Tree-ring based weather regime reconstructions over the past millennium for climate scenario development in the Western United States

Abstract:

Extreme hydrologic variability has long posed difficulties for water systems planning and management in the Western U.S. and is only projected to become more variable in a changing climate. Recently, planners have focused on developing robust systems that perform well under a range of plausible future climate scenarios derived from a variety of methods, including global circulation model (GCM) projections and stochastic weather generators. Weather-regime based stochastic weather generation has emerged as a method to systematically explore water system response to specific mechanistic sources of change in the climate system. Regime-based generators can create scenarios useful for linking system performance to local weather and associated patterns of large-scale atmospheric flow (i.e., weather regimes) that are credibly projected by GCMs. However, the identification and simulation of weather regimes is often conditioned on short instrumental records, which begs the question: are the scenarios that these methods produce the best representation of natural climate variability that could be experienced in the future?

We propose a new method to incorporate natural variability into future climate scenarios by reconstructing weather regimes based on tree-ring based moisture proxies across the Western U.S. We introduce a novel, multi-objective framework to identify latent low and high frequency signals that best explain multivariate weather regime dynamics across the Pacific/North American sector and that are maximally correlated with the tree-ring paleodata. We then develop a method to reconstruct these latent signals back to 1400 CE and couple them with a Non-Homogenous Hidden Markov Model (NHMM) to simulate new time series of weather regimes over the past millennium. The reconstruction identifies a prominent 15-year cycle of weather regimes consistent with previously observed multi-year droughts and pluvials over central California as well as tree-ring reconstructed megadrought behavior at the end of the 15^th and 16^th centuries. Ultimately, we demonstrate that the NHMM model, when coupled with a weather-regime based weather generator, can simulate new daily scenarios of temperature and precipitation to inform bottom-up vulnerability assessments for water systems in the Western U.S.