Figure 5.1: In situ correlations between IDDI and midday 1000 hPa NCEP/NCAR reanalysis
temperatures for February. Values significant at the 5% level are shown in bold. See equivalent
fields for April and June, and the 700 hPa field for April.
The effects of airborne dust on the thermal properties of the atmosphere were investigated by examining the relationships between local daily IDDI values, representing dust concentrations, and NCEP/NCAR reanalysis temperature values. Daily values of these parameters were interpolated (reanalysis) or averaged (IDDI) onto spatially coincident 5 degree latitude x 5 degree longitude geographical grids. For each grid-square, the daily data for each month were pooled over the period 1984-1993, so that each month was represented by a series of some three hundred values (e.g. daily values from all the Januarys from 1984-1993, pooled into a single 300-value series for each grid square). For each grid-square, the 10-year daily IDDI series was correlated with the corresponding temperature series.
The correlations were performed for temperature data representing the 1000, 850, 700, 600 and 200 hPa pressure levels. While these data represent particular altitude ranges (centred at approximately 0-80 m, 1.5 km, 3 km, 4 km and 12 km respectively), the IDDI data represent the measured reduction in brightness temperature of the Earth-atmosphere system due to the presence of aerosols throughout the atmospheric column – the IDDI data therefore are not altitude-specific. The IDDI data represent the situation at midday, whereas reanalysis data are available for 00:00, 06:00, 12:00 and 18:00 hours. IDDI-temperature correlations were performed for all four times of day, although the interpretation concentrated on the midday fields. Correlations at other times are by definition lagged, although a consideration of transport times (McTainsh, 1980) suggests that dust is likely to remain in the same grid box between the times represented by the IDDI and reanalysis data over the lags involved. When IDDI data were correlated with reanalysis data representing 00:00 and 06:00 hours, the series were lagged such that the IDDI values preceded the temperature values by one day, resulting in lags of 12 and 18 hours respectively. The statistical significance of the correlations was assessed using a Monte Carlo randomisation procedure.
Data visualisation
Fields of IDDI-reanalysis correlations were plotted for each month,
for each reanalysis time. The fields were divided into the 5 degree x 5
degree grid cells used in the analysis, and the value of the correlation
for each grid-cell was displayed in that cell, with correlations significant
at the 5 per cent level marked in bold (Figure 1).
It is desirable to visualise the vertical variations in the relationships between IDDI and reanalysis data values, in order to infer how the presence of dust affects different levels of the atmosphere. The area containing the gridded values was divided into the Sahel (10-20 degrees N) and the Sahara (20-30 degrees N). For each of these regions, correlation values for a given reanalysis level and a given time of day were pooled over 3-month seasonal periods (JFM, AMJ, JAS and OND), resulting in 72 correlation values (13 columns x 2 rows - see Figure 1) over each region for any given season. Box plots of the correlations (defined by the median and upper and lower quartile and centile values) were then plotted against level for each season for the Sahel and for the Sahara (Figure 2).
Figure 5.1: In situ correlations between IDDI and midday 1000 hPa NCEP/NCAR reanalysis
temperatures for February. Values significant at the 5% level are shown in bold. See equivalent
fields for April and June, and the 700 hPa field for April.
Figure 2: Box-plots of IDDI-temperature correlations for the Sahel in Spring (AMJ) and summer
(JAS) at midday and 06:00 hours. Positive values indicate warming associated with the presence
of airborne dust, and negative values indicate dust-related cooling. Values outside of the dashed
vertical lines represent statistically significant relationships at the 5 % level. See equivalent plots
for the Sahel in autumn and winter, the Sahara in spring and summer, and the Sahara in autumn and winter.
Fields such as that reproduced in Figure 1 indicate that atmospheric dust is associated with widespread perturbations to atmospheric temperatures. This effect is most pronounced at 850 hPa, where extensive negative correlations between IDDI values and reanalysis temperatures suggest a widespread dust-induced cooling. This cooling is most evident over the Sahel, and often extends to 25 degrees north. At higher latitudes, correlations are often not significant or positive, indicating a different relationship between dust and temperature over Saharan regions. The 1000 hPa fields tend to resemble the 850 fields, although correlations become more positive in some fields at night-time, suggesting a near-surface warming which is absent at 850 hPa. Coherent patterns of significant correlations at 700 and 600 hPa are generally less extensive than at lower altitudes, and the distribution of positive and negative values is more complex. Significant positive correlations are more prevalent at these levels. At 200 hPa, coherent regions of significant values are very rare, suggesting little coupling between dust at temperature near the tropopause.
The negative correlations at 1000 and 850 hPa over the Sahel and southern Sahara indicate that the presence of atmospheric dust over these regions is generally associated with cooling in the first 1.5 km of the atmosphere. Cooling at these levels is likely to be caused by the presence of a dust layer above 1.5 km, reducing incoming solar radiation and causing surface cooling and reduced outgoing longwave radiation (OLR). This explanation is consistent with studies by other authors (e.g. Alpert et al., 1998), who describe long-range dust transport over northern Africa and the eastern tropical Atlantic as occuring above 1.5 km. Dust originating in the Sahara is transported in the Saharan Air Layer (SAL), which lies above the cooler Atlantic airmass. During spring and summer, the Atlantic airmass move northwards with the development of the West African Monsoon; elevated transport of Saharan dust over the Sahel will therefore be particularly pronounced in these periods.
Absorption of both incoming solar radiation and OLR by a dust layer will result in a localised greenhouse warming in the vicinity of the dust layer. Warming due to the absorption of upwelling longwave radiation will be be particularly pronounced if the dust consists of a large number of particles with dimensions of the same order as the wavelength of the radiation (around 10 microns). The positive correlations between dust and temperature at 700 and 600 hPa are therefore likely to be due to a local greenhouse warming of the dust layer. The distribution of the correlations may therefore yield information concerning altitudes of dust transport.
If a dust event consists of very fine particles (of equivalent diameter less than 1 micron), it may significantly reduce the amount of incoming solar radiation without directly affecting the OLR (Maley, 1982). Thus very fine dust hazes near the ground may cause cooling which is not offest by a local greenhouse warming. Such events may offer an explanation for some of the apparent dust-related cooling at 1000 and 850 hPa, although this is speculative.
The positive correlations over the Saharan regions at 1000 and 850 hPa are likely to be due to the presence of relatively large particles (capable of absorbing OLR) in the vicinity of and below an altitude of 1.5 km. Larger particle sizes and lower altitudes are consistent with airborne dust close to its source, before particles are disaggregated by sandblasting and transported to high altitudes.
Vertical correlation profiles - box-plots.
The box-plots of IDDI-temperature correlation distributions (Figure 2) indicate a widespread and significant daytime cooling around and below 1.5 km over the Sahel, associated with the presence of atmospheric dust. This effect is most pronounced in autumn and winter (when it is also apparent at night), but is also a major feature of the spring and summer months. In spring and summer (the monsoon onset and wet seasons respectively), a less pronounced, but still significant, general cooling is evident at 700 hPa. This suggests dust transport above about 3 km, consistent with the higher altitude of the SAL as is undercut by the monsoon airmass. A dominance of warming at 600 hPa suggests transport in the vicinity of 4 km. This pattern of cooling at and below 700 hPa and warming at 600 hPa is more pronounced in the July-September (the core wet-season months) period than in the April-June period. The same general pattern is apparent in January-March, whereas warming extends down to 700 hPa in October-December, indicating lower altitutes of dust transport.
The impact of dust on the vertical temperature structure of the atmosphere
over the Sahara is not significant, except in January-March, when a significant
dust-related cooling is apparent at 1000 and 850 hPa. It is speculated
that this is due to the presence of dust closer to the surface in the Sahara,
and/or less structure in the vertical distribution of dust over Saharan
regions.
The slackening of the vertical temperature gradient in the presence of atmospheric dust, and the consequent potential for dust to suppress rainfall, suggests that increased dust concentrations may have played some role in reinforcing drought conditions over the last few decades of the twentieth century. The existence of drought-reinforcing positive feedback processes is suggested by the high autocorrelation in interannual rainfall values in the latter half of the twentieth century when compared with the corresponding autocorrelations for 1901-1950 and also as generated in general circulation model (GCM) simulations (Hulme, 1998). Such high autocorrelation values are also absent from the series of mean northern hemisphere minus mean southern hemisphere sea surface temperatures (SSTs), representing the global SST dipole which has been been associated with Sahelian drought. (Dry conditions are more prevalent in the Sahel when the southern hemisphere oceans and the entire Indian Ocean are warmer than normal, and the northern hemisphere oceans are cooler than normal). Other such positive feedback mechanisms have been postulated, generally involving changes in land-atmosphere interaction resulting from the removal of vegetation over large areas of the Sahel (e.g. Charney et al., 1975). Such hypotheses have been supported by modelling studies of land-atmosphere interactions which have tended to apply large changes in albedo, soil moisture or vegetation cover over wide areas. However, there is no empirical evidence for such large-scale changes in land-surface properties on a regional scale in the Sahel (Nicholson and Tucker, 1998; see also Chapter 6), although the possibility that changes in the land surface may have occurred between the 1950s and the 1980s (spanning the period for which satellite data are scarce) in some areas should not be discounted. Whereas the role of land surface changes in the modulation of the Sahelian climate is highly speculative, increases in atmospheric dust loadings have been well documented, and the analyses described above demonstrate that dust aerosols have the capacity to modulate the Sahelian atmosphere in such a way that a resulting reinforcement of drought conditions is plausible. It should be noted that, while the relationship between reduced rainfall and atmospheric dust concentrations may be interpreted in terms of a positive feedback, the mechanisms driving this feedback are likely to operate within the atmosphere, and may not require changes in the land surface (Chapter 6) - the most important quantity in the model suggested in Chapter 6 is the rate of removal of dust from the atmosphere, not the rate of mobilisation of aerosol particles from the land surface.