Stochastic Simulation of Storm Occurrence, Depth, Duration, and Within-Storm Intensities

Newer watershed models require detailed continuous temporal precipitation to drive modeled hydrologic processes.The variability of precipitation inputs to models is a major source of variability in watershed flows that is often evaluatedin the context of risks. However, available short-term-increment rainfall data are not adequate. A storm-generator model(StormGen) that synthesizes storms directly (several storms per day to several days per storm) was developed and tested usingprecipitation data from Coshocton, Ohio. The concepts for modeling the four elements of actual storms (storm occurrence,storm duration and depth, and within-storm intensities) and storm-model characterization and parameterization using anempirical and statistical approach are presented. Supporting studies have been conducted to help with practical parameterizationin ungauged areas. Times between storms (TBS) are represented and simulated well by exponential distributions. Stormdurations are characterized by empirical distributions of durations in a month. Storm depths require conditional simulationgiven storm durations. Within-storm intensities utilize probabilistic information contained in Huff curves. Initial performanceevaluation of three of the elements of the model for a 200-year simulation shows that the model works well. TBS wasmodeled best (monthly average deviation of -1.3%, with values ranging from -3.2% to 3.0%). Monthly average storm durationdeviations ranged from -6.2% to +1.6% with an average of -1.3%. Monthly average storm depth deviations ranged from-17.1% to +0.1% with an average of -9.2%, although actual magnitudes ranged only from -2.2 to 0 mm. Average deviationsbetween simulated and measured average monthly precipitation was +1.7%, ranging from -8.6% to +12.6%. Correspondingdepth differences ranged from -10.0 to 10.8 mm, with an average of +0.5 mm. Total simulated precipitation for the entireperiod deviated from measured precipitation by +0.6%, corresponding to +0.5 mm. The Poisson assumption for statisticalindependence of storms in the model is validated using Coshocton data, with the ratio of average storm duration to averageTBS of only 0.107. The results of this study suggest that the StormGen model is promising and that modeling concepts andcharacterization deserve further investigation.