NCAS - Climate, University of Reading, UK

Atmosphere-land surface interactions and their influence on extreme rainfall and potential abrupt climate change over southern Africa

C. J. R. Williams*

NCAS - Climate, University of Reading, UK

D. R. Kniveton

School of Global Studies, University of Sussex, UK

* Corresponding author address:

NCAS - Climate

Room 2U18, Walker Institute
Department of Meteorology
University of Reading, Earley Gate

Reading

RG6 6BB
United Kingdom

Tel: (+44) (0) 118 378 5586

Fax: (+44) (0) 118 931 8316

Email:

Online Supplement

1. ASSOCIATED LARGE-SCALE CIRCULATION FEATURES

This online supplement presents further results from the RCM and GCM experiments, focusing on the differences between the vegetation experiments and the control for various other associated fields. The purpose of this was to determine whether the observed rainfall response to changing vegetation cover is associated with larger-scale atmosphericcirculation features.

1.1. RCM

Figure S1 shows the mean near-surface (850mb) geopotential height differences during DJF where, compared to the control run, a widespread increase in pressure can be seen over all of southern Africa in the desert run (Fig. S1a). In contrast, decreased pressure relative to the control is shown in the savanna run, and this is particularly evident over southwestern Africa corresponding to the semi-permanent (summertime) Angolan low. This region of low pressure is particularly important for southern African summer rainfall, as its position influences the location of the tropical temperate troughs (TTT) whichthe main rain producing systems over the region; for example, when the low is situated over the Namibian–Angolan border, the cloud-band (and therefore rainfall) occurs more over central southern Africa, as opposed to further east. These changes in near-surface pressure are also highlighted by the mean 850mb vector wind differences, with increased anticyclonic flow over much of southern Africa in the desert run compared to an increase in cyclonic winds in the savanna run (Figs. S2a and S2b, respectively). All of these near-surface pressure and wind changes are consistent with the decreases (increases) in mean rainfall shown by the desert (savanna) run. The differences in near-surface moisture flux, representing the combination of vector winds and humidity, are very similar to differences in winds (not shown). However, when the moisture flux differences are integrated over all vertical levels, a different pattern emerges (Fig. S3). In the desert run, a clear increase in northeasterly flow is shown relative to the control, whereas an increase in southwesterly flow is shown in the savanna run (Fig. S3a and S3b, respectively). This is slightly contradictory to what might be expected, where increased rainfall over southern Africa is normally associated with an increase in northerly and northeasterly flow from the moister Congo basin and western Indian Ocean, and so might suggest that in these experiments changes in vertical motion and/or changes in moisture flux convergence might be more important than changes in the direction of near surface and upper level moisture fluxes.

Altering the vegetation cover over southern Africa also induces a temperature change. Figure S4 shows the mean surface temperature differences during DJF where, relative to the control run, a large increase in temperature is shown across the entire subcontinent in the desert run (Fig. S4a). This increase clearly corresponds to the vegetation anomaly used to force the model, with the boundary of the temperature increase (at the Equator) corresponding exactly to the boundary of the region of desertification. In the savanna run, Fig. S4b, increases in temperature are also shown throughout southern Africa, although these are mostly smaller in magnitude than the desert run. The largest increases are shown throughout southwestern Africa, such as over the Angola/Namibia border, although further south over southern Namibia and South Africa small decreases in temperature are shown (Fig. S4b). Temperature changes were also analysed at various vertical levels, such as 1.5m and 850mb, but these produced almost identical temperature differences to those at the surface (not shown).

In addition to surface temperature differences, there are also associated changes in surface evaporation and vertical uplift. Figure S5 shows the differences in mean near-surface evaporation rate during DJF. Relative to the control, in the desert run a widespread reduction in evaporation is shown across all of southern Africa, and this is particularly evident over extratropical southern Africa where evaporation is reduced by over 2 mm day-1 (Fig. S5a). Conversely, increases in evaporation are shown across southern Africa in the savanna run, particularly over southwestern Africa and equatorial regions (corresponding to the areas with the largest vegetation change relative to the control), although these are smaller in magnitude than the decreases shown in the desert run (Fig. S5b). In terms of vertical motion, the differences in mean daily omega (averaged longitudinally) during DJF are shown in Figure S6, where in the desert run decreased ascent (compared to the control) can be seen at approximately 500mb over most of central southern Africa and especially around 15°S (Fig. S6a). In the savanna run, small increases in ascent are shown over tropical regions, both at the surface and higher up in the atmosphere (Fig. S6b). The larger changes in omega at the boundaries of Figure S6can be disregarded, as these relate to boundary artefacts of the RCM and are not shown in the GCM experiments.

1.2. GCM

Figure S7 shows the mean near-surface (850mb) geopotential height differences during DJF where, similar to the RCM results, an increase in pressure can be seen across southern Africa, relative to the control run. A widespread strong decrease in pressure is shown further south. Mean 850mb vector wind differences correspond to the observed pressure increases over southern Africa, with increased anticyclonic flow particularly over tropical and extratropical regions (Fig. S8). Near-surface moisture flux differences between the desert and the control run are again very similar to the near-surface wind differences (not shown), however when vertically integrated the moisture flux shows a clear difference, with increased easterlies in the desert run over most of tropical southern Africa (Fig. S9).

A clear change in surface temperature is again shown in the GCM desert run relative to the control, with a large increase (more than 3oC in some areas) over the majority of southern Africa (Fig. S10). The boundary of the vegetation anomaly used to force the model is again clearly shown in the temperature changes. Higher up in the atmosphere, changes in temperature again reflect those at the surface (not shown). A widespread decrease in evaporation rateduring DJF, relative to the control, is shown in Figure S11, and this is particularly strong over southeastern Africa. Associated with these changes in surface temperature and evaporation rate is again a change in vertical motion, with the desert run showing decreased ascent (relative to the control) over all of Africa south of the Equator and throughout the atmosphere (Fig. S12).

1