Urban Stormwater Modeling Offers Room for Improvement
As reported in last week’s blog, the March/April 2010 issue of Stormwater magazine includes several useful, provocative articles.
We wanted to offer some thoughts on several of these.
First off, let’s look at Improving the Practice of Modeling Urban Hydrology (Erik G. Peters)
From the outset, Peters acknowledges - perhaps confesses - that, although stormwater professionals would all maintain that they are committed to protection of water resources and prevention of flooding or erosion of downstream properties, we have been less than successful. Studies demonstrate that very low levels of urban development (5% to 15% impervious surfaces) have resulted in degraded streams (Booth and Jackson 1997, Wang et al. 1997, Short et al. 2005). Other studies indicate that the quality of our waters is still drifting downward. What are we doing wrong?
Reducing urbanization effects on our water resources is a challenge. Many studies have confirmed that urbanization has a greater impact on frequent events than on the rare flood events (ASCE 1993). Urbanization’s effect on hydrology typically results in the following:
1. The size of precipitation event necessary to generate runoff is lowered.
2. There is an increase in the peak runoff rate and volume, particularly from the smaller more frequent precipitation events.
3. There is an increase in runoff pollutant concentrations and loading (mass/surface area/time) to water bodies
Peters argues that what’s needed is to improve the current state of engineering practice - particularly urban stormwater modeling - through a better understanding of the challenges/assumptions behind the most common urban stormwater models (e.g., the Rational Method and the Curve Number or CN Method) and through providing design guidance for developing a hydrologic model using the CN Method (more specifically, developing curve numbers). The author also argues for greater focus on Time of Concentration, “…the second most influential parameter in the CN Method, behind selection of CN values.” Peters points out - and this is something that’s been discussed in this Blog in past weeks (refer to the article for the referenced figures):
The CNs were originally developed to predict runoff from relatively uniform agricultural landscapes, based on research conducted largely in the eastern and midwestern US (Woodward et al. 2002). These areas of the country receive almost all of their annual precipitation in the form of rainfall. A moderately sized storm event in these regions is large enough (e.g., typical two-year, 24-hour storm event exceeds 2.5 inches or 64 mm) that CNs typically approach a constant value with rainfall depth (Figure 1). In other words, the entire watershed is contributing runoff, initial losses have been satisfied, and runoff contribution from vegetated areas is a significant contribution to the total runoff volume from a watershed. Approximately 70% of watersheds fit the pattern of Figure 1 (Hawkins 1993).
When trying to model runoff from smaller storm events or landscapes that don’t meet the above criteria, the engineer, designer, or reviewer must be more careful in the approach. When only a portion of the watershed is contributing runoff, then the CN for the overall watershed (composite CN) will vary, typically decreasing with rainfall depth as shown on the left half of Figure 1. To address this scenario, the best and most defensible method is to break up a watershed into subwatersheds of similar runoff-generating potential. This is frequently referred to as the distributed CN approach.
Peters spends a considerabe amount of time pointing out the dangers of using a weighted or composite curve number approach, arguing instead for a distributed curve number modeling approach.
There are two methods or approaches for estimating CN for watersheds having more than one hydrologic soil-cover complex. The two are commonly referred to as the composite CN and the distributed CN approach. The National Engineering Handbook, Part 630, Hydrology (NEH 630), Chapter 10, refers to the two approaches as the weighted-CN and the weighted-Q respectively.
A composite CN is an area-weighted average CN calculated for an entire watershed. In a distributed approach, polygons within a watershed are broken out based on runoff-generating potential. There is no CN averaging; rather, separate CNs are developed for each polygon and separate runoff values calculated (Grove
et al. 1998).
The most common approach is the composite CN. However, employing the distributed CN approach is necessary to avoid significantly underestimating runoff volumes when differences in CN values within a watershed are large or precipitation depth is small. The underestimation of runoff using the composite CN approach is a result of the nonlinear relationship between CN and runoff depth (Figure 2).
As the focus of stormwater management increasingly includes smaller storms, use of weighted vs distributed curve numbers becomes an even greater problem. Peters comments that although composite curve numbers can continue to be a reasonable approach for very large storm flooding analyses, “…a composite CN approach for site specific stormwater management design within a proposed development won’t be appropriate for most communities.” distributed curve numbers are vastly preferable and more accurate. The article includes a variety of additional practical recommendations, designed to improve modeling applications.
You can access the complete article here.
Our next post will look at A Better Way of Measuring BMP Effectiveness…
