Remote-Sensing Data in Water Quality Investigations
Using satellite technology to calculate impervious surface area in the watershed
By
Jeffrey N. Rogers,
Marcus M. Quigley,
Steven P. Roy,
Tommy Liddell
One of the critical environmental problems affecting the urban areas of the United States and many other developed countries is the impact of both existing and new commercial, residential, and industrial development on surface water quality from nonpoint source (NPS) pollution. Nonpoint sources can include stormwater runoff from parking lots, roads, building roofs, sidewalks, and other surfaces and can contain many types of pollutants, including solids, heavy metals, nutrients, litter, oils and greases, road salt, and manmade chemical compounds.
Although substantial progress has been made in addressing point sources, much work remains in assessing and controlling nonpoint sources. Determining where and how much impervious surface is present within a watershed at large scales to a level of detail needed for accurate planning and modeling can be difficult and costly. Watershed and stormwater utility managers require accurate determinations of impervious surfaces to develop stormwater quality and quantity models, conduct planning-level assessments, and identify and prioritize locations where stormwater management facilities are best implemented.
An increasing number of municipalities and counties are forming or have formed stormwater utilities to address the increasing fiscal burden of meeting water quality regulatory requirements and to protect valued resources. Quantification of the location and extent of development and its surrogate‹imperviousness‹is an important component in setting up and managing these utilities. Specifically, detailed imperviousness data are an increasingly important component in stormwater planning and priority setting, and in establishing rate structures. Through revenue generated from user fees, many stormwater utilities fund investigations and BMP implementation based on the amount of impervious surface at the parcel level.
For cash-strapped municipalities, the use of remote-sensing and satellite imagery to accurately determine the amount of impervious surface area can provide significant cost savings as compared to more traditional and manual methods of mapping these impervious areas by digitizing data from standard aerial photographs. At regional scales, these cost savings can be quite significant. Savings result from the lower cost of satellite imagery as opposed to aerial surveys, as well as in reduced staff labor time when using satellite imagery to determine the level of imperviousness.
Remote Sensing and Water Quality
Remote-sensing technologies can be used in many ways to support stormwater programs. Applications include impervious surface studies, direct water quality assessment, evapotranspiration studies, urban growth and sprawl studies, land-use-change analysis, quality and quantity modeling, and many others. These technical elements are important components of stormwater programs developed to meet a host of both regulatory and nonregulatory drivers such as total maximum daily loads (TMDL), the National Pollutant Discharge Elimination System (NPDES), drinking water source protection, stormwater utilities, and land-use planning. High-quality datasets are essential to improving stormwater management programs. The more substantial the planning dataset available to the manager for decision-making, the more robust the Phase II program (Armstrong 2001).
Remote-Sensing Data Sources: Historical Overview and the State of the Practice
 |
| Figure 1a. An early Landsat image shows insufficient detail to be very useful in detecting impervious features. |
 |
| Figure 1b. A recent QuickBird image of the identical region shows good detail. |
Under the Landsat program, NASA and the US Geological Survey have been operating low Earth orbiting satellites, providing a continuous stream of Earth observations, for more than 30 years. Beginning in the late 1980s and into the 1990s, many watershed managers began seeing the potential of detecting and analyzing impervious areas and classifying land use at regional scales by using remote-sensing approaches with the hope of using the results as planning and monitoring tools. Many studies of this nature were conducted with the help of research institutions and universities, with varying and limited success. Although imperviousness could be detected in this manner, the results were too coarse and not very useful for planning because of Landsat's coarse pixel size (30 x 30 m) (Figure 1a).
Since the launch of Space Imaging's IKONOS satellite in 1999 and Digital Globe's QuickBird satellite in 2001, high-resolution imagery (1 m and 0.6 m pixel size, respectively) from a space-based platform has become commercially available and competitively priced (Figure 1b). During this same period, increased computer hardware performance, software pricing, and new desktop computer feature-based extraction tools have led to conditions in which impervious surface extraction and analyses that result in sub-parcel-level detail at regional scales has become a viable, mainstream component of stormwater planning. Pricing for the minimum capture size (typically 50 km2) can vary depending on product options, the vendor, and the vendor's backlog, but generally ranges from $2,000-5,000. Archived images may be available for many areas and can usually be found through an online browser. These images, which are quite useful to the stormwater planner, can be purchased at a significant discount compared to new capture prices because they are several months to several years old. Imagery can be purchased directly through the satellite companies or through resellers such as Harris/ImageLinks. According to Justin Kusterer, major account manager at Harris Corporation, "Harris Corporation views stormwater management as a major emerging market for remotely sensed data. We provide high-resolution four-band fused products created using our ImageLinks product's superior fusion technology. Our fusion technology preserves the spatial and spectral information within the imagery, allowing users to quickly, accurately, and easily perform their classification."
Processing and Accuracy
Commercial satellite imagery from the two providers listed above is composed of two parts: a high-resolution, black-and-white scene and a multispectral (color plus near-infrared) scene. To accurately locate ground details, the image is often orthorectified into a map-quality product. Using a process called fusion, the high-resolution black-and-white data can be merged with the lower-resolution color data to create a new high-resolution, multispectral image. The identification of impervious features can be performed using feature-based extraction software such as Visual Learning System's Feature Analyst extension for ESRI's ArcGIS 8.
Satellite images are often captured during the growing season with leaves on the trees. Impervious extraction under these circumstances results in the identification of "impervious surfaces without canopy" (ISWoC). ISWoC are impervious surfaces that are not covered or obscured by mature trees or other vegetation. Although further research needs to be conducted to fully assess ISWoC relative to typical imperviousness assessments, under some conditions, ISWoC may provide a more hydrologically relevant assessment of impervious area. It has been shown that in many circumstances healthy tree canopy can significantly reduce stormwater runoff even when the ground below is impervious (Keating 2002).
 |
 |
| Figure 2. Hand-digitized technique (top) and remote-sensing technique (bottom) used in assessing accuracy and time requirements |
Depending on the climatic region of the study area and the time of year, the percent imperviousness estimated using ISWoC might differ. The procedure for evaluating impervious areas using remote sensing often works best, particularly in arid regions, when some emerged vegetation (grasses and leaves) is present rather than the traditional "leaves off" approach used for capturing aerial photography. This is because of similarities in color of natural materials such as bare earth, desert pavement, dry dead grass, leafless trees, and the color of manmade impervious structures such as roofs, decks, concrete, and pavement. The amount of imperviousness identified using this process differs slightly from imperviousness calculated from digitizing features from site plans or flyovers due to the concealment of some of the surfaces. The ISWoC approach estimates the types of surface a raindrop will initially fall on, and may be more realistic for modeling smaller, more frequent rain events (events that in some regions may dominate water quality hydrology, such as the Pacific Northwest). Specifically, ISWoC account for the increased surface area provided by canopied trees for interception of rainfall.
The accuracy of the remote-sensing approach makes this process a viable alternative to manual digitizing. In an evaluation conducted by the authors, impervious areas identified with the remote-sensing technique were compared with a hand-digitized version using the same images (Figure 2). The results of this evaluation, outlined in Table 1, indicate that this method can be successfully applied to detailed stormwater planning on a regional scale. If the hand-digitized dataset is assumed to have an accuracy of 100%, the remote-sensing method had an estimated accuracy of 91-95%. It is important to note that, because of factors including geographic information system (GIS) operator differences, technician fatigue, and level of detail (how small are the digitized features such as median strips, courtyards, sidewalks, and planter beds), hand-digitization can be subject to a significant degree of error (particularly when working on regional assessments). Nevertheless, it has traditionally been the best method available, with the exception of field investigations that are several orders of magnitude more costly than hand digitizing. The authors have found that, with a resolution of 60 cm (2 ft.), it is possible to automatically identify sidewalks and individual trees and shrubs within parking lots using this remote-sensing process.
When performing manual digitizing, technicians often use black-and-white aerial photography or, in fewer cases, true color. The fourth band (near infrared) contained in the satellite imagery provides some of the most useful information for determining impervious surfaces. Not only was the remote-sensing method determined to be sufficiently accurate, but it also required only a small fraction of the labor effort needed for hand digitization. The time needed to manually digitize the study area amounted to approximately 25 times greater than that of using the remote-sensing process (see Table 1) to generate the same dataset.
Case Studies
The remote-sensing method for impervious-surface calculations and land-use planning is playing an increasingly important role in watershed and stormwater planning. Studies conducted by the County of San Diego (California), North Carolina's Center for Geographic Information and Analysis, and the New York City Department of Environmental Protection (DEP) confirm that feature extraction is a viable alternative to manual digitization at a significant cost savings, and support findings by the authors of accuracies of around 95%. Many communities are using remote sensing to evaluate not only what is currently on the ground but also what will likely be there in the future. For example, once an assessment is conducted of a built-out area using remote sensing, the information obtained can be projected to other undeveloped areas that are zoned for similar development. In other words, remote sensing can be used to game or simulate the likely future development condition based on an accurate depiction of the existing development and projecting that same type and intensity onto undeveloped areas. The following case studies represent additional examples of the uses of this technology for stormwater planning.
Santa Barbara County, CA
 |
| Figure 3. Study regions in Santa Barbara County are identified in magenta. |
The County of Santa Barbara, CA, is conducting an assessment of impervious areas at a parcel and sub-parcel scale for use in watershed and subwatershed BMP implementation and planning for over 325 km2 of unincorporated, developed areas under county jurisdiction (Figure 3). The approach chosen to meet these objectives was an innovative application of remote sensing from high-resolution satellite imagery combined with GIS analysis. The benefits of this method included the ability to perform accurate BMP implementation assessments for large areas in a relatively short time frame, and at a much lower cost than previously possible. "The satellite imaging technique used in our project by the GeoSyntec team is a cost-effective, accurate means of determining 'hardscape' at an individual parcel level," says Rob Almy, manager of the Santa Barbara County Water Agency. "We intend to use the remote-sensing data, coupled with other themes from the county's GIS database, to model urban areas at the subwatershed level in our development and implementation of effective stormwater BMPs."
 |
| Figure 4. Impervious surfaces for Vandenburg Village, CA, extracted from the imagery are displayed in yellow. |
 |
| Figure 5. Note the difference in mean imperviousness between lots on the left and those on the right near the golf course. |
 |
| Figure 6. Satellite image and ground photo of a high school in Vandenburg Village, CA, identifying common features. |
 |
| Figure 7. Satellite image and ground photo of the Camino Real Shopping Plaza in Goleta, CA, identifying common features. |
This project used a combination of existing archived aerial imagery for two areas within Santa Barbara County (Vandenberg Village/Mission Hills and Orcutt), and approximately 240 km2 of new satellite imagery for another area in the county (South Coast region) collected by the QuickBird satellite (owned and operated by Digital Globe). Based on the results of the remote-sensing imperviousness determination and an analysis of subwatershed delineations and detailed land-use data, several useful BMP planning layers were developed (Figure 4). Included in these layers are maps of percent imperviousness by land-use category and subwatershed. Although these layers and maps are not a new concept to stormwater planners, in this case they are based on site-specific imperviousness, not imperviousness drawn from land-use-based reference tables or sampling and statistical extrapolation techniques. This kind of detailed assessment can help planners solve local stormwater problems by directing resources to problematic areas. Without this type of analysis, results such as those shown by Figure 5 would not be possible. Using available land-use classification data developed by the county, single-family residential areas were delineated by neighborhood block. Figure 5 shows the results of the remote-sensing imperviousness analysis for each of these blocks. It can be seen that the left cluster has twice the imperviousness of the right cluster. The basis for this discrepancy appears to be the mean lot size of each cluster, with the left cluster lots approximately half the size of right cluster lots near the golf course. Figures 6 and 7 display ground-level photos illustrating the detail captured in the satellite imagery. Stormwater modeling based on traditional land-use classifications and their associated curve numbers would have produced different and less-accurate results for these areas. Although methods such as incorporating factors like lot size into imperviousness estimates are used frequently and in many cases explanatory variables can be incorporated to improve land-use—based imperviousness assessments, only a lot-by-lot field assessment or manual or automated feature extraction from remote sensing can provide site-specific data.
For planning-level mapping, identification of areas of high relative imperviousness can be better identified using an area-weighted averaging approach. The result of this method is a continuous (as opposed to binary) data layer depicting percent imperviousness. Maps resulting from this imperviousness analysis clearly identify "hot spots" of imperviousness and can be thematically displayed with user-defined impervious thresholds (percentage), thereby facilitating efforts to target specific BMP implementation where they might be most effective.
Santa Barbara County is particularly interested in being able to assess imperviousness at the parcel and sub-parcel levels. Such information would prove useful for analyzing and potentially setting fee structures should the county choose to establish a stormwater utility. However, development of this data layer proved difficult in this case due to an unrectified, historic county parcel layer not aligning with the rectified satellite imagery. The existing parcel boundaries were often as much as 30 ft. off, prompting imperviousness delineations associated with rights of way to be included in some lot-based calculations. This discrepancy will ultimately be corrected as the county upgrades the existing parcel data in the coming years. To collect parcel-level estimates without degrading the new dataset's accuracy or re-rectifying the countywide parcels to the new imagery, the area-weighted average imperviousness was used. This process permitted the parcel database to be populated with a site-specific estimate of imperviousness while reducing the effects of rights of way on actual parcel estimates. Although not ideal, it is a good first step toward a countywide database of parcel imperviousness. The satellite data purchased for this project may have other applications. Already, the county is looking at using the data for analysis of its dedicated landscape meter program to set evapotranspiration-based irrigation budgets for large landscapes.
Mill River Watershed, MA
 |
| Figure 8. Mill River Watershed Study region in Massachusetts is identified in magenta. |
In Massachusetts, the state's Executive Office of Environmental Affairs is conducting a pilot remote-sensing impervious-surface analysis of the Mill River watershed (approximately 154 km2 primarily located in the cities of Springfield and Wilbraham) to help implement the action items outlined in the Connecticut River Watershed Action Plan (Figure 8).
Using high-resolution satellite imagery and remote-sensing software, impervious surfaces will be extracted for further analysis and planning. Restoration of riparian corridors is a primary goal of the project, as is surface-water and groundwater protection. Impervious areas will be assessed within 200 ft. (50 ft. in urban areas) for the Mill River and its tributary streams, within 303(d)-listed waterbody drainage basins, and within wellhead protection zones.These areas will be assessed for impacts both cumulatively and at a parcel level. Potential parcel owners can then be approached by various stakeholders to develop future streambank restoration projects.
This analysis will not only assist the state in coordinating restoration efforts but also will help fulfill the requirements of several TMDLs and assist the individual communities with their Phase II programs by guiding the development, the location, and the types of water quality BMPs.
Parcel layers of imperviousness produced as part of this project might one day help these communities establish stormwater utilities. It is the goal that this type of analysis will be extended to other regions of the state and perhaps statewide in the coming years.
Conclusions Drawn and Future Directions Promised
Advertisement
As the availability of high-quality, low-cost remote-sensing data has come within reach of even modest budgets, significant benefits are being realized by proactive agencies (e.g., Santa Barbara County and the Massachusetts Executive Office of Environmental Affairs) willing to embrace innovative approaches to meeting complex water quality planning and management issues. Remote-sensing analysis can provide a cost-effective tool to evaluate stormwater and watershed issues. Change analysis can be conducted to quickly determine where the problem stormwater areas are located in a large watershed. This can assist the watershed manager to better site stormwater controls targeted at water quality improvements. Just as watershed managers have become comfortable with the use of GIS in watershed assessment and planning, so they will need to become knowledgeable in the benefits derived from the use of remote-sensing data.
Although the focus of the remote-sensing case studies discussed in this article has been to establish detailed GIS layers primarily for stormwater planning, entities that are investing in, understanding, and applying remote-sensing approaches realize that they are making a significant step toward addressing future needs. As remote-sensing approaches continue to spread and are making their way into stormwater programs around the country, the leading edge of the technology continues to progress. The detailed data stemming from remote-sensing studies are becoming the bases for dramatically improved water quality and quantity modeling, micro-scale stormwater management studies, improved flood analysis, vegetation assessment, water balance studies, and a host of other innovative approaches that were previously not able to be implemented because of the high cost of site-specific data. In coming years and decades, remote-sensing approaches promise to be the major source of data for large-scale water quality and quantity studies.
Author's Bio: Jeffrey N. Rogers, GISP, is a geologist and GIS manager with GeoSyntec Consultants Inc. in Acton, MA.
Author's Bio: Marcus Quigley, P.E., associate, is with Geosyntec Consultants in Portland, OR, and Acton, MA.
Author's Bio: Steven P. Roy is an associate with GeoSyntec Consultants Inc. in Acton, MA.
Author's Bio: Tommy Liddell is an engineering technician with Santa Barbara County's Project Clean Water in Santa Barbara, CA.
July -August 2004
Remote-Sensing Data in Water Quality Investigations
Using satellite technology to calculate impervious surface area in the watershed
By
Jeffrey N. Rogers,
Marcus M. Quigley,
Steven P. Roy,
Tommy Liddell
One of the critical environmental problems affecting the urban areas of the United States and many other developed countries is the impact of both existing and new commercial, residential, and industrial development on surface water quality from nonpoint source (NPS) pollution. Nonpoint sources can include stormwater runoff from parking lots, roads, building roofs, sidewalks, and other surfaces and can contain many types of pollutants, including solids, heavy metals, nutrients, litter, oils and greases, road salt, and manmade chemical compounds.
Although substantial progress has been made in addressing point sources, much work remains in assessing and controlling nonpoint sources. Determining where and how much impervious surface is present within a watershed at large scales to a level of detail needed for accurate planning and modeling can be difficult and costly. Watershed and stormwater utility managers require accurate determinations of impervious surfaces to develop stormwater quality and quantity models, conduct planning-level assessments, and identify and prioritize locations where stormwater management facilities are best implemented.
An increasing number of municipalities and counties are forming or have formed stormwater utilities to address the increasing fiscal burden of meeting water quality regulatory requirements and to protect valued resources. Quantification of the location and extent of development and its surrogate‹imperviousness‹is an important component in setting up and managing these utilities. Specifically, detailed imperviousness data are an increasingly important component in stormwater planning and priority setting, and in establishing rate structures. Through revenue generated from user fees, many stormwater utilities fund investigations and BMP implementation based on the amount of impervious surface at the parcel level.
For cash-strapped municipalities, the use of remote-sensing and satellite imagery to accurately determine the amount of impervious surface area can provide significant cost savings as compared to more traditional and manual methods of mapping these impervious areas by digitizing data from standard aerial photographs. At regional scales, these cost savings can be quite significant. Savings result from the lower cost of satellite imagery as opposed to aerial surveys, as well as in reduced staff labor time when using satellite imagery to determine the level of imperviousness.
Remote Sensing and Water Quality
Remote-sensing technologies can be used in many ways to support stormwater programs. Applications include impervious surface studies, direct water quality assessment, evapotranspiration studies, urban growth and sprawl studies, land-use-change analysis, quality and quantity modeling, and many others. These technical elements are important components of stormwater programs developed to meet a host of both regulatory and nonregulatory drivers such as total maximum daily loads (TMDL), the National Pollutant Discharge Elimination System (NPDES), drinking water source protection, stormwater utilities, and land-use planning. High-quality datasets are essential to improving stormwater management programs. The more substantial the planning dataset available to the manager for decision-making, the more robust the Phase II program (Armstrong 2001).
Remote-Sensing Data Sources: Historical Overview and the State of the Practice
|
| Figure 1a. An early Landsat image shows insufficient detail to be very useful in detecting impervious features. |
|
| Figure 1b. A recent QuickBird image of the identical region shows good detail. |
Under the Landsat program, NASA and the US Geological Survey have been operating low Earth orbiting satellites, providing a continuous stream of Earth observations, for more than 30 years. Beginning in the late 1980s and into the 1990s, many watershed managers began seeing the potential of detecting and analyzing impervious areas and classifying land use at regional scales by using remote-sensing approaches with the hope of using the results as planning and monitoring tools. Many studies of this nature were conducted with the help of research institutions and universities, with varying and limited success. Although imperviousness could be detected in this manner, the results were too coarse and not very useful for planning because of Landsat's coarse pixel size (30 x 30 m) (Figure 1a).
Since the launch of Space Imaging's IKONOS satellite in 1999 and Digital Globe's QuickBird satellite in 2001, high-resolution imagery (1 m and 0.6 m pixel size, respectively) from a space-based platform has become commercially available and competitively priced (Figure 1b). During this same period, increased computer hardware performance, software pricing, and new desktop computer feature-based extraction tools have led to conditions in which impervious surface extraction and analyses that result in sub-parcel-level detail at regional scales has become a viable, mainstream component of stormwater planning. Pricing for the minimum capture size (typically 50 km2) can vary depending on product options, the vendor, and the vendor's backlog, but generally ranges from $2,000-5,000. Archived images may be available for many areas and can usually be found through an online browser. These images, which are quite useful to the stormwater planner, can be purchased at a significant discount compared to new capture prices because they are several months to several years old. Imagery can be purchased directly through the satellite companies or through resellers such as Harris/ImageLinks. According to Justin Kusterer, major account manager at Harris Corporation, "Harris Corporation views stormwater management as a major emerging market for remotely sensed data. We provide high-resolution four-band fused products created using our ImageLinks product's superior fusion technology. Our fusion technology preserves the spatial and spectral information within the imagery, allowing users to quickly, accurately, and easily perform their classification."
Processing and Accuracy
Commercial satellite imagery from the two providers listed above is composed of two parts: a high-resolution, black-and-white scene and a multispectral (color plus near-infrared) scene. To accurately locate ground details, the image is often orthorectified into a map-quality product. Using a process called fusion, the high-resolution black-and-white data can be merged with the lower-resolution color data to create a new high-resolution, multispectral image. The identification of impervious features can be performed using feature-based extraction software such as Visual Learning System's Feature Analyst extension for ESRI's ArcGIS 8.
Satellite images are often captured during the growing season with leaves on the trees. Impervious extraction under these circumstances results in the identification of "impervious surfaces without canopy" (ISWoC). ISWoC are impervious surfaces that are not covered or obscured by mature trees or other vegetation. Although further research needs to be conducted to fully assess ISWoC relative to typical imperviousness assessments, under some conditions, ISWoC may provide a more hydrologically relevant assessment of impervious area. It has been shown that in many circumstances healthy tree canopy can significantly reduce stormwater runoff even when the ground below is impervious (Keating 2002).
|
|
| Figure 2. Hand-digitized technique (top) and remote-sensing technique (bottom) used in assessing accuracy and time requirements |
Depending on the climatic region of the study area and the time of year, the percent imperviousness estimated using ISWoC might differ. The procedure for evaluating impervious areas using remote sensing often works best, particularly in arid regions, when some emerged vegetation (grasses and leaves) is present rather than the traditional "leaves off" approach used for capturing aerial photography. This is because of similarities in color of natural materials such as bare earth, desert pavement, dry dead grass, leafless trees, and the color of manmade impervious structures such as roofs, decks, concrete, and pavement. The amount of imperviousness identified using this process differs slightly from imperviousness calculated from digitizing features from site plans or flyovers due to the concealment of some of the surfaces. The ISWoC approach estimates the types of surface a raindrop will initially fall on, and may be more realistic for modeling smaller, more frequent rain events (events that in some regions may dominate water quality hydrology, such as the Pacific Northwest). Specifically, ISWoC account for the increased surface area provided by canopied trees for interception of rainfall.
The accuracy of the remote-sensing approach makes this process a viable alternative to manual digitizing. In an evaluation conducted by the authors, impervious areas identified with the remote-sensing technique were compared with a hand-digitized version using the same images (Figure 2). The results of this evaluation, outlined in Table 1, indicate that this method can be successfully applied to detailed stormwater planning on a regional scale. If the hand-digitized dataset is assumed to have an accuracy of 100%, the remote-sensing method had an estimated accuracy of 91-95%. It is important to note that, because of factors including geographic information system (GIS) operator differences, technician fatigue, and level of detail (how small are the digitized features such as median strips, courtyards, sidewalks, and planter beds), hand-digitization can be subject to a significant degree of error (particularly when working on regional assessments). Nevertheless, it has traditionally been the best method available, with the exception of field investigations that are several orders of magnitude more costly than hand digitizing. The authors have found that, with a resolution of 60 cm (2 ft.), it is possible to automatically identify sidewalks and individual trees and shrubs within parking lots using this remote-sensing process.
When performing manual digitizing, technicians often use black-and-white aerial photography or, in fewer cases, true color. The fourth band (near infrared) contained in the satellite imagery provides some of the most useful information for determining impervious surfaces. Not only was the remote-sensing method determined to be sufficiently accurate, but it also required only a small fraction of the labor effort needed for hand digitization. The time needed to manually digitize the study area amounted to approximately 25 times greater than that of using the remote-sensing process (see Table 1) to generate the same dataset.
Case Studies
The remote-sensing method for impervious-surface calculations and land-use planning is playing an increasingly important role in watershed and stormwater planning. Studies conducted by the County of San Diego (California), North Carolina's Center for Geographic Information and Analysis, and the New York City Department of Environmental Protection (DEP) confirm that feature extraction is a viable alternative to manual digitization at a significant cost savings, and support findings by the authors of accuracies of around 95%. Many communities are using remote sensing to evaluate not only what is currently on the ground but also what will likely be there in the future. For example, once an assessment is conducted of a built-out area using remote sensing, the information obtained can be projected to other undeveloped areas that are zoned for similar development. In other words, remote sensing can be used to game or simulate the likely future development condition based on an accurate depiction of the existing development and projecting that same type and intensity onto undeveloped areas. The following case studies represent additional examples of the uses of this technology for stormwater planning.
Santa Barbara County, CA
|
| Figure 3. Study regions in Santa Barbara County are identified in magenta. |
The County of Santa Barbara, CA, is conducting an assessment of impervious areas at a parcel and sub-parcel scale for use in watershed and subwatershed BMP implementation and planning for over 325 km2 of unincorporated, developed areas under county jurisdiction (Figure 3). The approach chosen to meet these objectives was an innovative application of remote sensing from high-resolution satellite imagery combined with GIS analysis. The benefits of this method included the ability to perform accurate BMP implementation assessments for large areas in a relatively short time frame, and at a much lower cost than previously possible. "The satellite imaging technique used in our project by the GeoSyntec team is a cost-effective, accurate means of determining 'hardscape' at an individual parcel level," says Rob Almy, manager of the Santa Barbara County Water Agency. "We intend to use the remote-sensing data, coupled with other themes from the county's GIS database, to model urban areas at the subwatershed level in our development and implementation of effective stormwater BMPs."
|
| Figure 4. Impervious surfaces for Vandenburg Village, CA, extracted from the imagery are displayed in yellow. |
|
| Figure 5. Note the difference in mean imperviousness between lots on the left and those on the right near the golf course. |
|
| Figure 6. Satellite image and ground photo of a high school in Vandenburg Village, CA, identifying common features. |
|
| Figure 7. Satellite image and ground photo of the Camino Real Shopping Plaza in Goleta, CA, identifying common features. |
This project used a combination of existing archived aerial imagery for two areas within Santa Barbara County (Vandenberg Village/Mission Hills and Orcutt), and approximately 240 km2 of new satellite imagery for another area in the county (South Coast region) collected by the QuickBird satellite (owned and operated by Digital Globe). Based on the results of the remote-sensing imperviousness determination and an analysis of subwatershed delineations and detailed land-use data, several useful BMP planning layers were developed (Figure 4). Included in these layers are maps of percent imperviousness by land-use category and subwatershed. Although these layers and maps are not a new concept to stormwater planners, in this case they are based on site-specific imperviousness, not imperviousness drawn from land-use-based reference tables or sampling and statistical extrapolation techniques. This kind of detailed assessment can help planners solve local stormwater problems by directing resources to problematic areas. Without this type of analysis, results such as those shown by Figure 5 would not be possible. Using available land-use classification data developed by the county, single-family residential areas were delineated by neighborhood block. Figure 5 shows the results of the remote-sensing imperviousness analysis for each of these blocks. It can be seen that the left cluster has twice the imperviousness of the right cluster. The basis for this discrepancy appears to be the mean lot size of each cluster, with the left cluster lots approximately half the size of right cluster lots near the golf course. Figures 6 and 7 display ground-level photos illustrating the detail captured in the satellite imagery. Stormwater modeling based on traditional land-use classifications and their associated curve numbers would have produced different and less-accurate results for these areas. Although methods such as incorporating factors like lot size into imperviousness estimates are used frequently and in many cases explanatory variables can be incorporated to improve land-use—based imperviousness assessments, only a lot-by-lot field assessment or manual or automated feature extraction from remote sensing can provide site-specific data.
For planning-level mapping, identification of areas of high relative imperviousness can be better identified using an area-weighted averaging approach. The result of this method is a continuous (as opposed to binary) data layer depicting percent imperviousness. Maps resulting from this imperviousness analysis clearly identify "hot spots" of imperviousness and can be thematically displayed with user-defined impervious thresholds (percentage), thereby facilitating efforts to target specific BMP implementation where they might be most effective.
Santa Barbara County is particularly interested in being able to assess imperviousness at the parcel and sub-parcel levels. Such information would prove useful for analyzing and potentially setting fee structures should the county choose to establish a stormwater utility. However, development of this data layer proved difficult in this case due to an unrectified, historic county parcel layer not aligning with the rectified satellite imagery. The existing parcel boundaries were often as much as 30 ft. off, prompting imperviousness delineations associated with rights of way to be included in some lot-based calculations. This discrepancy will ultimately be corrected as the county upgrades the existing parcel data in the coming years. To collect parcel-level estimates without degrading the new dataset's accuracy or re-rectifying the countywide parcels to the new imagery, the area-weighted average imperviousness was used. This process permitted the parcel database to be populated with a site-specific estimate of imperviousness while reducing the effects of rights of way on actual parcel estimates. Although not ideal, it is a good first step toward a countywide database of parcel imperviousness. The satellite data purchased for this project may have other applications. Already, the county is looking at using the data for analysis of its dedicated landscape meter program to set evapotranspiration-based irrigation budgets for large landscapes.
Mill River Watershed, MA
|
| Figure 8. Mill River Watershed Study region in Massachusetts is identified in magenta. |
In Massachusetts, the state's Executive Office of Environmental Affairs is conducting a pilot remote-sensing impervious-surface analysis of the Mill River watershed (approximately 154 km2 primarily located in the cities of Springfield and Wilbraham) to help implement the action items outlined in the Connecticut River Watershed Action Plan (Figure 8).
Using high-resolution satellite imagery and remote-sensing software, impervious surfaces will be extracted for further analysis and planning. Restoration of riparian corridors is a primary goal of the project, as is surface-water and groundwater protection. Impervious areas will be assessed within 200 ft. (50 ft. in urban areas) for the Mill River and its tributary streams, within 303(d)-listed waterbody drainage basins, and within wellhead protection zones.These areas will be assessed for impacts both cumulatively and at a parcel level. Potential parcel owners can then be approached by various stakeholders to develop future streambank restoration projects.
This analysis will not only assist the state in coordinating restoration efforts but also will help fulfill the requirements of several TMDLs and assist the individual communities with their Phase II programs by guiding the development, the location, and the types of water quality BMPs.
Parcel layers of imperviousness produced as part of this project might one day help these communities establish stormwater utilities. It is the goal that this type of analysis will be extended to other regions of the state and perhaps statewide in the coming years.
Conclusions Drawn and Future Directions Promised
As the availability of high-quality, low-cost remote-sensing data has come within reach of even modest budgets, significant benefits are being realized by proactive agencies (e.g., Santa Barbara County and the Massachusetts Executive Office of Environmental Affairs) willing to embrace innovative approaches to meeting complex water quality planning and management issues. Remote-sensing analysis can provide a cost-effective tool to evaluate stormwater and watershed issues. Change analysis can be conducted to quickly determine where the problem stormwater areas are located in a large watershed. This can assist the watershed manager to better site stormwater controls targeted at water quality improvements. Just as watershed managers have become comfortable with the use of GIS in watershed assessment and planning, so they will need to become knowledgeable in the benefits derived from the use of remote-sensing data.
Although the focus of the remote-sensing case studies discussed in this article has been to establish detailed GIS layers primarily for stormwater planning, entities that are investing in, understanding, and applying remote-sensing approaches realize that they are making a significant step toward addressing future needs. As remote-sensing approaches continue to spread and are making their way into stormwater programs around the country, the leading edge of the technology continues to progress. The detailed data stemming from remote-sensing studies are becoming the bases for dramatically improved water quality and quantity modeling, micro-scale stormwater management studies, improved flood analysis, vegetation assessment, water balance studies, and a host of other innovative approaches that were previously not able to be implemented because of the high cost of site-specific data. In coming years and decades, remote-sensing approaches promise to be the major source of data for large-scale water quality and quantity studies.