Remote Sensing Techniques to detect Surface Water ...

Remote Sensing Techniques to detect Surface Water Quality Constituents in Coastal and Inland Water Bodies from Point or Non Point Pollution Sources Alfonso Blanco, P.E* and William E. Roper, Ph.D., P.E.** *U.S. Environmental Protection Agency Office of Wastewater Management Washington, DC 20460 **Chief Environmental Officer and Director of the Environmental Services Department, Arlington County Arlington, Virginia 22201 ABSTRACT This study demonstrates that point and non-point pollution sources in coastal and inland water bodies can be identified and monitored using remote sensing techniques so early corrective action can be take t opr e ve ntormi ni mi z eHa r mf ulAl g a eBl ooms( HAB’ s ) . By tracking these changes in the remote sensing images could be very beneficial in establishing cleanup and restoration efforts for improving water quality on a watershed basis. Water quality collection protocol has its limitation because of poor accessibility to the sites. The proposed remote sensing technique will validate, and in some cases may replace some conventional methods of water based hydrological monitoring and analysis. Images are analyzed using several algorithms integrated into the Environmental Visualization software such as Partial Unmixing, Minimum Noise Fraction Transformation, and Spectral Angle Mapper. Spectral angle mapper was the principal method used in this study. The method will complement other airborne or satellite remote sensing technologies being developed to locate the origin of point and non point sources entering the upper Chesapeake Bay Watershed. Federal, State, and Local agencies, as well as Non-Profit organizations need accurate information for identifying impaired waters, prioritizing response actions, and developing long term restoration plans. This study shows a baseline approach that would allow agencies to update their water quality information, using satellite images, track changes and perform trend analysis on a continuous basis. KEYWORDS Remote sensing, Hyperspectral sensors, water pollution, image analysis, water quality monitoring, spectral analysis, spatial analysis INTRODUCTION Current techniques for measuring water quality involve in situ measurements and/or the collection of water samples for subsequent laboratory analyses. While these technologies provide accurate measurements for points in time and space, they are expensive, and do

1

not provide either the spatial or temporal view of water quality needed for monitoring, assessing, or managing water quality for an individual water body or for multiple water bodies across the watershed. Remote sensing of water quality indicators offers the potential of relatively inexpensive, frequent, and synoptic measurements using airborne or satellite sensors. Federal, State, and local environmental agencies need to collect water quality data on a large geographical framework in order to plan and design watershed restoration plans. Collection of water quality data by conventional sampling methods is not very cost effective and time consuming. Remote sensing can be very successful because spatially data can be collected faster and less expensively than by conventional methods. Remote sensing techniques can provide for coverage of large geographical areas for identifying and characterizing impacts at a variety of study scales with data that are consistent, accurate, and objective. Currently, the assessment of spatial and temporal variations is impractical in many watershed areas because of limited access and high costs of monitoring. Potential immediate applications of remote sensing technology could be linked to ongoing projects in the Chesapeake Bay Watershed, since nutrient loadings are a product of point and non-point sources. Preliminary contacts with several agencies (US Army, MDDNR, MDER, USGS, and Chesapeake Bay Program) [Aberdeen, 2007] have revealed a high priority for the use of remote sensing technology in water quality identification and characterization because these techniques will meet a well-defined need and will be potentially transferable to other watersheds. One of the major expected benefits for the user of the remote sensing database is that agencies can input their water quality data, synchronize them with remote sensing images with the actual water quality data results and share the data with other agencies aiding in the restoration efforts by tracking the health conditions of those watersheds. BACKGROUND The Chesapeake Bay Watershed (CBW) was the first watershed to be targeted for restoration and protection by the EPA under the Clean Water Act resulting in listing s e ve r a ls e g me nt sa ndi t st i da lr i ve r sa s“ i mpa i r e dwa t e r s ” .[Chesapeake Bay Foundation, 2007] These impaired segments included the Gunpowder and Bush River in the upper western shore of the Chesapeake Bay. The CBW is a very dynamic environment because their, biogeochemical processes mold the marine environment, they also alter the optical water properties. These effects can be observed in the watercolor, or the surface reflectance. Excess sediment has been identified as the major factor affecting water clarity in the Bay. The major pollution sources are point (PS) and non-point sources (NPS). Point sources are direct treated discharges from wastewater treatment plants, power plants or failing septic systems. Nonpoint sources are uncontrolled discharges such as agricultural run-off, Confined Animal Fe e dl ot s( CAFO’ s ) ,e r os i ona nds e di me nt a t i on,i ndus t r i a la c t i vi t i e s ,l a ndde ve l opme nt and other uncontrolled activities. Locating and characterizing the chemistry of

2

discharges on the receiving streams can be difficult, time consuming, and costly due to inaccessibility to the sampling sites. Study Area Carroll Island is located in the upper western shore of the Chesapeake Bay Watershed south of the Aberdeen Proving Grounds Military Army Base at latitude 39.3225N and longitude-76.3425W3 [Water Resources, 2007]. Carroll Island is located in the town Edgewood in southeastern Baltimore County as shown on Figure 1 and 2 below [Carroll Island, 2007]. This island has a surface area of 855 acres and is located between Saltpeter and Seneca Creeks which connects on the west side to separate the island from the mainland. The island is flat, low lying area, inhabited by wildlife and covered with forest, open fields, and marshland. Carroll Island was at one time a testing site for chemical agents. The Gunpowder River is located on the north side of the Carroll Island, and it is classified as oligahaline (low salinity or brackish waters) with an SAV acreage restoration goal of 572 acres at a 2-meter Secchi Application Depth. Water quality and Submerged Aquatic Vegetation (SAV) distribution and abundance are a good indicator for the health of the ecosystem. The US Army Environmental Staff at Aberdeen Proving Grounds had found thirteen species of SAV, but there are three most commonly found milfoil, cottontail, and wild celery [Williams et al., 2000]. Figure 1- General Study Area

Figure 2- Aberdeen Proving Grounds Site Map

GOALS AND OBJECTIVES The purpose of this study is to demonstrate remote sensing techniques that use spectral reflectance technology for identifying and characterizing surface waters affected by point or non-point sources. These methods are used to extract surface reflectance for the identification and characterization of water quality constituents such as Chlorophyll a, Suspended Sediments, and Colored Dissolved Organic Carbon (CDOM). Spectral data from both ground and aerial or satellite mounted sensors can be collected during summer 3

months at selected sites. Water quality samples at each of the sites can be collected concurrently with the spectral data and the orbit cycles or flyovers from the satellites or airborne sensors. The ground truth data can be used to verify that spectral reflectance is capable of differentiating the different water column constituents. Specific objectives are: (1) To correlate aerial surface spectral reflectance imaging data with ground spectral reflectance data to verify that the images observed from the air are comparable to those recorded on the ground; (2) To quantify the relation between key water quality measurements collected in the water and the over head collected spectral reflectance data. ASSESSMENT OF THE PROBLEM The EPA CBW Program established a watershed wide nutrient reduction goal to achieve a forty per cent (40%) reduction of phosphorus and nitrogen discharges into the Chesapeake bay by the year 2010 [Langland et al., 2000]. The Chesapeake Bay Report of 2003 suggested that pollution loads exceeded in concentration amounts by over 2 ½ times for nitrogen and almost two times for phosphorus The main sources of nutrient pollution are derived from manmade activities such as Point Sources which are treated, partially treated, or untreated discharges from a wastewater treatment plant or septic tank into a water body. Non-Point Sources (NPS) which include agricultural run off (fertilizers and pesticides) animal feedlots, septic tank failures, stormwater runoff, erosion from land development activities and other related surface discharges. If excess nutrients are present in the water column such as nitrogen and phosphorus, it will clog the SAVs causing algae a bloom, preventing sunlight from reaching the aquatic plants and animals. Increases in nutrients such as Phosphates and Nitrates, as well as Chlorophyll a, Turbidity, Total Suspended Matter, and Colored Dissolved Organic Matter are symptoms of eutrophication [NRC, 2001].Total Suspended matter affect water quality because it absorbs nitrogen, phosphorus, and organic compounds that can be water quality indicators This decline in water quality primarily is due to nutrients and sediments which is characterized by excess concentration of Chlorophyll a (Chl-a) reducing water clarity, and depleting the dissolved oxygen. In order to asses this eutrophication process ; specific criteria such as a reduction in water clarity, low dissolved oxygen, food supply imbalance, proliferation of species will need to be established for designated uses [Bukata et al., 1995]. Biogeochemical Processes Eutrophication is defined as the state of having high nutrient content and high organic production. The upper Chesapeake Bay is affected by these conditions of oxygen depletion, especially where there are SAV beds. This eutrophication process decreases water quality by promoting excessive growth. It has been shown that in the upper Chesapeake Bay a reduction in water clarity and low dissolved oxygen conditions improves when excess phytoplankton or algae bloom, measured as Chl-a is greatly reduced. Sunlight penetrates the water column to a certain depth called phonic depth, meaning the sunrays penetrate the vertical distance from the water surface to 1% photic subsurface irradiance level [Lunetta et al., 2006]. The phytoplankton within this phonic depth consumes nutrients and converts them into organic matter by photosynthesis, which

4

is called the primary production level. Zooplankton consumes the phytoplankton creating organic matter and the bacterioplankton decomposes the organic matter even further, introducing Colored Dissolved Organic Matter (CDOM) in the water body. The breakdown of phytoplankton cell yields carbon dioxide (CO2), inorganic nitrogen, and sulfur and phosphorus compounds. The more these phytoplankton are produced, the greater the release of this CDOM. Humic substances once they are released will have a colorant agent in the water column called yellow substances or gelbstoff, which can impact the absorption and the scattering of light in the water column, as well as changing the color of the water. CDOM and Chl-a can be used to further predict the occurrences of Harmful Algal Bloom (HAB), indicate the input and distribution of organic matter, and contribute to the biochemical processes (see figure 3). Figure 3- Components of the Chesapeake Bay Ecosystem [USGS, 2006]

Chlorophyll-a (Chl-a). Chl-ac r i t e r i aha sbe e nde f i ne da s“ t hos ec onc e nt r a t i onsofChl -a in free-floating microscopic aquatic plants (algae) that shall not exceed levels that results in ecologically undesirable consequences. Chl-a relates to phytoplankton biomass and is the single most responsive water quality indicator of other surrogates substances like Ni t r og e na ndPhos phor usf oundi nt heChe s a pe a keBa ywa t e r s he d”[USGS, 2006]. Chlorophyll when introduced into pure water changes the spectral reflectance characteristics as shown on Figure 4. shows how the spectral signature of plankton in the water is different from clear water sample [Maryland, 2006]. Chlorophyll is found in phytoplankton, which absorbs most of the incident blue radiant flux, causing the photosynthetic portion to appear dark. Phytoplankton contains carbon and they sink to the bottom of a water body once they die, and other sediments will cover the carbon in the dead phytoplankton. Phytoplankton uses carbon dioxide and produce oxygen during the

5

photosynthetic process. Water bodies act as a carbon sink. All the phytoplankton in the water bodies contain photosynthetically active pigment Chl-a. However, other chlorophyll types could be present at different depths (Chlorophyll c, d, and e) [USGS, 2006]. The r ec oul dbeot he rphy t opl a nkt on’ sphot os y nt he s i z i nga ge nt spr e s e nts uc ha s Carotenoids and Phycobilins [Water Quality, 2007] that suggest that Chlorophyll a may be considered a reasonable surrogate for organic components [Chesapeake Bay, 2004]. Be c a us et he s ephy t opl a nkt on’ sha v edi f f e r e ntc onc e nt r a t i onsofChl -a, they tend to appear as different colors to sensitive remote sensors. A pure chlorophyll-a spectral data maybe able to be separated from t hei nt e r f e r e nc eofTSM’ s .Spe c i a la t mos phe r i cc or r e c t i on techniques to separate the spectral data from Chl-a and a complex multiple-component extraction methodology [Shafique et al., 2002] or another technique, which uses a derivatives spectra techniques. Studies have shown a correlation of local in-situ measurements of Total Suspended Matter, Chlorophyll-a, etc with the spectral data acquired from remote sensing. There are local algorithms that are only good for a specific location and cannot usually be transferred through space or time. Transportable algorithms are the ones that can be spatially and temporarily invariant, meaning that these algorithms can work mostly anywhere and at anytime, so these algorithms can be applied to any satellite sensors to produce a map of Chlorophyll-a and TSM Figure 4-Typical Spectral Signatures of concentrations on a routine basis. In Coastal Waters ( Clear Blue Waters (Bl), this research, study a comparison Blue-Green Waters (BIG), Green Waters will be made of the images from (GR), and Brownish-Green Waters (BrG) different sensors at times which will [Reef et al., 1997] tell what changes occurred over time and the processes at work. 18 It has been documented that Chl-a reflects 2% between 400 and 500 nm [Jensen, 2000a] and drops gradually to 1 % at wavelength beyond 710 nm in the Visible-Near Infrared (VNIR). It is documented that in algae contaminated water there are four (4) scattering/absorption features of Chlorophyll-a [Harding, 1997] 1. Chl-a absorption of blue light between 400 to 500 nm 2. Chl absorption of red light at 675 nm (VNIR) 3. Maximum reflection about 550 nm (Green Peak) caused by relatively lower absorption of green light by algae. The Case 2 waters surrounding the Carroll Island have been observes as sediment-laden waters, which has a brownish-green color as seen in figure 5. 4. Surface reflectance peak around 690 to 700 nm caused by an interaction of algaecell scattering and minimum combined effect of pigment and water absorption

6

As the Chl-a concentration increases in the water column, a significant decrease is reflected in the blue and red regions of the electromagnetic spectrum, but an increase in green wavelength reflectance occurs. It is expected that Chl-aa ndTSM’ swill be present in the water column a tt heSAV’ sl oc a t i ons[Jensen, 2000a] which will produce a different spectral signature than a pure chlorophyll water sample. It has been documented that in an algae contaminated water, the peak surface reflectance in the visible region will shift from 547 nm (green light) at 0 mg/l to 596 nm (orange light) at 500 mg/l. obtained accurate estimates of algal chlorophyll pigment amount in water surface using a ratio of NIR 705 / Red (670) nm when Chl-a concentrations was low. Computing the first derivative of reflectance around 690 nm produced the best results that Chl-a concentration was high [Bukata et al., 1995]. Turbidity. Turbidity has been defined as a unit of measurement that quantifies the degree to which light travels through a water column and is scattered by organic and inorganic particles including the algae. Turbidity is related to Total Suspended Matter SM and Dissolved Matter, like clay, silt, finely divided organic matter plankton and other microscopic organisms, organic acids and dyes. The more suspended particles found in the water column, the more scattering of the sunlight will occur. The waters close to Carroll Island maybe sediment contaminated waters and they are referred as turbid .Turbidity can be measured by sampling it and analyzing with a turbidity meter or by a Secchi disk. A Secchi disk measures the transparency of the water body. Nevertheless, as impurities entered or are formed in the water body, its spectral characteristics change. Sediments can enter from natural sources or man made activities as previously mentioned and they consist of erosion of silts and clays, run off, which can be suspended on the water surface. Turbidity was measured at those SAV locations by MDDNR and the US Army personnel from Aberdeen Proving Grounds (APG). Figure 5 -Turbidity on the Gunpowder and Bush Rivers Gunpowder River

Bush River

Carroll Island

Submerged Aquatic Vegetation. SAV’ sorUnde r wa t e rBa yGr a s s e s ,a st he ya r ec a l l e d, are plants that grow underwater and are found through out the Chesapeake Bay

7

Watershed include tidal and non-tidal waters. About 13 classes of grasses have been identified in the CBW. [Maryland Bay Grasses, 2006] Figure 6 SAV Sampling Locations by APG

Figure 7 APG SAV Areas

Carroll Island

The s eSAV’ ss e r vema nyi mpor t a n te c ol og i c a lf unc t i onss uc ha si mpr ovi ngwa t e rqua l i t y , s uppl i e sf ooda nds he l t e rf orpl a nt sa nda qua t i cl i f e .SAV’ sr e l e a s eoxy ge ndur i ng photosynthesis by absorbing the nutrients (Nitrogen and Phosphorus) and converting t he mi nt oor g a ni cma t e r i a lt ha ti sus e dbyt hepl a nt .SAV’ si nhi bi twa vea c t i ont ha t erodes shorelines, so it stabilizes the sediment at the bottom sediment. Water quality decreases because of excess nutrients and sediments, which causes significant losses of SAV’ s .Ther e s t or a t i onf ort heMa r y l a ndDe pa r t me ntofNa t ur a lRe s our c e s ,a swe l la st he US Army APG is a priority of the Chesapeake Bay Restoration Program. The US Army at APG has been performing water quality sampling a tt hel oc a t i onswhe r eSAV’ sha ve been found. The Carroll Island area has a history of poor water clarity, which continues to degrade the SAV’ sbe ds ,a ndt hi si st her e a s onf ors e l e c t i ngt hi spa r t i c ul a rs i t ebe c a us et hi smode lc a n be duplicated across the CBW and in other similar watersheds. It is expected than during the validation of spectral signatures surface reflectance for Chlorophyll a is expected to be the highest at the SAV sampling locations. It has been documented that in the Chesapeake Bay the absorption band depths were 681 nm and at 574 nm were most pronounced for Milfoil and Wild Celery. It was determined that this difference is due to the interaction of the SAV plant canopy and sunlight .In this research study, five (5) sampling locations will be georefenced and spectral analysis will be performed. Ground spectroscopy measurements will be obtained and direct water quality readings for Chlorophyll-a, Total Suspended matter, will be performed. TECHNICAL APPROACH Principles of Aquatic Optics

8

When sunlight penetrates the water column, the light intensity decreases exponentially as the depth increases and this is called Attenuation. In the visible light region, which is the red portion, attenuates more rapidly than the shorter wavelength of the blue region. Spectral radiance recorded by a sensor is dependent on the surface reflectance and the depth (see figure 8). Attenuation result from absorption and back scattering. Phytoplankton in water converts the electromagnetic energy into photosynthesis. Other absorbers are inorganic and organics resulting from the breakdown of plant tissue, and water itself, which absorbs red light and has a smaller effect on shorter wavelength (blue color of clear water). Chlorophyll in algae will look green because reflects in the central region of the visible spectrum (green) and absorbs strongly at either end of the spectrum. Dissolved organic compounds absorb at the shorter wavelengths (blue) and strongly reflect in the yellow-red-end. Electromagnetic radiation may interact with TSM in the water column and change its direction. This scattering process is caused by inorganic and organic constituents and will increase with turbidity. In water quality applications, we do not have neither transmittance nor reflectance .Waters surrounding Carroll Island are not pure, like ocean waters because they have organic and inorganic constituents, especially at the surface which causes interference of the sunrays. The reason been is that these constituents will cause NIR surface reflect and subsurface volumetric scattering that the amount of NIR radiant flux leaving the water surface will increase. TSM will cause significant scattering and reflectance of the radiant flux to the sensor causing brightness. This is called sunglint, which needs to be removed from the image [Jensen, 2000b]. Figure 8- Electromagnetic Spectrum [33]

When the light photons enter the water column, they are influenced by absorption and scattering by pure water like Case 1 waters and by scattering, reflection, and diffraction by particles that are suspended on the water like in Case 2 waters. Scattering in Case 1 water are small with relation to wavelength. Comparisons can be made between open ocean wand coastal –inland waters. Maximum transmittance of light in clear water will occur between 440 and 540 nanometers, peaking at 500 nanometers [Zhang, 2005]. The color of the water is determined by volume scattering rather than by surface reflectance, spectral properties are determined by transmittance rather than by surface characteristics alone. Penetration of sunlight maximum in the blue region (400 to 446 nm) but at a slightly longer wavelength, the penetration is much greater, for recording the bottom of

9

the water body. However, at longer wavelengths, red region, the absorption of sunlight is much greater, and only shallow features can be detected. Finally, in the Near Infrared Region (NIR) 700 nm absorption is so great that only land-water distinction can be made. Coastal –Inland (Case 2) waters contain suspended sediments, dissolved organic matter. Coastal zones are mostly classified Case 2 category. Case 2 waters is the classification of the water bodies surrounding Carroll Island, creates a difficulty due to complexity of substance mixtures, unstable atmospheric residue and noise. Water-leaving radiances used for the extraction of upper ocean constituents represent no more than 10% of the total radiance captured by satellite optical sensors. The atmosphere dominates the readings. An atmospheric correction may be performed to remove the contribution of the atmosphere in satellite measurements. In Case 2 waters, there is a significant water leaving radiance in both, the visible and near-infrared regions. Current operational ocean color algorithms are not suited to analyze Case 2 waters. Some algorithms additionally apply a turbid water test on 555 nm channel normalized water leaving radiances after the full atmospheric correction. Then, they just label isolated Case 2 waters because chlorophyll estimates in Case 2 zones are unreliable [Gitelson, 1992]. The overall brightness will increase in the visible region so the water ceases to be a dark object and becomes bright as sediments concentration increases, the wavelength then shifts from the blue to the green region of the electromagnetic spectrum. For different classes of water depth the intensity of radiation will decrease exponentially with depth, in other words brightness decreases as depth increases .Dark waters mean bottom is deep, brighter is shallow. For ocean waters, the definition of Color Index is a qualitative measure that is the ratio of NADIR Radiance in the water at the blue wavelength to that of the green wavelength of the spectrum. Therefore, Chl-a can be estimated meaning that from ratios of spectral radiance, a retrieval algorithms can be developed The algorithm for Chl-a is as follow [Bukata, 2005]. 2 Chl-a = X1 [L ( λ 1)] X [L( λ 2)]

Log Chl-a = aLog L λ 1 Lλ 2+ b

Chl-a = Chlorophyll concentration L( λ 1) = Upwelling radiance L( λ 2) = X1 and X2 are empirical constants

The retrieval algorithm Cl = aX + b Cl is the concentration for Chl-a, suspended matter, Colored Dissolved Organic matter (CDOM) a and b are constants X = Channel ratio For inland and coastal waters (Case 2 waters), a relationship can be established between bulk and specific inherent optical properties. Total absorption coefficient a ( λ )=a w( λ )+a Фλ+a d( λ )+a g ( λ ) w = water Ф =phy t opl a nkt on d = detritus y = CDOM (gelbstoff)

10

In order to describe optical properties of absorption and backscattering of Case 2 waters, the ratio of the retrieval algorithm for 2 wavelengths that the ratio for the above retrieval algorithm cannot provide for accurate data. The approach to use multispectral information is by using linear regression analysis and neural network algorithm. For the research approach it is necessary to identify the organic and inorganic constituents in the water column such as Chl-a and TSM by separating the subsurface volumetric radiance This equation is interpreted as the reflected energy Er is equal to the EI ( λ ) - EA that is either absorbed or transmitted. Lv = Lt –(Lp + Ls + Lb)

Lt = Lp + Ls + Lv + Lb

Atmospherically correction of the sensor data is needed by removing the atmospheric attenuation (Lp) surface glint as previously mentioned and other surface reflectance (Ls) an bottom reflectance. (Lb) [Bukata, 2005] This Subsurface Volumetric Radiance (Lv) is the product from the down swelling solar and sky radiation that actually penetrated the air-water interface, interacting with the water and organic/inorganic constituents and then exits the water column towards the sensor without even reaching the bottom surface. Lv is a function of the concentration of pure water (w) ,inorganic suspended minerals,(SM), organic Chlorophyll a (Chl-a) ,dissolved organic material (DOM) and the total amount of attenuation of absorption and scattering taking place inside the water column. Lv = fwc ( λ ,SMc( λ ,Chla( λ ) ,DOMc( λ )[Bukata, 2005] Figure 9- Water Leaving Radiances [Jensen, 2000c]

This is the portion of the radiance which is been recorded at the sensor which results from the downswelling solar (Esum) and Sky (Esky).radiance which does not reach the water surface. Ls is the radiance from the downswelling solar and sky radiation that reaches air-water interface (surface layer or boundary layer. Only penetrates about 1 mm into the water column and is reflected back (figure 9). Lv is from the downswelling solar and sky radiation that do penetrate the air-water interface, intersects with the water constituents and exists the water column without touching the bottom. This is what is called interface volumetric radiance [Bukata, 2005] . Lv = Lt –(Lp + Ls + Lb).Lb

This is the portion of the radiance been recorded which results from the downswelling solar and sky radiation that penetrates the air-water interface and reaches the bottom and is propagated back through the water column, and finally. In aquatic remote sensing, to extract the radiance from all of the components

11

recorded by the sensor, the organic and inorganic water column c ons t i t ue nt s ’spectra will be collected. By correcting the sensor data radiometrically, the atmospheric attenuation (Lp), surface sunglint and bottom reflectance (Lb) be removed [Jensen, 2000c] Hyperspectral Imaging Systems Hyperspectral imaging overcomes spectral and spatial limitation that other remote as not bands or the bandwidth are too broad, or between individual or landscape scale features, or modifications of these features based on spectral signatures ,size, and spectral configuration high resolution remote sensing provide unique capabilities in detecting a variety of features like the type of chlorophyll ,suspended matter, water column constituents, potential sources of non-point sources pollution [Green et al., 2000]. Figure 10 shows an illustration of the hyperspectral imaging concept where the hyperspectral imaged data consisting of over a hundred contiguous spectral bands forms a three – dimensional (two spatial dimensions and one spectral dimension) image cube. Each pixel is associated with a complete spectrum of the imaged area that will identify the materials in the pixel. The high spectral resolution of hyperspectral images enables better identification of the images. A hyperspectral imaging passive sensor was used to analyze the collected images at Carroll Island. The Airborne Visible Infrared Imaging Spectrometer (AVIRIS), is ma na g e dbyNASA’ sJ e tPr opul s i onLa b, and is an optical sensor that uses a scanning mirror to sweep back and forth (whiskbroom) producing 614 pixels that delivers calibrated images in 224 contiguous spectral channels or bands with wavelengths ranging from 400 to 2500 nm, with an spectral resolution of about 10 nm, and a ground pixel size of 18m x 18m. This instrument flies aboard a NASA ER2 airplane at about 20,000 m above sea level [Green et Figure 10- Hyperspectral Imaging Concepts al., 2000]. (Courtesy of JPL) The main objective of AVIRIS is to identify, measure, and monitor c ons t i t ue nt soft heEa r t h’ s surface and atmosphere based on molecular absorption and particle scattering signatures When the data from each sensor is plotted, it yields a spectrum of which is compared with the spectrum of known spectral libraries such as JPL, ASTER, John Hopkins revealing the information about the composition of the area being viewed by the instrument. Each pixel covers 20 m2 with some overlap yielding a ground swath of 11 km wide. The data is processed and stored on the ground; yielding about 140MB for every 512 scans or lines of data, corresponding to about 10 km long on the ground. AVIRIS

12

records several runs of data known as flight lines and each flight line has 1 to 25 runs and each run has 1 to 40 scenes. Figure 11 shows a flight line over Carroll Island in Aberdeen Proving Grounds. The AVIRIS scene of the Carroll Island in the Aberdeen Proving The AVIRIS image was acquired by the NASA Jet Propulsion Laboratory (JPL) in November 21, 2002 for Chesapeake Bay with start-latitude N + 39.31 and stop-latitude N+ 39.327 and start-longitude -76.328 W and stop-longitude. -76.365.W. The flight number was f020912t01p00r03. The image was provided by Robert Green Jet Propulsion Lab. The image shows darker a r e a sonl a ndwhi c hi sve g e t a t i ona ndi nt hewa t e r ,i ta ppe a ra sSAV’ s .The r ea r es e ve r a l access roads thru the center of the island. .Sediment plumes and water movements are shown on the Figure 11 below. A power plant is located west of the island and connected by an accesses road. Multispectral Imaging Systems Multispectral data was acquired from the Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER), which is an advanced multispectral imager, launched by NASA in 1999aboard the TERRA spacecraft. ASTER has 14 multispectral bands from the visible to the thermal infrared with a high spatial, spectral and radiometric resolution with three (3) Figure 11- AVIRIS Image of Carroll Island hyperspectral bands 1, 2, and 3 sensitive to green, red, and near-infrared, respectively ,operating in the VNIR range. It has an additional backward – looking near-infrared band, which gives stereo coverage. The spatial resolution varies with wavelength, 15 m for the visible wavelength and Near-Infrared (VNIR), 30 m in the Short Wave Infrared (SWIR), and 90 m in the Thermal Infrared (TIR). ASTER has a Field of View of 60 x 60 km. TERRA spacecraft flies in a circular, near –polar orbit at an altitude of 705 km in a sun-synchronous with equatorial crossing at local time of 10:30 a.m. returning to the same orbit every 16 days. The Terra orbits will be downloaded from web site http://ssee.miscredu/datacenter/terra.na2006_11_37_331_gy. The satellite orbits are synchronized with overpass over the study site within an Instantaneous Field of View (IFOV) of 15 degrees. It would be necessary to establish a sequence of overpasses with climatic conditions for the area ahead of deploying equipment and manpower. This effort will be tracked using a 5-day weathercast for the area thru the Aberdeen Proving Grounds Weather or the Baltimore-Washington International Airport (BWI). The TERRA satellite orbits the globe every 16 days.

13

Images will be selected from the orbit at approximate time that the satellite passes over the Chesapeake area allowing for an Infield of Vision (IFOV) of 15 degrees. The ASTER images will be accessed from the USGS Earth Resources Observation and Science (EROS) Land Processes Distributed Active Archive Center Datapool@LPDACC which is a website hosted by USGS-NASA . This data pool is an online images archive, which provides direct FTP access to LP DAAC data products. The Data Pool contains ASTER L1B V003 data covering the US and global MODIS data. ASTER products are kept in the data archive for 2 years. The ASTER L1B data is registered radiance at the sensor, which can be directly downloaded to a desk or laptop. This product contains radiometric ally calibrated and geometrically co-registered data for the acquired bands of the three different telescopes of Level 1A data. ASTER Level 1B image needs to be georeferenced to True North by using ENVI 4.3 [ENVI, 2007] version by rotating the image Figure 12- ASTER VNIR 15 m resolution of Carroll using the value of the Island [USGS, 2007] Map Orientation angle. This ASTER L1B 15 m resolution image is only shown in the red and green bands because this is due to historically reasons because early remote sensing work used infrared-sensitive film because healthy vegetation strongly reflects those wavelengths. The human eye cannot see infrared, but some visible color has to be used to represent it if the images are going to be of some use. For the infrared-sensitive film, the color is red, and red is used to represent the infrared region since, even for digital images that use no film like ASTER. The second reason is that ASTER does not have a band that detects blue light because bl uel i g h t st e ndst obes c a t t e r e dmos tofi tt ot hea t mos phe r e ,t he r e f or ear e a l“ na t ur a l c ol or ”i ma g ei snotpos s i bl e .Thet hi r dr e a s oni st ha tt hi si st hewa yt hec ol ora s s i g nme nt s are made. The color red is assigned to band which is sensitive to the part of the infrared spectrum, green is assigned to band, which is sensitive to red, and blue is assigned to band 1, which is sensitive to green. In the above image, the ground reflects highly in band 3 appearing bright red, the water image, which reflects highly in band 2, will appear bright green, and band 1 is not visible. Bands 3 and 2 predominates this image DATA COLLECTION PROCEDURES

14

For field collection the surround vegetation, physical characteristics of the surface water, weather data such as wind direction, ambient air temperature, etc were entered in a log book. The weather data information was collected from the Baltimore-Washington Airport or the US Army at Aberdeen Proving Grounds. Using this weather data from the same dates and times that the satellite images are acquired. Certain equipment were used for the data collection GPS Unit; Fixed Secchi Disc, YSI Instrument; Spectrometer with a light calibration target; Camera with 400 ASA Film; Tape Measure. All the data records from the GPS File Name will be cross referenced with the spectral data which corresponds with it, even though, the data is been entered electronically. A file naming system will be established before going to the field. Each water quality measurement, Sampling Point Number, and the time collected were also entered into a logbook. A Data Dictionary will be set up for the GPS and YSI in order identify predetermined features. The Secchi Disc will be lowered into the water until is no longer visible. Measurements are critical for monitoring temporal changes in clarity, but provide little information on spatial variations in clarity across and around the water body. Water quality measurements for the water constituents will be performed using a direct reading instrument such as the YSI instrument Model 6600 V2-4 Stoned with 4 optical ports which can measure Chlorophyll, Blue-Green Algae, Turbidity, Dissolved Oxygen. The YSI Model 6600 equipment has been installed by the Maryland Department of Natural Resources at a nearby location and is measuring Chlorophyll and Total Suspended Matter continuously. The advantage of this equipment is that several water quality parameters can be analyzed at the same time instead of preserving, transporting, and analyzing the samples to a laboratory [YSI, 2007]. Water quality measurements will bec ol l e c t e dwhe r et hel oc a t i onso ft heSubme r g e dAqua t i cVe ge t a t i on( SAV’ s )ha sbe e n located on the SAV Location map supplied by APG as previously shown on Figures 6 and 7. During this procedure, spectral signatures of the surface water will be taken using a Spectrometer Field Spec (ASD) to match these signatures with the spectral signatures extracted from the ASTER and AVIRIS hyperspectral image. Correlation between the groundtruth measurements, water quality sampling, and the ASTER and AVIRIS mages will be analyzed. Field collection will involved a Secchi disks, GPS units, Field Spectrometer and the fixed YSI equipment Ground Spectroscopy Measurements The Spectrometer is manufactured by Analytical Spectral Devices, Inc (ASD) [Analytical Spectral Devices, 2007] and will be provided by George Mason University Physics Laboratory. The instrument will collect surface reflectance of the water between 325 and 1075 nanometers (nm). Correlation between the ground truth measurements, direct water quality readings, and the remote sensing images will be analyzed for validation purposes. Collected field data will include latitude, longitude, Secchi disc depth, chlorophyll-a, and total suspended matter (TSM). Field spectra measurements would need to be collected 2 hours before and 2 hours at NADIR. The reason been is that the sun angle changes around noon time , so it is better to limit the data collection to that time window from 10 A.M to 2 P.M. However, weather conditions can make this limitation somewhat

15

impractical. Ideally, spectral measurements will be taken at the time that the satellite crosses over the exact latitude and longitude of the sampling points to determine how large a difference it would make the ground truth measurements and the remote sensing data. The minimum size of these target areas should exceed the pixel resolution of the imaging from 9 to 25 pixels or even more. Calibration of the instrument will be done using a flat aluminum plate called Spectralon. These three readings are repeated which will finish the collection procedures at each site. This synchronization may not be possible since there maybe variations with factors such as weather, location, and the exact timing. Spectral measurements may change if there are cloud covers, which create for a variation in the intensity of the sunlight. In case a cloud cover is present during field measurements, calibration will need to be performed every10mi nut e si na c c or da nc et ot hema nuf a c t ur e r ’ s recommendation. Consequently, this technique will increase the quality of the data measurements in order to prevent a drift and recalibration of the instrument for changes in the sun angles. Measurements will need to be made early in the day, if cloud conditions are expected in the afternoon and correct for the solar angles. Weather and meteorological data will be collected. Shadows will be avoided during field measurements because shadows will cause variations in the overall brightness of the reflected spectral signatures. It is expected that the SAV sampling locations should not have any obstructions since the location is facing open water. These Sampling locations may or may not need to be changed on the field in order to obtain maximum surface reflectance. The GPS locations will be obtained and twenty five (25) Spectrometer light readings [Zaraeri, 2007] will be collected as the first step. Then light readings consists of an open sky readings, one with the direct sun blocked (technician holding a shade object) and a final reading of just the reflected light from the technician and no shade. Spectral measurements of bright calibration targets areas within the study site. General notes will be taken during the readings of physical features surrounding the sampling site. Water clarity readings are to be taken with the Secchi Disc (Recording the depth below water surface that the disc is no longer visible) and a turbidity meter reading is also taken simultaneously. The Secchi depth will be used to determine the depth from the surface for recording the Spectrometer data. Spectrometer will be used to obtain single downward facing spectra, directly over the water. Before this surface reading, a calibration (white) reading will be obtained using a special reflective target called Spectralon In order to develop a true spectral signature without any interference; a sample of the water will be collected in a clear flask and a spectral reading taken with the FieldSpec (ASD) instrument. Additional spectral readings will be taken of the flask with distilled water and the empty flask. Readings will be taken of the 3 flasks in order to deduct and obtain a true spectral signature of the water [Roper, 2005]. The following photos (figures 13, 14, 15 and 16) were taken on a trial run for recognizance of the sampling area and go over the sampling protocol.

16

Figure 14-Col l e c t i ngs ampl e sofSAV’ s

Figure 13- SAV Sampling site

Building the Spectral Library Direct spectral readings will be taken from a boat and they will be collected at the SAV sampling locations. The surface water reflectance spectra and underwater radiance or irradiance spectra will be collected by the FieldASD. The pistol of the ASD will be held approximately 18 inches from the water surface and spectra will be collected every 25 seconds. Figure 15- FieldSpec Spectrometer

Figure 16- Taking Spectral readings on the water surface

The Field Spec will take about 25 spectra readings and average them at the same time that the water quality readings are been collected by the YSI Instrument. Spectral signature profiles will be extracted and analyzed in-situ with the Field Spec that will register the surface reflectance of Chlorophyll a and TSM. A Spectral Library will be built for Chlorophyll a and Total Suspended Matter and the spectra data will be saved by the instrument for transferal later on to a database. A reference spectral signature of the water sample will be collected in a clean, clear flask and the spectral signatures of the empty flask and of the water sample flask will be measured. A true spectral signature will

17

be the difference of both spectral signatures. The spectral readings will be performed once per month from June 2007 to September 2007 because this is the season the chlorophyll is active. After September, the concentrations decline with the lowering of the water temperatures. The equipment used includes the Field Spec Spectrometer (ASD), Secchi Disc, YSI Model 6600,, Camera, and GPS unit. Each will be synchronized for recording the time and date of the data collection. A trophic state analysis on the field data will be performed to give a background understanding to the water quality parameters being extracted. Analysis of Results Results will be plotted and data will be compared with historical water quality data collected by the Maryland Department of Natural Resources and APG that can be downloaded from their website [Chesapeake Bay, 2007]. A linear regression analysis and Neural Networks will be performed and graphs plotted of the water quality concentrations for Chl-a and TSM. The concentrations will be plotted on the maps. INNOVATIVE ASPECTS OF THIS STUDY. This study will establish a historical water quality database using several multispectral and hyperspectral sensors such as ASTER, and AVIRIS; however, at a later date this may be include Hyperion 15 m spectral resolutions. By using a variety of hyperspectral techniques, analysis will provide innovative new methods for identifying and characterizing concentrations of the water quality constituents, Chlorophyll a and Total Suspended Matter. Single spectral bands and combination of multiple bands will be used to perform the linear regression analysis. The imagery will be analyzed with Environmental Visualization Software (ENVI) version 4.3 produced by ITT. First step would be to mask the water surface area. The Unsupervised Image Classification technique will be used, K Means, to cluster imagery into spectrally similar categories. Instead of the Supervised Classification Technique which is also available in ENVI. This technique because the identifications made by the unsupervised technique are made based on human sight, which is limited to the visible wavelength . Scatter plots will then be created between spectrally classified image and the collected ground truth data, based on the linear trends, simple linear regressions will be used to determine the relationships between single and a combination of bands and the water quality parameters. Then the entire image will be converted into a water quality maps showing the concentrations of Chl-a and Total Suspended matter. All of the hyperspectral and multispectral bands will be tested for establishing the relationship with the water quality parameters until it can be found which bands and parameters correlate well with each other. First of all, since there is no spectral library built in the ENVI software for Chlorophyll a and Total Suspended matter, it would have to be built on the field. It is very possible that an already established spectral library will be imported into ENVI for matching the spectral signatures. This research will develop new methods for detecting and quantifying chlorophyll a using more advanced signal processing techniques, especially the applications involving detection and mapping. The

18

spatial resolution of space borne hyperspectral sensors is generally coarser compared to airborne sensors because of higher altitudes. ASTER sensors provide spatial resolution ranging from 15 m VNIR, 60 m SWIR, and 90 m TIR ASTER only has 3 hyperspectral bands in the Visible Near Infrared regions in comparison with AVIRIS, which has 224 hyperspectral bands. The nearest water quality monitoring stations to Carroll Island is the Maryland Department of Natural Resources located at Gunpowder River as indicated at Otter Point Creek Marina where continuous YSI samplers have been installed. The Department of the Army has provided water quality data collected over the past 8 years at locations where the Submerged Aquatic Vegetation has been identified. The reason for sampling at these locations is that the SAV produces the highest concentration of Chlorophyll a and an analysis can be performed on how the nutrients are impacting the SAV’ s . The historical data for Chlorophyll-a, Total Suspended Matter, and Turbidity will be correlated to historical ASTER images (used as a baseline) in order to asses water quality changes in the watershed. The data collected from the multispectral images will be analyzed using the Environment Visualization software (ENVI) version 4.3 for several algorithms Spectral Angle Mapper. [MicroImages, 2007]. PREPROCESSING The Aster Level 1 b image received from EROS Data Center and the AVIRIS hyperspectral image from Jet Propulsion Labs (JPL) have been radiometrically and geometrically corrected. The images will be transformed to conform to the Maryland State Plane Coordinate System. Furthermore, if accurate geographical location of an area ont hei ma gene e dst obeknown,gr oundc ont r olpoi nt s( GCP’ s )wi l lbeus e dt or e g i s t e r the image to a precise map (geo-referencing). Since the study area is rather flat, Digital Elevation Modeling (DEM) is not necessary during georeferencing, neither orthorectification will be needed. Atmospheric Correction Radiometric correction consists of correcting the image for conditions in the atmosphere that intercepts incoming solar radiation, the intensity or the frequency of the reflected. .Radiometrically corrected data will be converted to radiance. One of the approaches for reducing calibrated radiance data to reflectance will be atmospheric modeling. Atmospheric modeling will be performed with the ENVI software version 4.3 using the Fast Line-of-Sight Atmospheric Analysis of Hyperspectral cubes (FLAASH) atmospheric correction algorithm /code.. The FLASSH software will model the backscattering and the absorption process in the atmosphere by accessing user –supplied parameters [ENVI, 2007]. During the field data collection, atmospheric conditions will be entered into a log book for Ambient Temperature, Humidity, Haze or Aerosols, Wind Speed, and Direction, Incident Solar Radiation, and a data logger, with a computer link for downloading. Geometric Correction

19

Geometric distortion are i nf l ue nc e dbya i r c r a f ta ndt hes e ns or ’ svi e w.Ge ome t r i c distortions may appear on images as the scale changes over the image, irregularities of the angles and the displacement of the objects in an image. The altitude of the sensor platform or the pitch, roll and yaw of the aircraft can account for changes in the spacing of the scan line, lateral shifts in the scan line positions or scan lines that are not parallel. For each flight line, the pixel widths along the scan line will be calculated and added up to obtain the distance in meters to the edges of the image. This distance will then be converted to latitude and longitude and the flight lines will be transformed using WGS-84 datum and units of degrees latitude and longitude with the polynomial warping function in ENVI. The corresponding pixels in the image are then mapped [ENVI, 2007]. SPECTRAL ANALYSIS This Research will utilize imaging spectroscopy as an analytical tool in the identification of environmental pollutants and in the technology transfer from Research and Development (R&D) remote sensing methods to operational remote sensing methods. The key to the selection of the operational sensor will be the development of the complete reference spectrum for the reflectance and absorption of energy related to the various phenomenons. Once this is developed, the key hyperspectral bands in the AVIRIS and ASTER Sensor data will be used to select the operational remote sensing and/or image processing method that will be most sensitive to that same part of the spectrum. A process flow diagram of the data collection and analysis in the field and with the ENVI software is shown below on Figure 17. Figure 17- Spectral Analysis Methodology [Salem et al., 2004] SPECTRAL DATA

ALGORITHM

DATA

MEASUREMENTS

IMAGE CLASSIFICATION

SPECTRAL LIBRARY

VISUALIZATION

This project demonstrates the fundamental utility of spectral remote sensing, as well as, a basic process for technology transfer to a much lower cost, operational remote sensing method than the one been presently used. The image (figure 18) for Carroll Island was analyzed with ENVI version 4.3 software and the spectral profile was extracted.

20

Classification of Chl-a and TSM can be very complicated because their optical properties interfere in the water column. The signal to noise hyperspectral data of the AVIRIS sensor can be complimented by new data reduction and processing techniques permits unambiguous identification of these water quality constituents and spectral unmixing of subpixel targets ;subtle spectral differences which have been enhanced in the data . This will allow for classification and validation for modeling water-leaving radiances. The analysis techniques will focus on classification of each pixel into a single class by identifying the main component in the pixel [John et al., 1999]. Figure 18-Processed AVIRIS Image of Carroll Island with ENVI Software

21

Extracting Chlorophyll-a and Total Suspended Matter Two or three images will be obtained for the analysis, one from AVIRIS, ASTER, and Hyperion. All the three images will be used to help determine the availability and repeatability of the procedure. If the process cannot be repeated, an attempt will be made to determine what the constraints that caused to fail were. Chlorophyll-a and Total Suspended Matter will be gathered once per month at the selected sampling locations. The expected ranges for Chl-a and Total Suspended matter based on 2003 data from the APG water quality data are as follow: Gunpowder River (Carroll Island) [Zetina, 2007] Chl-a Total range 9.15 to 43.68µg/L Total Suspended Matter Total Range 8.0 to 31 mg/L Bush River Data [Besert, 2007] Chl-a Total range 16.70 to 23.30 µg/L Total Suspended Matter Total Range 12.6 to 18.8 mg/L Standard multispectral image classification techniques were generally developed to classify multispectral images into broad categories. Hyperspectral imagery provides an opportunity for more detailed image analysis. For example, using hyperspectral data, spectrally similar materials can be distinguished, and sub-pixel scale information can be extracted. In this application several separation analysis passes were used with the ENVI software system to separate the specific spectral signature for Chlorophyll-a in the study area. The results of this analysis are shown in figure 20. Figure 20- Spectral separation of Chlorophyll-a signature using the ENVI spectral analysis system

There are many unique image analysis algorithms that have been developed to exploit the extensive information contained in hyperspectral imagery. Most of these algorithms also provide accurate, although more limited, analyses of multispectral data. Spectral analysis

22

methods usually compare pixel spectra with a reference spectrum (often called a target). Target spectra can be derived from a variety of sources, including spectral libraries, regions of interest within a spectral image, or individual pixels within a spectral image. The most commonly used hyperspectral/multispectral image analysis methods that are provided by ENVI are described below. Whole pixel analysis methods attempt to determine whether one or more target materials are abundant within each pixel in a multispectral or hyperspectral image on the basis of the spectral similarity between the pixel and target spectra. Whole-pixel scale tools include standard supervised classifiers such as Minimum Distance or Maximum Likelihood, as well as tools developed specifically for hyperspectral imagery such as Spectral Angle Mapper and Spectral Feature Fitting. Another approach to matching target and pixel spectra is by examining specific absorption features in the spectra. An advanced example of this method, called Tetracorder, has been developed by the U.S. Geological Survey (Clark et al., 2001). A relatively simple form of this method, called Spectral Feature Fitting, is available as part of ENVI. In Spectral Feature Fitting the user specifies a range of wavelengths within which a unique absorption feature exists for the chosen target. The pixel spectra are then compared to the target spectrum using two measurements: 1) the depth of the feature in the pixel is compared to the depth of the feature in the target, and 2) the shape of the feature in the pixel is compared to the shape of the feature in the target (using a leastsquares technique). This method was used to develop the image shown in figure 21 of chlorophyll-a concentrations in the waters near Carroll Island. Figure 21- Representative Chlorophyll-a concentrations in open water near Carroll Island using unsupervised spectral signature classification from the AVIRIS sensor data

23

CONSLUSIONS The products resulting from the research will be maps based on the ASTER and AVIRIS, satellite imagery combined with results from the water quality model, as well as, maps generated. ASTER has a fairly good spectral resolution of 15 m, but only 3 multispectral bands; on the other hand the airborne AVIRIS system collects 224 spectral narrow frequency bands and a range of spatial resolutions of 18m to 25m. Specific products will include: 1). A database containing satellite imagery for study sites; 2 products for above imagery; The expected results will be used to identify areas where pollution effects to the CBW living resources occur or have the potential to occur. Specific products and continued areas of study on this project include: Contribute to data analysis, processing, and validation, including development of advanced image processing sequences to produce a HSI library and archiving products and techniques for future use by the science community; Identify areas sensitive to change and potential contamination Provide new tools and expertise on the processing of the hyperspectral and multispectral data, allow researchers to detect the presence of small, abundant chlorophyll-a and total s us pe n de dmi ne r a l swhi c hc a n’ tbede t e c t e df r om multispectral coverage or water sampling analysis. Develop and test reflectance models that aid in monitoring, assessing and quantifying the spatial and temporal distributions of water quality parameters using remotely sensed data (i.e., airplane or satellite). Development of optimal separation algorithms for mineral identification of highlevel data products. Constraints on the suspended sediments composition, and water subsurface materials will be derived using a variety of new sub pixel spectral unmixing and delectability techniques. The resulted suspended sediments mapping can be used to develop GIS data layer for pollution finding for low present data and information. Demonstrate the effectiveness of restoration/ remediation at the watershed level using remote sensing techniques Generate recommendations for future application of multispectral and hyperspectral water quality models in the Chesapeake bay Watershed Develop and Spectral Library so that more water quality parameters can be extracted Improve upon and partially replace expensive, labor-intensive shipboard field sampling, and allow for economical sampling and mapping of large geographic areas. Measuring Chl-a and TSM and compared these parameters with field measurements .Address data gaps and advance the science of ecological monitoring. FUTURE WORK. Future efforts will focus on more complete implementation of modeling systems to better analyze spectral information from a variety of sensors and identify the spatial location of water contaminates. Five areas currently planned for the future include:

24

1 Develop and deploy a web-based system that will manage a water quality database in a Microsoft Access Database Management System (DBMS) linked to historical remote sensing images. Data share this web base system with environmental agencies. 2, Develop new algorithms for identifying and characterizing nutrient loadings like nitrates, phosphates, sulfates etc. Build up an Spectral Library for these nutrients, inorganics. 3. Perform bench scale tests of algae laden water taken at different times of the year and suing an Spectrometer equipment analyze and separate for different constituents that interfere with Chlorophyll a ,such as Total Suspended Matter, other Chlorophyll types, Colored Dissolved Organic Matter (CDOM) , heavy metals, and other organics and inorganics. 4. Obtain image from other hyperspectral sensors such as CASI, HYMAP, and multispectral such as MERIS, Establish performance data for these and other sensors with respect to Chlorophyll-a , TSM, CDOM, etc. 5. Continue to develop more accurate Chl-a algorithms ACKNOWLEDGEMENTS We want to thank Dr John J Qu, Professor of Remote Sensing at George Mason University and my Thesis Advisor for his guidance and support during this study. This project could not have been accomplished without the equipment used for Groundtruthing, so we thank Dr Amin Jazeiri of George Mason University Physics Department. We would like to thank another important person from George Mason University College of Sciences, Earth Systems and Geoinformation Systems, Dr Ruixin Yang for providing us with the Environmental Visualization Software. Without this software the analysis would not have been possible. We want to thank the personnel at the Aberdeen Proving Grounds in Edgewood, Maryland especially to Steve Wampler, Chesapeake Bay Restoration Manager for the US Army who facilitated personnel and equipment to carry the tasks in this study. Also Steve Getlein and Todd Besert of the APG for their facilitation of the water quality data. The Maryland Department of the Environment, specially Dr Richard Eskin and Elinor Zetina who also provided us with the water quality database. Thank You to Ross Lunetta, David Williams of EPA, Research Triangle Park, North Carolina for providing me with documentation. Also Robert K Hall of EPA Region 9 for his guidance and support on the project and Darryl Keith of EPA Land Ecology Group for providing me with documentation related to this investigation. Finally many thanks to Dr Robert Green and Sarah Lundeen of NASA Jet Propulsion Laboratory for providing me the Hyperspectral image used in this study.

25

REFERENCES Aberdeen Proving Grounds Meeting (February 2007), Chesapeake Bay Restoration Group Amin Zaraeri, Meeting (June 2007), George Mason University, Department of Physics and Astronomy Analytical Spectral Devices (2007), http://www.asdi.com Boardman, J. W., (1998), Leveraging the high dimensionality of AVIRIS data for improved sub-pixel target unmixing and rejection of false positives: mixture tuned matched filtering, Summaries of the Seventh Annual JPL Airborne Geoscience Workshop, Pasadena, CA, p. 55. Bukata. R.P., J.H. Jerome, K.Y. Kondratyo and D.V. Pozdrvehov, (1995), Optical Properties and Remote Sensing of Inland and Coastal Waters, New York, CRC Press, p. 362 Bukata Robert (2005), Satellite Monitoring of Inland and Coastal Water Quality , retrospection, Introspection, Future Directions, Taylor and Francis, Pg 53 Carroll Inland (2007), http://www.apg.army.mil/apghome/sites/directorates /restor/html/carroll_island.html Chesapeake Bay Foundation (2004), State of the Bay Report, http://www.cbf.org Chesapeake Bay Foundation (2004), State of the Bay report. p 2, Chesapeake Bay, (2007), http://mddnr.chesapeakebay.net/eyesonthebay/index.cfm Clark, R. N., Swayze, G. A., King, T. V. V., (2001), Imaging Spectroscopy: A Tool for Earth and Planetary System Science Remote Sensing with the USGS Tetracorder Algorithm, Journal of Geophysical Research (submitted). Da vi dJ . Wi l l i a ms ,Na nc yB.Ry bi c ki ,Al f ons oV.Lomba na ,Ti m O’ Br i e n(October 2000), Preliminary Investigation of Submerged Aquatic Vegetation mapping using Hyperspectral Remote Sensing Dr. Foudan Salem ,Dr Menas Kafatos, (2004), Hyperspectral Partial Unmixing Technique for Oil Spill Target Identification, George Mason University, Center for Earth Observing and Space Research Elinor Zetina (2007), Water Quality data provided by of Maryland Department of the Environment

26

ENVI, Environmental Visualization Software (2007), http://www.ittvis.com/envi/ E.P. Green, P.J. Murphy, A.J. Edwards, and C, D, Clark, (2000), Remote Sensing Handbook for Tropical Coastal Management, UNESCO, France Gitelson, A.A. (1992), The Peak Near 400 nm on Radiance Spectra of Algae and Water: Relationships of its magnitude and position with chlorophyll concentration, International Journal of Remote Sensing, 13:3367-3373. Harding. L., Jr and E.S. Perry, (1997), Long Term increase of Phytoplankton biomass in Chesapeake Bay, 1950-1994, Main Ecology Progress Series 157, 39-52 Jensen, John R.(2000a), Remote Sensing of the Environment. An earth resource prospective, Prentice Hall Series in Geographic Information System Jensen, John R.(2000b), Remote Sensing of the Environment. An earth resource prospective, Prentice Hall Series in Geographic Information System, p. 388. Jensen, John R.(2000c), Remote Sensing of the Environment. An earth resource prospective, Prentice Hall Series in Geographic Information System, p. 376 Langland, M.J., Phillips, S.W., Raffensperger, J.P., and Moyer, D.L., (2005), Changes in streamflow and water quality in selected nontidal sites in the Chesapeake Bay Basin, 1985-2003: U.S. Geological Survey Scientific Investigations Report 20045259, 50 p Maryland Bay Grasses (2006), http://www.dnr.maryland.gov/bay/sav/bgic/ Maryland (2006), Water Quality Criteria specific to designated uses, http://wwdsd.state.md.us/comar/26/26.08.02.03-3.htm MicroImages (2007), Introduction to Hyperspectral Images, http://www.microimages.com/getstart/pdf/hyprspec.pdf National Research Council(2001), Report on Water Quality in Chesapeake Bay Reef Han. L and D.C. Rundquist, (1997), Comparison of NIR/RED Ratio and First Derivative of reflectance in Estimating Algae-Chlorophyll concentration: A Case Study in a Turbid Reservoir, Remote Sensing of the Environment, 62:253-261 Richards John, Jia Xiunping, (1999), Remote Sensing Digital Image Analysis, p. 323 Roper, W.E. (2005), Environmental Indicator Assessment for Smart Growth, Int. J. Environmental Technology and Management, Vol. 5, Nos. 2/3, pp. 162-182.

27

Ross S. Lunetta, Joseph F. Knight, Hans W. Parcel, John J. Streicher, Benjamin L Peireres, Tom Gallo, John Lyon, Thomas Mace, Christopher Buzzelli (2006), Measurements of Case II Water Color using AVIRIS Imagery in Pamlico Sound, North Carolina Shafique, N.A., Fulk, F., Autrey, B.C., and Flotermerch J. (2002), Hyperspectral Remote Sensing of Water Quality Parameters of Large Rivers in the Ohio River Basin, http://www.tucson.ars.ag.gov Todd Besert (2007), Water Quality Data provided by of the Aberdeen Proving Grounds USGS (2006), An Overview of the US Geological Survey Chesapeake Bay Ecosystem Survey, http://chesapeake.usgs.gov/overview-cbcp.html USGS, (2007), http://glovis.usgs.gov, http://LPDAAC.usgs.gov/atapool/datapool.asp Water Quality Criteria specific to designated uses (2007), http://wwdsd.state.md.us/comar/26/26.08.02.03-3.htm WaterResources (2007), RestorationProjects.pdf,http://www.harfordcountymd.gov/ /downloads/ 4. http://www.apg.army.mil/apghome/sites/directorates/restor /html/carroll_island.html, Carroll Island YSI (2007), http://www.ysi.com Yuanzhi Zhang (2005), Surface Water Quality Estimation using Remote Sensing in the Gukf of Finland and the Finnish Archipelago Sea, May 2005, Report 55, Pg 16

28