Section 8: Validation of Results from the Chosen Method
Design flows estimated with any method used should be confirmed and validated. This may be achieved by:
- Anchor: #GKHMJHEJ
- Comparing the predicted
design flow for the selected frequency with observed flows to assess reasonableness.
The binomial distribution, which is available as a function in most
spreadsheet software, is helpful for this assessment. It computes
the probability of y exceedances in a given period of n years of
a design event that has an
AEP of p as:
- Anchor: #FKMIKIEJ
- P = probability (0 to 1) of y exceedances in n years for design level p Anchor: #EIMKJJFF
- y = number of design exceedances Anchor: #FMJMKMNK
- n = number of years Anchor: #HFKFHFEI
- p = Design AEP (0 to 1)
Note that p in this equation is AEP and ranges from 0 to 1. So for this equation, p is the selected design frequency divided by 100.
Suppose, for example, that in a 20-year long record of observed flows, the computed 1% AEP flow was found to have been exceeded in three years. With the binomial distribution, the probability of this is computed as 0.001 or about one chance in 1000. This is so unlikely that it raises doubt about the estimate of the 1% chance flow, suggesting that the computed design flow is too low. Fewer exceedances would be reasonable.
Similarly, suppose that the 10% AEP design flow was not exceeded at all in a 30-year-long record. The binomial distribution shows that the probability of no exceedances of the 10% AEP (10-year) flow in 30 years is 0.04—again an unlikely scenario. This suggests that the 10% flow predicted is too high; more exceedances would be reasonable.
- Comparing the design flow computed with the selected method with those computed for the same AEP for watersheds with similar properties in other studies in the region. A “flow per unit area” comparison is useful. Significant differences should be investigated and explained. Anchor: #KMNMEMEL
- Comparing the results of other methods, if those are appropriate. For example, in some cases, both the rational method and the regression equations will be acceptable for design flow computation. Or if a frequency function is fitted with statistical methods, the design flows can be transposed from the gauged site to the location of interest, using methods described later in this manual. Anchor: #OOLIGKFE
- Comparing the design flow computed with the selected method to design flow for the same frequency computed by other agencies with different methods. These may include local public works agencies, the United States Army Corps of Engineers (USACE), the FEMA, and the Natural Resource Conservation Service (NRCS). Significant differences should be investigated and explained.
The results of these alternative methods can be compared. Again, significant differences should be investigated and explained.