Evaluation of mesoscale simulations of dust sources, sinks and transport over the Middle East

Abstract

Desert dust is ubiquitous in the Earth's atmosphere, with measurable concentrations found everywhere from the Middle East to Antarctica. It has the ability to alter the global radiation balance and affects the African Easterly Jet, along with hypothesized roles in hurricane formation and past climate change. In addition to these large-scale effects, dust can have a deleterious effect on regional hydrology, air quality and visibility. Recent military activity in the Middle East has led to increased interest in looking beyond global models that are chiefly concerned with long distance transport, to mesoscale models capable of resolving smaller scales and incorporating more elaborate radiation and microphysics schemes. Following the work of Stokowski (2005), who introduced a modification that allowed dust to be radiatively active, a dust emission and deposition module has been added to the CSU Regional Atmospheric Modeling System [RAMS]. The new scheme utilizes a global map of potential sources and parameterizes emissions based on model wind speed and soil moisture. Though many existing models incorporate dust as an offline tracer, RAMS allows it to advect and feed back into the model, affecting the long and shortwave radiative balance at each timestep. Removal is done via Brownian diffusion, gravitational deposition and rain scavenging. This study focuses on the testing of the new scheme, with emphasis on its ability to accurately model the spatial extent of dust as well as the column optical thickness and surface visibility. Validation is done through a combination of NCEP reanalyzed meteorological maps, Aeronet AOT and back-trajectory data and surface observations. RAMS did well in predicting the areal extent of elevated dust as well as visibility within the higher resolution nested grid. Prediction of visibility and optical thickness values in the coarse grid exhibited larger absolute errors, though the latter was generally predicted within a factor of two. A persistent dry bias was found through comparison of sounding data with model output. This is thought to be due to the initialization data, and resulted in sometimes-poor simulation of precipitation areal extent, explaining several of the erroneous forecast values. The work done in this study paves the way for future improvements to the RAMS model, the most important being the addition of dust effects on model microphysics. The structure of the dust scheme also allows for parameters such as in-situ size distribution to be tuned to known regional values, thus presumably improving forecast ability. Several such data sets are already in existence although they were not available for this study.