Shawn M. Landry, Kyle N. Campbell and Naomi Oliver
Hillsborough County Florida, like so many other communities throughout the United States, faced the need to efficiently manage its stormwater infrastructure to comply with government regulations, respond to citizens' needs and flooding concerns, and ensure the long-term functionality of its stormwater drainage system assets. With the assistance of the Florida Center for Community Design and Research, a service-oriented University of South Florida research center, Hillsborough County completed a rapid inventory of all stormwater pipes and structures within the jurisdictional boundaries of Hillsborough County. This paper highlights the cooperative effort undertaken by the County and the University that resulted in the GPS-based inventory of over 200 square miles in one year, the consolidation of over 800 linear miles of pipe and over 100,000 stormwater structures into an ArcInfo GIS database, and the development of a geographic interface to an existing asset management system to enable non-GIS users full access to inventory information to fill a variety of needs.
Municipalities across the United States have been developing comprehensive and consolidated GIS-based infrastructure inventory databases for many years. These inventory efforts have included such techniques as digitization of paper-based inventory maps, field sampling using traditional survey methods, and various GPS-based efforts. Despite a wide variety of approaches to conducting infrastructure inventories, the primary impetus behind conducting a stormwater infrastructure inventory have been compliance with Federal National Pollution Discharge Elimination System (NPDES) regulations, hydrologic modeling related to planning efforts, and maintenance of stormwater drainage facilities.
Hillsborough County is located in west central Florida at approximately 28 degrees north latitude. The Hillsborough County land area is under the separate jurisdictional control of the City of Tampa, City of Plant City, City of Temple Terrace, and unincorporated Hillsborough County. Each of the four jurisdictional areas is responsible for the maintenance of stormwater infrastructure within its boundaries. The inventory effort described in this paper was conducted with unincorporated Hillsborough County. Of the 1,048 square miles of land area within Hillsborough County, 931 square miles are under the jurisdictional control of unincorporated Hillsborough County. According to the 2000 census, the population of the entire land area of the County was 998,948. Between 1990 and 2000, the population of the County increased by 20%, with the majority of this increase occurring within the jurisdictional area of unincorporated Hillsborough County.
Historically, roads and other infrastructure built during development activities within the County became the property and responsibility of the County soon after completion of the new development. As a requirement, in order for the County to assume ownership of this infrastructure, the developer submitted an "as-built" plan document that detailed the location and type of each asset, including the location of the stormwater drainage system. For many years, these as-built documents provided the only available information related to the location of County-maintained stormwater infrastructure.
Beginning in 1994 and driven primarily by the need to accurately map stormwater infrastructure for maintenance purposes and for NPDES permitting, Hillsborough County embarked on a field sampling effort to develop a stormwater infrastructure inventory, one structure and pipe at a time. This initial inventory effort involved locating stormwater structure positions relative to roadways, transferring positions to 1:200 scale rectified aerial photographs, and then digitizing data into GenaMap. This early inventory process often involved 3-4 separate teams of people and often required several months, from start to finish, to complete the inventory of a single development. By 1999, five years later, Hillsborough County had inventoried approximately 75% of the total land area of 931 square miles.
In 1999, faced with approaching deadlines to remain compliant with NPDES permit requirements, the County realized that it would need assistance to complete the Countywide inventory within a reasonable time period. In an effort to utilize the years of experience of the existing field team staff, the County entered into an interlocal agreement with the University of South Florida to develop more efficient field inventory and data processing protocols using GPS-based technology and advanced GIS processing techniques using the Esri ArcInfo platform. The goals of the project were several:
The following paper highlights the approach taken by Hillsborough County Florida to rapidly develop a stormwater infrastructure inventory in ArcInfo GIS and to provide access for non-GIS users to inventory data via a geographic interface.
The effort was organized into three phases, including: the GPS-based field inventory, GIS data processing, and development of a GIS interface to the stormwater asset management system. The following section highlights the methods used during these phases.
In recent years, affordable, user-friendly real-time differential global positioning system (GPS) equipment has become available that provides the capability to map a real-world position with a horizontal accuracy approaching 1 meter, often in less than 30 seconds. Since vertical positions of primary stormwater conveyance systems were being obtained as part of a separate watershed management planning effort, only horizontal position was required for this stormwater inventory effort. As a result, differential GPS units with a demonstrated horizontal accuracy (in Hillsborough County) approaching 1 meter were chosen to record the location of each stormwater asset. The Trimble ProXR Differential GPS unit with the TSC1 data logger was chosen for this project.
In any data collection effort, a well-designed data dictionary is critical to ensure that accurate and comparable attribute information is recorded (Robbins 1996; Dewberry and Davis 2000). Since the data dictionary developed for the GPS-based protocol would have to be comparable to the data dictionary that was developed 5 years prior, previously collected data elements were retained and a few additional attributes were added to increase the amount of information obtained during the GPS-based field inventory effort. Stormwater assets inventoried during this effort included: inlets, outfalls, manholes and junction boxes, control structures, maintained stormwater ponds, and all pipes. Where possible, data constraints were added to individual parameters to ensure data integrity. A list of selected parameters is included in Table 1.
Table 1. Selected attribute parameters collected during GPS inventory.
| Structures | Pipes |
|---|---|
| Asset type | Pipe type |
| Condition rating | Material type |
| Condition description | Shape, Width and Height |
| Digital photograph | Condition rating |
| Inspection date | Condition description |
| Inspection date Direction of flow |
Although most of the attributes collected are typical for all stormwater inventory efforts, a few of the data elements deserve mention, such as digital photograph, inspection date, and condition. Although field staff was trained to identify the general type of stormwater structure (e.g. curb inlet), staff had not originally been trained to identify the engineering-level detail about specific asset sub-types. Rather than train field staff to identify these details, a photograph was taken of every structure so that office-based engineering staff could identify the specific sub-type of structure if necessary. Photographs also served to document the general condition of the structure as well as the condition of the surrounding roadway or drainage area. During the prior inventory efforts, field staff were required to rate the condition of each asset as "good", "fair", "poor"; vague information at best. The GPS data dictionary was improved to include inspection date, and a more detailed description of asset condition (i.e. cavitation, obstructed flow, corroded, damaged ends, etc). In addition to recording stormwater assets, inventory teams were tasked with identifying the location of any suspected illicit discharge (unlawful release of regulated materials) to the stormwater system, to satisfy requirements of the NPDES permit. As these locations were identified, GPS data would be sent via email to the appropriate County staff whose task it would be to further investigate these locations.
During the field inventory effort, emphasis was placed on rapid data collection. Field teams systematically visited one section-township-range at a time, recording the location and attributes assets using the GPS data logger. On the rare occasions when GPS positions could not be acquired (approximately 2.5% of all assets), field teams recorded all attributes using the data logger but recorded the location on a 1:200 scale aerial photograph for later digitizing. Although this technique resulted in a reduction in accuracy of stormwater asset locations, such a reduction was considered acceptable in order to ensure completion of the field inventory within one-year. It is important to note that each asset in the final database was tagged to identify the relative accuracy of each asset.
Field staff from the University and the County formed three teams of two persons each; one University staff and one County staff. At the beginning of the one-year field inventory period, University staff operated the GPS units while County staff identified stormwater asset attributes. After six months of field data collection, County staff was given a Trimble certified GPS class and then trained to operate the GPS unit in the field. By the end of the field inventory period, County staff was fully trained in the use of the GPS unit for the stormwater inventory and were able to continue the inventory (e.g. new development) after the University staff were removed from the field team.
One of the basic requirements for most utility related GIS databases is that features connected in the real world (e.g. a curb inlet to a pipe) are connected in the GIS dataset (see Figure 1). To comply with existing data standards within the County, an ArcInfo coverage was chosen as the final data format for the stormwater infrastructure inventory. As a result, all stormwater pipes would be stored as arcs and each arc would be connected to exactly two nodes, one at each end. In a real-world stormwater system, pipes within a network are typically connected on each end to a junction box, inlet, control structure, or some other structure. In this example, each pipe was defined as an arc in ArcInfo while each structure was defined as a node. In other cases, pipes (such as driveway culverts) may not have a structure at either end. In this example, and because every pipe requires connection to two nodes to exist within an ArcInfo coverage, an end-point structure called "end treatment" was created to maintain the correct topological relationships. In other words, every pipe end was either connected to a stormwater structure or it was connected to an "end treatment" structure.
Figure 1. View of stormwater structures and pipes illustrating topological relationships.
The GPS unit chosen for this project further enforced the need to require nodes at the endpoints of every pipe. The Trimble ProXR, like most GPS units, records point locations (e.g. nodes) by averaging multiple position values over a period of time. In general (albeit an oversimplification), the accuracy of each GPS point location increases (with limits) as the number of individual positions increases. GPS line features, by contrast, are composed of individual positions for each node. In short, GPS point locations are typically much more accurate than GPS line positions. As a result, most GPS-based protocols use only point locations and therefore establish methods to identify pipes according to the points at each end. The method developed as part of this protocol required field staff to record each pipe as a line and assign all attributes to this line. At each end of a pipe, a point location was recorded either as a real structure such as a curb inlet or as an "end treatment". During GIS processing, the endpoint of each pipe was "snapped" to the more accurate location of the appropriate point location.
Another important consideration when developing a stormwater infrastructure inventory is the accurate determination of the direction of water flow through the system. Since stormwater modeling efforts would be most easily conducted using topologically accurate GIS data, it was important the direction of flow be included with the dataset. During field sampling, staff were instructed to record pipes according to direction of flow whenever possible. As a quality control measure, staff was also tasked to identify the cardinal direction of flow (e.g. north, south, east, west). During GIS processing, a turntable technique was employed to match the inherent direction of flow as identified by the "from node" and "to node" with the recorded cardinal direction of flow. If necessary, the arc was "flipped" to correct pipes not originally recorded in the proper direction.
As previously mentioned, the final stormwater infrastructure inventory database required consolidation of the GPS-based inventory and the prior digitized data. A major effort was undertaken to convert Genamap formatted digitized data to ArcInfo coverage formatted data; a task that required substantial automated and manual quality control. In addition, attributes had to be translated into a common data dictionary and all partial stormwater system data from these multiple sources had to be connected according to the appropriate topological constraints. For example, a pipe from Genamap GIS would have been connected to a structure collected during GPS efforts if those assets were connected under real-world conditions. During consolidation of all sources, quality assurance attributes were added to each record (e.g. asset) to identify the original method used to collect each record and the approximate positional accuracy based on a post-inventory field sampling of stormwater assets throughout the County (using a stratified random sampling technique). These quality assurance attributes would provide a means of qualifying the level of trust the user would have in the accuracy of each stormwater asset location.
The final procedure utilized after consolidation of all datasets into a comprehensive ArcInfo coverage, was an automated and manual quality control procedure applied to the entire dataset. Questionable data were tagged to indicate that future field verification efforts were required. Table 2 summarizes the total number of assets recorded during the project and included with the final ArcInfo coverage as well as the method used to locate these assets (GPS versus digitized). Based upon the total time period required for each method, 14,555 assets were recorded per year using the prior digitizing method while 31,504 assets were recorded during one-year using the GPS method. As illustrated by these data, the GPS-based rapid field sampling technique developed by the University was instrumental in enabling the County to meet the deadline to create a comprehensive stormwater infrastructure inventory.
Table 2. Summary of generalized stormwater asset types. End Treatments include headwalls, mitered ends, and other pipe end structures.
| Structure Type | Digitized | GPS | Total |
|---|---|---|---|
| Inlets | 14,530 | 6,976 | 21,506 |
| Manhole/Junction Box | 2,257 | 1,206 | 3,463 |
| Control Structures | 606 | 380 | 986 |
| End Treatments | 55,382 | 22,942 | 78,324 |
| Total | 72,775 | 31,504 | 104,279 |
A comprehensive asset management system provides not only an inventory of all assets, but also a means of issuing and recording maintenance events, prioritizing annual maintenance and budgets, responding to citizen needs such as flooding, and reporting assets as part of County financial information based upon GASB guidelines. Prior to the stormwater infrastructure inventory, Hillsborough County committed to using the Hansen Information Systems asset management system for all public works infrastructure (i.e. roads, signs, bridges, stormwater, etc). While this system accommodates many of the service requests and financial reporting related issues, it does not provide a GIS-based graphical interface to view and query inventory data. To compliment the existing Hansen system, the University developed an ArcView GIS application to allow users (including non-GIS trained users) to access inventory data via a geographic interface. While a more suitable long-term solution might utilize a "thin-client" interface such as ArcIMS, the County chose ArcView GIS version 3.2 as an interim solution since it already owned enough ArcView licenses to serve all necessary users.
ArcView GIS has been a powerful desktop application for many years, enabling "semi-skilled" GIS users the ability to view, query, and edit GIS datasets. However, most non-GIS trained users have difficulty understanding the functionality included with the software and cannot reliably use the application "out of the box" to serve their needs. Fortunately, the ArcView interface can be customized using the Avenue programming language. Prior to customizing the interface, several meetings with County staff were held to identify specific needs. As a result of these meetings several important conclusions were drawn regarding specific functionality required for non-GIS users to effectively utilize the GIS inventory data:
Using the Avenue programming language, Seagate Crystal Reports, Adobe Acrobat and by including appropriate reference GIS datasets, all of the above mentioned functionality was incorporated into an ArcView interface. Separate menu choices provided users the ability to run pre-coded queries, select assets using a custom query form, print reports of selected assets, generate preformatted printable maps, and view photographs and attribute information by selecting individual assets (see Figure 2). As part of a separate but coordinated effort, a GIS dataset of historic information related to past flooding events were mapped (geocoded) by address. These locations were added as an ArcView shapefile to the interface and a tool was developed to allow users to select each point location and view a scanned Adobe Acrobat version of the original document describing the event. The final product of these efforts was an ArcView GIS interface configured to be accessible to all users to enable non-GIS staff the ability to effectively access the stormwater inventory information using a geographic interface.
Figure 2. Screen-shot of customized ArcView interface showing stormwater assets. Note the customized menu items and symbology.
The overall success of the project was evaluated in terms of the original goals set forth in 1999 at the start of the project: The GPS-based rapid sampling technique provided a means to complete the stormwater inventory with acceptable accuracy; transfer of technology has enabled County staff to continue to update the field inventory; stormwater inventory data were consolidated; and non-GIS users were provided with a tool to effectively utilize GIS data in conjunction with the existing asset management system. Despite accomplishing the original goals of the project, important conclusions can be drawn to help improve future efforts of this kind.
Quality data collection efforts begin in the field with the personnel tasked with field sampling (Phipps and Robbins 1997). Several common inconsistencies were noticed during quality control procedures, such as incorrect determination of direction of flow and invalid pipe diameter. While it is not reasonable to expect zero data inconsistencies in an effort of this magnitude (e.g. over 50,000 assets recorded in one year), additional training and/or periodic evaluation and re-training could have prevented the above mentioned data errors.
The time required for GIS data conversion and quality control efforts are often underestimated when planning projects of this type (Mochty 2000); such was the case with this project. During the conversion of data from Genamap GIS to the ArcInfo coverage format, data inconsistencies and ambiguous data attributes were found that posed problems for data conversion. These problems contributed to a project delay of several months. Since data inconsistencies should be expected, additional time should be budgeted to prevent project delays from occurring.
The completeness of an infrastructure inventory should be determined by how one will utilize the inventory data. In terms of asset management, the final product of this stormwater inventory project was mostly complete since most maintained assets were inventoried. However, in terms of hydrologic modeling, the final GIS dataset was not complete because natural conveyance systems such as canals and ditches were not inventoried. In order to establish a truly complete inventory to use for asset management as well as modeling, all conveyance systems would need to be included and connected based upon appropriate topology within the GIS database.
Finally, it is important to remember that so long as development and maintenance activities continue, an infrastructure inventory is never complete. With nearly 1,000 square miles of land area under the jurisdiction of Hillsborough County, the installation of new stormwater infrastructure can easily outpace the ability for the County to record, let alone manage this infrastructure. The maintenance of an up to date infrastructure inventory is a never-ending effort, a fact that should not be overlooked during annual budget processes.
Dewberry and Davis, Inc. 2000. Stormwater Infrastructure and Conveyance System Inventory Project Phase IIIA - Citywide Inventory. City of Greensboro Stormwater Services, Greensboro, North Carolina. Retrieved May 5, 2001, from the world wide web: http://www.ci.greensboro.nc.us/stormwater/Documents/phase3a_report.pdf.
Mochty, Tom. April 2000. A To-Do List for A Successful GPS Utility Inventory. Public Works. Retrieved May 5, 2001, from the world wide web: http://www.woolpert.com/news/articles/mochty.pdf.
Phipps, Stephen P. and Clarence H. Robbins. January 1997. New Tools for Better Stormwater Planning. CE News. Retrieved May 5, 2001, from the world wide web: http://www.woolpert.com/news/articles/ar0197.html.
Robbins, Clarence H. July 1996. Stormwater Infrastructure Inventory Development. Public Works Magazine. Retrieved May 5, 2001, from the world wide web: http://www.woolpert.com/news/articles/ar0796.html.
Hillsborough County Public Works Department, Stormwater Section
Steven Reader, PhD., Geography Department, University of South Florida
Shawn M. Landry: Co-Author and Presenter
Senior Scientist
landry@arch.usf.edu
Kyle N. Campbell: Co-Author
Director
campbell@arch.usf.edu
Naomi M. Oliver: Co-Author
Assistant in Research
oliver@arch.usf.edu
Florida Center for Community Design and Research
University of South Florida
School of Architecture and Community Design
3702 Spectrum Boulevard, Suite 180
Tampa, Florida 33612
Phone: (813) 974-4042
Fax: (813) 974-6023
Website: www.fccdr.usf.edu