**1.3. Chesapeake Bay Watershed**

The Chesapeake Bay—the largest estuary in the United States—is an incredibly complex ecosystem that includes important habitats and food webs. The Chesapeake Bay watershed spans more than 16,575,900 ha and encompasses parts of six states—Delaware, Maryland, New York, Pennsylvania, Virginia, and West Virginia—and the entire District of Columbia. Almost 18 million people live in the Chesapeake Bay watershed. The Chesapeake Bay's landto-water ratio is 14:1: the largest of any coastal water body in the world. It is a very productive ecosystem that supports important recreational and commercial fisheries [43]. As shown in **Figure 3**, anthropogenic activities on land have a large impact on the Bay's health, most importantly; the water quality in Chesapeake Bay has declined significantly in recent decades, driven by the human population growth and changes in land use that together increased the nutrient loads [44].

**Figure 3.** Chesapeake Bay Watershed land use and land-cover map showing majority land uses for forest, grass and agricultural activities and residential development (data from MD DRN LULC 2010 census), and the surface Chlorophyll-*a* concentration in the Chesapeake Bay by aircraft remote sensing in 2013 [45].

Soil and water quality protection are major challenges that must be addressed in modern day agricultural systems. Fifty billion liters of water flow into the Chesapeake Bay from over 150 streams and rivers. It supports more than 2700 species of plants and animals and produces 227 million kg of seafood each year. However, today, various factors have negatively impacted water quality and the deterioration of the overall health of the bay ecosystem. The Chesapeake basin water is highly impacted by agricultural development, residential development, effluent discharge from sewage treatment plants, and residential development by Pennsylvania Association of Conservation Districts, Inc. [46]. Point and non-point source (NPS) pollutants have caused ecosystem eutrophication which stimulates hypoxia, anoxia, frequent fish kills, increased turbidity, loss of submerged aquatic vegetation, and changes in food web structure [21]. Excess nutrients and sediment pollution originated by the agricultural activities in Delmarva's tributaries are a major concern.

**1.3. Chesapeake Bay Watershed**

176 Water Quality

nutrient loads [44].

The Chesapeake Bay—the largest estuary in the United States—is an incredibly complex ecosystem that includes important habitats and food webs. The Chesapeake Bay watershed spans more than 16,575,900 ha and encompasses parts of six states—Delaware, Maryland, New York, Pennsylvania, Virginia, and West Virginia—and the entire District of Columbia. Almost 18 million people live in the Chesapeake Bay watershed. The Chesapeake Bay's landto-water ratio is 14:1: the largest of any coastal water body in the world. It is a very productive ecosystem that supports important recreational and commercial fisheries [43]. As shown in **Figure 3**, anthropogenic activities on land have a large impact on the Bay's health, most importantly; the water quality in Chesapeake Bay has declined significantly in recent decades, driven by the human population growth and changes in land use that together increased the

Soil and water quality protection are major challenges that must be addressed in modern day agricultural systems. Fifty billion liters of water flow into the Chesapeake Bay from over 150 streams and rivers. It supports more than 2700 species of plants and animals and produces 227 million kg of seafood each year. However, today, various factors have negatively impacted water quality and the deterioration of the overall health of the bay ecosystem. The Chesapeake basin water is highly impacted by agricultural development, residential development, effluent discharge from sewage treatment plants, and residential development

**Figure 3.** Chesapeake Bay Watershed land use and land-cover map showing majority land uses for forest, grass and agricultural activities and residential development (data from MD DRN LULC 2010 census), and the surface

Chlorophyll-*a* concentration in the Chesapeake Bay by aircraft remote sensing in 2013 [45].

According to Chesapeake Bay Foundation Report [47], the three major contributors to the poor health of the Chesapeake Bay include nitrogen, phosphorus, and sediment. As stated in the report, excess nitrogen and phosphorus fuel unnaturally high levels of algae growth in the water, blocking sunlight from reaching submerged aquatic vegetation such as grass bed that serve as food and habitat. In addition to hindering the growth of aquatic vegetation, algal die-off results in an increase in bacterial population that decomposes the dead algae and aquatic vegetation, as well as consumption of dissolved oxygen in the water. The Chesapeake Bay Foundation Report [47] states that fertilizers, wastewater, septic tank discharges, air pollution, and runoff from farms, cities, and suburbs result in excessive nutrients.

Erosion, coastal alteration, and construction-related development result in excessive amounts of sediment. Excessive sedimentation from tiny particles of dirt, sand, and clay floating in the water clouds the water, blocking sunlight from reaching submerged aquatic vegetation. Aquatic species such as oysters, a keystone species, and other bottom-dwelling organisms can be smothered when that sediment finally settles to the bottom. Major sources of pollution in the Chesapeake Bay include agricultural runoff by 41%, air pollution by 25%, wastewater treatment and factories by 16%, urban and suburban storm water runoff by 15%, and septic tank failures and leaks by 3%. As the major contributor to the pollution, agricultural runoff includes animal waste and fertilizers which wash off agricultural land or contaminate groundwater, which in turn pollute rivers and streams, and the Bay. **Figure 4** illustrates these pollutants and their origins [47].

**Figure 4.** Illustration shows the sources of nitrogen pollution in the Chesapeake Bay [47].

Specifically, the waste from poultry production has raised serious concerns about treatment and disposal of the chicken manure along the shores of the Chesapeake Bay. With over 12 million cubic meters of chicken waste produced by over 523 million chickens produced each year in just Maryland and Delaware alone, serious risks in the coastal states have been raised. Diminishing crop lands have led to an increased use in manure load limits for cropland soil enrichment, resulting in excess nutrients from the poultry farms to flow into the ditches, streams, and rivers feeding into the Chesapeake Bay [48]. According to Ogejo [49], regardless of the size or type of farm, animal, livestock, and poultry producers need to manage manure for better economic returns and environmental protection. Similar to other basins in the mid-Atlantic, the Chesapeake water basin is excessively impacted by agricultural development, residential development, habitat loss, and effluent discharge from sewage treatment plants [50].

### **1.4. Remote sensing: multispectral and hyperspectral analyses**

During the last few decades, remote sensing has allowed continuous monitoring opportunities of areas and objects over longer period and larger scale on Earth [51]. According to NOAA [52], remote sensing basically is the science of obtaining information about the areas or objects from a distance such as a satellite or an aircraft. As they stated, remote sensors, which can be on satellites or mounted on aircraft, collect data by detecting the energy that is reflected from Earth. Those sensors are either passive or active sensors. The most common source of radiation detected by passive sensors is reflected sunlight, while active sensors such as a laserbeam remote-sensing system use internal stimuli to collect data about Earth [52]. **Figure 5** shows an example of the Chesapeake Bay: Land Remote Sensing Image by using A Landsat 8 Surface Reflectance Mosaic by United States Geological Survey [53].

**Figure 5.** Chesapeake Bay land remote-sensing image by using A Landsat 8 Surface Reflectance Mosaic on April 4, 2016 [53].

Coastal waters are under increasing pressures from anthropogenic disturbances at various temporal and spatial scales [54, 55]. Water quality monitoring is vital for assessing such impacts, and further provides important information for sustainable water resource managements. Water quality analysis is an extremely important tool used to determine the overall health of an aquatic ecosystem, as a healthy environment will impact aquatic life forms and affect how they interact with the environment and each other. We are identifying how land use affects water quality, especially as it pertains to direct human interaction and indirect runoff from agricultural fields. Estuaries receive continuous inputs of biogeochemical constituents from their fresh water sources. For example, discharge of nutrient-rich water from urban, forest, salt marshes, agriculture, and ranching areas increases organic and inorganic sediments accumulation in Blackbird Creek. Nutrients can be measured by monitoring water quality and estimating distributions of matters in water bodies [56]. Although a wealth of new knowledge generated over the last several decades about these ecosystems, the spatial and temporal patterns of biologic and physical processes as well as anthropogenic influences are not fully understood.

Specifically, the waste from poultry production has raised serious concerns about treatment and disposal of the chicken manure along the shores of the Chesapeake Bay. With over 12 million cubic meters of chicken waste produced by over 523 million chickens produced each year in just Maryland and Delaware alone, serious risks in the coastal states have been raised. Diminishing crop lands have led to an increased use in manure load limits for cropland soil enrichment, resulting in excess nutrients from the poultry farms to flow into the ditches, streams, and rivers feeding into the Chesapeake Bay [48]. According to Ogejo [49], regardless of the size or type of farm, animal, livestock, and poultry producers need to manage manure for better economic returns and environmental protection. Similar to other basins in the mid-Atlantic, the Chesapeake water basin is excessively impacted by agricultural development, residential development, habitat loss, and effluent discharge from sewage treatment plants [50].

During the last few decades, remote sensing has allowed continuous monitoring opportunities of areas and objects over longer period and larger scale on Earth [51]. According to NOAA [52], remote sensing basically is the science of obtaining information about the areas or objects from a distance such as a satellite or an aircraft. As they stated, remote sensors, which can be on satellites or mounted on aircraft, collect data by detecting the energy that is reflected from Earth. Those sensors are either passive or active sensors. The most common source of radiation detected by passive sensors is reflected sunlight, while active sensors such as a laserbeam remote-sensing system use internal stimuli to collect data about Earth [52]. **Figure 5** shows an example of the Chesapeake Bay: Land Remote Sensing Image by using A Landsat 8

**Figure 5.** Chesapeake Bay land remote-sensing image by using A Landsat 8 Surface Reflectance Mosaic on April 4, 2016 [53].

**1.4. Remote sensing: multispectral and hyperspectral analyses**

178 Water Quality

Surface Reflectance Mosaic by United States Geological Survey [53].

While traditional monitoring programs can provide valuable water quality information, these programs are time and labor intensive, and lack the necessary temporal or spatial information needed for better decision making. Remote sensing offers the most effective means for frequent, synoptic water quality measurements [57–59]. Remote sensing offers a unique perspective because of the synoptic view and those quantitative algorithms can be used to extract geophysical and biophysical information [60]. Optically complex estuarine environment such as Blackbird Creek can be accurately mapped, measured, and characterized using remotesensing techniques in order to develop models to map water quality characteristics of estuaries and freshwater tidal wetlands.

Ozbay et al. [12] discussed how recording the presence of *P. australis* within a tidal marsh zone presents challenges in the physical execution of a vegetation survey in an environment with restricted mobility, an overall study region with an area measurement on the order of tens to hundreds or even thousands of square kilometers, as well as the accuracy of species identification via sensor capabilities and various classification methodologies. Remote sensing utilizing satellite or aerial imagery enables one to more to easily collect data from afar without having to be physically present in harsh environments. Likewise, remotely sensed imagery can be collected at various resolutions, providing detail on a multitude of scales. Lastly, various algorithms and sensor capabilities provide varying degrees of accuracy for species identification.

According to GIS Geography [61], multispectral and hyperspectral remote sensing are being used for many applications in the fields of agriculture, ecology, oil and gas, oceanography, and atmospheric studies to better understand the world we live in [61]. However, their applications are depending on the information needed. Having a finer spectral detail in hyperspectral images gives better capability to see the unseen in finer details such as specific fish population in water. Technically, the main difference between multispectral and hyperspectral is the number of bands and how narrow the bands are to one another. Per GIS Geography [61] "multispectral imagery generally refers to 3 to 10 bands that are represented in pixels. Each band is acquired using a remote sensing radiometer. Hyperspectral imagery consists of much narrower bands (10-20 nm). A hyperspectral image could have hundreds of thousands of bands. This uses an imaging spectrometer." According to Fisher et al. [62], "a collection of several monochrome images of the same scene taken by a different sensor makes a multispectral image that we referred them as a band. Multi-band image also called as multi spectral is a RGB color image that consists of a red, a green and a blue image taken with a sensor sensitive to a different wavelength". Example is provided as hyperspectral remote sensing distinguishes between three minerals because of its high spectral resolution while the multispectral Landsat Thematics Mapper could not.

An example given by Ozbay et al. [12] provided a comparison of the application of multispectral versus hyperspectral imaging with an important factor in the classification or monitoring of a wetland species, such as *P. australis*, where utilizing remotely sensed data involved the concept of scale. The extent of an intended study area has a great impact on the necessary available data to analyze. As discussed by Ozbay et al. [12], a sensor with high spatial resolution provides finer detail compared to a medium-resolution sensor over the same geographic area. If a specific area has a large geographic extent, the use of satellite imagery with a coarse resolution (10–30 m), such as *Landsat* Thematic Mapper (TM) or Le Systeme Pour l'Observation de la Terre (SPOT), may be sufficient. However, if the primary interest is on the order of several hundred meters or a few kilometers in lineal extent, aerial photography or the use of high-resolution (1–4-m) imagery, such as IKONOS or *QuickBird*, may be more appropriate or beneficial. The use of imagery with high spatial resolution may not necessarily result in improved detection or classification accuracy.

In general, multispectral data can be useful to determine broad vegetation classes and hyperspectral data can be effective to differentiate vegetation classes at the species level [12]. Authors noted that high-resolution imagery or hyperspectral data can lose effectiveness because of different vegetation species with similar biochemical and biophysical properties [63]. They also discussed that results may suffer due to spectral variations in the same species as a result of age differences, soil and water background, or stresses most notably in the near- to midinfrared bands issues.

Classification of various wetland vegetation species including *P. australis* poses a complex challenge utilizing remote-sensing techniques. As stated by Shippert [64], standard multispectral image classification techniques are generally developed to classify multispectral images into broad categories while hyperspectral imagery provides an opportunity for more detailed image analysis for specific narrow range. Hyperspectral data-derived vegetation classifications generally are possible due to the ability to distinguish species using known reference spectra or spectral libraries.

As stated by Ozbay et al. [12], the cost of field validation of multi- or hyperspectral imagery is relatively high, limiting the use of remote-sensing data due to the uncertainty of the classifications required to manage natural resources [65]. A small wetland site may be adequately and reasonably covered with high-resolution satellite imaging. On the other hand, if the size of a study site becomes larger, it will become more cost-effective to utilize medium-resolution satellite imagery over high-resolution satellite imagery. As stated by the authors "Whereas aerial photography may be suitable for smaller wetlands or non-frequent observations, mapping or monitoring on a regional scale and increased frequency with such imagery would be costly and time-consuming to process."

Each band is acquired using a remote sensing radiometer. Hyperspectral imagery consists of much narrower bands (10-20 nm). A hyperspectral image could have hundreds of thousands of bands. This uses an imaging spectrometer." According to Fisher et al. [62], "a collection of several monochrome images of the same scene taken by a different sensor makes a multispectral image that we referred them as a band. Multi-band image also called as multi spectral is a RGB color image that consists of a red, a green and a blue image taken with a sensor sensitive to a different wavelength". Example is provided as hyperspectral remote sensing distinguishes between three minerals because of its high spectral resolution while the multispectral

An example given by Ozbay et al. [12] provided a comparison of the application of multispectral versus hyperspectral imaging with an important factor in the classification or monitoring of a wetland species, such as *P. australis*, where utilizing remotely sensed data involved the concept of scale. The extent of an intended study area has a great impact on the necessary available data to analyze. As discussed by Ozbay et al. [12], a sensor with high spatial resolution provides finer detail compared to a medium-resolution sensor over the same geographic area. If a specific area has a large geographic extent, the use of satellite imagery with a coarse resolution (10–30 m), such as *Landsat* Thematic Mapper (TM) or Le Systeme Pour l'Observation de la Terre (SPOT), may be sufficient. However, if the primary interest is on the order of several hundred meters or a few kilometers in lineal extent, aerial photography or the use of high-resolution (1–4-m) imagery, such as IKONOS or *QuickBird*, may be more appropriate or beneficial. The use of imagery with high spatial resolution may not necessarily

In general, multispectral data can be useful to determine broad vegetation classes and hyperspectral data can be effective to differentiate vegetation classes at the species level [12]. Authors noted that high-resolution imagery or hyperspectral data can lose effectiveness because of different vegetation species with similar biochemical and biophysical properties [63]. They also discussed that results may suffer due to spectral variations in the same species as a result of age differences, soil and water background, or stresses most notably in the near- to mid-

Classification of various wetland vegetation species including *P. australis* poses a complex challenge utilizing remote-sensing techniques. As stated by Shippert [64], standard multispectral image classification techniques are generally developed to classify multispectral images into broad categories while hyperspectral imagery provides an opportunity for more detailed image analysis for specific narrow range. Hyperspectral data-derived vegetation classifications generally are possible due to the ability to distinguish species using known

As stated by Ozbay et al. [12], the cost of field validation of multi- or hyperspectral imagery is relatively high, limiting the use of remote-sensing data due to the uncertainty of the classifications required to manage natural resources [65]. A small wetland site may be adequately and reasonably covered with high-resolution satellite imaging. On the other hand, if the size of a study site becomes larger, it will become more cost-effective to utilize medium-resolution

Landsat Thematics Mapper could not.

180 Water Quality

result in improved detection or classification accuracy.

infrared bands issues.

reference spectra or spectral libraries.

Hyperspectral remote sensing provides a new class of Earth observation data for improved Earth surface characterization. For example, it has been used accurately to characterize complex coastal environments [66, 67] and has the potential for mapping and detecting harmful algal blooms [68], major sediment pigments of benthic substrates [69], and suspended particulate matter [70]. The high spectral resolution provides the opportunity to develop and evaluate advanced methods of spectral shape analysis, such as derivative spectroscopy that can accurately distinguish subtle features in spectra and may be critical for discriminating optically significant water constituents [71, 72].

Various empirical, semi-analytical, and analytical ocean color models have been developed last few decades to derive the water quality parameters of interest such as concentrations of total suspended solids (TSS), chlorophyll-*a* (Chl-*a*), and the absorption of colored dissolved organic matter (CDOM)) [73]. However, coastal and estuarine waters are optically complex and the signal that a remote-sensing detector collects is a mixed signal composed of various water optically active constituents from different sources [59]. For the case of the mixing of two waters, attempts to apply the ocean color models could result in poor predictive ability in retrieval of various water quality proprieties [74].

Jo et al. (unpublished data) [75] discussed the advantage of using high spectral and spatial resolutions as they illustrated in **Figure 6** where **Figure 6A** shows chlorophyll-*a* concentration in the Delaware Bay, calculated from MODIS data (500-m spatial resolutions), **Figure 6B** shows the Enhanced Thematic Mapper Plus (ETM+) (30-m spatial resolutions) on the same day, and **Figure 6C** shows an aerial photograph from the Helikite (Helium Kite, http://www. allsopp.co.uk/) at around 30-cm spatial resolution to monitor specific locations such as oyster beds continuously [76].

**Figure 6.** (A) Chlorophyll-a concentration from MODIS at 500-m spatial resolution; (B) combined three bands from ETM+ of Landsat 7 at 30-m spatial resolution; and (C) aerial photo of Helikite at 30-cm spatial resolution in August 31, 2011 (Jo et al. unpublished data) [75].
