Influence of land surface processes on seasonal to decadal variability of dust and climate in the NOAA/GFDL CM3 model

Abstract:

Observational records show substantial interannual to inter-decadal variability in dust concentrations, which have been related to changes in source extent and intensity. GCMs, however, have long struggled to reproduce this low frequency variability. A possible cause for this lack of variability is that GCM dust emission parameterizations frequently are not embedded within dynamic land models with prognostic vegetation and land use changes. These changes vary more slowly in time than surface winds, creating memory in the model, and providing pathways through which dust emission can feedback on itself via dust induced changes in rainfall. Here we use a new dust emission model to explore this structural uncertainty, and demonstrate that inclusion of a prognostic land surface in the dust emission parameterization is important to simulating low frequency dust variability.

We use an updated version of GFDL’s Land Model 3 (LM3) running within Coupled Atmosphere-Ocean-Land-Sea ice Models version 3 (CM3) and compare its output to the CM3 runs in the CMIP5 archive. In the CMIP5 control run, dust emission has a prescribed spatial pattern and seasonal cycle that is modulated only by surface wind speed. In our updated version of LM3, dust emission depends not only on wind speed, but also on the soil water and ice, snow cover, leaf and stem area indices, and land use type, all of which, with the exception of land use, are prognostic. We run 500 years of 1860 conditions for the updated LM3 and compare to the CM3 1860 control run in the CMIP5 archive. By allowing soil water and vegetation to respond to climate and in turn affect dust emission, we increase the inter-decadal variability in dust that was lacking. The radiative forcing from the dust then increases the low frequency variability in regional precipitation and climate indices such as the AMO. Our findings suggest that accounting for land surface conditions is an important step toward accurately modeling both the observed climate variability of the recent past and simulating projected changes in the future.