INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 30: 774–789 (2010) Published online 5 May 2009 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/joc.1925 Synoptic-based evaluation of climatic response to vegetation change over southern Africa Neil MacKellar, a,b * Mark Tadross a and Bruce Hewitson a a Climate Systems Analysis Group, Department of Environmental and Geographical Science, University of Cape Town, Private Bag X3, Rondebosch 7701, South Africa b Danish Climate Centre, Danish Meteorological Institute, Lyngbyvej 100, 2100 Copenhagen Ø, Denmark ABSTRACT: The results of regional climate model (RCM) simulations of the effects of vegetation change in southern Africa are analysed to asses the role of synoptic forcing in land–atmosphere interactions. A self-organizing map (SOM) is used to identify the dominant large-scale features in the atmospheric boundary conditions used to force the RCM. The fields used to characterize the large-scale circulation are geopotential height at 850 and 500 hPa and total precipitable water between these two levels. For each of the patterns (nodes) identified in these variables by the SOM, the mean RCM-simulated response to vegetation change is evaluated. Notable differences are seen in the response of precipitation, near-surface temperature and geopotential heights to the land surface change between different nodes. Conditions characterized by strong sub-tropical anticyclones and low atmospheric moisture show the greatest temperature and geopotential height changes and are most sensitive to changes in radiative fluxes, whereas precipitation and surface hydrological processes are more sensitive under conditions of weak subsidence and high levels of atmospheric moisture. Copyright  2009 Royal Meteorological Society KEY WORDS land–atmosphere interaction; self-organizing maps; southern Africa Received 23 November 2007; Revised 12 November 2008; Accepted 14 March 2009 1. Introduction Changes in vegetation can have a marked impact on local and regional climate (e.g. Charney et al., 1977; Lean and Warrilow, 1989; Zeng et al., 1996; Bonan, 1997; Pielke, 2001; Narisma and Pitman, 2003; Marshall et al., 2004). Most modelling experiments that investigate the effects of vegetation change on climate are concerned with the mean response of the atmosphere to land sur- face perturbations over monthly or seasonal timescales. This is appropriate, as the seasonal cycle can modulate the effects of land surface change by altering the radia- tion, moisture and circulation regimes under which sur- face processes operate. For example, Narisma and Pitman (2003) show large differences in mid-summer (January) and mid-winter (July) responses to land cover change over Australia. January experiences much larger reduc- tions in latent heat (LE) flux than July, which leads to significant changes in temperature for the summer month. This is consistent with the suggestion of Koster et al. (2006) that coupling between hydrological land surface components and the atmosphere should be strongest in summer because evaporation rates are at a maximum. Observations by Rabin et al. (1990) reveal that influ- ences of the land surface on cumulus cloud formation * Correspondence to: Neil MacKellar, Danish Climate Centre, Dan- ish Meteorological Institute, Lyngbyvej 100, 2100 Copenhagen Ø, Denmark. E-mail: ncm@dmi.dk are favoured under conditions of maximum incoming radiation and weak atmospheric forcing. Numerical simu- lations by Wang et al. (1996) indicate that a stable atmo- sphere and strong synoptic winds suppress mesoscale circulations induced by a thermally heterogeneous land surface. Similarly, Chen and Avissar (1994) note that large-scale background wind can inhibit landscape-driven convective precipitation. In contrast, Weaver and Avissar (2001) demonstrate that landscape-induced circulations are not necessarily dependent on synoptic winds, and can in fact be strengthened depending on the orienta- tion of the large-scale flow. Findell and Eltahir (2003a) employed a one-dimensional model to investigate how early morning temperature and humidity profiles are crit- ical in determining the response of the boundary layer to altered land surface fluxes. Findell and Eltahir (2003b) extend this to three-dimensions to show that differ- ences in low-level wind strength and wind shear can suppress or enhance soil moisture–precipitation feed- backs and Findell and Eltahir (2003c) describe how the strength and direction of these feedbacks vary accord- ing to climatic region. It was also noted in the latter study that interannual variability is an important factor in determining potential land surface feedbacks. Studies of land–atmosphere coupling in general circulation mod- els also reveal much regional variability in the strength of this coupling (Koster et al., 2006; Guo et al., 2006). Transitional zones between wet and dry regions (e.g. the Sahel) are shown to be particularly prone to land surface Copyright  2009 Royal Meteorological Society