Remotely sensed satellite images have the advantage of relatively high spatial resolution compared to point data collected by on-site instruments. However, higher spatial resolution satellite image data, such as Landsat MSS and TM, generally have the disadvantage of relatively poor temporal resolution compared to data collected by on-site instruments. A goal within this project is to develop a monitoring procedure that takes advantage of both types of data (e.g., the data collected by on-site instruments will be used to calibrate and check the accuracy of results generated using remotely sensed satellite images). Within the next few years there are plans by private industry to launch six to eight satellites with identical sensors having approximately two to five meters resolution, giving a site revisit time of every other day. Under these conditions the temporal revisit frequency problem and possible spatial resolution concerns of current satellite imaging systems will be reduced.
In general, the reflectance of water increases with increased suspended sediment concentrations (positive correlation) and decreases with increased salinity (negative correlation). Mapping suspended sediment concentrations, cholorphyll-a, and salinity using satellite images has been described in previous studies (Gordon and others, 1980; Johnson and Harris, 1980; Khorram, 1981; Smith and Baker, 1982; Lathrop and Lillesand, 1986; Topliss and others, 1990; Bhargava and Mariam, 1992; Khan and others, 1992; Froidefond and others, 1993; Tasan, 1993). However, most previous studies have not attempted to tackle the 'temporal' comparison and change detection of the water parameters, mainly because of the difficulty associated with absolute radiometric calibration of the satellite images. From a monitoring point of view change detection, therefore calibration, is a critical aspect of ecosystem evaluation and management.
One of the goals in this project is to use a satellite calibration and radiometric correction model which we have developed to convert satellite digital numbers (DNs) to ground reflectances to allow the mapping of the desired water parameters on a temporal basis without the need for water sampling during each overflight. Our current radiometric correction model makes corrections for sensor gains and offsets, spectral irradiance, solar elevation, atmospheric scattering and absorption (additive and multiplicative effects), and Earth-Sun distance (Chavez, 1996). We are using geometrically and radiometrically corrected satellite images to generate digital images that represent the desired water parameters. Currently, to have the images represent the water parameters in an absolute sense, the satellite image DNs, or ground reflectance values computed from the satellite image DNs, must be correlated to values derived from water samples collected during the overflight of the satellite when the image is recorded. The water parameter values derived from the water samples are used to apply regression analysis with the satellite image or reflectance DNs. The relationships generated using regression analysis are used to transform the satellite or reflectance image into the water parameters of interest.
In our work several multitemporal Landsat MSS and TM images have been used to extract information about the water. With historical data we have looked at, and compared, sediment volumes on a relative basis, as well as looked at the sediment transport patterns that can be seen in the images. Using a more recent Landsat TM image collected on October 24, 1995 we were able to generate suspended particle matter (SPM), chlorophyl, salinity, and temperature images in an absolute, rather than relative, sense. As discussed above values for these water parameters were derived for water samples collected during the satellite overflight by Water Resources Division (James Cloern and Brian Cole). The satellite image DNs and the water parameter values were used as input to regression analysis. The resulting relationships were used to transform the satellite image DNs to the equivalent digital water parameter images. Shown in Figure 1 are the plots between the SPM and Temperature values of the water samples and corresponding satellite DN values. In this case the sum of TM spectral bands 1 (blue), 2 (green), and 3 (red) were used in the regression analysis. Note that the non-water pixels have been set to zero and show up as black in the images. Also, the water pixels in the image are split into about equal amounts in the Bay and coastal/open ocean areas. In these image results, the higher levels of SPM occur within Suisun Bay and the entrance of the Petaluma River into San Pablo Bay, as well as at extreme South Bay where Coyote Creek enters the bay (30 to 46 mg/liter). Most of the deeper channel in San Pablo Bay and most of South Bay have low SPM values; the lowest values (1 to 6 mg/liter) occur in Central Bay. The temperature image shows that the warmer waters are in South Bay, with waters in Suisun Bay and along the channel connecting it with San Pablo Bay also rather warm (16.5 to 19 degrees C). Central Bay and the open ocean have the colder waters (10 to 15 degrees C). For these data it was assumed that the reflectance values correspond to only surface water information and no sub-bottom affects are present.
(Click on image for a larger
|Figure 1. Shown above are the plots of the satellite digtial numbers (DNs) versus the values computed from the water samples taken during the satellite overflight for left) suspended particle matter (SPM-mg/liter), and right) temperature (degrees C). Below the plots are the corresponding SPM and TEMPERATURE images generated from the regression analysis done between the water samples and Landsat TM DN values. Note, non-water pixels were set to zero and show up as black in the images.|
Note, in Figure 1 there is a point in the temperature plot with a DN value of about 107 and temperature of 16.8 degrees that is isolated from the other points. The reason for this is clouds, so the point was not used in the regression analysis. Other image results generated in this project using multitemporal Landsat MSS and TM images can be seen on the World Wide Web pages at:
Some of the issues and research that still needs to be looked at are:
Chavez, P.S., Jr., 1996. Image-Based Atmospheric Corrections - Revisited and Improved, Photogrammetric Engineering and Remote Sensing, Vol. 62, No. 9, pp. 1025-1036.
Froidefond, J.M., P. Castaing, J.M. Jouanneau, R. Prudhomme, and A. Dinet, 1993. Method for the quantification of suspended sediments from AVHRR NOAA-11 satellite data, International Journal of Remote Sensing, Vol. 14, No. 5, pp. 885-894.
Gordon, H.R., D.K. Clark, J.L. Mueller, and W.A. Hovis, 1980. Phytoplankton pigments from the Nimbus-7 Coastal Zone Color Scanner: Comparisons with surface measurements, Science, Vol. 210, No. 3, pp. 63-66.
Johnson, R.W., and R.C. Harris, 1980. Remote sensing for water quality and biological measurements in coastal waters, Photogrammetric Engineering and Remote Sensing, Vol. 46, No. 1, pp. 77-85.
Khan, M.A., Y.H. Fadlallah, and K.G. Al-Hinai, 1992. Thematic mapping of subtidal coastal habitats in the western Arabian Gulf using Landsat TM data - Abu Ali Bay, Saudi Arabia, International Journal of Remote Sensing, Vol. 13, No. 4, pp. 605-614.
Khorram, S., 1981. Use of ocean color scanner data in water quality mapping, Photogrammetric Engineering and Remote Sensing, Vol. 47, No. 5, pp. 667-676.
Lathrop, R.G., Jr., and T.M. Lillesand, 1986. Use of Thematic Mapper data to assess water quality in Green Bay and Central Lake Michigan, Photogrammetric Engineering and Remote Sensing, Vol. 52, No. 5, pp. 671-680.
Smith, R.C., and K.S. Baker, 1982. Oceanic chlorophyll concentrations as determined by satellite (Nimbus-7 Coastal Zone Color Scanner), Marine Biology, Vol. 66, pp. 269-279.
Tassan, S. 1993. An improved in-water algorithm for the determination of chlorophyll and suspended sediment concentration from Thematic Mapper data in coastal waters, International Journal of Remote Sensing, Vol. 14, No. 6, pp. 1221-1229.
Topliss, B.J., C.L. Almos, and P.R. Hill, 1990. Algorithms for remote sensing of high concentration inorganic suspended sediment, International Journal of Remote Sensing, Vol. 11, No. 6, pp. 947-966.
AutobiographyChavez has used remotely sensed images for earth science research and applications for over 25 years. Areas of research have included radiometric calibration (relative and absolute) of satellite multispectral images and automatic change detection using multitemporal satellite images. Also, the application of image processing to sidescan sonar, including digital mosaicking, for seafloor and coastal regions.