Lab 02: Estimating Flood Frequencies in Urban Oregon with the NFF Model

Lab 02: Estimating Urban Flood Frequencies in Oregon

with the NFF and Rational Models

Geog 422/522, River Mapping and Modeling

Department of Geography, University of Oregon

Goals of this Exercise:

This exercise introduces you to some of techniques used to estimate urban flood hazard in the United States. Specific goals of the lab and readings are to improve your understanding of:

• Processes controlling flood occurrence in urban areas - especially inOregon,

• Modeling approaches and applications– in this case, the NFF program used by the U.SG.S. and the Rational Model for urban locations,

• Data acquisition –how to find and download data for peak flows from a gage site, as well as data included with the NFF documentation, and

• Model evaluation - through a comparison of the NFF estimates to estimates based on gage data and estimates from the Rational Model.

Required Software

The NFF (National Flood Frequency program), version 3, software for this exercise has been developed by the U.S. Geological Survey and is available over the web at no cost from: The software is loaded on the SSIL server at the Geog422/Data directory. Note: the national data base that accompanies the NFF program requires Microsoft Access in order to run.

The Rational Model that we will also be using in this exercise does not require any software, although you might want to use a spreadsheet for calculations.

Background Reading Required for Lab 02

A number of references on peak flow estimation approaches, peak flow estimates in Oregon, and the NFF modeling approach are listed at the end of this exercise. The NFF Help File model provides a good overview of the history and basis for the flood estimation approaches used by the program. I strongly recommend reading through the Help Files as you do this exercise.

Reading 1: A brief description of the NFF program (Perl and Ries, 2002) with links to documentation is at IF the USGS server is down, I have also loaded a pdf version to the lab folder and to the class web page. Read the description before proceeding with the lab.

Reading 2: A brief description of the Rational Model and its underlying assumptions is in pages 298-305 of:

Dunne, T., and Leopold, L.B., 1978. Water in Environmental Planning. New York, W.H. Freeman and Co., 818 p.(read pages 298-305).

This reading in on reserve in the standard locations.

Evaluation of Lab 02

As you complete this lab, please keep a list of improvements that could make the exercise a better learning experience. Turn in the list with the lab when you have completed it.

Part 1. Factors Controlling Peak Discharges in Urban Environments

In Table 1 below, check (X)the boxes to indicate which variables at different locations are used in the multiple regressions to estimateurban flood magnitude. You can access this information by: (1) looking up the individual state files on the USGS web site at: ( or (2) bringing up each region within the NFF program and observing which variables are requested. NOTE:the National Urban scenario won’t even run for many states unless you first specify a Rural scenario location for the state. This is because the urban calculations need to know which region (e.g., Willamette) the calculations apply to.

1-A. Fill in Table 1on the Answer sheet with an X in boxes where variables are used.

1-B. Which variable(s) is/are included in all the flood estimates? Whatmakes this/thesevariable(s) so critical to runoff processes, regardless of location?

1-C . On last week’s lab, you speculated about regional variability in hydrological processes that led toregional variations in the variables included in the multiple regressions. What non-process factor or factors might be affecting which parameters are included in the urban regressions? What implications might this have for model consistency from location to location with the urban equations?

1-D. Take some time to cruise around the urban equations/documentation for different states. You will notice that many states do not have any equations for urban areas – even where the state may have significant urban populations. Why do so many states not have local urban equations?

Part 2. NFF Model – Urban Application and Sensitivity Analysis

Log on to the NFF model. Read through the urban section of the NFF Manual (found by clicking on Help on the NFF menu bar).

As a planner with Environmental Services, City of Portland, you need to estimate the peak discharge in upper Fanno Creek as part of the restoration efforts associated with the Fanno Creek Watershed Plan ( Determining the range of likely flows is a critical component of designing channels, determining floodway corridors, and planning retention basins. Your plan is to work in upper Fanno Creek above the 56th Ave. USGS gaging station (hydrologic site identification number:14206900). The upper portion of the Fanno Creek watershed is shown on the USGS Beaverton Quad and orthophoto available in the SSIL folder for this lab and on the class web site.

As you can see from looking at the map and orthophoto, Fanno Creek above 56th Ave. is a small watershed in the greater Portland, Oregon area. Fanno Creek was the first watershed to be urbanized in the TulatinRiver catchment. Because it has been suburbanized for a long time, trees have grown back into yards and clear areas, giving the catchment the appearance on air photos of being relatively undeveloped with lots of park land. In actuality, only 13% of the total basin area is still undeveloped. Portions of the Fanno Creek channel have been set aside as greenways. The combination of greenways and tree growth in residential areas creates the illusion of a relatively “natural” channel and watershed in a highly urbanized area. However, the combination of residential-urban development, steep slopes, and high precipitation have led to significant channel change and water quality problems. Specifically, increases in peak runoff coupled with loss of vegetation cover have made sediment loading and bank erosion a major problem along significant portions of Fanno Creek. In addition, loss of riparian buffers, failure of septic systems, and pollution from fertilizers and automobiles have led to the standard gamut of residential urban water quality woes: high temperatures, high nutrient loads, and bacterial contamination.

In order to estimate the peak runoff volumes from upper Fanno Creek, you decide to use the NFF urban model. You have two options with this model: the National Urban equations (Sauer et al., 1983) and the equations for the Portland, OR – Vancouver, WA region (Laenen, 1980 – on reserve in the standard locations). Because the Portland, OR equations were developed for this locale, you are almost certain this approach will give you the more accurate answer. However, you also decide to run the National Urban equations to gain a sense of the possible range of answers.

As you know from Table 1, the Portland, OR equations require data on: drainage basin area;sum of Land Use Types I and II within the basin; gutter length per unit area; and water storage. You could visit the Portland planning office, get basin area from the U.S.G.S. site, or do your own measurements to get these values. Alternatively, you might check page 13 of Laenen (1980), where you discover to your delight that he has already compiled values for a number of key runoff controls in Fanno Creek at 56th avenue (Table 2).

2-A. Useappropriate values from Table 2 with the Portland-Vancouver equations in NFF to calculate flood recurrence intervals in Fanno Creek. Use these NFF-based values to fill in the first blank column of Table 3.

The problem with using values from Laenen (1980) is that urban landcover variables may have changed since 1980. Channel slope and contributing basin area probably will not have changed unless there have been major channelization efforts, diversion of flow out of the basin (which oftenoccurs in urban areas where storm sewers take flow out of the basin), or addition of retention structures and features. None of these factors have been significantly altered in the upper Fanno Creek basin since 1980.

However, gutter length and percent area in Types I and II land cover may have changed since 1980. Rather than re-measuring the variables, which would require significant time, you decide to run a sensitivity analysis. A sensitivity analysis examines howmodel outputs respond to changes in input parameters. A sensitivity analysis is conducted by varying one or more parameters while holding the others constant. For the purposes of this analysis, the sensitivity analysis should be conducted for the range of values that might reasonably occur in Fanno Creek in the post-1980 period. One way to estimate a reasonable range of values for the sensitivity analysis is to work within the range of values for Portland documented by Laenen (1980), while also using our personal knowledge of the Fanno Creek area (gained from visits, land records, air photos, maps, etc.) Tables 4 through 6 below list a range of potential variations in landcover and gutter lengths based on this approach..

Input variables might increase or decrease in value from those of 1980. For example, increased area in landcover type I might occur if some buildings were torn down and the lots were left vacant. Alternatively, a more likely scenario is that landcover in types I and II has decreased since 1980 due to residential development.

2-B. Use NFF with the Portland-Vancouver equations to conduct a sensitivity analysis. Fill in the values in Tables 4 and 5 below to create the data for the analysis. Keep all the other variables constant at the 1980 level while varying the parameter of interest.

Note: I highly recommend using a naming convention that helps you keep track of files in NFF. For example, you might name your urban files for the land use sensitivity analysis LU = 20, LU = 15, LU = 10, etc.

2-C. In technical reports, sensitivity analyses of this sort are usually reported in graphical form, which makes them easier to interpret. Create one graph for Table 4 and one graph for Table 5. The graphs should have land cover or gutter length as the x variable, and discharge for each recurrence interval as the y variable. Use the same symbol for all the 2 year recurrence discharges, a different symbol for all the 5 year recurrence discharges, yet another symbol for 10 year recurrence intervals, and so on. Attach the graphs to the end of the Answer Sheet.

2-D. Based on: (1) your results in Tables 4 and 5 and the accompanying graphs; (2), the uncertainty in the model (check out the standard error); (3) the maps and aerial photos of Fanno Creek, and (4) other knowledge you acquire about Fanno Creek from web sites and your research, how critical is it to re-measure the parameters of land cover and/or gutter length? What factors might play into your decision on this matter?

Alternatively, rather than only one parameter changing at a time, it is more likely that gutter lengths and land cover in Types I & II will vary in an inverse manner. For example, as land is converted to residential use, the area of land in Types I II will decrease, while the gutter length will increase. Therefore it is important to also evaluate the sensitivity of the Portland-Vancouver equations to having both variables change concurrent.

2-E. Use NFF with the Portland-Vancouver equations to conduct a sensitivity analysis where land cover and gutter length vary concurrently. To run the National Urban NFF module you must first run the Rural module. Be sure to clear out any previous rural scenarios you have in NFF (the program is very finicky this way and can easily “choose” the wrong rural scenario to calibrate the urban results). With only the correct rural scenario open in NFF, run the National Urban scenario for Fanno Creek, using a 2-yr-2-hour precipitation value of 1 inc. Keep all the other variables (basin area, storage) constant at the 1980 levels. Fill the values in Table 6 below.

2-F. Do your values in Table 6 above change your thinking about your answer to 2-D in which you speculated on whether new measurements were necessary? Why or why not?

You will notice that NFF also provides a National Urban equation (Sauer et al., 1983) for areas where local regression estimates are not available. The National Urban model in NFF can be applied using the data in Table 2 and the guidance in the Urban Flood-Frequency Estimating Techniques chapter of the NFF Manual (accessed under the Help button on the main NFF menu bar). The Basin Development Factor (BDF) for Fanno Creek can be estimated using the aerial photos and map, although ideally an on-site visit and survey should be conducted.

2-G. Use the National Urban equations in NFF to estimate the flood frequencies for Fanno Creek at 56th Ave. Fill in the discharges for the National Urban column in Table 3.

2-H. What processes or factors do you think are leading to the discrepancy between the National Model and the Portland-Vancouver model results? Explain your reasoning.

Finally, a frequent goal of urban stream naturalization projects is to reduce peak discharges – sometimes to pre-urbanization levels. This can be done through many methods; one of the more commonly used structural approaches is to create small retention ponds and holding areas to increase storage. Almost all U.S. cities now require such structures when parking lots or large buildings are built. Creating storage after the fact, however, can be costly and difficult to implement because of land ownership issues, pre-existing land uses, and a host of other complications.

Use NFF to determine how much storage would be needed to bring peak discharges back to pre-urbanization levels. Do this by using the rural equation for the Fanno Creek watershed to estimate peak discharges prior to urbanization. Then use the Portland-Vancouver equations to estimate runoff with varying amounts of storage until discharges are approximately equal to the rural discharges..

2-I. Focusing on the 2 through 25 year recurrence intervals, approximately how much storage did the NFF model indicate is required in order to bring urban peak discharges in line with rural ones? How accurate do you believe this answer is and why?