Representative Period Selection
Previous Topic  Next Topic 

While detailed collection system models can today practically be run for one or several years, it remains in most cases impractical to run such models for simulations of a decade or longer. However, robust statistics on combined sewer system overflow frequency, for example, can only be obtained by running models in continuous simulation for one or several years. Judicious selection of the representative period can best be achieved by running a planning level model such as NetSTORM to characterize the relevant parameter. While it is fairly simple to select a representative year based on precipitation statistics (e.g. a single selected year with precipitation matching the long-term mean), the representative period for a CSO model should properly be based on obtaining CSO volume and frequency statistics that are representative of the long-term mean behavior.

 

This difference can be readily understood through a simple example. Consider a city where average precipitation is 1200 mm (47 inches) with 40 storms exceeding 6 mm (0.24) inches. A representative year might be selected where the total precipitation and the number of storms are both close to the long-term means. However, this provides no guarantee that the selected year yields the long-term mean CSO. At one hypothetical extreme, each of the 40 storms could consist of a single hour of 30 mm (1.2 in) of rainfall. In a small collection system, this could result in very large CSO volumes. At the other extreme, each hypothetical storm could be 20 hours long, so that the rainfall intensity never exceeded 1.5 mm/hr (0.05 in/hr), likely yielding no CSO. While the analysis might be refined to consider the average precipitation intensity, or even the 10th percentile intensity, any precipitation-based method is at best a fair surrogate for an overflow simulation model. Even a runoff model alone is inadequate, as CSO is driven by duration of intensity exceeding a threshold, as well as by timing of high intensity rainfall within storms.

 

NetSTORMs STORM module may be used to rapidly characterize CSO on a citywide basis once basic parameters of the collection system are known. Table 1 shows STORM results for the City of Hartford Connecticut based on a 40-year simulation (model input parameters were not calibrated for this analysis). The model results demonstrate the pitfalls of selecting a single year to represent “typical” conditions based on precipitation data. The situation would be improved but not completely remedied were a representative period of three to five years selected. Were the representative year selected based on annual precipitation totals and number of storm events, either 1984 or 1986 would be selected, as 1974 has considerably more storms than the average year (52 vs. 43). However, when the STORM results are considered, 1984 and 1986 are shown to have annual CSO well below average. 1974 would be the best representative year based on CSO events and volume, as its CSO volume is within 3% of the average year, and includes close to the average number of CSO events. The table also shows that the years with close to average CSO consistently have above average precipitation. This is because annual CSO is not normally distributed, but tends to be positively skewed, as CSO increases non-linearly with rainfall due to the effects of runoff from pervious areas.

 

 

 

Precipitation

CSO

Year

mm

inches

Storms

Events

m× 10-3

million gallons

Years closest to mean precipitation

1984

1090

42.9

45

28

1094

289

1974

1102

43.4

52

31

2048

541

1986

1125

44.3

41

23

1298

343

Years closest to mean CSO (also 1974)                                

1979

1153

45.4

43

27

1995

527

1977

1369

53.9

46

37

2120

560

1959

1161

45.7

47

28

2290

605

Statistics

Median

1090

42.9

43

27

1904

503

Mean

1097

43.2

43

27

2124

561

80th Percentile

1288

50.7

48

29

2760

729

Note: A storm is defined as a precipitation event with at least 6.4 mm (0.25 inches) depth

 

Table 1. Use STORM for Selecting Representative Year