.
A general procedure to calibrate a hydrologic model would follow something along the following procedure:
1) Obtain all watershed, meteorological, and gauge data.
2) Develop conceptual model (links, nodes, etc - if link-node model framework)
3) Analyze results for meteorological period-of-record
4) Compare model predicted analysis results to gauge data
5) Adjust watershed parameters/variables to improve “fit” of model predicted analysis results to gauge data
6) Repeat steps 4 and 5 until acceptable “fit” obtained
7) Obtain additional meteorological and gauge data for sufficient period
8) Analyze results for additional meteorological period-of-record
9) Compare model predicted analysis results to gauge data for additional meteorological period-of-record
10) Verify the model produced reasonable results (acceptable “fit”) for additional meteorological period-of-record
If you need/want to calibrate hydraulics and water quality, you need to include appropriate measurement data for that (as well as employ a model which addresses these components).
As you can imagine, this is a fair amount of data needed for calibration. The watershed and meteorological data is available for most areas of the United States, but generally far less in other areas. The meteorological data available is usually representative of a given general nearby region. It is, however, of a specific site. As such, you can't apply it to another site as the driving variables in calibration. There is simply way too much temporal and spatial variability in this data. (I am sure you have seen how it will rain heavily in one location and not a drop will fall a couple miles away.) Site/watershed specific precipitation data is needed for that (temperature, barometric pressure, humidity, and other variables can generally be applied over a wider area if these are utilized in the model). So, that means you have to instrument the site/watershed and collect the data (while maintaining the equipment) for each site/watershed. This could add weeks, months, or years to the analysis. Some projects will allow for that and it is wise to do complete this work. For other projects, it is way overboard.
Now, also remember that event models require knowledge/information of antecedent conditions - basically the watershed condition at the time that a rain event starts. As the site/watershed varies constantly due to weather, the calibration of an event model is only good for those specific antecedent conditions. (An example of such knowledge/information is the antecedent moisture condition in the NRCS/SCS runoff method - I, II, III). Hence, event model calibration is a very limited type of calibration with a very limited application. I have seen NRCS/SCS Curve Numbers specified for a particular watershed. These were developed through “calibration”, but this is highly misleading as the actual “Curve Number” varies from storm event to storm event; even of the same rainfall depth as the intensity distribution during the storm varies from storm to storm and the site/watershed antecedent conditions differ from storm to storm. In other words, a single event model is tough to calibrate (if not impossible in a practical sense).
Beyond these issues, calibration is only possible for existing conditions, not future conditions. Proposed sites/watersheds (what you are designing) cannot be calibrated, regardless of model type.
Continuous simulation models remove many of the problems of event models and allow estimates of runoff statistics such as return periods. Direct estimates of runoff return periods are helpful as event models “force” the assumption that runoff return period is equivalent to runoff return period, but that is not the case. For example, a “100 year storm” does not necessarily produce a “100 year flood”. It would, in fact, be highly unusual if that was the case. You can imagine a “100 year storm” falling on very dry ground might produce a 50 year return period runoff event (or something like that) while a “100 year storm” falling on very wet ground might produce a 150 year return period runoff event (or something like that).
Given all that (and several items I haven't covered), I focus efforts on utilizing the best available model appropriate for the task at hand and applying the best available data to that model. Sometimes the analysis involves calibration, but not usually. I, again, suggest SWMM5.
As a side note, many stream gauging stations are calibrated for larger flows as well as smaller and mid-sized flow. This includes “overbank” flows which are generally exceeded on average at least once every 3 months to 5 years (depending on channel and watershed type).
.
tsgrue: site engineering, stormwater
management, landscape design, ecosystem
rehabilitation, mathematical simulation