Water Demand Allocation Using GIS

Craig Scott, Jerry Edwards

Abstract

This paper presents different methods of allocating existing and projected water demand to hydraulic model nodes using ArcView GIS™ and MSAccess™. These methods were developed for Water Master Plans, Integrated Resource Plans, and General Plan Updates and have also been used to monitor the impact of new development on water supplies and infrastructure. Additionally, the paper presents demand allocation methods suitable for input to modern hydraulic models incorporating different water use patterns for a single node and improved calibration techniques. The methods presented incorporate procedures for optimizing the use of a variety of available, however, imperfect, data.


Introduction

Demand allocation methods, with the aid of GIS technology, have evolved to meet changing water utility needs.The methods used to allocate existing and future water demands to hydraulic model nodes have become increasingly sophisticated. These methods incorporate improved data sources including GIS, metered water use, and SCADA data and link to advanced hydraulic modeling software.

Early computer modeling of hydraulic networks typically relied on population projections by census tract, using per capita water use, to distribute water demands uniformly to the model nodes. Models tended to be skelotonized approximations of the systems which did not require detailed demand distributions. Tools available to engineers included paper maps of the census tracts with model nodes manually superimposed. Areas were planimetered or obtained by counting grids.

Today, detailed or “all-pipe” models are common to identify distribution piping and water quality improvements. Water utilities rely on modeling to provide justification for new service approvals and establishing water service rates. Improvements in modeling software and the availability of GIS tools and GIS data, have enabled demand allocation methods to evolve and meet these needs.

The following sections describe the role of demand allocation in the master planning process and the data sources typically available for establishing demands and the allocation of these demands. A description of alternative demand allocation methods are then presented followed by a discussion of the coordination of the demand databases with the hydraulic model. In conclusion, the overall considerations for selecting a demand allocation method and recommendations for improving data sources are presented.

Demand Allocation as Part of the Infrastructure Master Planning Process

The first steps in preparing a Water Master Plan are to derive existing and future demands and to develop the hydraulic model. Water demands are derived for current and build-out conditions and are reviewed before being “adopted.” The demands are then allocated to the model nodes and applicable peaking factors and diurnal water use patterns applied. The model is then used to determine system improvements needed to correct existing and projected future deficiencies.

The establishment of existing and projected future demands is an early and critical step and a miscalculation can derail the entire master planning process. The demands have many uses besides distribution system modeling including supply planning and setting treatment plant and transmission main capacities. Therefore, it is important the demands are consistent with accepted water use characteristics and available data and undergo a thorough review process.

Databases used to determine the demand allocations have been used, after completion of the master plan, to identify demands attributed to Specific Plan areas or even individual parcels. This ability enables utilities to track actual demand growth against planning assumptions.

Data Sources

Data useful for demand development and demand allocation fall into two categories: 1) regional planning data including population projections and general plans and 2) utility specific data including metered billing records and water production reports. The quality (completeness) and accessibility of this data as well as the goals of the utility help to determine the allocation method to be used. These attributes are described below in more detail for each data source. Suggested improvements to the data sources are also included.

Regional Planning Data

Regional Planning data include General and Specific Plans prepared by the respective County and population projections provided by the Census Bureau and regional agencies such as the Association of Bay Area Governments (ABAG). These figures are sometimes further refined for a specific area into grids for more localized planning purposes.

These data sources are available to nearly all water utilities and are used to help define the intensity and extent of future water demands. Estimates of current population are used with water demand data to establish unit water use factors which are applied to future population and land use estimates to project future demand. Increasingly, these data sources are becoming available in electronic and even GIS formats.

Metered Billing Records

Metered billing records are an invaluable resource for determining and allocating demands because they are actual metered water use. However, this data rarely has spatial attributes stored in the database relating the water use data to its point of use. Typically, the spatial attribute is limited to the entire water service area, individual pressure zones, or user type areas.

This data is commonly only available in summary form as a monthly report listing billed water use by meter size and/or user type. However, recent advances in desktop database software has enabled the meter use data, stored on mainframe computers, to be downloaded to a desktop PC and manipulated for demand analysis. The actual metered demands are typically adjusted to:

- Normalize water use to a dry year

- Adjust to a design system-wide UAW percentage, and

- Match accepted meter useage summary reports.

These adjustments should be kept as small as possible (0 to 5%) to maintain the integrity of the original data.

Whether the available data are summary reports or the meter use data, determines how the data can best be used for input to the hydraulic model.

Summary Billing Reports

Summary billing reports are nearly always available and list total monthly billed water use by meter size, customer category (residential, industrial, etc.), and often times by water pressure zone. These demands can be evenly distributed by user type if an accurate existing land use map or parcel map with attributes are available.

Metered Use Data

The metered use data are actual meter reads which are read approximately every month or bimonthly. These data typically do not contain spatial data other than a billing address which may or may not be the same as the service address. These addresses can be converted to a location using ArcView’s geocoding to approximate the meter locations. Alternatively, cadastral data containing parcels’ addresses can be joined with the billing data to obtain a meter location. Additional data manipulation is required to summarize the water use by service connection by month and year and to correct for meter rollovers, manual billing adjustments, and meter reader error.

A water utilities’ infrastructure maps often denote meter locations but with no attributes assigned to the CAD element. These locations could be obtained through a GPS survey, use of meter readers inspecting the infrastructure maps, or adding an APN field to the meter database to link with parcel maps. This data would also enhance a utilities’ customer notification and emergency services. A direct link to the customer billing database or a routine transfer of data to the demand allocation application is useful in maintaining a current model.

Water Production Records

Supervisory Collection and Data Acquisition (SCADA) systems record and archive pumping flows and reservoir levels which can be used to derive hourly demands within each pressure zone. When compared to the meter billing data, an estimate can be made of the water loss or Unaccounted for Water (UAW) in each zone. Other flows to account for in the demand calculation are flows to other systems, flows between zones through PRVs, and treatment plant backwash water. Use of annual production flows eliminates the need to quantify operational maintenance activities influencing water demands including reservoir draining and refilling for maintenance and main flushing.

Cadastral Data

Parcel maps available in GIS format from Counties provide locational data to allocate the existing and future demands. Geocoding and parcel maps used together can be expected to match up to 85% of the metered services. Parcel attribute data maintained by the county or the municipalities indicate whether the parcel’s developed or vacant and the intended land use. These attributes can be used to distribute the unmatched existing meters and future projected demands in the distribution system.

Aerial Photos

Recent aerial photos can be used to verify existing development and vacant parcels. By overlaying updated aerial photos, using ArcView’s Image Analyst ™, new developments are identified and used to update the demand allocation.

Database and GIS Tools

The data sources used for determining and allocating demands are large and complex and database and GIS software are needed to refine the data, define relationships between data sources and perform demand calculations. MSAccess ™ , Excel™, and ArcView™ software were used in the examples referred to in this paper. These software enable links between one another and have the ability to import and export data in a variety of formats. Formulas programmed into MSAccess ™ document how demands are generated directly from the source data and ArcView™ is used to manipulate the area and point themes to ultimately assign each water use to a model node.

Alternative Methods

Alternative methods fall into two groups:

1. Area Methods, and

2. Point Methods

There are variances of each method depending on the quality and type of available data, and the intended use of the planning document. For each method to be effective, it must make the most use of available data, and meet the following requirements:

- Match actual production records and metered consumption

- Allocate Future demands to vacant and redevelopment areas

- Provide demand components for each model demand node.

For example, a single model might use the point method for existing demands and area method for future demands. The following is a list utilities across the United States which have recently updated their master plans and used a variety of source data and demand allocation methods for existing and future demands. The demand methods used are described in the next section and examples from actual planning projects are provided.

Example Utilities (Utility, Year 2000 Population)

Contra Costa Water District, California, 212,050
Water and Sewer Authority of Cabarrus County, North Carolina, 134,100
Beverly Hills, California, 41,600
Petaluma, California, 55,800

Area Methods

Area methods use areas as the basis of the demand distribution. Water demands are derived from population or land use types which are specified in either Census tracts or Land Use maps, respectively. The areas are broken down into subareas for allocation to model nodes by intersecting them with pressure zone boundaries, grids, or customized nodal influence areas. A nodal influence area is an area drawn around each demand node and all demands generated within that area are assigned to that node.

For area based methods, an additional area GIS coverage is needed to differentiate existing from vacant areas. This coverage is also intersected with the other area coverages for existing and future demands to be calculated and assigned to the model nodes. GIS tools are used to intersect area coverages and recalculate acres and population for each new subarea.

The Contra Costa Water District and the City of Petaluma used area methods for their water master plans. These methods are presented as examples of the area method.

Contra Costa Water District (CCWD)

A land use based method was used for Contra Costa Water District’s 1996 Model Update. Derived demands were matched to actual consumption and production using an iterative process starting with assumed water use factors (by acre and land use type) then comparing result to actual records. The water use factors were adjusted until a match was obtained between the derived and actual records. Actual water use and meter locations were used for the top 50 users.

Future demands were assigned based on identifying vacant land uses from aerial photos and accounting for planned redevelopments. This work is revisited when development plans are reviewed by CCWD. The review determines whether development demands were accounted for in the master planning work. This level of scrutiny was not provided for or required of prior master plan updates.

City of Petaluma

The City of Petaluma used parcel map data to allocate future demands to its model nodes. The City identified potential future residential units and/or nonresidential square footage by each parcel. Future demands were calculated based on the water use factors and placed at the centroids of the parcels.

Point Based Methods

Point based methods are used where actual meter locations are known or can be obtained cost effectively. This method uses actual historical demand by meter, adjusted to match the accepted existing base demands, and uses the ArcView “spatial join: function to assign each meter to the closest demand model node. For areas nearing buildout this method significantly reduces the uncertainty of predicting demands since demands are only estimated for the remaining vacant and undeveloped parcels.

Petaluma

Meter use data was geocoded and address matched to parcel data to obtain an 85 percent match of its 19,000 services. Water use for the remaining 15 percent unmatched metered services was assigned by user type to unmatched developed parcels.

The formulas are documented in the MSAccess ™ database and can be adjusted to perform system calibration and sensitivity analysis. Although it was not feasible to individually check every meter read, general checks were made including an individually check of each of the top 25 users. This data also provided insights into future land use water uses by determining average water use by street or neighborhood and identifying water use trends. These water use characteristics can be applied to future development areas and can be allowed to vary from unit uses in more established areas. The database lets the user query by customer or parcel for the demand attributed to that parcel for the model.

Hydraulic Model

Often the hydraulic model is developed independently of the demand allocation process. The combining of the two efforts for the finished product requires additional steps and checks:

- Both model and demand GIS coverages should be on same coordinate system

- Model should have existing and future pipes mapped out and fall within an established study area boundary

- Identify “demand” nodes

- Simplify demand types

- Input demand patterns and peaking factors

Establishing a well defined study area boundary should be done before developing the model or demand allocations. This reduces confusion in developing the piping network and the demands. For example, mismatched geocoded meters are easily spotted landing outside the study area boundary and can rematched. Additionally. a grid of future pipelines can be mapped out to fill in the area between the existing network and the boundary.

Demands should not be allocated to reservoir, pumping station, and other supply nodes. This mistake can occur when demand points are matched to nodes in “batch” mode. These nodes should be removed from the node point file before performing the shape join.

Although the demand allocation methods typically record many different water use types, these types should be generalized to no more than three demand types to simplify the modeling. Patterns can be given for many different categories but it is useful to limit this to as few categories as possible to keep it simple. Residential, nonresidential and unaccounted for water are three appropriate categories to use.

Conclusion

Database and GIS software provide tools to combine the different large databases and GIS coverages into a single application for queries and updates of the demand allocation. The demand database documents demand calculations and assumptions and provides a direct link of those calculations to the hydraulic model.

The best method for a specific system depends on the quality and type of data, agency preferences and past methods, planned uses, and the available budget. Ideally, these details have been specified in a project’s scope of work. Typically a modeling effort is started with less than ideal data but, with creative solutions, fill the gaps and provide a demand allocation representative of current conditions and future projections.

Area methods for demand allocation apply when there is no meter or parcel specific water use data available. Population or land use areas provide the distribution, or spatial component, of the demands. Accurate meter billing data with address information makes a point based method feasible for existing demand allocation. Significant time savings and improved accuracy results with preexisting locational data including GPS’d meters or an APN field in the meter database with a parcel GIS coverage. Future demand allocations typically use the area method because future demands are usually based on population or vacant area projections. Area methods require population, land use, and aerial photos to confirm current vacant status and future development potential.

Acknowledgments

The Author's wish to thank the following people and their respective agencies for permission to reference the water system modeling projects used in this report:

Jerry D. Brown, Contra Costa Water District

Pamela Tuft, City of Petaluma

Mark Lambert, Water and Sewer Authority of Cabarrus County

Rob Beste, City of Beverly Hills

The contents of this paper reflect the view of the authors and do not necessarily reflect the official policy of these agencies.


Authors

Craig A. Scott
Associate Engineer
Engineering and Transportation Department
City of Modesto
Phone (913) 458-3151
E-mail: cscott@modestogov.com

Jerry Edwards
Project Engineer
Black & Veatch
Phone (913) 458-3151
E-mail: jaedwards@bv.com