Rainfall disaggregation

Concerning the discussion about rainfall disaggregation, I would just like to point out that Prof. Van Nguyen at McGill University has been working on that for a while. I will try to locate the numerous recent papers he wrote and put them on the list on tuesday when I get back to the office.

He has been very successful at coming down to 1 hour from daily rainfall data (which is useful in many developing countries that have only daily rainfall data). He is still struggling however to go down from hourly rainfall data to sub-hourly. From some of the conversations we had last year on that, he suspects that the statistics of sub-hourly data are somewhat very different from hourly and that it would be indeed very difficult to go down with a good accuracy without a reference station. I think obviously that the best approach, as suggested by Dr. Huber, is to use available 15-min. data for a nearby station and synthesize the 15-min for the other location.

In Canada, I do not think we have 15-min. data available at many stations and for durations like 25 years as described by Dr. Huber. The City of Montreal does have an enormous amount of complete rainfall data that is gathering dust on the shelves. Prof. Van Nguyen will get a graduate student to look at that and with this data base (they have been operating many stations in the Montreal area for many years), I guess he would be able to come up with disaggregation algorithms and to test them.

I think the availabilty of a robust and adequate procedure to get 15- min. rainfall data from hourly data would without any doubt be of very high interest for anyone interested urban runoff management. I imagine it would be a significant step toward the paradigm shift of using continuous simulation on practical projects (as has been advocated for many years by Dr. James).

I am glad to read that Dr. Huber pointed out that hourly rainfall totals smooth out precipitation peaks and I would add that the use of hourly data does indeed result in lower predicted peak flows in urban areas. About two years ago, we tried to use continuous simulation with hourly rainfall data in a model for an urban area (about 200 ha), with RUNOFF and EXTRAN to simulate overflows controls. We also had about 50 rainfall events discretized to 5 min. interval and covering the same period as the continuous rainfall simulation. The main results was that the peak discharges were largely underestimated with the hourly data, so much so that the overflows were totally not realistic. I would therefore advise against using hourly rainfall for simulation on typical urban catchments as a casual practice and to be very cautious about the results thereby obtained. The only possible application of the hourly rainfall as they are readily available would be for a fairly large rural catchment.

By the way, about a year ago somebody in England said on this list that he had a methodology to do disaggregation; he has not completed his discussion however and I would be interested (if this person is still out there listening) to hear about his approach.

Concerning the trash racks, I think good references are the small report from ASCE on outlets from detention ponds (1985?) and the drainage manual for Denver (Wright and McLaughlin). Both documents have a good discussion on that topic.

Finally, some time ago there was a discussion to name important reference books. I just ordered a book called Hydraulic Structures by C. Smith, at the University of Saskatchewan. I had an old version of 1978 and this new version of 1995 is great. For non-canadians out there, I think this extremely practical book would make a very good reference for many concepts used in stormwater management. Prof. Smith, who recently retired, has produced over the years (first edition in 1971) an oustanding work. The book is available directly from the University of Saskatchewan bookstore.

This is a very useful topic for all who take urban hydrology seriously. Any methodology which aims to provide useful disaggregation from 1 hour rainfall data should be tested against a statistically significant set of 15 minute rainfall data. I suspect we should also be sensitive about applying a methodology across different geographic/climate zones.

Of course, if a nearby 15-minute station IS available, then the pattern can & should be superimposed onto the hourly station. Again, one such simple Fortran utility is offered at http://www.nvpdc.state.va.us/4MileRun/4mr-swmm.htm#util although this one happens to superimpose HOURLY rainfall onto a DAILY record station.

And again, for those interested in single-event synthetic hyetograph generation, a newly de-bugged utility based on the Keifer & Chu methodology is available at the same URL. The nice thing about this method is that the user can specify short rainfall intensity intervals, load them into SWMM-RUNOFF, and get very sharp, well-defined peak flows. The not-so-nice thing is that unless you have a very long & complete rainfall record, including boatloads of short-duration peaks (preferably as short as a 5-minute interval), your synthetic storm may be meaningless.

In fact, as long as a record is complete enough to have confidence in a synthetic single-event storm, why not just go ahead and run the model in multi-year continuous simulation and let the rainfall data speak for themselves? I'm glad folks like Dr. James are steering us away from single-event "design storms", especially now that computing power/time is less of an issue. (A standard run of the Four Mile Run "supermodel" under a 3-second EXTRAN timestep & a 4-hour design storm takes a half hour to run on my old 486DX2-50, versus a 2 minute runtime on my new P-233MHz PC.) Still, single-event simulations will remain a popular option for quickly determining relative differences in storm sewer design analysis, evaluating land use impacts, etc.

Don Waye
NVPDC

[minazivar]

Dear Don Waye,
http://www.nvpdc.state.va.us/4MileRun/4mr-swmm.htm is not accessible would you please refresh the link to be used again?