Topic Thread

Expand all | Collapse all

Please provide your opinion about CN values for VERY LARGE RAINFALL events

  • 1.  Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 03-31-2017 03:42 PM

    To All,

    I am involved with a local group trying to better understand the hydrology of the October 2015 flood that impacted South Carolina from roughly Columbia to the coast.   Different sources indicate the rain was on the order of a 1,000 event.  During computer simulation studies to recreate the flooding and streamflow, modelers have used CN values as low as 65 and place high confidence in runoff volumes and peaks simulated by the models they are using.  As FYI, these models use standard CN and compute Ia=0.2S.  Said CN values were applied for the entire storm duration.

    The following are snippets from one report about the event.

    Hourly rainfall rates at the Forest Acres Richland County Emergency Services Gills Creek automated gage (Forest Drive and I-77) recorded 1.76 inches from 2-3:00 a.m., 3.76 inches from 3-4:00 a.m., 3.00 inches from 4-5:00 a.m. and 2.12 inches from 5-6:00 a.m. yielding an unprecedented 10.64 inches over four hours.   At noon, the Forest Acres Gills Creek gage rainfall amount had risen to an incredible 15.51 inches for the past twelve hour period. (FYI, I live in the Forest Acres area and across the road from Gills Creek.

    Please provide your thoughts/opinions about CN application to this event.  Should the CN value be increased during the event as the ground became saturated?

    Help me, please.



    ------------------------------
    Michael Meadows P.E., F.ASCE
    Assoc Professor
    Univ Of South Carolina
    Columbia SC
    ------------------------------


  • 2.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 03-31-2017 03:47 PM

    Mike,

    I have successfully replicated a recorded stage in a reservoir with a CN and a precipitation of about 12 inches as I remember.  So I would assume that CN hydrology would work in you case. I used standard NRCS procedures.



    ------------------------------
    Donald Woodward P.E., F.ASCE
    Prof Hydrologist
    Gaithersburg MD
    ------------------------------



  • 3.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 03-31-2017 03:50 PM

    Hi Mike and all,

    I took a quick look at an 8000 acre region in Lexington/Richland County just below Columbia with Web Soil Survey. There appear to be many A/B soils with some significant C/D soils in the region. Slopes were zero to 2% for the most part. Available (0 to 60 inch) storage ran about 10 inches (the major block of Congaree loam of 2000 acres). Part of this storage was no doubt filled before the event began. The layer K value was roughly 22 inches per day with lots of other higher values. I don’t know the situation below 60 inches but it is probably much less permeable.

    With the series of rainfalls you outlined, I don’t believe you experienced complacent/violent behavior but I would speculate that you were losing storage S as the event unfolded. This would in effect result in increasing CN values  as you are thinking.

    This could be at least a significant footnote for the ongoing revision. Thanks for sharing this.



    ------------------------------
    Ernest Tollner P.E., M.ASCE
    Professor
    University of Georgia
    Athens GA
    ------------------------------



  • 4.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 03-31-2017 04:06 PM

    Mike

    I agree with Bill… even not knowing the details, like how much runoff (Q) was there?

    May not have had the lead-in Complacent limb, but surely approaching Violent response. That is 1:1 with dQ/dP

    At the VERY large event we don’t have much experience.  Expectations are that it’s a different game.  Well past the curvy part of the CN’s  P:Q plot.

    Plus, if an instrumented research  installation encounters one, it often blows out the site and leaves no record.

    Assuming that it‘s a Curve Number world out there….The 65 CN  doesn’t look impossible as a start-of-storm CN.  depending on the site, but the rain surely way beyond experience limits.   

    Averaging  the 2 rainfalls comes to about 13 inches. On a CN(20) of 65, that gives P/S=2.4 . which is  well into the development of runoff processes.

    Don’t know if it’s useful perspective, but in the 46 years of record at Coshocton ws 26020, [with cn=70] the biggest P/S they ever got was 1.1.  

    Continuing with the CN belief system, the CN 65 and 13 inches of rain  gives a contributing area fraction (dQ/dP)  of about 90%.

    All above assumes Ia/S=0.20

     

    This is also a problem that comes up in PMFs for spillways: P= mean + 15 standard deviations!  Really big storms.

    Also, there may be as much uncertainty with the hydrographing in the model as with the CNs?   In my thoughts they’re as suspect as the CNs. What model was being used?

    Hope this was of some help. Let me/us know how it comes out.

     

     

    RHH



    ------------------------------
    Richard Hawkins Ph.D., P.E., F.EWRI, F.ASCE
    Professor Emeritus
    Tucson AZ
    ------------------------------



  • 5.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 03-31-2017 04:11 PM

    One question …

    Are you more interested in Q (volume) or q (peak discharge)?

    To follow on to Pete’s discussion.  Q for CN=65 with 10.64” of P = 6.12” or about 57% of P.

    Assuming an AMC III at the start of the storm, CN goes up to 82.  Q for CN=82 with 10.64” of P = 8.78 inches or about 83% of P.

    Runoff volume will certainly affect peak discharge, but more of the effect on peak is probably seen in the rainfall distribution, the unit hydrograph, and the Tc used in the model.

    With the largest volume increments towards the middle to end of the storm, the system is essentially primed before the heavy rains arrive and we can expect larger qs than if the heavy rain occurred during the beginning of the storm.

    All of these parts have to play together to get the whole picture.

    Are there gages to do some comparison? I’m guessing not, but it would certainly be helpful to have something to calibrate to.  High water marks might be a potential.



    ------------------------------
    Claudia Hoeft P.E., F.ASCE
    Nat'l Hydraulic Engr
    USDA - NRCS
    Washington DC
    ------------------------------



  • 6.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-03-2017 09:50 AM

    I hate to pour cold water on your desire to model extreme flood events with curve number (CN) hydrology, but I have to ask the question: Why are you using this method?

    CN hydrology was invented when we were using slide rules to calculate flood events. I still own a slide rule (I have been doing flood studies since the early 1970s), but I don’t use it any more. This simplified method does not take into account the impact of changing soil moisture conditions on runoff timing, volume, and magnitude. A soil that has a curve number of 65 in average antecedent soil moisture conditions (whatever they are) is not going to act the same during a 100-year or 1000-year event. Trying to guess what the curve number should be and how it changes during large storm events is a pointless activity. Today with computers there are better ways to model hydrology. Specifically, I am talking about continuous simulation hydrology. Continuous simulation hydrology is not new and it is not too complicated. I describe the argument for using continuous simulation hydrology instead of the old, single-event hydrology methods in my paper “Dumbing Down Hydrology” published in the July/August 2016 issue of Stormwater magazine. The paper is also available on my LinkedIn page. Also, I am happy to send a copy of it to anyone who is interested. The bottom line is: today we have better hydrologic modeling tools than CN hydrology. As professionals we should be using them.




    ------------------------------
    Douglas Beyerlein P.E., D.WRE, M.ASCE
    Principal Engineer
    Clear Creek Solutions, Inc.
    Mill Creek WA
    ------------------------------



  • 7.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-03-2017 12:08 PM
    I am of the same opinion as Douglas on this one. Empirical models like the NRCS (SCS) derived methods using curve numbers blow up in extreme events, larger watersheds and steep terrain and become non-translatable across watersheds and rainfall volumes, indicating to me that they are fitting the model to the data for the single event only and will not scale. A time-based infiltration and loss model such as Green and Ampt seems more appropriate. 

    ------------------------------
    Michael Anderson P.E., M.ASCE
    Owner
    AEU
    Carson City NV
    ------------------------------



  • 8.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 10:26 AM
    One of the things we forget as hydrologists is that we are extrapolating beyond the data when we make statements like it was a 1000-year event.  In this we need to be careful.  The best measure of the rainfall-runoff response is some sort of stream gage.  Given that the event that you are discussing was regional, look for a stream gage or number of gages, and see how the flow rate in those stream gages responded to the rainfall that the watershed received.  I am guessing that the FIS books for your floodplain models have already done this for events up to the 500-year event.  For most areas in the US, the difference between a 500-year and 1000-year event is an extremely small amount of rainfall which all becomes runoff. 

    One of the conceptual problems with the curve number system is that it was developed for the implementation of irrigation on farm lands for a relatively small area.  Farm lands have good infiltration.  We as hydrologists liked the simplicity of the math, and so adopted it as a predictor for everything.  Curve numbers don't work well in urban environments where everything is sewered, and forested environments where things are not.  They also do not work well in areas of steeper slopes.

    It is relatively easy to break down a curve number model into a continuous model.  This is another reason why hydrologists chose it as a system for modeling every storm event.  I would always choose to model the pervious and impervious areas separately, rather than weighted averaging them to get a net curve number.  1 acre of impervious area for the mentioned storm equals 1.25 ac-ft of direct runoff.   

    I enjoy reading hydrology reports that report everything to 3 decimal places, but the authors have modeled everything with 1 parameter.  Most hydrologic models should state that the degree of accuracy is +/- 30%.  When you consider the accuracy, using a curve number methodology in a continuous storm model is not much different that using Clark's Unit Hydrograph and Green Ampt Losses. (how we model in Houston)  You need to tailor the method to the data that you have to define the model parameters.   If all you know is type of soil, type of vegetation, and approximate land slopes, curve number is no worst than any of the other methods for estimating runoff response.

    I would take a long look at the assumptions in Ia=0.2S.  This has been shown to be very area dependent.  It also has a bias towards undeveloped or agricultural watersheds.
     

    ------------------------------
    Dwayne Culp Ph.D., P.E., P.Eng, M.ASCE
    Culp Engineering, LLC
    Richmond TX
    ------------------------------



  • 9.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 10:26 AM
    Hello Doug, Sometimes quick and dirty method would have a place if one is out in woods, out of reach of modern equipment and facing information void. However, this doesn't seem to be the case, as Prof. M. indicated "I am involved with a local group trying to better understand the hydrology of the October 2015 flood that impacted South Carolina from roughly Columbia to the coast.   Different sources indicate the rain was on the order of a 1,000 event.  During computer simulation studies to recreate the flooding and streamflow, modelers have used CN values as low as 65 and place high confidence in runoff volumes and peaks simulated by the models they are using."
    Trying to understand a significant historic event in the order of 1,000 yr by means of "computer simulation studies" using CN values is totally unacceptable.

    ------------------------------
    Peter Ko Ph.D., P.E., F.ASCE
    Water Res. Consult.
    Brossard QC
    ------------------------------



  • 10.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 11:15 AM
    Great article, and thanks so much for emphasizing on the "Continuous-based Hydrologic Simulations" for today's engineering designs. 

    Mehdi

    ------------------------------
    Mehdi Rezaeianzadeh S.M.ASCE
    Graduate Research Assistant (GRA)
    Auburn AL
    ------------------------------



  • 11.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 12:28 PM
    I wonder about low flow events? I do water rights engineering in Colorado and this is the critical event. I have used the CN method on large basins and the antecedent soil moisture is usually assumed to be zero. But I found assigning the CN values to be daunting given the varying soil types and ground cover. Dividing the basin into subbasins helped but I am curious about the CN applicability and whether another model is more appropriate for low flows? Not a usual hydrologic investigation.
    Thank you

    ------------------------------
    Cheryl Signs P.E., F.ASCE
    Owner
    Cheryl Signs Engineering
    Denver CO
    ------------------------------



  • 12.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 01:13 PM
    Hello Douglas,

    I think we need to understand that actual physical processes of infiltration and runoff response are extremely complicated, and actual flood events come in different ways, so prediction is problematic.  Here in Pima County (Tucson Az)  have a major watercourse, that has had similar large flood peaks from rain on snow, tropical storm, and convective monsoon-type events.

    We use the CN here, because Pete Hawkins and I can take measured rainfall depth (P) and runoff depth (Q) from small rangeland watersheds, rank each and get asymptotic CN value with an r2 of 0.9+ with 40 years of data.  From what we know, we don't have the 'violent CN response' that Pete mentioned where CN increases with rainfall depth for higher rainfall depths.

    I spent a good chunk of my graduate school time trying to calculate saturated hydraulic conductivity on small rangeland watersheds using the Smith-Parlange (like G & A, but accounting for a distribution of Ks), and concluded that each event was differently hydrologically, and I was trying to 'measure with a micrometer and cut with an axe.'

    I agree with Dwayne here.  There is a tremendous amount of uncertainty in prediction, not the least of which is measuring rainfall itself (in this case over the whole state of South Carolina).  Then we have a soils map that provides crop yields, aggregate suitability and 1000 other characteristics, prepared by someone with little knowledge of hydrology, and the odds are stacked against us.

    So the CN method has a lot of warts (only way to know what it really it is to measure P & Q on a watershed), but I am not confident alternatives will improve our capability to predict large floods (which can have a variety of forcing precip events) and our methods (basically TR 55) produce peaks similar to USGS envelope curves.

    ------------------------------
    Evan Canfield, PE M. ASCE
    Pima County Flood Control District
    Tucson AZ
    P.E., M.ASCE
    ------------------------------



  • 13.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-05-2017 09:32 AM
    Modelling tools are great for simulating various conditions but are subject to misinterpretation and error.  They need calibration and adjustment to  parameters, similar to CN numbers, though often more extensive.  Some models even use CN values.  The Rational Method and CN hydrology are a good check for model results regardless since they bring it back to real world understanding. No student should be without this basic understanding.

    Slide rules went out of date for much different reasons.

    ------------------------------
    Harlan Kelly P.E., M.ASCE
    President
    KNE2 Ltd
    Vancouver BC
    ------------------------------



  • 14.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-06-2017 11:22 AM
    I disagree that CN and Rational Method calculations are a good check on model results.  They are not the standard by which we should be judging model results.  We may be comfortable using curve numbers and Q = CIA since they have been in use since pre-computer days, but that doesn't mean that they produce accurate results.  Rather we should be checking our model results on real world, observed or recorded flow data.  This is especially true in extreme events where we should be able to use high water marks as a check of what the model says happened.  This was our standard approach when doing FEMA flood insurance studies back in the '80s.  Comparing model simulated flow results with USGS or local streamflow gaging records is also a good check of model results.

    ------------------------------
    Douglas Beyerlein P.E., D.WRE, M.ASCE, HG
    Principal Engineer
    Clear Creek Solutions, Inc.
    Mill Creek WA
    ------------------------------



  • 15.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 03:16 PM
    I working on research now to estimate the surface runoff. I am using SWAT model which uses the CN method. My results are acceptable so far. I think CN method is helpful.

    ------------------------------
    Majid Mohamod S.M.ASCE
    Missouri University S&T
    Rolla MO
    ------------------------------



  • 16.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-04-2017 03:16 PM
    As other posters have noted for large events emperical methods don't work very well.
    In australia we use a variety of non empherical methods for large events - one which which has just happened here in northern Queensland this week also - cyclone Debbie

    I have listed a relevant link below

    A Guide to Using the Method: Revised Generalised Tropical Storm Method: Probable Maximum Precipitation: Design Rainfalls: Water Information: Bureau of Meteorology




    ------------------------------
    John Harris CPEng, A.M.ASCE
    John Harris
    Brisbane
    ------------------------------



  • 17.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-05-2017 09:32 AM
    Prof. Meadows - DOP, 

    To get back to the original post in this discussion, if you choose to stick to the CN approach, clearly there is a need to adjust CN as soil conditions change during the duration of this event. This will have a major impact on predicted runoff and, in turn, stream flow.

    I took a quick look at the Web Soil Survey for the Gills Creek area, and I saw several soils with hydrologic soil group A/D and C/D (Johnston loam, Pelion loamy sand, and others). If no other measures are taken to account for the effects of the 1000-yr event when using the CN approach, at least the hydrologic soil group should change from undrained to drained conditions sometime during the event which accumulated rainfall over the better part of 5 days. 

    I would recommend approximating the antecedent soil conditions as "drained" everywhere (perhaps with the exception of areas very near water bodies), but at some point during the event the soils surely behave as "undrained." The "switch" in hydrologic soil group probably means that the event must be modeled in two stages (in time) - the time before the soil becomes "undrained" and the time thereafter.

    Since spatial distribution of rainfall isn't being considered, to select the threshold of when the soils "switch" perhaps simply looking at the time required to reach some level of saturation of the soils would be appropriate? This can easily be done by evaluating the infiltration rate of each soil in your area of interest. This will surely not give you perfectly matching results with observations, but it will be an improvement; I do not think "perfect" should be in hydrologists' vocabulary. 

    Now with the updated hydrologic soil group, updated CN values will will be used for the second stage of modeling such an extreme event resulting in higher runoff rates being predicted. 


    ------------------------------
    Hassan Ismail EIT, A.M.ASCE
    Ph.D. Candidate, Water Resources
    University of South Caorlina
    Columbia SC
    ------------------------------



  • 18.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-06-2017 11:22 AM
    As I often say to students in a Water Resources Engineering course I teach at Missouri S&T, "..all models are wrong, but some are better than others and the only way to truly assess that is with real data."  Unfortunately, to really look at model parameterization, data from small watershed is the best to utilize but the small watershed streamgages are not as prevalent as we would all like (subject of another discussion).  For the Hurricane Matthew induced flooding in NC and SC, USGS has published a preliminary report on the data from our streamgages. See   https://pubs.usgs.gov/of/2016/1205/ofr20161205.pdf  In that report, I note there are three gages in SC that have drainage areas less than 52 square miles. However, there are a couple streamgages on less than 10 square miles in NC, so maybe if the soil/landuse conditions are not too much dissimilar to the area in SC, using a NC streamgage location may suffice to at least investigate your CN assumption issue.  If you have not already done so, you can connect with the USGS office in  SC to determine if they have ran any small watershed peak Q determinations on ungaged watersheds that I may not be aware of.  Good luck.

    Bob Holmes 

    ------------------------------
    Robert Holmes Ph.D., P.E., D.WRE, F.EWRI, F.ASCE
    National Flood Hazard Coordinator
    US Geological Survey

    ------------------------------



  • 19.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events

    Posted 04-07-2017 09:38 AM
    As a Civil I specilize in septic systems, this means that I have looked at a lot of soil. We also work on site development from small to large sites. One thing that most Engineers don't fully have an understanding on is the soils. They are site specific and change of short distances. This can mean that when trying to evaluate a large area, like a State, it can be difficult. The USGS soils survey is not accurate in some areas and perfect in others. 
    My point is that when trying to determine stormwater events and do any comparison, you really need to know your soil absorption rates as well. Averaging is not a good indicator. Sands and Loamy Sands can absorb a lot of water during a storm event while Clay will have almost none. This also applies to the runoff and erosion that is happening. Clay will hold along with rock while loam and sand will easily move. 

    Storm events can be modeled, but make sure that you take into account all the variables. In short you never can, only take your best shot. 

    ------------------------------
    George Miles P.E., M.ASCE
    President
    Alligator Engineering Inc
    New Smyrna Beach FL
    ------------------------------



  • 20.  RE: Please provide your opinion about CN values for VERY LARGE RAINFALL events