THE CONTINUUM OF DROUGHT IN WESTERN NORTH AMERICA

Abstract

The continuum of western North American hydroclimate during the last millennium is analyzed here using instrumental records, proxy data, and global climate model (GCM) simulations. We find that variance at long timescales (low frequencies) is generally more substantial than variance at short timescales (high frequencies). We find that local sources of autocorrelation (e.g., soil moisture storage) likely explain the tendency for variance to increase from monthly to interannual timescales, but that variance at longer timescales requires remote climate sources of variability. Our analysis of global climate model data indicates that at least one fully coupled GCM can reproduce the characteristics of the continuum on short (interannual) and long (multicentury) timescales, but that proxy spectra and GCM spectra disagree about the amount of variance present on intermediate (decadal to centennial) timescales. Since instrumental records, as well as multiple independent types of paleoclimate records, provide evidence that variance increases with timescale at these frequencies, and because numerical experiments indicate that local autocorrelation is not a likely source of variance at these timescales, we argue that climate model simulations underestimate the full range of low-frequency drought variability. Moreover, the models may also underestimate the risk of future megadroughts, which we attempt to quantify using a new method that combines frequency information from observational data with projections of 21st century hydroclimate. Our results indicate that the risk of a severe, decadal-scale drought during the coming century is at least 1-in-10 for most of the US Southwest, and may be as high as 1-in-3. These findings should be incorporated into adaptation and mitigation strategies to cope with regional climate variability and climate change.