A COMPARISON OF INDEX-SEQUENTIAL AND AR(1) GENERATED HYDROLOGIC SEQUENCES

Many water agencies in California and the rest of the United States utilize sequences of wrapped historic hydrology rather than synthetic streamflows in their simulation models. The wrapped historic procedure is also known as the index-sequential method. This paper compares results from an annual regulation model of the Colorado River system which used both historic hydrologic sequences generated by the index-sequential method, and autoregressive, order 1 (AR(1)) synthetically generated streamflows as inputs. The synthetic AR(1) log normal model also included parameter uncertainty. Reservoir storage exceedance probabilities developed for Lake Powell and Lake Mead from a 32-year simulation show that AR(1) model generated flows yield slightly higher storage levels at 50% exceedance. This is in part the result of the parameter estimation procedure of the AR(1) model, and also to the fact that AR(1) models do not reproduce critical period severity with the same frequency as that evidenced in the index sequential method. Thus, the index-sequential method appears to be a conservative approach for average storage predictions. However, for predictions involving higher exceedance probabilities, (i.e. the tails of the distribution) synthetic streamflows may be more appropriate.