Report No FR0318



Sept 1992



The testing programme reported here, and earlier testing, show that the Stochastic Rainfall Generator (SRG) model can provide long series of hourly rainfall data representative of any site in the UK. These data can be used to drive sewer flow quality, quantity and river impact models for investigation and planning purposes.


To assess the characteristics of SRG rainfall and compare them to equivalent historical series.

To report on the accuracy of spill/storage predictions when SRG rainfall is used in simplified sewer flow models.

To test the accuracy of a disaggregation routine that converts hourly rainfall data to 5 minutely data which are suitable for models such as WALLRUS.


Rainfall time series provide the user with a means of assessing the hydraulic performance and pollution impact of sewer systems. They allow the frequency, rates, volumes and durations of discharges to be quantified.

The methods currently available for creating suitable rainfall input to models are inadequate. The SRG model is the result of a research programme designed to develop a model for the synthesis of rainfall time series (to a resolution of 5 minutes) for urban areas in the UK. The efficacy of this product needs to be tested in practical terms.


  1. The annual mean rainfall produced by the regional SRG was accurate to within 5% of the historical rainfall.
  2. Using a minimum interevent dry period (MIEDP) of 1 hour, events generated with the SRG tend to be shorter and smaller than those in the historical series. This effect can, in part, be compensated by defining events with a 4- hour MIEDP.
  3. When rainfall events (defined using a 1 hour MIEDP) were run through sewer models, the SRG series tended to indicate smaller storage requirements (for specified spill frequency standards) than the historical series.
  4. No east or west geographical bias was found in any of the basic rainfall statistics comparisons or in the storage volume analysis.
  5. The overall performance of the regional SRG was found to be good. $he variations observed from historic records are those which are associated with any regionalisation procedure.
  6. The disaggregation model showed good results. Differences in spill volumes for individual events were generally within 5%.
  7. The SRG and Disaggregator models are not currently available as supportable software packages. Hence the scope for practical application by industry users is limited.


The recommended interim procedures for using the SRG are:

  1. If a very long (15 years or more) local historical hourly record is available, this should be used in preference to the regional SRG.
  2. If over 20 years of daily rainfall data are available, the daily regionalised SRG could be used in preference to the regional SRG. This version of the SRG should provide a more accurate regionalisation method, however it has not been tested in this r esearch programme.
  3. When using the SRG at least 20 years of data should be generated to avoid any bias.
  4. When defining rainfall events a 1 hour MIEDP should be used.
  5. When using the regional SRG to estimate storage volumes to meet a given spill frequency standard, 5% should be added to the volume estimate. This pragmatic approach should compensate for the bias detected in the testing programme. The resultant storage volume can then be anticipated to be between 10% of that which would have been specified with the use of historic rainfall data.

The disaggregation model provides an accurate method of producing 5 minutely values from either SRG or historical hourly rainfall data.

The SRG and disaggregation models should be packaged in a form which is applicable to users in the water industry.

One important aspect which has not been fully addressed is the most suitable technique for sampling from the rainfall record. Further guidance on this, and on how to use the SRG and disaggregation models, will have to be completed before the practising engineer can adopt the technique.

The SRG model is amenable to further development should practical application experience suggest it to be necessary. Such work should focus on the simulation capabilities in respect of rainfall events of return periods in excess of two years. The accuracy of the simulation parameters could be improved, to reduce the existing error band, with further development. In addition, an effort should be made to improve knowledge of the spatial variation in rainfall to the point where it can be incorporated into a practical design methodology.


This report describes the testing undertaken on the SRG and disaggregation models.

Copies of the report are available from FWR, price 25.00, less 20% to FWR Members.