Precipitation inputs for testing ground-water-resources management in the Santa Clara-Calleguas basin were synthesized using methods that preserve historical cycles and randomness. Historical water levels indicate that the ground-water system is particularly sensitive to interannual and interdecadal forcings. Southern California precipitation data are readily decomposed into a few apparently "reliable" cycles on these low-frequency time scales. Therefore, the synthetic precipitation inputs were designed to project these cycles smoothly into the future without discontinuities from the historical record. The synthesis combines singular-spectrum analysis (SSA) prediction methods to preserve the cyclic climate components, with inputs of white and residual noise to preserve realistic randomness. Autoregressive (AR) models were fitted to oscillatory SSA components with periods of about 22 years, 5.3 years, and 2.2-2.9 years, as found in the 1905-93 precipitation record. The AR models were projected 50 years beyond the historical record, with perturbations by white noise to prevent reversion to the historical mean. The AR projections were added and the result was further perturbed by residuals from the non-oscillatory parts of the historical record. The synthesized series are initially dominated by the AR influence of the recent precipitation record and vary relatively little from series to series. Interestingly, the measured precipitation totals corresponding to the first 3 years of the synthesized series (1994-96) also have shared these earliest variations, indicating the short-term reliability of the climatic cycles. After about 5 years, the synthetic series diverge into strikingly different projections of future precipitation.
A regional ground-water model of the Santa Clara-Calleguas basin was driven by one typical synthetic precipitation series by using a regression scheme that links precipitation to streamflow and infiltration of seasonal precipitation, which in turn become recharge into the ground-water flow model. The resulting ground-water simulation displays nonlinear responses to the variations in recharge and streamflow from the synthetic precipitation series. The transition between the trends and fluctuations in the historical simulations and the predicted values using the new precipitation-synthesis approach was smoother than was possible with either repetition of the historical climate record or long-term mean forcings. The nonlinear ground-water responses to wet and dry climatic cycles, in turn, affect the simulated outcomes of proposed additional diversions, recharge, and pump-back of streamflow and imported water as part of seawater-intrusion and land-subsidence abatement plans.