HydroMod, a combination of HYDROLIGHT and MODTRAN has been assembled and tested. A water analysis laboratory has been established at RIT to support the next phase of modeling.

We have continued contacting people to gather information on all types of water quality issues. New contacts this quarter include the Utah Agriculture Dept., and Texas Agricultural Experiment Station, SVS Inc, and the Spectral Information Technology Application Center (SITAC). A table of water quality applications and the corresponding HSI requirements was assembled.

Some additional hyperspectral data sets with accompanying water truth were identified, but getting quality data remains a high priority.

A Hyperspectral Water Quality Monitoring web page was established.

In lieu of the previously identified West Coast AVIRIS collection, we have since found two AVIRIS data sets of Narragansett Bay, Rhode Island. These data sets seem better for our study based on the water characteristics and existing truth data.

Contact name: John Mustard, Assistant Professor, Department of Geological Sciences, Box 1846, Brown University, Providence RI, 02912 (401) 863-1264 john_mustard@brown.edu http://www.planetary.brown.edu/~mustard/mustard_info.html

f980711t01p02_r08 8 Narragansett Bay 2, RI

+41-21.0 +41-45.0 -71-13.0 -71-13.0 1426 1431

f980711t01p02_r09 9 Narragansett Bay 1, RI

+41-51.0 +41-21.0 -71-21.0 -71-23.0 1437 1443

We have also found a littoral data set called Island Radiance that was acquired by a hyperspectral imager flown by SETS in Hawaii. This imagery and accompanying water truth data has been used to study the health of coral reefs in Hawaii. We expect to be able to get this data for review.

HydroMod is a highly accurate tool for calculating radiance distributions using realistic environmental conditions. It was specifically designed for studies involving remote sensing of water quality parameters and consists of four main modules. The four modules calculate: (1) input radiance distribution; (2) the transition through the water surface (both in to and out of the water); (3) propagation and reflection underwater; and (4) propagation back through the atmosphere to the sensor. Each module emphasizes realistic environmental conditions and accuracy. For instance, the input radiance distribution module uses the Air Force Research Laboratory's MODTRAN as a base to calculate the radiance in to a point of interest from each hemispherical direction using a 5° x 5° grid. MODTRAN is one of the most accurate radiative transfer computer models available; it allows great flexibility in atmospheric conditions including the use of measured data. MODTRAN is an excellent tool in and of itself. However, HydroMod goes one step further by building on the MODTRAN base: with HydroMod, clouds can be added to any number of grid points and the clouds can vary in size, shape, and type.

Similar enhancements are used in the underwater module. Here, invariant imbedding methods as described by Dr. Curtis Mobley in "Light and Water"¹ are used to solve the radiative transfer equations. In fact, variations of Dr. Mobley's original HYDROLIGHT code are used to perform the core calculations. Using that core, HydroMod allows flexible modifications in water quality parameters in near real time. These modifications are made whether using HydroMod defaults or user-supplied absorption and scattering profiles. Virtually any water body can be modeled. The underwater module calculates the radiance distribution exiting the water, but it also calculates the radiance distribution at any user-desired depths. Thus, HydroMod can effectively be used to study light distribution from within the water body to above the water surface and even to a remote sensing platform orbiting the earth.

Current studies using HydroMod include the characterization of the impact of neighboring clouds on SeaWiFS data and the evaluation of underwater light field distributions. A sample data set from HydroMod is shown in Figure 2. The two windows represent the water-reflected radiance field (right) and a profile of that field across the middle. The image on the right represents the amount of radiance in each direction for declination angles of zero (center of the window) to 90° (at the circle edges) and azimuth angles from 0° to 360° with 0° representing due North at the plot top and positive counterclockwise. The data are displayed using either a log scale (as shown here) or a linear scale. In this sample, the sun was located at 40° declination at 180° azimuth and a single fairly bright cloud was located at 50° declination and 100° azimuth. Noticeably brighter areas result 180° away from these locations. A wind speed of 5m/s was used which tends to diffuse the specular bounce. SeaWiFS Band 5 centered at 555 nm is displayed.

For the purposes of the current studies that use HydroMod, the code is considered valid and operational. However, further development and expansion is already underway. The current phase of code development is the conversion from a Windows platform to a Unix platform. The conversion should be complete by the beginning of summer. Other modifications that are being considered range from the simple update to the next MODTRAN version to the more involved vectorization of the water portion of the code.

Full end-to-end validation of HydroMod was not possible due to the lack of "ground truth" data. However, other methods are possible. The HydroMod validation program included step-by-step validation and module-by-module validation. At each step (input, output, calculations,…) the code was tested and validated versus known expectations. Most of these were accomplished as the code was developed and modified. The code eventually passed all tests.

Each module was tested individually using specific test input values; HydroMod provided adequate output in all cases. For instance, by dividing the reflected component by the sky input radiance component with no wind, the Fresnel reflection curves are obtained as expected. Changes in wind speed modified that result appropriately. Another example is the water-leaving radiance variations for changes in depth and changes in water quality parameters. These were validated based on the magnitude of the changes in the correct spatial directions. With deeper water, the water-leaving radiance changed appropriately (darker bottom resulting in a brighter water leaving radiance and a lighter bottom resulting in a darker water-leaving radiance). The original HYDROLIGHT code also came with a few test cases. These cases were used to validate the original code. Once MODTRAN was added for the sky input radiance, the original HYDROLIGHT sample cases were no longer valid. (The original samples were calculated based on another sky radiance model.) Other test cases included using pure water; zeroing the sky radiance (so that only the sun was used); multiple scaled suns and constant sky radiance; using band ratios; and other similar scenarios.

In preparation spring data collection, a water sampling and analysis laboratory was assembled. The required filtration apparatus, sampling bottles and a spectrophotometer were added to R.I.T.’s Digital Imaging and Remote Sensing (DIRS) facilities.

Several more of the individuals identified as original participants in our user group were visited. In addition, two other visits were made to sites where hyperspectral imaging of water has been studied. One of our visits was to Mark Quilter of the Utah Dept. of Agriculture. His main interest is in water run-off from farms. He indicated that his number one issue is silt from erosion. His second issue in order of importance is waste run-off from farms. Figures 3 and 4, digital photographs taken of problem sites, show both problems. The fecal run-off from farms is a dark brown compared to the lighter colored silt. This should show up in hyperspectral imagery, but the imagery will have to be of high spatial resolution (on the order of a meter), since the creeks of interest are relatively small. Mark indicated that one obstacle to getting government agencies such as agricultural departments to use imagery in solving their problems, is that it requires allocation of extra funds for purchase of imagery. The mind-set of many such organizations is that they already have the people, so they would rather send them out to inefficiently collect the data rather than procuring new funds. Unrelated to the water quality issue, Mark said that for precision farming applications, the break-even point of doing ground assessment of crops verses purchasing aerial imagery is $20 per acre.

Another visit was made with Dennis Hoffman of the Texas Agricultural Experiment Station, Blackland Research Center, Temple, Texas. His number one problem is also monitoring of sediments, particularly from Ft. Hood. Other problems in descending order of importance are monitoring of nitrates and phosphates, fecal run-off, and heavy metals. Other mentioned issues are monitoring of hydrilla and other invasive plants, septic tank monitoring, atrozine (a herbicide) monitoring, and industrial waste running into drinking water. The last two are monitored in parts per billion, making them impossible to detect with HSI imagery. The run-off problems are best surveyed right after a rainstorm.

We visited Lee Rickard of the Naval Research Labs. Needless to say, the Navy has great interest in water, and has been doing hyperspectral research with their Hyperspectral Digital Imagery Collection Experiment (HYDICE) instrument. The Navy’s interests include biological constituents, water clarity, mine detection, hazards, detecting currents, and sea surface temperatures. Lee suggested that we contact Curt Davis of the NEMO program regarding getting additional hyperspectral data, which we have done.

Greg Pavlin, of SITAC, a hyperspectral research arm of the Central MASINT Organization, was also kind enough to sit down with us and describe activities in his shop relative to hyperspectral imaging. Through this visit, we learned of a set of hyperspectral data called Island Radiance taken of coral reefs in Hawaii. This data was acquired by the Advanced Airborne Hyperspectral Imaging System (AAHIS), a hyperspectral sensor fielded by SETS. Greg also put us in touch with Fred Portigal of SVS Inc. in Albuquerque, N.M. who has used the Island Radiance data to determine the health of coral reefs. Fred has demonstrated that the hyperspectral data can distinguish between coral and the algae that often kills it. We plan to obtain the Island Radiance data for closer study. Fred also suggested that we contact Gary Borestat of Borestat & Associates on Vancouver Island, who has done monitoring of effluents with the Compact Airborne Spectrographic Imager (CASI) hyperspectral data.

Travel costs have been kept low on most of the above-mentioned trips by making the visits in conjunction with other business in the respective locations.

5. Application and Requirements Assessment

The following table is our best estimate of hyperspectral parameters necessary to satisfy the various water quality applications that we have identified. We expect to refine this table as we assemble more data from our analysis.

A Hyperspectral Water Quality Monitoring Research web page was developed to facilitate communications between the research team and potential users. The web site is: http://www.cis.rit.edu/research/dirs/research/hywater/ This site includes contact information, the First Quarterly Report, and a sampling of slides from the first user group meeting. The site also includes links to the NASA EOCAP, Eastman Kodak, and Digital Imaging and Remote Sensing Laboratory web sites. The page will be periodically updated as the research progresses.