Preliminary analyses using 500 m MODIS data indicate that only around 100 Minnesota lakes will be able to be assessed using MODIS 500 m data. This is around 1 percent of the lakes that can be assessed using Landsat imagery.
At 250 m resolution, the most useful one for lake monitoring, the spectral sensitivity of MODIS is low, with only two bands, one in the red portion of the spectrum, and one in the near infrared. Spectral sensitivity increases as spatial resolution decreases. Unfortunately, the number of lakes that can be studied with each of the resolutions also decreases sharply. Click on the picture below to view the difference between Landsat ETM+, MODIS 250, 500 and 1000 m data. We can roughly estimate the number of lakes than can be monitored at each of the MODIS resolutions by considering that a lake should have a bare minimum total area greater than the area of a 4x4 pixel block in order to contain any useful, non-contaminated pixels. Using this method, we conclude that, out of more than 10,000 lakes in Wisconsin, only 502, 118 and 22 lakes can be monitored with 250 m, 500 m, and 1000 m resolution MODIS imagery, respectively. The practical number of lakes is actually lower, because some lakes have complex shapes or include islands, and have therefore a higher probability of containing pixels contaminated by land. Additionally, the nominal resolution of the bands corresponds to nadir data collection, with ground resolutions essentially doubling at the edges of the 2330 km wide MODIS swaths.
|
 |
To explore the capabilities of MODIS data for regional lake water quality monitoring, we selected an image in which Wisconsin was near the nadir of the sensor, cloud cover was low and a sufficient number of ground-based Secchi disk depth observations was available. The selected image was acquired on September 17, 2000, and contained 32 ground based Secchi disk depth observations in cloud free areas. Of these, only 17 were used in our analysis, because lakes smaller than 160 ha in surface area were excluded. The 250 m resolution MODIS image was sampled manually to extract the pixel with the minimum radiance inside the lake of interest; then, observed Secchi disk depth observations were regressed against minimum radiance in the red and near-infrared bands. The best model used only radiance in the red band, and "raw" (not log-transformed) data. The relationship presented below is only slightly lower than the results we have seen using Landsat data with an R2 = 0.75 and standard error of estimate (SEE) = 49 cm. MODIS holds promise for monitoring lake water clarity of relatively large lakes on a large regional scale. |
 |
|
)