Identifying a Representative Year of Precipitation in Support of a Wet Weather Plan
In support of a regional wet weather plan, computer models have been developed to simulate baseline conditions to establish the frequency and duration of combined sewer overflow (CSO) and sanitary sewer overflow (SSO) discharges, the potential water quality impacts they produce, and to develop and assess alternative control strategies. Much of the uncertainty in a carefully constructed hydrologic and hydraulic model is derived from uncertainty in the rainfall record. Therefore, in-creasing the level of detail of the rainfall input, both spatially and temporally, increases the accuracy and precision of the model results. Careful attention to rainfall collection and analysis is critical to the modeling effort. The refinement of precipitation data becomes an im-portant process because precipitation is the driving force that increases wastewater flow along sewers and transports pollutants via CSO and SSO discharges to receiving waters.
The U.S. Environmental Protection Agency CSO Control Policy (1994) requires the characterization of the combined sewer system area and evaluation of control measure performance using the com-plete rainfall record for the geographic area of its existing combined sewer systems (CSS) using sound statistical procedures. However, for most US cities historical precipitation data is available for periods any-where from fifty to hundred years. It is not possible to run complex and large models for all the years for which precipitation data is available and hence it is imperative that a short period of few years or a typical year, which is truly representative of long-term hydrological condi-tions, be selected from the larger precipitation dataset.
The selected representative year, based upon quality assured long-term local precipitation data, should be able to produce annual CSO sta-tistics such as volume, duration and event frequency that closely match the long-term historical average. Ideally, representative periods would be selected by running the model for a long-term simulation based upon long-term high-quality local precipitation data and selecting periods that produce annual CSO volume, duration and number of CSO events closest to the mean. The CSO events should also have a representative seasonal distribution in order to assure the validity of applying the results to receiving water studies. Since it would be time-consuming to run the complex model for many years, surrogate techniques can be used in the selection process.
CSO occurrence is considered to be a complex function of storm-event characteristics such as total volume, duration, peak intensity and length of antecedent dry period or inter-event time (IET). Continuous 12 month periods selected from the recent quality assured radar rainfall data were evaluated against the long-term record based on storm-event characteristics such as annual event frequency, total annual rainfall volume and event peak hourly rainfall intensity. Statistical analyses were conducted to determine adjustments to the actual 12 month rain-fall that were necessary to eliminate bias against historical record average values.
This chapter presents a statistical analysis approach including dou-ble-mass regression and cumulative distribution frequency (CDF) analysis of long-term regional rain gauge data and high resolution short-term radar rainfall data to establish a representative one-year precipitation record.
This paper is only available in PDF Format:
View full text PDF