Environmental Products
AAI uses its QSC Water Quality application to retrieve the water composition (column-integrated concentrations of suspended chlorophyll, suspended minerals, and colored dissolved organic matter) and water clarity (wavelength-dependent turbidity and subsurface sighting ranges, and Secchi Depth) for each water pixel in a multispectral image. This can be done for a wide range of water conditions, including deep-ocean, coastal, and inland waters with widely varying water quality characteristics. QSC is fully automatic and it autonomously adapts to varying water quality characteristics both within scenes and scene-to-scene. Measurements are quantitative and derived independently for each pixel, yielding compositional and water clarity (turbidity) "maps." Because of their quantitative precision and accuracy, the spatial patterns of the various water composition and clarity parameter values can be analyzed (e.g., correlations and gradients) to, for example, identify areas of non-point-source pollution and locate their apparent sources (e.g., shoreline sediment influx from agricultural run-off, organic influx from septic system failure, etc.). They can also be used for assessment and monitoring of sediment loading, transport, and deposition, as well as provide a means for quantitative assessment of progress against water quality restoration targets.
How the process works. The QSC Water Quality application is based on solid science, and it has been fully described in a peer-reviewed journal article (download PDF version). The process requires an overhead spectral image with a minimum of four spectral bands at blue, green, red, and very-near infrared wavelengths, respectively. Before using the QSC Water Quality process, the image undergoes an automatic sequence of pre-processing steps. The image is first automatically calibrated from its native Digital Number units to units of material reflectance through removal of atmospheric and sensor contributions using AAI's iCee™ Atmospheric Correction process. Pre-processing next automatically subsets water from land, (including subpixel water separation from mixed pixels) using AAI's Mixed Material Classifier and Material Identifier processes. Next in the pre-processing sequence is the removal of sky reflections and glint from the pixels in the water subset image using AAI's Glint Remover process. The output of the pre-processing sequence is a water subset image in which the water pixels are free of sensor and atmospheric contributions, land and haze contributions, and sky and sun reflections. What remains are the water column and bottom material contributions to the pixel spectrum in calibrated units of material reflectance. The QSC Water Quality application then operates on these pixels to simultaneously retrieve the water composition and clarity characteristics, as well as depth and bottom material spectrum. It does so using a detailed radiation transfer equation as the engine to create a look-up table of predicted water column reflectance values based on combinations of scattering and absorption attenuation coefficients, depths, and bottom spectra. Each image pixel spectrum is compared to the computed look-up table spectra to find the "best-fit" look-up table entry. The process then reports the water composition (three material concentrations), water clarity (vertical and horizontal subsurface sighting ranges, Secchi depth, and turbidity), depth and bottom spectrum that corresponds to the selected "best-fit" look-up table entry.
Landsat 7 Thematic Mapper image-derived suspended chlorophyll concentrations (mg/m3, or µg/l ) in Hongze Lake, China (20 April 2004). The QSC Water Quality application retrieves the concentration of suspended total chlorophyll from each water pixel, yielding a compositional "map" (above).
Each water pixel also has a simultaneously retrieved concentration of colored dissolved organic matter (above), reported as carbon (gC/m3, or mgC/l ). Concentration patterns are products of both organic material influx and in situ degradation of suspended and submerged vegetation. A third simultaneously retrieved parameter is the concentration of suspended minerals (below) in units of g/m3 (or mg/l). Concentration gradients arise primarily from runoff, influx from sediment-laden rivers and streams, and entrained bottom sediments.
The three simultaneously retrieved compositional parameters (concentrations of suspended chlorophyll, colored dissolved organic carbon, and suspended minerals) are used by the radiation transfer equations in QSC to derive the attenuation characteristics of the pixel water column. QSC reports the vertical subsurface sighting range (m), horizontal subsurface sighting range (m), and turbidity (m-1) for each of the four image spectral bands for each pixel. It also reports the Secchi Depth (m) for each pixel.
QuickBird image of coastal Trinidad, CA (26 October 2003), showing the image-derived Secchi Depth (above). The image-derived vertical subsurface sighting range at one of the four spectral band wavelengths (560nm) is shown below for the same image.
Another image-retrieved water quality parameter is Turbidity Confidence, a reported "goodness of fit" parameter comparing the actual pixel reflectance spectrum to the computed reflectance spectrum for the pixel (computed by the QSC radiation transfer formula). For coastal estuaries, where oceanic salt water meets inland fresh water, Turbidity Confidence can be effectively used to estimate the salinity gradient. Salinity systematically affects the pixel spectrum so that salinity gradients can be estimated. The images below show a portion of the Patuxent River, MD. Salt water appears as light blue, while fresh water is dark blue. A close-up of the poly-haline wedge (light blue) is shown on the right. Note that the wedge (light blue) was apparently dipping below the fresh water (dark blue) before re-emerging at its northern extent.
Complementarity to field measurements. The image-derived water quality maps are highly complementary to field measurements. Both are in good quantitative agreement when the field samples are spatially and temporally comparable to the image measurements. Typically agreement is better than 1mg/l for the concentration of suspended chlorophyll, 1mgC/l for colored dissolved organic carbon, and 1mg/l for suspended minerals when the measurement conditions are spatially and temporally comparable. The image-derived concentrations allow the field measurements to be put in spatial and statistical context. They also can be used to assess how representative the field measurements are of the conditions across the water body. There are typically millions of image-derived measurements for every field sample location, and the image coverage is complete, allowing assessment of how representative the field measurements are. The spatial patterns of the image-derived concentrations can also be used to identify potential point and non-point pollution sources and cue sampling locations. Field measurements typically include extra parameters (e.g., fecal coliform bacterial counts, nitrate and phosphate concentrations, etc.) not directly retrievable from the imagery, and they remain a critical component of most water quality assessments. Correlations of these "extra" parameters (e.g., fecal coliform counts) with the parameters common to both (e.g., the concentration of colored dissolved organic carbon) can be used with caution to spatially extrapolate or extend the field measurements to, for example, identify non-point sources associated with leaking septic systems.
| Date | Chl (mg/l) | CDOC (mgC/l) | *TSS (mg/l) |
| 4/10/07 | 5.37 | 6.79 | 30.0 |
| 5/10/07 | 5.90 | 3.89 | 30.8 |
| 6/06/07 | 6.13 | 3.83 | 30.8 |
| 7/04/07 | 6.36 | 3.85 | 34.5 |
| 8/07/07 | 5.76 | 3.92 | 32.6 |
| 9/04/07 | 5.68 | 4.10 | 37.1 |
| 5.87 | 4.40 | *32.6 |
| Date | Chl (mg/l) | CDOC (mgC/l) | SM (mg/l) |
| 4/20/04 | 7.31±2.27 | 5.74±3.62 | 13.21±5.16 |
| 7/09/04 | 4.93±2.54 | 3.74±2.53 | 15.06±8.19 |
| 7/31/06 | 5.24±2.32 | 3.83±1.88 | 5.24±2.32 |
| 5.82±2.38 | 4.44±2.68 | *11.17±5.22 |
Comparison (above) of field-sampled and fully remote image-derived water quality parameters in Hongze Lake, China. Water samples were collected monthly from April to September, 2007 from boats at four locations on Hongze lake by personnel from the Chinese Environmental Protection Agency. Two of the parameters were directly comparable to the image-derived parameters (suspended chlorophyll and colored dissolved organic carbon). *In contrast, the field sample parameter TSS (Total Suspended Solids) included more than SM (Suspended Minerals), consisting of the dry weight of all dissolved and suspended material, and cannot be directly compared to the image-derived sediment-only SM concentrations. The similarity of the monthly 2007 Chl (suspended chlorophyll) and CDOC (colored dissolved organic carbon) field-sampled values to the image-derived values in 2004 and 2006 revealed the annual consistency of overall lake conditions.
Illustration (above) of the utility of the image-derived compositional measurements as a strategic planning and assessment tool for field sampling. The field samples were collected by the Maryland Dept. of Natural Resources at 7 strategically selected locations in four primary reaches of the Patuxent River on 12 October 2004 coincident with the image overpass. Continuous monitoring data (every 15 minutes) were collected at 4 additional fixed locations in the four reaches. Image data statistics for the four reaches represent a million times more measurements, and revealed a trend consistent with the known tidal mixing. Field samples generally lie within 1-2 standard deviations of the image-derived mean Chl concentration (i.e., samples were generally, but not uniformly representative).
Comparison of the image-derived suspended chlorophyll concentrations to the continuous monitoring station results for chlorophyll at the Chesapeake Biological Lab Site XCF9029 in Lord Baltimore's Bay of the Patuxent River on 12 October 2004. The field measurements were taken within 4 minutes of the time of the satellite overpass, and the concentrations agreed
(-.356mg/m3 difference) to within the typical < 1mg/m3 difference limits.
Environmental Products List