**2.2. Application of AISA hyperspectral analysis in investigation of total suspended solids and chlorophyll-***a* **in the Blackbird Creek, Delaware**

The many biogeochemical processes taking places in estuarine environments maintain ecosystem health. Estuaries have a resilience to change, as long as the disturbance has not reached the threshold of where changes can no longer be reversed. A report by the Department of Interior, in conjunction with the U.S. Fish and Wildlife Service Dahl [114], states that over the past two centuries the continental United States has seen a 53% reduction in total wetland acreage, with Delaware losing approximately 54% of its original wetland habitat. It is generally accepted that this loss of habitat cannot be reclaimed or fully restored to historic structure and function in the near future, and thus represents a constant reminder of the importance of the ongoing management of the remaining wetlands throughout the United States [79].

The amount of direct physical alteration can be estimated by monitoring the distribution of turbidity, suspended solids, each of which is coupled with physical processes, including tidal activity and wind forcing in the Blackbird Creek, where farming and agriculture practices may generate considerable runoff of fertilizers and pesticides [115]. The strength of hyperspectral data analysis techniques lies in their ability to provide both spectral and spatial views of surface water quality parameters that are typically not possible from in situ measurements. Hyperspectral Airborne Imaging Spectroradiometer for Applications (AISA) airborne sensors with high spectral and spatial resolution can be used for monitoring environmental changes in optically complex turbid environments.

Hyperspectral remote sensing has the potential to provide accurate synoptic views of water quality conditions over a large spatial extent, particularly the spatial distributions of total suspended solids (TSS) and phytoplankton that can be retrieved using remotely sensed data. This technique was applied in Blackbird Creek. Our second goal was to develop a procedure for the total suspended solids, turbidity, and total chlorophyll-*a* using method of hyperspectral data (Salem et al. unpublished data) [116]. Those characteristics were used to develop models to map water quality characteristics in optically complex waters. Airborne AISA images with high spatial and spectral capabilities were used in a spectral analysis processes. The spectral angle mapper (SAM) classifier model was applied to AISA data to map spatial distribution of total suspended solids, chlorophyll-*a*, and turbidity levels. Photographs were used for results validation. The spectral analysis of wavelengths provided information on the distribution and concentrations of turbidity and Chl-*a* in Blackbird Creek.

Value increases in water quality parameters such as chlorophyll-*a*, turbidity, total suspended solids, and nutrients are symptomatic of eutrophic conditions. Concentrations of total suspended solids, chlorophyll-*a*, and turbidity parameters can provide insight on the extent of eutrophication and the potential impact on aquatic biota and overall water quality. Suspended solids may serve as a surrogate contaminant in agricultural watersheds since phosphorus, pesticides, and metals adhere to fine sediment particles and can be suspended in tidal estuaries. Estuaries receive continuous inputs of biogeochemical constituents from their fresh water sources and nutrients can be measured by monitoring water quality by estimating distributions of those materials [117]. There is a strong relationship between nutrient input, temperature, and phytoplankton biomass [118] but long-term monitoring of this relationship over a large scale provides information necessary for resource managers. Monitoring changes and impacts of land-use activities using remote-sensing technology provides the amount of data and information resource managers need to prepare management plans and policies for the watersheds.

**2.2. Application of AISA hyperspectral analysis in investigation of total suspended solids** 

3−) concentration for outgoing tide water [2].

The many biogeochemical processes taking places in estuarine environments maintain ecosystem health. Estuaries have a resilience to change, as long as the disturbance has not reached the threshold of where changes can no longer be reversed. A report by the Department of Interior, in conjunction with the U.S. Fish and Wildlife Service Dahl [114], states that over the past two centuries the continental United States has seen a 53% reduction in total wetland acreage, with Delaware losing approximately 54% of its original wetland habitat. It is generally accepted that this loss of habitat cannot be reclaimed or fully restored to historic structure and function in the near future, and thus represents a constant reminder of the importance of the ongoing management of the remaining wetlands throughout the United States [79].

The amount of direct physical alteration can be estimated by monitoring the distribution of turbidity, suspended solids, each of which is coupled with physical processes, including tidal activity and wind forcing in the Blackbird Creek, where farming and agriculture practices may generate considerable runoff of fertilizers and pesticides [115]. The strength of hyperspectral data analysis techniques lies in their ability to provide both spectral and spatial views of surface water quality parameters that are typically not possible from in situ measurements. Hyperspectral Airborne Imaging Spectroradiometer for Applications (AISA) airborne sensors with high spectral and spatial resolution can be used for monitoring environmental changes

Hyperspectral remote sensing has the potential to provide accurate synoptic views of water quality conditions over a large spatial extent, particularly the spatial distributions of total suspended solids (TSS) and phytoplankton that can be retrieved using remotely sensed data. This technique was applied in Blackbird Creek. Our second goal was to develop a procedure for the total suspended solids, turbidity, and total chlorophyll-*a* using method of hyperspectral data (Salem et al. unpublished data) [116]. Those characteristics were used to develop models

**and chlorophyll-***a* **in the Blackbird Creek, Delaware**

**Figure 10.** Correlations between turbidity and orthophosphate (PO4

188 Water Quality

in optically complex turbid environments.

The relationships between the spectral features of water reflectance and water quality parameters were investigated in Blackbird Creek. The research objective was to establish an empirical model that could be used for retrieval of water quality parameters from airborne hyperspectral AISA data in the Blackbird Creek Watershed. Hyperspectral AISA data provide a rapid and effective water quality-monitoring technique over a large spatial area for effective management of the watershed. Remote-sensing-monitoring results were coupled with the fieldtesting results. Four years of water quality-monitoring data from Blackbird Creek provided insight in relation to the land-use practices, especially agricultural activities in the watershed, and the presence of invasive *Phragmites*. Flyover photographs of the creek were taken from a small Delaware State University aircraft to help select locations in this research. Areas susceptible to sediment loads and chlorophylls were identified. This also allowed ground truth validation. Previous research by [119] provided the best fit spectral reflectance between 700 and 800 nm for total suspended solids. The scattering peak at ~700 nm was found to be strongly correlated with TSS and turbidity by Härmä et al. [120], Senay et al. [73], and Ammenberg et al. [121]. The image used includes 35 bands from 440 to 865 nm in the visible and infrared wavelengths in this research.

Aircraft-mounted hyperspectral spectrometer, the airborne imaging spectroradiometer for applications, was used to collect landscape images with 35 bands in Blackbird Creek. This type of spectroradiometer mounted in small aircraft can collect landscape images with a high spatial of 3.00 m × 3.00-m pixel size and a spectral resolution of 225 bands. Three segments of AISA images were selected to create a spectral library for phytoplankton (chlorophyll-*a*) in tidal water with high turbidity in the estuaries such as Blackbird Creek (**Figure 11**). Representative spectra for total suspended solids and chlorophyll-*a* in turbid water were selected in the visible and near infrared region (NIR) for areas with similar characteristics. These types of spectral signatures are very useful in either small or large water bodies to monitor chlorophyll concentration (heavy, moderate, or light). AISA data were also used to measure water turbidity in Blackbird Creek and how this related to the primary productivity other than clarity of water. According to Ritchie and Cooper [93], AISA provides a better set of spectral bands in different wavelengths than what Landsat can provide. The spectral angle mapper classification model was used to monitor chlorophylls and total suspended solids based on its chemical composition using spectral signatures for each component (Salem et al. unpublished data) [116].

**Figure 11.** The three segments of AISA data used in the Blackbird Creek and Delaware Bay (Salem et al. unpublished data) [116].

The reflectance spectra of water with high chlorophyll concentration showed the characteristic absorption at the blue (~400 to ~500 nm) and red (~600 to ~680 nm) wavelength regions, which results in low reflectance, especially at ~680 nm wavelength (see **Figure 12**). Also, there is a phytoplankton-scattering peak at ~700 nm; this could be the result from a combination of chlorophyll fluorescence and water absorption at wavelength longer than 700 nm. The TSS reflectance appeared in all wavelengths from ~430 to ~860 nm, but TSS reflectance increases in clear water more than turbid water in the infrared range of the spectrum, and reflectance decreases in the infrared range due to high absorption of light by sediments in turbid water. Turbid water dominated the spectra on the visible (VIS) range of the spectrum, the high reflectance of total suspended solids appeared between (~460 and ~680 nm) and the sediments absorption of light appeared in the infrared portion between 680 and 860 nm which shows that sediments in turbid water absorb light in infrared range and reflect it on the visible range. The algae bloom spectra in turbid water show that chlorophyll absorption of light of phytoplankton bloom has no effect in the signatures taken away from the shoreline due to high turbidity which reduce the absorption effect of chlorophyll.

Relationship between Land Use and Water Quality and its Assessment Using Hyperspectral Remote... http://dx.doi.org/10.5772/66620 191

selected in the visible and near infrared region (NIR) for areas with similar characteristics. These types of spectral signatures are very useful in either small or large water bodies to monitor chlorophyll concentration (heavy, moderate, or light). AISA data were also used to measure water turbidity in Blackbird Creek and how this related to the primary productivity other than clarity of water. According to Ritchie and Cooper [93], AISA provides a better set of spectral bands in different wavelengths than what Landsat can provide. The spectral angle mapper classification model was used to monitor chlorophylls and total suspended solids based on its chemical composition using spectral signatures for each component (Salem et al.

The reflectance spectra of water with high chlorophyll concentration showed the characteristic absorption at the blue (~400 to ~500 nm) and red (~600 to ~680 nm) wavelength regions, which results in low reflectance, especially at ~680 nm wavelength (see **Figure 12**). Also, there is a phytoplankton-scattering peak at ~700 nm; this could be the result from a combination of chlorophyll fluorescence and water absorption at wavelength longer than 700 nm. The TSS reflectance appeared in all wavelengths from ~430 to ~860 nm, but TSS reflectance increases in clear water more than turbid water in the infrared range of the spectrum, and reflectance decreases in the infrared range due to high absorption of light by sediments in turbid water. Turbid water dominated the spectra on the visible (VIS) range of the spectrum, the high reflectance of total suspended solids appeared between (~460 and ~680 nm) and the sediments absorption of light appeared in the infrared portion between 680 and 860 nm which shows that sediments in turbid water absorb light in infrared range and reflect it on the visible range. The algae bloom spectra in turbid water show that chlorophyll absorption of light of phytoplankton bloom has no effect in the signatures taken away from the shoreline due to high

**Figure 11.** The three segments of AISA data used in the Blackbird Creek and Delaware Bay (Salem et al. unpublished

turbidity which reduce the absorption effect of chlorophyll.

unpublished data) [116].

190 Water Quality

data) [116].

**Figure 12.** Spectral reflectance of chlorophyll and turbid water on the Blackbird Creek (BBC) and Delaware Bay (Salem et al. unpublished data) [116].

The signatures selected from the image of Blackbird Creek appeared in blue in **Figure 12** show low reflectance in the visible range of the spectra between (430 and 684 nm) due to absorption effect of the blue-green through red light of Chl-*a* and also show high reflectance of TSS reflectance in the near infrared range (727–813 nm) due to less turbid water in the Blackbird Creek than in the Delaware Bay. Low turbid water results in high absorption effect of chlorophylls in the visible range but low reflectance effect of TSS in the infrared range of the turbid water spectrum selected from Blackbird Creek.

**Figure 13** shows classified image of different levels of suspended solids in the Blackbird Creek where high concentration of total suspended solids in relatively clear water is shown in the red color and vegetation in green color. The red signature in the spectral collection showed very high water reflectance and low impact on Chl-*a* absorption in the visible range from 430 to 680 nm because low turbid water increases the reflectance in the infrared range between 710 and 860 nm. The reflectance in yellow shows high absorption of Chl-*a* due to relatively clear water. Suspended solids and Chl-*a* show more increase in reflectance and absorption features because of less turbid water.

Results by Salem et al. (unpublished data) [116] indicated that high turbidity influences spectral reflectance scattering at ~700 nm of suspended solids, masking the spectral absorption of Chl-*a* due to the effect of high total suspended solids. These results reveal a positive correlation between water clarity and Chl-*a* concentration with the reflectance troughs caused by chlorophyll absorption at ~680 nm. This case study has demonstrated the feasibility to estimate the relative Chl-*a* levels and total suspended solids concentrations in the Delaware Bay and Blackbird Creek with limited ground-truthing data available (Salem et al. unpublished data). Through the use of remote sensing, hyperspectral imagery analysis combined with aerial photographs of the area and a general understanding of the conditions in the region can be estimated by monitoring relative Chl-*a* levels and total suspended solids concentration in applications of water quality management and policy decision.

### **2.3. Hyperspectral remote sensing of phytoplankton taxonomy in estuarine waters**

Harmful algal blooms (HABs) have been observed to occur throughout the marine and aquatic environments of Earth during many different seasons, and under widely differing physical/ chemical conditions. These blooms involve individual organisms which range in size from m*icroplankton (>20 μm), large nanoplankton (8–20 μm), small nanoplankton (1–8 μm), and picoplankton (<1 μm)* [122]. HABs are not known to have simple, direct correlations with location, seasonality, temperature, salinity, pH, insulation, nutrient concentrations, or other broadly collected physical oceanographic data [123]. Given the increasing number of occurrences of HABs in economically significant waters of the world, methods to be developed for early detection of the predominant type of bloom organism, which when used with other environmental and historic data, can alert authorities of a potential HAB versus a benign bloom condition.

Normally, the study of HABs is conducted by boats and point sampling, which is time consuming, expensive, and limited in spatial scale. Conversely, remote sensing has the capability of acquiring synoptic, regional-scale information, and is the only available technique that combines the possibility of frequent, large spatial coverage [124, 125] at reasonable cost. Ocean color imagery and products, from space-borne instruments such as Coastal Zone Color Scanner (CZCS) on Nimbus-7, Sea-viewing Wide Field-of-view Sensor (SeaWiFS) on board OrbView2, MODerate Resolution Imaging Spectrometer (MODIS) on EOS Terra and Aqua satellites, and the Medium-Resolution Imaging Spectrometer (MERIS) on ENVISAT, have been used to study ocean surface phenomena for many years [126–129]. Most of these sensors focused on the detection of chlorophyll-a, the pigment common to all algae [130].

The techniques required to identify and quantify algal taxonomic composition are typically by time-consuming cell counts or more recently by high-performance liquid chromatography (HPLC) [131]. Many algae have accessory pigments that are taxonomically specific. Therefore, the detection of specific accessory pigments can often reveal the taxonomy of algae present in the aquatic system [132, 133]. Furthermore, each individual pigment has its own light absorption spectral features, so the detection of these optical features by remote sensing can discriminate the pigments and thus allow determination of the algal taxonomic composition. However, most of the sensors currently in ocean remote sensing cannot detect the algal accessory pigments because of a small number of spectral bands, that is, eight bands for SeaWiFS, six bands for MODIS, and the wide bandwidth associated with these limited numbers of spectral bands. Therefore, increasing the spectral resolution is critical to improve the capability of discriminating different algal groups by remote sensing. An (hyperspectral) imaging spectrometer is often defined as an instrument composed of numerous, contiguous, narrow spectral bands used to identify a material by its unique spectral features, and is the instrument of choice to identify algal groups as an imaging spectrometer.

Another challenge of remote sensing of algal blooms and HABs is the optical complexity in coastal and estuarine waters. Previously, most of the remote-sensing studies were focused on offshore case 1 waters (clear oceanic water, those waters whose optical properties are determined primarily by phytoplankton (Chl-*a*) and related colored dissolved organic matter (CDOM) concentration and detritus degradation products) where optical properties are fully determined by phytoplankton populations (phytoplankton and bacterial plankton) and their retinue (associated detrital material and CDOM), which are well correlated with concentrations of chlorophylls. This is in contrast to coastal Case 2 waters (waters whose optical properties are significantly influenced by other constituents such as mineral particles, CDOM, or microbubbles, whose concentrations do not correlate with phytoplankton concentration), where other substances, for example, sediments and dissolved materials, bottom features, also influence radiation propagation [134, 66]. Studies on the biogeochemistry in the coastal regions using satellite sensors are lacking due to the optical complexity of the coastal environment and sensor characteristics. Darecki et al. [135] studied the optical characteristics of two contrasting case 2 waters and their influence on ocean color remote-sensing algorithms, and concluded that, for accurate determination of chlorophyll from satellite-borne sensors in different coastal waters, a wider choice of spectral bands than currently available was needed.

**2.3. Hyperspectral remote sensing of phytoplankton taxonomy in estuarine waters**

shown (Salem et al. unpublished data) [116].

192 Water Quality

historic data, can alert authorities of a potential HAB versus a benign bloom condition.

Normally, the study of HABs is conducted by boats and point sampling, which is time consuming, expensive, and limited in spatial scale. Conversely, remote sensing has the capability of acquiring synoptic, regional-scale information, and is the only available technique that combines the possibility of frequent, large spatial coverage [124, 125] at reasonable cost. Ocean color imagery and products, from space-borne instruments such as Coastal Zone Color Scanner (CZCS) on Nimbus-7, Sea-viewing Wide Field-of-view Sensor (SeaWiFS) on board OrbView2, MODerate Resolution Imaging Spectrometer (MODIS) on EOS Terra and Aqua satellites, and the Medium-Resolution Imaging Spectrometer (MERIS) on ENVISAT, have

Harmful algal blooms (HABs) have been observed to occur throughout the marine and aquatic environments of Earth during many different seasons, and under widely differing physical/ chemical conditions. These blooms involve individual organisms which range in size from m*icroplankton (>20 μm), large nanoplankton (8–20 μm), small nanoplankton (1–8 μm), and picoplankton (<1 μm)* [122]. HABs are not known to have simple, direct correlations with location, seasonality, temperature, salinity, pH, insulation, nutrient concentrations, or other broadly collected physical oceanographic data [123]. Given the increasing number of occurrences of HABs in economically significant waters of the world, methods to be developed for early detection of the predominant type of bloom organism, which when used with other environmental and

**Figure 13.** Classified image shows TSSs in red and vegetation in green and different levels of TSSs concentration are

Hyperspectral remote sensing better defines spectral signatures and provides more independent information to potentially resolve simultaneously phytoplankton, suspended matters, CDOM, and bottom contributions [136]. As discussed previously, multi-spectral remote-sensing instruments that measure chlorophyll-*a* concentration are valuable in determining the bloom distributions and variability of chlorophyll in the case 1 waters. To identify the dominant algal group of a bloom, it is necessary to quantify the absorption features by accessory or marker pigments beyond chlorophyll-*a* [137]. Hyperspectral measurements can resolve spectral features that are relatively narrow, such as those of the accessory pigments, and therefore provide the possibility of identifying phytoplankton by taxonomic groups. Phytoplankton pigments, chlorophyll fluorescence, and in situ absorption spectra have been combined in studies to characterize micro-algal properties [138–142].

A number of research groups have used in situ or laboratory-based instruments to study the spectral properties of specific HAB groups in recent years. Millie et al. [140] demonstrated the utility of using bio-optical parameters for detecting and characterizing HABs by studying the photosynthetic pigments of *Gymnodinium breve*, a red tide dinoflagellate frequently appearing in Florida Bay. They concluded that the hypothetical assemblages of *G. breve* with increased concentration can be discerned from the absorption and the fourth-derivative information; however, the absorption spectra and the fourth-derivative alone may not identify the contribution of a chlorophyll c-containing taxon to the composite spectrum of a mixed assemblage. Richardson and Kruse [143, 144] were also successful in separating major types of algae and other parameters in Florida Bay using an AVIRIS imaging spectrometer.

To characterize the bio-optical properties of the coastal regions, the U.S. government has funded field campaigns to collect large suites of measurements from instruments based on numerous platforms: airborne, shipborne as well as in situ instruments. Those related to hyperspectral technology include HyCODE (Hyperspectral Coupled Ocean Dynamics Experiments) funded by the Office of Naval Research (ONR) [145, 146] and CICORE (California Center for Integrative Coastal Observation, Research and Education) funded by NOAA, NSF, and ONR. These campaigns with their large databases provide scientists with possibilities to study coastal ecosystems as never before. Although these field efforts were not specifically designed to "catch" the HABs' formation, maintenance, and dissipation, they help to build the bases for science and technology advancements in coastal region research including HAB studies.

In order to study the ecology of HABs and to address HAB forecast, detection, monitoring, and mitigation, a systematic effort is needed to develop methodologies that can be used uniformly for all types of HABs. Detecting the taxonomic grouping of phytoplankton is the initial step in identifying those classes that are harmful per geographic region. This information ultimately leads to managers and the public being accurately informed in terms of ecological and health issues. The bio-optical properties of the algal groups and the subsequent development of remote-sensing algorithms help by connecting laboratory and in situ studies with remotesensing technology. These retrieval algorithms can further be incorporated into future operational satellite-monitoring systems to monitor specific taxonomic characteristics of HABs.

During a hyperspectral flyover mission supported by NOAA Environmental Cooperative Science Center (ECSC) in the Chesapeake Bay region in July, 2005 (**Figure 14**), over 20 GB of airborne hyperspectral imagery data were collected using the AISA (Airborne Imaging Spectrometer Applications)-Eagle hyperspectral scanner on an aircraft (Piper Saratoga, owned and operated by the University of Nebraska-Lincoln) at an altitude from 1525 to 3050 m, which provides 1- to 3-m spatial resolution for the target areas. The spectral range for the AISA-Eagle scanner is 400–1000 nm, and is programmable with regard to the number of individual spectral bands. The AISA-Eagle is a push-broom sensor with minimal smile and keystone. The hyperspectral data consisted of 95 discrete channels with band resolution of 2.5-nm full width at half maximum (FWHM) from the spectral range of 400–1000 nm.

Relationship between Land Use and Water Quality and its Assessment Using Hyperspectral Remote... http://dx.doi.org/10.5772/66620 195

provide the possibility of identifying phytoplankton by taxonomic groups. Phytoplankton pigments, chlorophyll fluorescence, and in situ absorption spectra have been combined in

A number of research groups have used in situ or laboratory-based instruments to study the spectral properties of specific HAB groups in recent years. Millie et al. [140] demonstrated the utility of using bio-optical parameters for detecting and characterizing HABs by studying the photosynthetic pigments of *Gymnodinium breve*, a red tide dinoflagellate frequently appearing in Florida Bay. They concluded that the hypothetical assemblages of *G. breve* with increased concentration can be discerned from the absorption and the fourth-derivative information; however, the absorption spectra and the fourth-derivative alone may not identify the contribution of a chlorophyll c-containing taxon to the composite spectrum of a mixed assemblage. Richardson and Kruse [143, 144] were also successful in separating major types of algae and

To characterize the bio-optical properties of the coastal regions, the U.S. government has funded field campaigns to collect large suites of measurements from instruments based on numerous platforms: airborne, shipborne as well as in situ instruments. Those related to hyperspectral technology include HyCODE (Hyperspectral Coupled Ocean Dynamics Experiments) funded by the Office of Naval Research (ONR) [145, 146] and CICORE (California Center for Integrative Coastal Observation, Research and Education) funded by NOAA, NSF, and ONR. These campaigns with their large databases provide scientists with possibilities to study coastal ecosystems as never before. Although these field efforts were not specifically designed to "catch" the HABs' formation, maintenance, and dissipation, they help to build the bases for science and technology advancements in coastal region research including HAB studies. In order to study the ecology of HABs and to address HAB forecast, detection, monitoring, and mitigation, a systematic effort is needed to develop methodologies that can be used uniformly for all types of HABs. Detecting the taxonomic grouping of phytoplankton is the initial step in identifying those classes that are harmful per geographic region. This information ultimately leads to managers and the public being accurately informed in terms of ecological and health issues. The bio-optical properties of the algal groups and the subsequent development of remote-sensing algorithms help by connecting laboratory and in situ studies with remotesensing technology. These retrieval algorithms can further be incorporated into future operational satellite-monitoring systems to monitor specific taxonomic characteristics of HABs.

During a hyperspectral flyover mission supported by NOAA Environmental Cooperative Science Center (ECSC) in the Chesapeake Bay region in July, 2005 (**Figure 14**), over 20 GB of airborne hyperspectral imagery data were collected using the AISA (Airborne Imaging Spectrometer Applications)-Eagle hyperspectral scanner on an aircraft (Piper Saratoga, owned and operated by the University of Nebraska-Lincoln) at an altitude from 1525 to 3050 m, which provides 1- to 3-m spatial resolution for the target areas. The spectral range for the AISA-Eagle scanner is 400–1000 nm, and is programmable with regard to the number of individual spectral bands. The AISA-Eagle is a push-broom sensor with minimal smile and keystone. The hyperspectral data consisted of 95 discrete channels with band resolution of

2.5-nm full width at half maximum (FWHM) from the spectral range of 400–1000 nm.

studies to characterize micro-algal properties [138–142].

194 Water Quality

other parameters in Florida Bay using an AVIRIS imaging spectrometer.

**Figure 14.** Map of the five estuarine systems located on the east and south coasts of US where the bio-optical data were collected from 2002 to 2008. The locations of each field station in the Chesapeake Bay observation from 2005 to 2008 are also shown in the upper left map [147].

The algorithms developed in the laboratory and field studies were applied to retrospectively analyze the aerial spectrometry imagery. The imagery was first atmospherically corrected using a recognized standard atmospheric correction program, FLAASH. Once the data have been converted into reflectance, further analysis was carried out using ENVI (the Environment for Visualizing Imagery, ITT Visual Information Solutions) software for algal population classification and biomass estimation. ENVI software includes a series of spectral analysis programs that can be used for such operations as locating the target spectra within the imagery by comparison with library of known spectra (end-member spectra) for specific algal groups. The results from imagery analysis were further compared to the ground-truthing data, which verifies the results of imagery analysis and further refines the approach for detecting HABs by hyperspectral remote sensing.

The results by Fan and Warner [147] show that the estuarine environment is optically complex, and chlorophyll, suspended sediments, and CDOM all influence the water reflectance. Chlorophyll-*a* has a prominent absorption trough near 675 nm, and the depth of the trough could be indicative of chlorophyll-*a* concentrations. High sediment concentration increases overall reflectance and amplifies pigment absorption features. The high CDOM concentration suppresses reflectance in the visual spectral range.

The Environment Cooperative Science Center (ECSC) data show no correlation between chlorophyll-*a* concentration and R490/R555 ratio [148]. An empirical algorithm was developed to more accurately predict chlorophyll-*a* concentrations in estuarine waters using hyperspectral remote sensing. The wavelength of chlorophyll-*a* absorption (675-nm band), the wavelengths sensitive to chlorophyll-*a* (650- and 700-nm bands) absorption, as well as the wavelengths reflecting the effects of CDOM (550-nm band) and other suspended sediments (440-nm band) were included in this model. This algorithm can account for 72% of the variance in the chlorophyll-a data across a large area of estuarine waters, a considerable improvement compared to the previous algorithm (**Figure 15B**). By using the NIR region of the spectrum (600–750 nm), our attempts to estimate chlorophyll-*a* concentration, as well as the ability to filter out the effects of suspended sediments and CDOM, are promising.

**Figure 15.** (A) Plot of the relationship between the model (R490/R555) and Chl-*a* concentration (μg/L) data collected across ECSC study sites; (B) plot of new algorithm (Ave (R650+R700)-R 675)/(Ave R440+ R550) versus Chl-*a* concentration (μg/L) collected across ECSC study sites [149].

Overall, our data suggest a great variability of optically active constituents (e.g., Chl-*a*, TSS, and CDOM) among the different estuarine systems, as well as the different field stations within a single estuarine system (**Figure 16**). The correlation analysis further indicates that the optically active water constituents are not related to each other (**Figure 16**) in this study. The determination coefficient (*R*2) of linear relationship between Chl-*a* and TSS is less than 0.01, suggesting that TSS and Chl-*a* did not co-vary together, and the major sources of TSS in the estuarine systems in this study might be from the land runoff. Similarly, no correlations were observed for concentrations of TSS and Chl-*a* versus CDOM absorbance, suggesting that the complex origins of these water optically active constituents, and phytoplankton (e.g., Chl*a*) are not the only driver controlling the water optical properties.

The water reflectance *R*(*λ*) (**Figure 17A**) measured in this study also displayed a large variation both in magnitude and in shape over the visible and NIR spectral regions. This variability is the direct result of largely uncorrelated optically active water constituents with various concentrations.

A common spectral pattern as shown in **Figure 17B** suggests that most of the spectra have a reflectance peak around 570 nm. This reflectance peak should be the results of the minimal absorption by Chl-*a* and the backscattering by particulates, such as TSS. A secondary reflectance peak was observed at red/NIR spectral range around 695 nm; this peak could be the result of Chl*a* absorption at 675 nm and the strong water absorption at wavelength longer than 700 nm. Also, as shown in **Figure 17B**, the standard deviation of reflectance had similar pattern as the mean of reflectance, and it is wavelength dependent with a larger variance at the green spectral region.

Relationship between Land Use and Water Quality and its Assessment Using Hyperspectral Remote... http://dx.doi.org/10.5772/66620 197

reflecting the effects of CDOM (550-nm band) and other suspended sediments (440-nm band) were included in this model. This algorithm can account for 72% of the variance in the chlorophyll-a data across a large area of estuarine waters, a considerable improvement compared to the previous algorithm (**Figure 15B**). By using the NIR region of the spectrum (600–750 nm), our attempts to estimate chlorophyll-*a* concentration, as well as the ability to filter out the

Overall, our data suggest a great variability of optically active constituents (e.g., Chl-*a*, TSS, and CDOM) among the different estuarine systems, as well as the different field stations within a single estuarine system (**Figure 16**). The correlation analysis further indicates that the optically active water constituents are not related to each other (**Figure 16**) in this study. The determination coefficient (*R*2) of linear relationship between Chl-*a* and TSS is less than 0.01, suggesting that TSS and Chl-*a* did not co-vary together, and the major sources of TSS in the estuarine systems in this study might be from the land runoff. Similarly, no correlations were observed for concentrations of TSS and Chl-*a* versus CDOM absorbance, suggesting that the complex origins of these water optically active constituents, and phytoplankton (e.g., Chl-

**Figure 15.** (A) Plot of the relationship between the model (R490/R555) and Chl-*a* concentration (μg/L) data collected across ECSC study sites; (B) plot of new algorithm (Ave (R650+R700)-R 675)/(Ave R440+ R550) versus Chl-*a* concentration

The water reflectance *R*(*λ*) (**Figure 17A**) measured in this study also displayed a large variation both in magnitude and in shape over the visible and NIR spectral regions. This variability is the direct result of largely uncorrelated optically active water constituents with various

A common spectral pattern as shown in **Figure 17B** suggests that most of the spectra have a reflectance peak around 570 nm. This reflectance peak should be the results of the minimal absorption by Chl-*a* and the backscattering by particulates, such as TSS. A secondary reflectance peak was observed at red/NIR spectral range around 695 nm; this peak could be the result of Chl*a* absorption at 675 nm and the strong water absorption at wavelength longer than 700 nm. Also, as shown in **Figure 17B**, the standard deviation of reflectance had similar pattern as the mean of reflectance, and it is wavelength dependent with a larger variance at the green spectral region.

effects of suspended sediments and CDOM, are promising.

*a*) are not the only driver controlling the water optical properties.

concentrations.

196 Water Quality

(μg/L) collected across ECSC study sites [149].

**Figure 16.** The relationships of bio-optical properties (e.g. Chl-*a* concentration, TSS concentration, and CDOM absorbance) of 151 field stations [147].

**Figure 17.** (A) Water reflectance *R*(*λ*) spectra measured at 151 field stations from five estuarine systems on the east and south coasts of US; (B) the mean and standard deviation of the spectra [147].

A principal component analysis (PCA) of the reflectance dataset yields three dominant principal components which represented more than 97% of the total variance of the in situ water irradiance reflectance *R*(*λ*) in our study (**Figure 18**). The first principal component accounts for 72.8% of the total variance, and displays the positive loadings across all wavelengths. Its spectral pattern is similar to the mean water reflectance observed in **Figure 17B**. This similar pattern suggests the backscattering by suspended particles controlling the overall water reflectance *R*(*λ*). The second principal component accounts for 20.4% of total variance in the dataset. Its spectral shape was negative in blue and green spectral regions, with the minimal loadings around 560 nm. This feature corresponds to the minimal absorption by phytoplankton (e.g., Chl-*a* absorption) at this spectral range. Furthermore, the positive loadings in the red and NIR region could be contributed to the Chl-*a* absorption at 675 nm as well as the water absorption at this spectral range. So, the second principal component could be described as the effects of Chl-*a* (phytoplankton populations) absorption on water reflectance.

Relationship between Land Use and Water Quality and its Assessment Using Hyperspectral Remote... http://dx.doi.org/10.5772/66620 199

**Figure 18.** The loadings of the first three principal components of the correlation matrix of the reflectance *R*(*λ*) dataset. The figure illustrates the weights by which the original spectral bands are weighted to construct the first PCs.

**Figure 19** shows the biplot of the first two axes from canonical correspondence analysis (CCA), and it shows how the water reflectance in different in situ stations is influenced by water optical constituents (OACs), for example, Chl-*a*, TSS, and CDOM. The influences of OACs are indicated by vectors whose directions describe the gradients and whose lengths are proportional to their importance. While both TSS and CDOM have significant (*p* < 0.01) effects on first spectral axis, Chl-*a* has significant correlation (*p* < 0.01) with second spectral axis and nearly perpendicular to the TSS-CDOM gradient. So, the first CCA axis could represent the backscattering of the water, and TSS will increase the backscattering, while CDOM will suppress it. The second CCA axis could represent the influence of Chl-*a* (phytoplankton) on water reflectance. Together, these two axes explain most of the variability (96%) in the reflectance dataset.

A principal component analysis (PCA) of the reflectance dataset yields three dominant principal components which represented more than 97% of the total variance of the in situ water irradiance reflectance *R*(*λ*) in our study (**Figure 18**). The first principal component accounts for 72.8% of the total variance, and displays the positive loadings across all wavelengths. Its spectral pattern is similar to the mean water reflectance observed in **Figure 17B**. This similar pattern suggests the backscattering by suspended particles controlling the overall water reflectance *R*(*λ*). The second principal component accounts for 20.4% of total variance in the dataset. Its spectral shape was negative in blue and green spectral regions, with the minimal loadings around 560 nm. This feature corresponds to the minimal absorption by phytoplankton (e.g., Chl-*a* absorption) at this spectral range. Furthermore, the positive loadings in the red and NIR region could be contributed to the Chl-*a* absorption at 675 nm as well as the water absorption at this spectral range. So, the second principal component could be described as

**Figure 17.** (A) Water reflectance *R*(*λ*) spectra measured at 151 field stations from five estuarine systems on the east and

south coasts of US; (B) the mean and standard deviation of the spectra [147].

198 Water Quality

the effects of Chl-*a* (phytoplankton populations) absorption on water reflectance.

**Figure 19.** Biplot of the first two spectral axes of canonical correspondence analysis and the field stations. The influence of Chl-*a*, TSS, and CDOM is indicated by the vectors on this biplot [147].

This research advanced the basis of hyperspectral remote sensing under the optically complex coastal waters. The results provided necessary scientific information for using hyperspectral remote sensing in coastal and estuarine waters to monitor algal blooms. It also provides information leading toward the identification of the predominant bloom organism in estuarine waters. Knowing the type of organism as well as the synoptic view is of critical importance for coastal managers in their decisions relating to blooms, as well as being a more timely and economic method for this determination.
