1National Center for Atmospheric Research, Boulder Colorado, USA
2Institute for Computational Earth System Science, University of California, Santa Barbara, CA, USA
3Department of Geography, University Kansas, Lawrence, Kansas, USA
4Department of Geography, McGill University, Montreal, Quebec, Canada
*now at: United States Geological Survey, Denver, Colorado, USA
Abstract. There is a large uncertainty in the relative roles of human land use, climate change and carbon dioxide fertilization in changing desert dust source strength over the past 100 years, and the overall sign of human impacts on dust is not known. We used visibility data from meteorological stations in dusty regions to assess the anthropogenic impact on long term trends in desert dust emissions. Visibility data are available at thousands of stations globally from 1900 to the present, but we focused on 359 stations with more than 30 years of data in regions where mineral aerosols play a dominant role in visibility observations. We evaluated the 1974 to 2003 time period because most of these stations have reliable records only during this time. We first evaluated the visibility data against AERONET aerosol optical depth data, and found that only in dusty regions are the two moderately correlated. Correlation coefficients between visibility derived variables and AERONET optical depths indicate a moderate correlation (~0.47), consistent with capturing about 20% of the variability in optical depths. Two visibility derived variables appear to compare the best with AERONET observations: the fraction of observations with visibility less than 5 km (VIS5) and the surface extinction (EXT). Regional trends show that in many dusty places, VIS5 and EXT are statistically significantly correlated with the palmer drought severity index (based on precipitation and temperature) or surface wind speeds, consistent with dust temporal variability being largely driven by meteorology. This is especially true for North African and Chinese dust sources, but less true in the Middle East, Australia or South America, where there are not consistent patterns in the correlations. Climate indices such as El Nino or the North Atlantic Oscillation are not correlated with visibility derived variables in this analysis. There are few stations where visibility measures are correlated with cultivation or grazing estimates on a temporal basis, although this may be a function of the very coarse temporal resolution of the land use datasets. On the other hand, spatial analysis of the visibility data suggests that natural topographic lows are not correlated with visibility, but land use is correlated at a moderate level. This analysis is consistent with land use being important in some regions, but meteorology driving interannual variability during 1974–2003.