GC11D-1038: Stochastic simulation and decadal prediction of hydroclimate in the Western Himalayas
Authors: Andrew W Robertson1, Mickael D Chekroun2, Edward Cook1, Rosanne D'Arrigo1, Michael Ghil2, Arthur M Greene1, Tracy Holsclaw3, Dmitri A Kondrashov2, Upmanu Lall1, Mengqian Lu1, Padhraic Smyth3
Author Institutions: 1. Columbia University, Palisades, NY, USA; 2. UCLA, Los Angeles, CA, USA; 3. UC Irvine, Irvine, CA, USA
Improved estimates of climate over the next 10 to 50 years are needed for long-term planning in water resource and flood management. However, the task of effectively incorporating the results of climate change research into decision-making face a “double conflict of scales”: the temporal scales of climate model projections are too long, while their usable spatial scales (global to planetary) are much larger than those needed for actual decision making (at the regional to local level). This work is designed to help tackle this “double conflict” in the context of water management over monsoonal Asia, based on dendroclimatic multi-century reconstructions of drought indices and river flows. We identify low-frequency modes of variability with time scales from interannual to interdecadal based on these series, and then generate future scenarios based on (a) empirical model decadal predictions, and (b) stochastic simulations generated with autoregressive models that reproduce the power spectrum of the data. Finally, we consider how such scenarios could be used to develop reservoir optimization models. Results will be presented based on multi-century Upper Indus river discharge reconstructions that exhibit a strong periodicity near 27 years that is shown to yield some retrospective forecasting skill over the 1700-2000 period, at a 15-yr yield time. Stochastic simulations of annual PDSI drought index values over the Upper Indus basin are constructed using Empirical Model Reduction; their power spectra are shown to be quite realistic, with spectral peaks near 5–8 years.