Climate Dynamics (2006) 26: 55–63 DOI 10.1007/s00382-005-0073-9

Wolfgang Knorr Æ Karl-Georg Schnitzler

Enhanced albedo feedback in North Africa from possible combined vegetation and soil-formation processes

Received: 26 May 2005 / Accepted: 25 August 2005 / Published online: 27 October 2005
Springer-Verlag 2005

Abstract It has long been recognized that albedo related vegetation feedbacks amplify climate variability in North Africa. Recent studies have revealed that areas of very high albedo associated with certain desert soil types contribute to the current dry climate of the region. We construct three scenarios of North African albedo, one based on satellite measurements, one where the highest albedo resembles that of soils in the desert transition zones, and one based on a vegetation map for the ‘‘green Sahara’’ state of the middle Holocene, ca. 6,000 years ago. Using a series of climate model simulations, we find that the additional amplitude of albedo change from the middle Holocene to the present caused by the very bright desert soils enhances the magnitude of the June-to-August precipitation change in the region of the present Sahara from 0.6 to 1.0 mm/day on average. We also find that albedo change has a larger effect on regional precipitation than changes in either the Earth’s orbit or sea surface temperatures between 6,000 years ago and today. Simulated precipitation agrees rather well with present observations and mid Holocene reconstructions. Our results suggest that there may exist an important climate feedback from soil formation processes that has so far not been recognized.

W. Knorr (&) Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queen’s Road, BS8 1RJ Bristol, UK E-mail: [email protected] Tel.: +44-117-3315133 Fax: +44-117-9253385 K.-G. Schnitzler Max-Planck-Institut fu¨r Meteorologie, Bundesstr. 53, 20146 Hamburg, Germany

1 Introduction The albedo of the Sahara and Arabian deserts displays a large degree of spatial heterogeneity (Pinty et al. 2000), with important consequences for the present dry climate of the region (Knorr et al. 2001). This present state contrasts sharply with the much wetter climate during the middle Holocene, around 6,000 years before present (Yu and Harrison 1996; Hoelzmann et al. 2000; Pachur and Hoelzmann 2000; Prentice et al. 2000). Reconstructions of the vegetation cover of what is now the Sahara desert show either steppe all the way to its northern boundary (Hoelzmann et al. 1998), or grassland and shrubland up to 23N or further (Jolly et al. 1998). Termination of this wet period appears to have been rather abrupt (deMenocal et al. 2000). It is now widely recognized that the transition was caused by slow changes in the Earth’s orbit and a subsequent southward retreat of the North African monsoon (Kutzbach and Street-Perrott 1985; de NobletDucoudre´ 2000). The abruptness of the change, however, strongly indicates the presence of positive feedback processes (Claussen and Gayler 1997; Brovkin et al. 1998; Claussen et al. 1999). The underlying effect that is assumed responsible for the climate changes follows largely from the theory of desert formation by Charney (1975), which states that high albedo associated with dry, non-vegetated conditions creates a regional minimum in net surface radiation, enhances sinking motion of air and thus further suppresses precipitation (Sud and Molod 1988; Lofgren 1995; Zheng and Eltahir 1998). In addition, changes in sea surface temperatures (SST) have been found to have a large (Kutzbach and Liu 1997; Hewitt and Mitchell 1998) or moderate effect (Ganopolski et al. 1998) on North African monsoon strength. The existence of large, very bright desert areas in North Africa and the Arabian peninsula and their impact on regional climate has been recognized more recently (Knorr et al. 2001). In the Western Sahara, exceptionally bright areas are observed (see Fig. 1)

Fig. 1 Broadband solar surface albedo used in the climate simulations. a Present albedo derived from Meteosat satellite observations (simulations starting with PRE). b Albedo for the middle Holocene based on reconstructed vegetation and land cover data (HOL). The albedo used for simulation BAR is the one under (a) with an upper limit of 0.35

around the area of Mreyye Erg and Aoukar Erg in Mauretania. The central Saharan area with high albedo is located northwest of Lake Tchad, known as ‘‘Bilma Erg’’ in Niger. It is noteworthy that some areas were found to have present albedo values of 0.5 and higher, while the albedo of the soil background in areas with sparse and seasonally changing vegetation cover does not typically exceed 0.35 (Pinty et al. 2000; Tsvetsinskaya et al. 2002). This suggests that the total albedo change from mid Holocene to present conditions was caused not only by the removal of vegetation and the exposure of underlying, generally brighter soils (cf. ‘‘vegetation induced albedo change’’ in Fig. 2), but also by some soil formation and degradation processes yet to be identified. A possible mechanism for this ‘‘enhanced albedo change’’ is suggested again in Fig. 2 (Hoelzmann personal. comm.): After all vegetation cover has been removed, soils in arid areas are exposed to four main known erosion processes, which change topography, texture and optical properties (including albedo), namely salinization combined with salt weathering (Doornkamp and Ibrahim 1990), siltation and surface crusting (Lal et al. 1989), formation of desert varnish and duricrusts by aeolian erosion (Mainguet 1999), and formation of fine sand fields and dunes by aeolian transport (Schultz 2000). Which process dominates depends on the underlying soil structure and the local climate. In case of the two bright areas observed in the Western and Central Sahara, the soil degradation processes were presumably driven by salinization and formation of fine sand planes, both resulting in bright soil textures (cf. Fig. 2).

Knorr and Schnitzler: Enhanced albedo feedback in North Africa

In order to quantify the climatic effect of this possible enhanced albedo feedback from soil formation, we postulate the existence of two types of albedo feedback mechanisms having operated at the land surface between the ‘‘green Sahara’’ state of the middle Holocene and the present climate: one that involves a change from the vegetated state to one where the albedo of the underlying soils was exposed, and one that includes formation of the very bright areas observed presently but not found in places where (sparse) vegetation grows. To describe the albedo during the middle Holocene, we use a map of reconstructed vegetation and assign typical values by vegetation type found in analogous present vegetation zones. Since we do not know the albedo of the underlying soils during that time, we simply assume that their albedo did not exceed 0.35 and use the presently observed albedo from satellite data with an upper limit of that value. This is also the value for exposed soils that was originally assumed in Charney’s theory of desert formation, and the maximum desert albedo assumed in the interactive simulations of Claussen (1997); Kubatzki and Claussen (1998). For the full present albedo state, we use the same satellite derived data without the cutoff. We call the first albedo change ‘‘vegetation induced’’, and the second ‘‘enhanced albedo change’’ (Fig. 2). More details are given in the next section. The purpose of the present study is twofold: first, to compare the climate effect of the enhanced albedo change with that of the postulated vegetation induced albedo change; and second, accepting the additional magnitude of the enhanced albedo change, to compare its climate effect with those of orbital and SST changes. Both is done with an atmospheric general circulation model (GCM) with appropriately chosen boundary conditions. It will be the purpose of later studies to further identify the mechanisms of soil formation that have been responsible for the formation of the very bright desert areas and thus to construct a complete

0.55 full desert exposure of bleached sands

Albedo

56

Enhanced albedo change

remobilization of sand dunes 0.35 barren land Vegetation induced albedo change lowering of water table 0.25

open shrubland, steppe

savanna

0.15

dry

wet

Fig. 2 Schematic diagram illustrating a possible mechanism for an enhanced albedo feedback. ‘‘Dry’’ and ‘‘wet’’ refer to the climate state, ‘‘albedo’’ to the resulting land surface albedo depending on vegetation and soils

Knorr and Schnitzler: Enhanced albedo feedback in North Africa

57

theory of the full albedo feedback that has operated in North Africa and on the Arabian Peninsula.

derived from the experiments ‘‘CTL2’’ and ‘‘6K’’ of Voss and Mikolajewicz (2001). A dominant feature of those simulations is a cooling of the South Atlantic by approximately 1C relative to the North Atlantic, and a cooling of ca. 0.5C of the Indian Ocean for conditions 6,000 years ago. Such rather small changes are difficult to assess against paleo proxy data of SST. However, Voss and Mikolajewicz (2001) could show that the coupled simulation ‘‘6K’’ enhanced precipitation over North Africa, leading to a better agreement with land based paleo proxies than for ‘‘CTL2’’. Orbital conditions were set according to the PMIP specifications (Joussaume and Taylor 1995). ECHAM-4 was run at T42 resolution (ca. 2.8 by 2.8) over 30 years, with the first 4 years disregarded to exclude the influence of initial conditions. Experiments PRE and BAR are the same as ‘‘MSA’’ and ‘‘CTL’’ of Knorr et al. (2001), except that simulations were longer. Simulations with orbital conditions of 6,000 years ago are marked by adding ‘+O’, and those using orbital and SST conditions of 6,000 years ago by ‘+OS’ (cf. Table 1).

2 Data and model simulations We define a series of seven model experiments (see Table 1) with the ECHAM-4 GCM (Roeckner et al. 1996). The three albedo states are denoted HOL (for mid Holocene), BAR (postulated ‘‘barren land’’ albedo, created by vegetation change alone) and PRE (present albedo). The first three experiments define a series of simulations to assess the strength of the two albedo feedback mechanism: HOL minus BAR (vegetation induced), and HOL minus PRE (enhanced). The next two simulations are designed to assess the climate effect of orbital and SST changes for present albedo conditions, and the last two for Holocene albedo conditions. The ‘‘Meteosat’’ albedo map (used with PRE) was derived from the 10-day albedo product for 1996 of Pinty et al. (2000), using the hemispheric albedo product for 30 solar zenith angle and applying a linear transformation from the Meteosat visible band to solar broadband spectral characteristics. The median of the 36 10-day values was then taken at each 2 km pixel, after which albedo was averaged to the T42 grid of ECHAM4. The use of the 30 albedo product instead of another product that takes account of the diurnal variation in solar zenith angle leads to a slightly lower, more conservative estimate. This map was inserted into the standard ECHAM-4 albedo data set (Roeckner et al. 1996). For the second albedo map (BAR) an upper cutoff value of 0.35 was applied. The third map used vegetation cover, lake and wetland fractions for the middle Holocene derived by (Hoelzmann et al. 1998). Based on modern equivalents found in the Meteosatderived data set, typical values are assigned by land cover type: 0.05 (lake), 0.15 (wetland and alluvial plains), 0.13 (forest), 0.2 (xerophytic woods, savanna), and 0.25 (steppe). The middle Holocene map has a 1 by 1 spatial resolution and extends 10N–31N and 17W– 60E. All three albedo maps were constructed at 1 by 1 spatial resolution (see Fig. 1) before they were converted to T42 resolution and inserted into ECHAM. Monthly mean SST for present day conditions are taken as the average of 1979 to 1994 from the AMIP project (Gates 1992). For the Holocene conditions of 6,000 years ago, SST anomalies were taken as the difference between two 100-year average climatologies Table 1 Description of the sensitivity experiments performed with ECHAM-4

3 The climate effect of the enhanced albedo change The first three GCM experiments differ solely in their albedo forcing, and the two difference fields of the vegetation induced albedo change (BAR minus HOL, cf. Table 1) and the enhanced albedo change (PRE minus HOL) are shown in Fig. (3a, b), respectively. Due to the averaging to the T42 resolution, the effect of the very bright areas is somewhat less pronounced than it would be suggested from Fig. 1. However, whereas most of the present Sahara shows an albedo change forcing of just above 0.1 for the vegetation induced case, large areas exceed 0.2 for the enhanced albedo change. For the Arabian region, the effect is less pronounced, but the area above 0.15 differential forcing is greatly increased. The average value of the Sahara albedo forcing is about 45% higher for the enhanced albedo change (with albedo values of 0.39 for PRE, 0.34 for BAR and 0.23 for HOL, as inferred from Table 2). For the Sahel, adjacent to the south, the albedo change is similar in magnitude, but the difference between the two cases is much smaller (Table 2, cf. Fig. 3a, b). Precipitation in today’s transition zone, the Sahel, is determined by the strength of the African Monsoon during the northern-hemisphere summer months. According to Charney (1975), low precipitation over the

Code

Description

Albedo

Orbit

SST

PRE BAR HOL PRE+O PRE+OS HOL+O HOL+OS

Present albedo Barren land albedo Holocene albedo Holocene orbit Holocene orbit & SST Holocene albedo & orbit Holocene albedo, orbit & SST

Meteosat Meteosat £ 0.35 Holocene Meteosat Meteosat Holocene Holocene

Present Present Present Holocene Holocene Holocene Holocene

Present Present Present Present Holocene Present Holocene

58

Knorr and Schnitzler: Enhanced albedo feedback in North Africa

Fig. 3 Difference of surface albedo and various climate output variables. Left column (a, c, e, g) simulations BAR minus PRE for vegetation induced albedo change. Right column (b, d, f, h) simulations HOL minus PRE for the enhanced albedo feedback.

(a, b) albedo, (c, d) net surface radiation in Wm–2, (e, f) June-toAugust precipitation in millimeters per day, (g, h) annual precipitation in millimeters per day

Table 2 Simulated surface radiation balance and simulated and observed precipitation for the months June–August Simulations

Albedo Incident SW Outgoing SW Incident LW Outgoing LW Net Radiation Precipitation Obs. Precip.

Sahel

Sahara

PRE

BAR-HOL

PRE-HOL

PRE

BAR-HOL

PRE-HOL

0.29 250.5 –74.7 411.8 –472.4 115.3 4.3 4.1

0.10 48.6 –31.9 –2.8 –10.4 3.5 –1.6

0.12 73.0 –43.2 –5.6 –18.9 5.4 –2.6

0.39 299.7 –116.4 392.8 –505.1 71.0 0.3 0.3

0.11 19.8 –35.6 –8.3 0.5 –23.6 –1.0

0.16 35.1 –56.6 –17.1 0.4 –38.1 –1.6

SW shortwave radiation, LW longwave radiation, both in Wm–2, precipitation in millimeters per day. Sahel is 11...17N, 10E...35W, Sahara 17...31N, 10E...35W

Knorr and Schnitzler: Enhanced albedo feedback in North Africa

59

Fig. 4 Simulated and observed annual precipitation in millimeters. a Fully modern simulation PRE. b Observed climatology over land. Contour lines are every 100 mm

present Sahara desert is induced by sinking motion created by a regional minimum of net radiation in the area (cf. above). Knorr et al. (2001) found that changing albedo from what is here BAR to PRE decreased the average June-to-August net radiation of the Sahara from 85 to 71 Wm 2, but had little effect for the Sahel (same definition as in Table 2), where net radiation is markedly higher with 115 Wm–2, despite of considerably lower incoming solar radiation (see Table 2). The same picture emerges for the two cases of albedo change considered here: in the Sahara, net radiation decreases by about 24 Wm 2 for the vegetation induced albedo change, and by as much as 38 Wm 2 for the enhanced albedo change. The second value comes close to the difference between the simulated net radiation between the present Sahara and Sahel (44 Wm 2, Table 2), which would be consistent with the notion of a ‘‘green Sahara’’ during the middle Holocene. The largest contribution to the decrease in net radiation comes from increased outgoing solar radiation, as would be expected, but over half of it is compensated by increased incoming solar radiation. As a results, a third to almost half of the net radiation change is caused by a decrease in the incoming long-wave radiation created by drier air and reduced cloud cover. Contrary to the Sahara, June–August net radiation changes very little, and that in the opposite direction, for the area that is defined as Sahel in Table 2. A marked increase in incoming solar radiation is compensated by increased outgoing solar and thermal radiation, as well as some decrease in incoming thermal radiation. As for the Sahara, precipitation decreases, which is consistent with the notion that cloudiness and air humidity decrease, too.

The regional north-south contrast is also evident in Fig. (3c, d): There is a general, large-scale decrease in net radiation over the entire (present) desert area, and the additional magnitude from the enhanced albedo change is distributed rather uniformly. To the south, there is an increase in net radiation that varies from the western Sahel and Sudanian (further south), where it is small, to the eastern part, and peaks over the Ethiopian Plateau. A comparison between Fig. 3b, d, f reveals a possible reason for the steep gradient in net radiation change: in the areas where albedo change is only moderate (ca. 10...12N in the central part), decrease in precipitation and associated increase in solar incoming radiation dominates and net radiation increases, whereas north of a line varying between 12 and 15N, the dominant effect comes from albedo change – net radiation decreases. The same is true for the vegetation induced albedo change (Fig. 3a, c, e), only that precipitation and net radiation change much less. The most important climate variable from both a biogeographic and human perspective is certainly annual average precipitation. It is, therefore, especially important that this particular quantity is simulated realistically. Figure 4 shows the simulated value using full present conditions (PRE) compared to the landbased observed climatology taken over a period of ca. 1930–1960 (Leemans and Cramer 1991, Cramer pers. comm.), averaged to the same spatial resolution as the GCM runs. There is a tendency of ECHAM-4 to generate slightly too much rainfall in the western portion of the Sahara desert, on the Ethiopian plateau and Yemen. In general, however, the simulations agree very well with observations, in particular the position of the desert transition zones (ca. 150 mm/year, see below). It is,

60 Fig. 5 Precipitation for the climate simulations with present (dashed lines, PRE) and middle Holocene albedo (solid lines, HOL). Red: present orbit and SST, green: middle Holocene orbit and present SST, blue: middle Holocene orbit and SST, black: observed climatology. a Isolines of 5 mm/day for June–August. b Isolines of 150 mm/year. (Colors as in Fig. 6.)

Knorr and Schnitzler: Enhanced albedo feedback in North Africa

a

b

therefore, appropriate to ask whether the simulated precipitation change resulting from the enhanced albedo change (Fig. 3g) is more realistic than the one resulting from the postulated vegetation-only effect (Fig. 3h). In the middle Holocene, a large lake existed between approximately 12 and 18N and 14 and 19E (Pachur and Rottinger 1997) (see the area of low albedo in Fig. 2b). At present, precipitation is less than 100 mm/ year in the north, and about 800 mm/year in the south of this area. We calculate an equilibrium evapotranspiration (McNaughton and Jarvis 1991; Knorr 1997) in this former lake area of 1980 mm/year. We note that changes in surface albedo alone can have precipitation change by 500–600 mm/year for the vegetation induced case (Fig. 3g), and by 700–800 mm/year for the enhanced albedo effect (Fig. 3h). Even without a detailed water balance calculation, it seems reasonable to assume that the additional about 200 mm/year of precipitation as a result of a larger albedo change from ‘‘dry’’ to ‘‘green’’ Sahara conditions would decrease the net water loss (evaporation minus precipitation) of such a lake considerably, making its existence more consistent with a precipitation change simulated as the result of the enhanced albedo feedback. The larger amplitude of the Holocene-to-modern albedo change implied by the satellite data seems, therefore, to generally improve simulations of the regional climate change in North Africa, at least with ECHAM-4. Note, however, the additional effect of orbital and SST changes as discussed in the following section.

4 Comparison between albedo, orbital and SST effects Accepting the enhanced albedo change as the more realistic one to describe part of the climate forcing that lead to the transition from the ‘‘green’’ to the present desert state in North Africa – acknowledging some remaining uncertainties about the albedo of 6,000 years

ago – we now compare its climate effect to the effect of what is commonly assumed to be the main external drivers of the regional climate change in the area. Figure 5a shows the observed 5 mm/day isoline for the months June to August in black (solid line). The current 5 mm/day isoline corresponds approximately to the Sahelian–Sudanian transition zone according to White (1983). The position of the observed line and the one of the fully modern simulation PRE (red, dashed line) agree rather well. In the following, we consider the effect of the various forcing changes in the inverse direction as in the previous section, going from present to mid Holocene conditions. Setting only orbital parameters to 6,000 years ago (green dashed line, PRE+O), or both orbit and SST (blue dashed line, PRE+OS) results in only a slight northward shift of the African monsoon. There is an eastward expansion of the monsoon as a result of SST, but not of orbital changes. If, however, only albedo is set to mid Holocene conditions (red solid line, HOL), the eastward expansion is as pronounced as for the combined orbital and SST change (PRE+OS). Once the albedo has been decreased to mid Holocene conditions, orbital changes have only very little effect on the simulated monsoon strength (compare the red and the green solid lines, HOL and HOL+O). Only SST changes (blue solid line, HOL+OS) lead to a further eastward shift of the monsoon into Yemen, but at the expense of the monsoon strength in most of North Africa. It appears that albedo changes dominate and strongly influence how either SST or orbital changes affect climate: in the case of modern albedo (dashed lines), the northward expansion of the monsoon seems to be effectively blocked, with orbital and SST changes having little effect. A comparison with Fig. 1a suggests that the northward expansion of precipitation is particularly suppressed in areas of exceptionally high albedo (around 5W and 15E), in agreement with Knorr et al. (2001), who found that the southern position of those bright

Knorr and Schnitzler: Enhanced albedo feedback in North Africa 1600 1400

Mid Holocene estimate s Modern observed PRE PRE+O PRE+OS HOL HOL+O HOL+OS

1200 mm

1000 800 600 400 200 0 5°N

10°N

15°N

20°N

25°N

61

(Hoelzmann et al. 2000). Agreement is generally much better for all simulations with assumed middle Holocene albedo. Orbital and SSTs conditions have only little impact on the results at the low albedo. Only the simulation with the middle Holocene albedo and modern orbital and SST conditions (HOL) shows slightly less precipitation at the northern boundary than HOL+O or HOL+OS. The dry-desert margin, assumed at about 20 mm/year (White 1983), is as far north as 28N in the other two cases. Agreement between the fully modern simulation (PRE) and modern observations is also good.

30°N

Fig. 6 Simulations, observations and paleo-estimates for annual precipitation along 28E

areas created an effective barrier for the African monsoon. Again following White (1983), we use 150 mm/year of precipitation as an approximate indicator for the position of the northern and southern Sahara transition zones (Fig. 5b). Agreement of the fully modern simulation (PRE, red dashed line) with observations (black line) is again rather good (cf. Fig. 4). There is a northward shift of the southern boundary of the dry zone (