hydrology for landfill rainfall-runoff vs regression

[digger95]

I am working on a flood study to evaluate the 100-yr BFE for a creek directly adjacent to a proposed landfill. (19.4 sq. mi drainage area, all agriculturual very flat, a lot of depressional storage in fields. The stream is highly channelized(4+ miles) and ungaged.

Using TR-20 flow calculated at 9,100 cfs using state regression equation (Clark Unit Hydrograph) flow is calculated at 4000+/- cfs.

I have read through FEMA's "Guidelines and Specifications for Flood Hazard Mapping Partners, Appendix C: Guidance for Riverine Flooding Analyses adn Mapping".

I Also read through the "Evaluation of Flood Frequency Estimates of Flood Frequency Estimated for Ungaged Watersheds", by the Hydrologic Frequency Analysis Workgroup(HFAWG), and the Minority Opinion.

It appears that the acceptable method is the USGS Regression Equation, unless the watershed does not fall into the specifed parameters of the USGS Regression Equations; however the Minority Report by the (HFAWG) indicates that using the Regression and standard error to determine if a Rainfall-Runfall model is acceptable, appears to rely to heavily on the Regression Equations at the Benchmark.


My concern: Due to the sensitive nature of the proposed site usage (landfill), I do not want to blindly use one method over another, however it seems to me that if one uses the regression equations, and one has to defend their position in court, it seems to me like someone could easily run the runoff-rainfall calcs and show a flow is nearly double what the regression equation shows (and subsequent increase in the BFE).

We have been collaborating with another consulting firm in town and they feel the regression equation is the way to go, based on the IDNR and FEMA's willingness to accept the regression equations, but I am not sold yet.

The acceptance of the regression equation seems counter intuitive when one can use a rainfall-runoff calculation and calculate a flow that is nearly double the regression flow.

My personal feeling is that if you are calculating the existing BFE for an unstudied stream and you know the potential use such as a school, landfill, residences, business, etc, it is easier to justify the additional earthwork, comp storage, etc. than it is to explain why the facility flooded because you used method x over method y to determine the flows...

 

[blueoak]

I typically think the regression equations are "closer" since there are often a lot of conservativeness in a hydrologic model. However, for something as frequent as the 100 yr, why not be conservative, especially when it is for a critical or hazardous application. Plus you know things will change in the future, especially when your flow is less than a cfs per acre. This is where I like to look at costs. I have seen too many fights over the hundred year only to notice the design doesn't even change.

The other item is I know a lot of people assume AMC II for curve numbers. My thought is what happens when your 100-yr rainfall happens after a week of rain, is your client really going to grasp the difference?

 

[francesca]

In a small watershed (500 acres), I used TR-55 to calculate my flows as I was fairly confident with my land uses, topography, times of concentration, etc.

In a large watershed (65 square miles) I used the regression equations, which still left me with a bit of a dilemma. Did I use rural (the majority of the watershed) or urban (the future condition for a large part of the watershed and the condition in the proximity of my stream)? And what to do with the +/- 50%?

I ended up running my HEC-RAS model under various flow conditions (rural/urban, high/mid-range flow value) and guestimating that a basin-wide storm event was more likely in winter than summer (in TN) when frozen/saturated ground was also a distinct possibility, but foliage would be less dense so higher velocities would occur on the overbanks.

I ended up with two flow scenarios and two roughness scenarios, which roughly equated to a summer and winter conditions, and at the end, the two differed in elevation by less than 1' at each cross-section.

It is tricky when dealing with significant depression storage, but the regression equations should take that into consideration. For instance, I used a regression equation for a karst geology and therefore didn't have to worry with trying to account for the thousands of sinkholes in my drainage basin. The problem you face with depression storage is that, in winter, evaporation is so low that your depressions may be full when the storm arrives, so, while its tempting to reduce your runoff, it might not be worth the risk.

The other problem you face with the flat topography is that a difference of water surface elevation of 1' probably corresponds to a significant increase in floodplain acreage. Make sure that you only designate a floodplain on your client's property, and that you add at least a foot or two to your final answer to set the minimum elevation of their landfill.

 

[digger95]

Thanks to both of you for your input, I spoke with Will Thomas of the HFAWG and he reccomended using engineering judgement to see if we can get within the +/- 50% error. The other consultant across town came up with a Q of approx 3,200 using the Clark Unit Hydrograph and we initally had 9100, it seem like we can adjust some of the input parameters to tighten up the TR-20 Calcs. By using the latest regression equation it appears that we will be able to come up with a Q (5,500 cfs) that will be within +/- 50% percent of each value so it appears that we will be able to defend the regression equation in court if necessary.

However I definitaely agree that any structures that are proposed will be at least at the BFE + 2 feet to provide a bit of extra insurance.

The question that I still cannot resolve is if the regression is so great, why not just say that if your watershed meets the parameters (slope, drainage area, cover, imperviousness, channelization, etc) why have the other methods listed as acceptable (as long as they are within the standard error)? Why not just say the regression is mandatory for these situations.

 

[blueoak]

Our regression equations are pretty sparse and have limited application so they are sometimes more of a check of reasonableness.

Would you then use the 95% confidence interval for the regression equation? Some of mine have an 82% error, and that is only for the limited information sometimes highly influenced by periods of drought. I have trouble depending on them alone when I cannot find a gage to calibrate my model to, how accurate is the regression then?

 

[RWF7437]

You need also to consider consistency with other, previous studies. Look at any nearby streams which have been studied in detail and see what those studies used to calculate flows. Also, if you have nearby, similar streams which are gaged get the gaged flows for those streams and compare them to your stream. Regression equations are based on stream flow data while most hydrology models are based on rainfall data, as you probably know. It is good to be conservative but you may be able to protect the landfill from flooding without being overly cautious, and expensive, on other reaches of the stream you are studying.

good luck

 

[queque]

I recommend you discuss the regression equation with the authors- USGS Surface Water experts within your State.
They are the ones that developed the regressions and know the extent of errors and their meaning.
Errors much greater than 75% are typical in the Western US, where insufficient data exists to develop menaingful relations between rainfall and runoff.