Att 4 - Flood_Management_benefits_complete.pdf

Flood management benefits of

USGS streamgaging program

October 19, 2006

National Hydrologic Warning Council

Front cover

Kettle Creek gaging station above US Air Force Academy, Pikes Peak in background (photo courtesy of USGS)

Table of contents

Executive summary

Users and uses of USGS data

What are the information needs for fl ood management in the US?

Where is that information found?

How is the information need met?

What is the benefi t of meeting the information need for fl oodplain mapping?

What is the benefi t of meeting the information need for fl ood management project design?

What is the benefi t of meeting the information need for fl ood warning and reservoir operation?

Findings

References

Appendix I. Methods for developing information to solve fl ood problems

Appendix II. Actions taken after fl ood warning

Appendix III. Findings from researchers outside of US

Appendix IV. Contributors to this report

List of figures

Flood damages in the United States, 1983-2002

Example regulatory fl oodplain map

Illustration of warning timeline

Typical NWSRFS confi guration

Day’s curve for estimating fl ood warning benefi ts

Number of daily mean and annual peak streamfl ow stations in Texas, 1808-1998

Cumulative loss of USGS streamgages with 30 or more years of data 1980-2005

Illustration of sensitivity of regulatory fl ood estimate to record length

Boundaries of regions for regression equations for Texas

List of tables

Standard errors for regional regression equations

Fastest growing counties with special fl ood hazard areas

Texas Department of Transportation design standards

Example regression equations

Record length equivalents

Situation

Making wise decisions to manage fl oods and their impacts requires information derived from data on stream behavior—both current and past. For more than 100 years, the US Geological Survey (USGS) has collected, managed, and disseminated these data, measuring and reporting on the behavior of US streams. The USGS currently operates and maintains a nationwide streamgaging network of about 7,400 gages at an annual cost of $114 million (costs given are for 2004).

The data are used by a variety of public and USGS gage station (photo courtesy of USGS) private users, including government agencies, researchers, and recreational interests. The data are used in a variety of applications, including emergency management, fl ood forecasting and control, hydropower, and watershed management. Using the data in these applications yields direct and indirect, tangible and intangible benefi ts. The tangible benefi ts represent real monetary savings to the public.

At the same time, USGS activities to collect, manage, disseminate, and analyze the data incur a real cost.

Data collection, transmission, and management equipment—some of which is highly specialized—must be purchased, operated, maintained, repaired, and replaced. Highly-skilled scientists, engineers, and technicians must be employed for these tasks and for the task of applying knowledge to convert the collected data into information that is useful to the broad user community.

From the perspective of national economic development, safety and social well-being, the logical question that arises then is this: Does the benefi t derived from the streamgage network, in whatever form, exceed the cost of building and operating the network? Equally important, as proposals such as the National Streamfl ow Information Program (NSIP) are put forward to expand the system is this question: Does the incremental benefi t of an expanded network equal or exceed the cost of the expansion?

Task

The questions posed are broad and diffi cult to answer, for the value of the gage network is not intrinsic.

Instead it is a value that is accrued when the network is integrated with appropriate analyses and actions.

Estimating that value is the task we have undertaken herein. Our goal is to fi nd, or at least defi ne a range for the value of streamfl ow data and the system that collects and disseminates them. While the uses of streamfl ow data are many, as noted by NHWC (2006), here we limit our consideration to fl ood-related uses.

Executive summary

Actions

To complete the task, we describe fi rst the data needed to solve fl ood problems and how those data are used for decision-making. In particular, we consider use of streamfl ow data for decision making regarding:

• Mapping fl oodplains for land use management.

• Planning and designing fl ood management systems and facilities.

• Flood warning and reservoir operating.

In each of these categories, we present examples, identifying the benefi t attributable to the use of streamfl ow data in context. For each of the decision-making uses identifi ed above, we present examples with real cost savings due to the availability of streamfl ow data. When reasonable, we expand from the examples to a general assessment. While we do not argue that the benefi t yielded is due only to the gages and the data available from them, we argue that without the gages, the benefi t would be signifi cantly less.

Findings

The benefi ts attributable to collecting the streamfl ow data cannot be separated from the benefi t of analyzing and using the data for better decision making—just as the success of our drive to the offi ce this morning cannot be attributed solely to a bolt that holds in place the steering wheel on the car. In that case, and in the case of the gages, the absence will surely preclude success, even if we cannot claim that success is only due to the presence.

However, we can infer values, based upon the economic benefi t due to wise fl ood management at both the local and national level. For example, we found that:

• For Folsom Dam, CA upgrade costs including increasing the height of the dam, expanding the outlet capacity and constructing an auxiliary spillway could have been avoided if a long record of fl ows was available. The potential savings is equivalent to $63 million annually.

• For Mecklenburg County, NC increased certainty in fl oodplain mapping for land use regulations could help prevent $330 million in potential damages. If that cost is spread over 50 years, with a discount rate of 6%, the benefi t is approximately $20 million annually.

The cost savings for just these 2 cases represents a signifi cant portion of the $114 million annual cost of operating the USGS streamgage network.

Extrapolating from the examples to a national scale, we found that:

• Accurate design of levee improvements, using a long record of fl ows, can save potentially $7 million/ mile. If just 20 miles of levees are repaired nationally, the savings is equivalent to $140 million. This cost savings of the 10,000 or more miles of federal project levees in the US exceeds the cost of operating the streamgage network.

...each of the uses yields benefi ts that exceed much of the cost of the USGS streamgaging network...

• FEMA suggests that approximately 64,000 fl ood maps need to be updated. FEMA values data in hydrologic analysis at $4,400 / map. Thus, the value of updating the fl ood maps is about $56 million annually when spread over the 5 years outlined in FEMA’s original fl ood map modernization project.

• In an earlier report, NHWC estimated the value of fl ood forecasts and successful reservoir operation at over $1 billion annually. If 3 to 5% of this benefi t is attributed to the streamgage network, the benefi t nationally is $30-50 million annually.

A common thread amongst all these examples is the benefi t of long, continuous records of streamfl ow data.

For example, to fi t properly a statistical model that might serve as the basis for fl oodplain development regulation, a minimum of 10 years of data is needed, and 30 years or more is preferred.

The total benefi t of the network is unknown. From consideration of the examples here, the benefi t clearly exceeds the estimated cost of operating and maintaining the network. Each of the uses that we consider herein yields benefi ts that exceed much of the cost, even when considered in individual cases. Nationwide, the benefi ts of reducing fl ood damages or of improving the effi ciency of measures designed to prevent damage and loss of life greatly exceed the cost of collecting the data through NSIP.

Flooding in Des Moines, IA (photo courtesy of A. Booher/FEMA)

Users and uses of USGS data

In a companion report, Benefi ts of USGS streamgaging program: Users and uses of USGS streamfl ow data (NHWC 2006), the attributable benefi ts of the 7,400 NSIP streamgage stations are identifi ed and described through presentation of information about users and uses. Findings from the companion report are as follows:

• Data from the network are valuable to public and private users in various applications, such as the following:

1. Planning, designing, operating, and maintaining the nation’s multipurpose water management systems.

2. Issuing fl ood warnings to protect lives and reduce property damage.

3. Designing highways and bridges.

4. Mapping fl oodplains.

5. Monitoring environmental conditions and protecting aquatic habitats.

6. Protecting water quality and regulating pollutant discharges.

7. Managing water rights and transboundary water issues.

8. Education and research.

9. Recreational uses.

Each of the 9 categories listed above has a range of data users, data uses, and benefi ciaries of the data. Each category provides benefi ts—either direct or indirect, tangible or intangible.

• The value of streamfl ow records increase over time. Streamgages with a long period of record are particularly valuable as they form a baseline for information about future changes.

• Online access to USGS streamgage records dramatically shortens the process for obtaining historical streamfl ow data. In the future, both the number of users and the ways in which the data are being used will increase, and the information’s value will increase accordingly.

• Streamgages serve multiple uses. The same gage may provide useful information for water diversions, water quality monitoring, or fl oodplain mapping. Often, users have a one-time need for data. These ad hoc uses are diffi cult to properly value, but produce additional benefi ts that can be quite large over the period of record.

In this report we focus on a subset of users: those related to fl ood management. We identify here more specifi c roles for the USGS streamgage data and we assess the tangible value of the data in that context.

Forecaster from Lower Colorado River Authority (photo courtesy of LCRA)

What are the information needs for flood management in the US?

Is fl ooding a problem?

Despite 100 years of effort to manage fl oods and to manage the vulnerability of people and property, fl ooding and fl ood damage continue to be a problem in the US. For example, Figure 1 shows total annual fl ood damage for 1983-2002. The average throughout this 20-year period has remained more or less constant, at about $5 billion. This is, in part, a consequence of citizens moving to lands subject to fl ooding and infrastructure and industry traversing or developing in areas at risk of inundation. In fact, fl ooding now causes more deaths and damage than any other weather-related phenomena, and three-quarters of all federally declared disaster declarations are due, at least in part, to fl ooding.

Clearly fl ooding is a problem in the US.

Figure 1. Flood damages in the United States, 1983-2002 (Pielke Jr. et al. 2002)

6.44 5.93

0.78

9.12

2.14

0.33

1.54

2.26 2.29

1.00

20.46

1.36

6.13 7.10

9.78

2.75

5.89

1.41

7.53

1.21

$0

$3

$6

$9

$12

$15

$18

$21

B ill io ns (2

Le ve ls

1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Fiscal Years

Average $4.8

How is the fl ood problem solved?

In the US, there is a long history of working to reduce fl ood problems, at all levels of government. The federal government acts to reduce fl ood damages that have an adverse impact on the national economy, state agencies focus on intrastate fl ood problems, and city and county governments seek to reduce adverse impacts of fl oods within their jurisdictions. These agencies plan, design, construct, and operate facilities to meet these goals.

The facilities constructed and actions taken to do so can be categorized broadly as:

• Construction and operation of structural (fl ood control) measures. These are constructed facilities that control water in some manner. Detention ponds, for example, are constructed to store water upstream of vulnerable sites, releasing that water at a rate and at a time in which the adverse impact is minimized. Levees, on the other hand, are built to accept the water as it comes, but to keep it away from people and property by blocking the fl ow path, thus reducing the water level at the property.

• Implementation of nonstructural measures. These measures are actions taken, perhaps with some minor construction, to modify the susceptibility of property to damage. They reduce damage in the very near future and in the longer-term future. In the fi rst case, for example, fl ood warning reduces fl ood damage in near real time by providing owners of property at imminent risk with timely information about the characteristics of that risk—how deep the water will be, when it will reach that depth, and so on. This advanced notice allows those owners to move or otherwise protect their property. Flood forecasting integrated with improved reservoir operation reduces fl ood damage in a similar manner, but by permitting operators to make better-informed decisions regarding storing or releasing water from a reservoir or the reservoirs of a system. Floodplain management also reduces damage, but by fostering wise use of property at risk. Floodplain land use ordinances—one tool of fl oodplain management—may effectively prohibit construction of residential housing in areas that have a 1 percent or greater chance of fl ooding in any year.

What information is needed to solve the fl ood problem?

Structural fl ood-control measure design needs

Major structural fl ood-damage reduction measures in the US are planned and designed to balance the cost of the protection provided by the measures with the risk-weighted damage incurred if the protection is not provided. For example, when decisions are taken by the US Army Corps of Engineers and local partners about raising Folsom Dam, upstream of Sacramento, California, planners will compare the cost of that raise with the cost of potential damage incurred by fl ooding in the region if the dam is not raised, weighing that fl ood cost by the long-term risk of the fl ooding. To do so, the planners require information about the magnitude of potential future hydrologic conditions—fl ows, volumes, water levels, and timing—and information about the risk or probability of those events. If an extreme event can occur, they must know the properties of that event. They must know also how likely that occurrence is. Is the risk 1 chance in 100 or 1 chance in 500? Greater investment is justifi ed for the former case than for the latter, as the risk-weighted damage is greater. This information comes as a consequence of statistical analyses of streamfl ow information.

For smaller-scale fl ood management facilities, such as the stormwater drainage pipes and channels in our urban neighborhoods or the drainage facilities of highways, a decision about acceptable risk often is made fi rst, then a cost-effective solution is sought. For example, culverts under roadways may be designed to carry safely fl ows that have a 1 in 50 chance of being exceeded. Such a decision is made considering implicitly the benefi t and cost of larger and smaller levels of risk, including the indirect and intangible costs.

A fl ow rate (or water level or volume) is determined for the selected risk, and the design of the least-costly, reliable solution proceeds.

Floodplain management needs

Similarly, decisions about fl oodplain management typically are made by defi ning fi rst the acceptable level of risk, then identifying the fl oodplain area that would be inundated. Within this area, construction and development are controlled.

The common standard in the US is to constrain land use within the fl oodplain area that has a 1 percent chance of inundation to any depth. Information needed to identify this is a relationship of water level and frequency or probability, plus a corresponding geographic representation of the boundaries of the fl ooded area when the water reaches this level. The latter can be developed conveniently with geographic information system (GIS) tools and hydraulic analyses. The former require some form of statistical analyses, much like that required for structural design.

Flood warning, fl ood operations, and fl ood emergency responses needs

Hydrologic information needs for fl ood warning, fl ood operations, and fl ood emergency responses are slightly different than that required for design and fl oodplain management. While knowledge of the long-term risk is of interest, the time scale for decision making really focuses attention on conditions that are most likely to occur within the next few hours or days, rather than on the risk over the long term. For operation of fl ood control space in reservoirs along the Lower Colorado River of Texas, managers need information about likely infl ows within the next few days; if high fl ows are expected, non-damaging releases now may be possible to empty storage space for future infl ows. For example, for fl ood warning in the fl ood prone areas of Fort Collins, Colorado, emergency managers and others require information about how high the water will rise in the next few hours, so that they can act to move people and property out of the potentially inundated area.

Flood inundation map, Fort Collins, CO (photo courtesy of David Ford Consulting Engineers)

Where is that information found?

As with most natural phenomena, watershed and channel processes that lead to fl ooding are complex and not perfectly understood. This complexity precludes analysis of all contributing factors as a method of developing the required information. For example, a mathematical model cannot be created of all that happens in a watershed and used to predict fl ooding. Our understanding of the process is not suffi cient, nor is our ability to measure or otherwise observe all the critical properties of the watershed. Consequently, empirical models are used: models that use information about past behavior to predict or forecast future behavior.

In fl ood management, these empirical procedures include:

• Statistical model fi tting for design fl ow estimation.

• Empirical models for design fl ow estimation.

• Rainfall-runoff modeling.

Each of these is described in more detail in Appendix I.

A common thread amongst these procedures is that they all rely on historical streamfl ow data for development, calibration, and validation of the predictive tools. For this, as illustrated and described in Appendix I, longer, stable, consistent data sets are needed. For example, to fi t properly a statistical model that might serve as the basis for fl oodplain development regulation, a minimum of 10 years of data is needed and 30 years or more is preferred.

Of course, engineers and hydrologists will make design, management, and operation decisions with less data. They do so almost every day. But they are less certain about the decisions as a consequence of the sparse data sets. In a nutshell, to estimate reliably the design fl ows of interest or to forecast well fl ood fl ows for emergency response a long, continuous streamfl ow data set is necessary.

Flooded underpass, Beaumont, TX (photo courtesy of L. Roll/FEMA)

...national investment in the streamgage network is appropriate, as the data and information provide a variety of national benefi ts...

How is the information need met?

Resolution of the problem requires information derived from streamfl ow data.

To provide these data for fl ood-related uses and other purposes, the USGS operates and maintains a nationwide streamgaging network of about 7,400 gages at an annual cost of $114 million (costs given are for 2004). At these gages, water level is measured continuously, along with other weather and environmental states. Subsequent analyses provide peak fl ow, low fl ow, seasonal variations, hydrographs, volumes, impacts of development and facilities, and other valuable information.

The network is supported by funding through the USGS’s Cooperative Water Program, the USGS, NSIP, other federal water and environmental agencies, and approximately 800 state and local funding partners.

USGS activities to collect, to manage and disseminate, and to analyze the data collected incur a real cost.

Data collection, transmission, and management equipment—some of which is highly specialized—must be purchased, operated, maintained, repaired, and replaced. Highly-skilled scientists, engineers, and technicians must be employed for these tasks and for the task of applying knowledge to convert the collected data into information that is useful to the broad user community. For example, expert statisticians and hydrologists are needed to complete the analyses to derive the often-used regional regression equations and modeling analyses.

Intuitively, national investment in the streamgage network seems appropriate, as the data and information provide a variety of national benefi ts. The companion report categorized these benefi ts as tangible or intangible, direct or indirect.

A National Research Council (NRC) study in 2004 characterized streamfl ow information as a public good because

(1) those who have not paid for the service (the majority of users) are not excluded from using it, and (2) the marginal cost of servicing additional individuals is zero (NRC 2004). For example, if individuals safely cross a stream during a fl ood, traveling on a roadway bridge that is properly sized to span the high water, no cost is incurred. Instead, the cost is borne as a component of the national investment in economic development and social well-being.

Direct measurements by USGS hydrographers (photo courtesy of USGS)

Other national-scale, competing investments also provide for national economic development and social well-being. Thus the logical question is this:

• Does the benefi t derived from the streamgage network exceed the cost of building, operating, and maintaining this network, thereby justifying the investment?

Equally important, as proposals such as NSIP are put forward to expand the system:

• Does the incremental benefi t of an expanded network equal or exceed the incremental cost of the expansion?

The questions posed are broad and diffi cult to answer, for the value of a single gage or the network itself is not separable. That is, outside of the context in which the data collected are transformed to information for decision-making, the gages and the data have no intrinsic value. Instead the benefi t from the gages accrues when the network is integrated with appropriate analyses and actions. Similarly, the analysis that uses the data—be that for design, fl oodplain management, operation, or emergency response—has no separable value absent the gage network.

Accordingly, we consider in the remainder of this report the benefi t of the data and the collection system, when integrated with appropriate evaluation and action. We limit our consideration to fl ood management here, noting that, in fact, the benefi ts extend to all uses identifi ed in the companion NHWC report.

USGS fi eld operations (photo courtesy of USGS)

What is the benefit of meeting the information need for floodplain mapping?

The National Flood Insurance Program and mapping

The NFIP, which is administered by the Federal Emergency Management Agency (FEMA), was introduced in 1968 as a way to assist families devastated by fl ooding and to promote safe development in fl oodplains.

Currently, nearly 20,000 communities is the US participate in this program. These communities receive taxpayer-fi nanced insurance subsidies on existing buildings, in exchange for which they must use their land and construct new building in ways that reduce the risk of fl ood damage (Hunter 2006). The program discourages communities from developing in areas subject to inundation with annual probability 0.01 or greater. This is commonly called the 100-year fl ood.

The area that would be inundated by the 100-year fl ood commonly is displayed in a map, as illustrated in Figure 2. The water surface is represented in the map by the shaded polygon adjacent to the stream channel; property within that polygon would be under the water if the event occurred. For example, land along Hitchcock Way in the fi gure will be inundated should the 100-year fl ood (or a larger fl ood) occur.

Development within this area will require that new buildings be elevated to at least the 100-year level, often with more strict local ordinances.

Figure 2. Example regulatory fl oodplain map

Benefi t from data

How the data are used

To create a map such as that shown in Figure 2, analysts commonly take the following steps:

1. Estimate the 100-year (p=0.01) discharge rate, using previously described methods.

2. With the 100-year fl ow from step 1 as input, compute, with a mathematical model of stream hydraulics, the water-surface elevation along the channel and in the adjacent fl oodplain if fl ow exceeds the channel capacity. This elevation, which is referred to as the base fl ood elevation (BFE), is the basis for regulation.

3. Combine the computed BFE with terrain data to develop the map. In general, land with elevation less than the BFE is shown within the regulatory fl oodplain, and land with elevation above is not. This step is accomplished conveniently with GIS tools. This step requires topographic data, which, in many cases, are collected, managed, and distributed by the USGS.

Error in the 100-year fl ow determination directly affects defi nition of the BFE, and this, in turn, affects delineation of land that is subject to regulation.

If the fl ow is underpredicted, the BFE also will be underpredicted. (This presumes that the hydraulics model and terrain data are without error.) If the BFE is underpredicted, the inundated area defi ned will be smaller than the area truly inundated by the regulatory fl ood. Property that is at risk will not be included in that case, and necessary restrictions on development will not be imposed. This will lead to increased property damage over the years and to increased risk for occupants of the structures.

Conversely, if the fl ow is overpredicted, the BFE also will be overpredicted, and the mapped inundation area will exceed the area that truly should be included. This, in turn, will lead to restrictions on development in areas that need not be restricted. Property there may decrease in value due to the limitations on construction. Development may shift elsewhere, or if construction does take place, it will be more costly as a consequence of the requirement that the structures be above the BFE.

Good quality data will reduce the likelihood of over- or underpredicting the regulatory fl ow, and hence the BFE and the inundated area. The improved accuracy, in turn, leads to economic benefi t due to (1) reduced

Flooding in Mecklenburg County, NC (photo courtesy of Mecklenburg County) fl ood damage to property properly kept out of harm’s way, and (2) maintained land values and reduced construction costs in areas truly at lower risk.

Errors with short records

Thomas (2004) related errors in statistical estimates of 100-year streamfl ows to the record length of the streamgage, and in turn, related the streamfl ow errors to a corresponding error in fl ood depth. Thomas demonstrated that statistically computed streamfl ow estimates using gage data with short histories have large standard errors, but also noted that the uncertainties using regional regression equations and rainfall-runoff modeling are even higher—a fi nding that is consistent with information from the Corps of Engineers (shown in Table 5). Thomas found that for a location in the semiarid western part of the US with record of less than 30 years, standard errors for statistical estimates of the 100-year streamfl ow exceed 100%; the corresponding standard errors in fl ood depth exceed 34%. Thus, for a stream depth of 10 feet, the error in defi ning the BFE could exceed 3.4 feet. Thomas’ analysis showed that standard errors for eastern streams, while lower, were still signifi cant. For locations with 30 years of record or less, the standard error for the 100-year fl ow is 30%, and the corresponding error in depth would be about 12%.

Lacking any streamfl ow record at locations of interest, or lacking a suffi ciently long record with which a statistical model can be fi tted reliably, many fl oodplain mapping studies use regional regression or rainfall-runoff modeling to establish the regulatory fl ow. The implications for land use regulation are signifi cant.

Standard errors for regional regression equations in the USGS National Flood Frequency program (Ries 2005) are summarized in Table 1. Due to the short record upon which those are based (following Thomas’ logic), about half of the regression equations will produce estimates of peak fl ows with errors in the 30 to 50% range. About 40% of the equations will produce errors of 50% or more for 100-year and 200-year peak fl ow estimates. Only about 1 in 8 equations will yield a 100-year fl ow estimate with a standard error of 30% or less.

This has a direct tangible economic “downside.”

Table 1. Standard errors for regional regression equations

...streamfl ow estimates using data with short histories have large errors ...

Benefi t of avoiding errors: Case study from Mecklenburg County

An example that illustrates the linkage between streamfl ow data, BFEs, and avoidable fl oodplain property damage comes from Mecklenburg County, North Carolina. That county, a leader in proactive fl oodplain management, funded a study in 2000 to determine the following:

• Increase in BFE due to land use changes since a 1975 study.

• Potential property damage from higher fl ood peaks due to future development - when property in a watershed is fully developed according to the land use master plan (a condition referred to as build-out).

The Mecklenburg study found that the average BFEs, based on ultimate build-out of the County’s 24 watersheds, were 4.3 feet higher than BFEs based on 1975 maps and land use. About half of the 4.3 feet increase was due to land use changes between 1975 and 1999, while half of the increase was projected from future development expected to occur between 1999 and build-out. The County’s economic analysis found that more than $330 million in fl ood damages to structures and contents would be avoided if the County adopted effective fl oodplain management measures and regulated future development to account for that increase (ASFPM 2004).

Thomas’ results of fl ow errors can be used to make the connection between streamgage data and these avoidable fl ood damages (Plasencia 2005). The Mecklenburg County study analyzed 17 watersheds and compared damages with the current building inventory to the damages that would occur in the future based on projected development. Of the $330 million in avoidable damages described above, 86% fell within 5 watersheds that have the highest projected growth. While the difference in BFEs in their studies was due to an increase in peak runoff from the effects of future development, differences of a similar magnitude would be expected from a hydrologic analysis that has errors consistent with those expected from analysis with 30 years of record.

The 100-year water surface profi les for the 5 streams in Mecklenburg County that comprise the bulk of potential fl ood damages show channel depths that range from 18—22 feet. From Thomas’ estimates for eastern streams with 30 years of record, the 12% error in fl ood depth corresponds to uncertainty of 2.2—2.6 feet for BFEs for these streams. These depth differences are comparable to the increase in BFEs that the County’s study predicted between 1999 and build-out. Therefore, it is inferred that the magnitude of potential economic benefi t due to reducing hydrologic uncertainty is comparable to the $330 million in damage avoided with enhanced regulations in Mecklenburg County over the lifetime of the project.

It should be noted that Mecklenburg County has made a substantial investment in streamgaging. The County funds 55 streamgages in the County or in the vicinity, with operation and maintenance provided by the USGS. The County has concluded that the benefi t from this investment offsets the cost, as the additional hydrologic records defi ne better the baseline conditions to support wise fl oodplain management, avoiding future fl ood damages.

...streamfl ow data yield accurate estimates of regulatory fl oods, avoiding $330 million in fl ood damages in North Carolina...

Benefi t of avoiding errors on a national scale

On a national scale, many fast-growing communities are facing similarly complex fl oodplain management decisions, and wise decision making for those will require accurate fl oodplain mapping.

Table 2 shows the 10 fastest growing counties between 2000 and 2005 that have special fl ood hazard areas. If the counties that need to update their fl ood insurance rate maps (FIRM) have cost savings similar to Mecklenburg County, more accurate estimation of the BFE possible with lengthened data sets may well exceed the cost of operating the streamgages.

How does FEMA value hydrologic data?

FEMA’s Blue Book (2002) identifi es monetary credits for technical contributions that cooperating partners make towards fl ood mapping. For detailed riverine hydrologic analysis, FEMA credits a cooperating agency $1,100 / stream mile to develop 100-year fl ow estimates with rainfall-runoff models, regional regression equation methods, or statistical analysis of annual fl ood peaks from gage data. (Note: This is not the total cost of delineating the fl oodplain. FEMA credits a cooperator $2,500 / mile and $1,400 / mile for computing the fl oodplain elevation and delineating the fl oodplain, respectively.)

As noted earlier, all three hydrologic analysis approaches have a foundation in streamfl ow data. We argue that without the data to develop or apply those approaches, we cannot estimate the fl ows and water levels for regulation. Without the regulatory water levels, we can produce no maps. Without the maps, we have no knowledge for wise fl oodplain land use regulation, and future damages will not be reduced to the extent possible.

To infer from this the benefi t, it is assumed that the value of estimating the regulatory fl ows equals the cost that FEMA is willing to incur to do so.

Table 2. Fastest growing counties1 with special fl ood hazard areas (US Census Bureau 2006; FEMA 2006)

A 2005 Government Accounting Offi ce (GAO) report addressed the need to update fl ood maps. The GAO report cited FEMA statistics showing that nearly 70% of the nation’s 92,000 fl ood maps or 64,000 maps are more than 10 years old and should be updated. In 2000, FEMA analysis of digital fl ood data for communities in the US showed an average of 4 miles / map panel. Using the FEMA credit for hydrologic analysis of $1,100 / mile, the cost of hydrologic analyses to estimate 100-year streamfl ows is $4,400 / map.

Thus the value of hydrologic analyses to develop 100-year base fl ood discharge for the maps that are to be updated is about $280 million.

Since updating fl ood maps is a multiyear process, we think it reasonable to spread the estimated value over 5 years, the original duration for FEMA fl ood map modernization. Ignoring the time value of money in this short period, we estimate the cost as $56 million / year. This benefi t alone represents approximately 50% of the $114 million annual cost (USGS 2006) of operating the streamgage network.

Example of updated Mecklenburg County FIRM

What is the benefit of meeting the information need for flood management project design?

Knowledge of fl ood volumes, peak fl ows and corresponding water levels, and the timing of the runoff from watersheds are basic requirements for planning, designing, and operating cost-effective fl ood management projects. These are projects with structures that keep damaging fl oodwaters away from property and people. The structures range from large dams upstream of major urban centers to small culverts beneath rural roadways.

How are these projects planned and designed?

Flood management projects commonly are planned and designed either to provide a selected level of protection from fl ood risk or to yield maximum economic return. In some special cases, the projects are designed to eliminate the risk to the extent practical.

Design for specifi ed risk

With this strategy, an acceptable risk is identifi ed as a matter of policy, and the facility is designed to provide that. For example, Table 3, from the Texas Department of Transportation (TxDOT) hydraulic design manual, shows that acceptable risk or level of protection varies for facilities designed and constructed by that agency. To plan and design a freeway culvert for TxDOT, for example, the project engineer determines the 50-year (p=0.02) discharge, and then sizes the culvert to pass safely that value. He or she will also estimate the 100-year fl ow and check performance of the facility with that.

Design for maximum benefi t

Larger fl ood management facilities are designed in a similar manner, using risk measures. However, that information often is used in a slightly different way. Rather than considering a single event and designing to avoid its consequences, this design standard considers the full range of likely events and their consequences and seeks to strike a balance between risk, benefi t, and cost.

Oroville Dam spillway (photo courtesy of California DWR)

For example, for design of a levee, the Corps of Engineers standard is to select from alternative designs the one that would maximize net benefi t without having an adverse impact on the environment. Net benefi t is the long-term benefi t attributable to the chosen alternative, less the cost of planning, designing, constructing, and operating that alternative. This long-term benefi t, in large part, is the economic value of inundation damage reduction due to the project.

Damage reduction in this case is not computed for a single fl ood event. Instead, it is computed as the statistical average for all likely fl ood events. Practically, that is accomplished by defi ning the complete fl ow or water level frequency relationship for the location(s) of interest, transforming that to a damage-frequency function using relationships between inundation depth and potential damage, then integrating the result.

The procedure for this is well known and well described in Corps of Engineers documents (USACE 1996) and computer software (USACE 1998).

Table 3. Texas Department of Transportation design standards (modifi ed from TxDOT 2004)

Design for negligible risk

Related to design for a specifi ed risk is design for the case in which risk should be at or approaching zero.

This is the standard for planning, designing, and constructing fl ood management structures for which failure would cause signifi cant loss of life and extensive property damage. Such standards are used, for example, for design of the spillway on a large dam.

In that case, rainfall-runoff models are used, with inputs derived through climate studies that seek to defi ne the probable maximum precipitation (PMP) for a watershed. This is defi ned by the World Meteorological Organization (1983) as a “…quantity of precipitation that is close to the physical upper limit for a given duration over a particular basin.” Chow et al (1988) suggest that this PMP can have exceedence probability from 0.001 to 0.000000002.

What information is needed for this planning and designing?

Methods

Planners and designers commonly acquire information needed for fl ood project design via statistical model fi tting, empirical models for design fl ow estimation, or rainfall-runoff modeling. Designs for specifi ed risk or designs for maximum benefi ts can be developed with any of these methods. PMP design, on the other hand, commonly relies on rainfall-runoff modeling, as the events of interest clearly exceed those experienced.

Role of streamfl ow data

The role of streamfl ow data in design for specifi ed risk has been described earlier: It provides the foundation for developing design fl ows. If a streamgage with adequate records is located at or near the site of a proposed structural measure, then a statistical model can be fi tted to the data. In the absence of that, regression equations can be developed with data from similar sites. If the information needs cannot otherwise be met, a rainfall-runoff model can be created. In that case, streamfl ow data provide the basis for calibrating and verifying the model.

Streamfl ow data play a critical role in estimating the PMP runoff too. For PMP runoff analyses, the rainfall-runoff-routing model that will be used must be calibrated, which is accomplished best by confi guring a model with estimates of parameters computing and comparing with historical observations, then adjusting iteratively to reach a good representation of the response of the watershed, channel, or reservoir system.

Benefi t from data

The benefi t of streamfl ow data for planning and designing fl ood projects is attributtable to costs avoided through good design based on adequately long records. This is illustrated here with two examples from California.

...the benefi t for planning and designing fl ood projects is attributable to costs avoided through good design based on adequately long records...

Flood storage in Folsom Reservoir

Folsom Dam and Reservoir, was constructed in the 1950s on the American River in California, east of Sacramento. This multipurpose reservoir was sized using hydrologic records collected from about 1905 until 1940.

Examination of the longer record available now shows that the 5 largest fl ood events on the American River have occurred since 1950.

Recent fl ood studies with the longer record indicate that the original fl ood reservation space, which was thought to provide protection against at least the p=0.004 fl ood, is, in fact, too small to provide that level of protection to the Sacramento metropolitan region.

To remedy this, the Corps of Engineers, US Bureau of Reclamation (USBR), and local governments have proposed measures that will enhance the protection, including increasing the height of the dam, expanding its outlet capacity, and constructing an auxiliary spillway. Estimated project costs to retrofi t Folsom Dam are on the order of $1 billion. To accommodate greater fl ood control releases, substantial levee improvements have been completed, and more are planned. The total authorized cost of these levee improvements exceeds $20 million.

As the design for Folsom Dam was optimized originally to provide the selected level of protection, we conclude that for this case, a longer record would have defi ned more accurately the true value. If so, the subsequent costly modifi cations could have been avoided, thus saving the $1 billion or more investment.

If that cost is amortized over 50 years, with a discount rate of 6%, the annual equivalent cost is about $63 million. Such a cost savings—for this project alone—offsets much of the annual cost of operating and maintaining the USGS streamgage network. Considering that the Corps alone operates 541 reservoirs and has interests in another 150 – 200 for fl ood control issues (USACE 1992), the potential savings in dam modifi cations alone may far exceed the cost of the entire streamgaging network.

Is it fair to attribute all this benefi t to the gages alone? Perhaps no more fair than attributing the success of our drive to the offi ce this morning to a bolt that holds in place the steering wheel on the car. In that case, and in the case of the gages, the absence will surely preclude success, even if we cannot claim that success is only due to the presence.

Folsom Dam (photo courtesy of USBR)

Central Valley levee system

The need to strengthen the fl ood defense system in California’s Central Valley provides an additional example of the value of streamgages and the benefi t attributable to long records of river stages and fl ows.

The Sacramento River Flood Control Project, planned, designed, and constructed in the early 1900s, protected agricultural areas with a network of levees and fl ood bypasses. With rapid urban development in recent decades, these same system levees now protect residential, commercial, and industrial property with high value. This growth, coupled with the larger fl ood events that have occurred in recent decades, has spurred investigation of the system reliability and capability to provide the desired level of fl ood protection.

Since 1986 the federal government, State of California, and local agencies have spent hundreds of millions of dollars to raise levees, construct deep slurry walls inside them, or place seepage berms alongside.

Representative unit costs for levee explorations and these repairs include the following (Mayer 2006):

• Geotechnical borings of levees cost approximately $100,000 / mile for explorations at the riverside toe, landside toe, and a deep boring from the levee crest.

• Levee bank protection ranges from $1,000 to $9,000 / foot.

• Seepage berms cost approximately $500,000 / mile with existing right-of-way;

however, some projects cost several million dollars / mile.

• Deep slurry walls cost at least $3 million / mile. A recent study to fi x 25 miles of levees in Sutter County estimated the cost at $7 million / mile (Dickey 2006).

• New levees cost approximately $20 million / mile.

Clearly with these high costs, decision makers do not want to err in determination of appropriate remedial actions. That is where the requirement for data enters the picture.

...overspending for 20 miles of levee repairs in a year would offset the annual cost of operating the USGS streamgage network...

Levee repairs in California’s Central Valley (photo courtesy of California DWR)

The appropriateness of a slurry wall or a seepage berm is determined from seepage computations. These consider the height of the water against the levee for the design event. This height is found, as it is for fl oodplain mapping, by fi nding fi rst the fl ow rate of specifi ed probability, then determining with a hydraulics model the corresponding water level.

Error in the design fl ood elevation, the likelihood of which increases with short or inadequate streamgage data, could lead to incorrect decisions about the need for repair or the best measure to use. The difference is signifi cant, with an additional $3 to 7 million / mile required for the slurry wall. Overspending for 20 miles of levee repairs nationwide in a year would offset the annual cost of operating the USGS streamgage network. Considering that the Corps has responsibility for maintenance and repair of almost 10,000 miles of levees nationwide, the cost savings could be tens or hundreds of millions of dollars. Similarly, failing to repair or strengthen levees for which that work is justifi ed could lead to failures and the consequent avoidable damage.

Sacramento area levee improvements (photo courtesy of SAFCA)

What is the benefit of meeting the information need for flood warning and reservoir operation?

Flood warning and successful reservoir fl ood operation share a common objective and a common requirement. Both seek to provide more time to make better decisions about responding to and managing fl oods. To do so, both fl ood warning and fl ood operations require timely and reliable precipitation and weather data for forecasting. Combined with timely and reliable data on current and recent streamfl ow these permit confi rming and adjusting the forecasts to conform to reality.

Where does the time go?

Figure 3 illustrates how time is spent responding to fl oods as they occur. The triangles represent milestones in the process, the last of which is exceedence of a threshold at which property is damaged, injuries occur, or lives are lost. If warning is available prior to that, some mitigative actions can be taken. The goal of a fl ood warning system is to ensure that this is so. Similarly, with fl ood operations of a reservoir, the goal is to provide more time to use better the storage available, perhaps releasing water now in anticipation of future infl ows.

Figure 3. Illustration of warning timeline (Carsell et al. 2004)

The maximum potential warning time is the time between the fi rst detectable or predictable precipitation and the time at which water level exceeds the threshold for damage or threat to life at a critical location.

Of course, this time varies from storm to storm and location to location. For example, if damageable property in a watershed is near the outlet, and if a short duration thunderstorm is centered near the outlet, the maximum potential warning time would be small. On the other hand, if the storm is centered at the far extent of the watershed or if a forecast of the precipitation is available before it actually occurs (a quantitative precipitation forecast), the maximum potential warning time for this same location would be greater. Similarly, the watershed state plays a role in determining the maximum potential warning time: If the watershed soils are saturated, the time between precipitation and runoff is less than if the watershed soils are dry. Snow pack conditions are important as well. Storms may add to or rapidly deplete the snow pack, thus causing greater potential for fl ooding.

Practically, response to a fl ood threat, either for reservoir fl ood operations or for fl ood emergency response, does not occur from the very onset of or prediction of precipitation.