Hydrologic Model Calibration/Uncertainty

State dam safety program responses to the following inquiry:

Montana Dam Safety has assembled a group of experts to evaluate our methodology for determining the inflow design flood for spillway analysis. This “Extreme Storm Working Group” is focusing on basin response parameters and uncertainty. They have asked us to conduct a survey of other states to better understand the national state of practice. We realize that there are other recent surveys dealing with inflow design flood determination. The purpose of this survey is to query states on how they deal with model calibration and uncertainty.

#1 Does your state have guidelines for determining basin response parameters (infiltration, unit hydrograph parameters, etc.)? A. If so, do your regulatory rules require your “guidelines” to be followed? A link to your guidelines would be appreciated.

#2 Montana is looking closely at how hydrologic models are pseudo-calibrated / verified. We rarely have sufficient gage data and need to improvise. We would be interested in hearing your thoughts. Specifically: A. Do you use regional (e.g. USGS Peak Flow) regression equations as a calibration tool? Or do you prefer that local regression equations be developed and utilized? B. What frequency storm do you require for calibration/verification (i.e. 100 year, 500 year etc.)? C. Other comments or advice?

#3 We are struggling with how best to pseudo-calibrate / verify models that have large drainage basins with diverse land cover. Do you have any guidance for model sub basin delineation? A. If yes, how do you calibrate/verify these models?

#4 With a risk based spillway standard, the uncertainty inherent in hydrology can be important. Our current line of thinking is that the modeler should strive for a best estimate, with an awareness of the uncertainty. This “awareness” comes from doing a sensitivity analysis. How do you deal with the large uncertainty associated with determining extreme storm flows in your state (for example, calibrating 1 standard deviation above a regression equation result)?

#5 What loss method do you generally see used (or require), and are there any loss methods that you will not accept?

#6 How do you deal with initial abstraction and antecedent moisture conditions? Note if you handle this differently during model calibration/verification. Any other comments, suggestions or advice?

The ASDSO surveys archive is available to members through the members only section of DamSafety.org