Opinions on Hydrology Model Applicability for Large Watersheds?

[debgallery]

A little bit of background...
I'm doing a peer review of a Stormwater Masterplan by another consultant. The watershed has ten sub-basins, six of which are included in FEMA's FIS for the region, meaning there are published peak flows for the 100-year flow (I'm focusing on the 100-year peak flow for simplicities sake in this post). These are based on regression equations developed by FEMA a long time ago. The consultant used PCSWMM to generate their own hydrographs for the basins, calibrating where gage data was available. The basins range in size from 600-acres to 18,000-acres and are between 20% and 95% developed. They're located in the mountains of Colorado with generally steep slopes.

When comparing the results of the peak flow analysis from PCSWMM to FEMA, five of them are surprisingly close but the peak flow in one basin doubled with the PCSWMM model (400-cfs to 800-cfs). This basin is 2,500-acres, is roughly 20% developed and is ungaged. I decided to run it through the USGS NSS software and got a peak 100-year flow of 137-cfs (73% above 7500 elevation).

Now my questions....

Is PCSWMM the right model to use for this scenario (strictly hydrologically speaking)? My understanding of EPA SWMM is that it is for primarily urban watersheds. The majority of these sub-basins are less than 40% developed.
Is USGS the right model to use? The documentation on the software says it is appropriate for natural streams, which they further define as being less than 10% developed. If the basins are more than 10% developed does it automatically eliminate NSS as an option?

Is there another method that might be more appropriate?

I'm thinking of maybe generating a hydrograph in NSS (I think you can do this) and then using it in the PCSWMM model for the sub-basins where it is appropriate.

The masterplan is recommending upgrading a lot of infrastructure based on these flows and I want to make sure they are balancing "being conservative" with reality.

Really looking forward to hearing your opinions.

 

[gbam]

Since you are using the FEMA Regression equation results as control wouldn't it be prudent to evaluate them first? What is the deviation for the equations used and are the model results within the tolerances of the Regression Equations when compared? Just because the model results are double that of the Regression results doesn't mean it is not correct. I have seen regression equations applied incorrectly before.

Since the consultant calibrated the model for some of the watersheds did they apply the calibrations with or without justification to the area(s) in question? I would think that there would be justification to or not to use the calibrated model.

As to your question, I have not used PCSWMM but there are so many different hydrologic models each one slightly different, it is important to be able to justify the use of a specific model and that it is accepted by the reviewing agency.

 

[cvg]

depending on when the FEMA study was done, the regression method used may have been OLS, WLS or GLS. Both OLS and WLS have significant issues which may result in inadequate estimation of the flow rates. In addition, the regression methods are not appropriate for developed watersheds. So, comparing a rainfall runoff model result to the regression equation result is not recommended. There is nothing wrong with SWMM, but I would look at the specific data and methodology used in the SWMM model. Specifically, combinations of rainfall - rain on snow - and snow melt; infiltration rates on frozen ground etc.

 

[debgallery]

Thanks for your responses. I am looking into the FEMA regression equations more. The reason I'm so concerned is that the community has been using the FEMA flows for design of several major infrastructure projects in recent years and this new study indicates that they are all undersized and require replacement (major $$ and a lot of explaining to the tax payers). I fully recognize that just because the new calculated flow is double the FEMA flow it doesn't mean its wrong, but since its the only basin where this occurred (the other five are actually fairly close) I thought it was worth examining further. This also happens to be the least developed basin and ungauged.

So let me ask another question - at what point do you consider a watershed "developed"?. 10%? 20%? 50%?

 

[SMIAH]

I don't trust a rainfall-runoff for a 2,500-acres watershed point.

You may want to (try to) find 2-3 similar watershed (as of size, average slope, developpment, location, etc.) with recorded flows (using peak flow factors if daily flow is recorded) and perform a statistical analysis and transposition for the 2,500-acres watershed (?).

Compare this to the results of a rainfall-runoff.
Check if there's other gage downstream of your location on the same stream?
Check 100-year flow with 2 cross section (uniform flow i.e. manning) to see if the peak flow is probable regarding WSEL and med velocity?

Voodoo Hydrology.

 

[beej67]

If I were you, I'd use this:

http://pubs.usgs.gov/sir/2009/5136/

Hydrology, Drainage Analysis, Flood Studies, and Complex Stormwater Litigation for Atlanta and the South East -

http://www.campbellcivil.com

 

[debgallery]

I did - NSS=National Streamflow Statistics.
I got 137-cfs, compared to 400-cfs (FEMA), and 812-cfs (new studies PCSWMM)

 

[TerryScan]

This is also a handy tool:

http://water.usgs.gov/osw/streamstats/colorado.html

 

[debgallery]

Anyone know where to find the Comparison of Nine Uncalibrated Runoff Models to Observed Flows in Two Small Urban Watersheds, (Zarriello, P, 1998) paper referred to in earlier posts? The links to the USGS page are no longer valid and a general google search has been getting me no-where.

 

[TerryScan]

I found it here:

http://acwi.gov/hydrology/mtsconfwkshops/conf_proceedings/first_fihmc_nevada.pdf

It was a VERY slow download (had to use "save as..") If you have trouble getting it there, let me know.

 

[cvg]

NSS is not suitable for developed watersheds, so you should not use that. Neither are regression methods unless they are based on regression analysis of suitable, similar watersheds. However, in general, not very good for developed areas such as your 95% developed watershed.

the new calculated flow is double the FEMA flow it doesn't mean its wrong, but since its the only basin where this occurred (the other five are actually fairly close)

the FEMA flows are based on regression equations, regression equations are not recommended for developed conditions, they were done a long long time ago, meaning they potentially used Ordinary or Weighted Least Squares regression methods, both of which can have significant skews. So it is clear that the FEMA flows may not be considered either accurate or adequate for design of urban drainage infrastructure. the fact that the SWMM model compares favorably with the regression results in some cases is also reason for concern as I would not expect them to be similar. I would have expected all of the rainfall runoff models to be different than the regression. Unless significant retention or diversions have been constructed (they would have to be very significant to have a measurable affect in the 18,000 acre watershed), I would expect peak flow rates would have likely increased due to the development.

you might check with the state engineers office library to locate the paper

http://water.state.co.us/Home/Pages/default.aspx

 

[beej67]

We have FEMA regression studies for both undeveloped and developed watersheds here in Georgia. One is calibrated to region, the other is calibrated to region and impervious cover ratio.

Hydrology, Drainage Analysis, Flood Studies, and Complex Stormwater Litigation for Atlanta and the South East -

http://www.campbellcivil.com