Climate impact of deforestation over South Sudan in a regional climate model

Authors


A. A. M. Salih, Department of Meteorology, Bert Bolin Centre for Climate Research, Stockholm University, Stockholm, Sweden. E-mail: abubakr@misu.su.se

Abstract

This study examines the sensitivity of climate to changes in vegetation cover and land use in South Sudan. The focus lies on the effect of deforestation on precipitation and surface temperature especially during the rainy season. Sensitivity experiments are performed with the third version of the Abdus Salam International Centre for Theoretical Physics Regional Climate Model (RegCM3) where the present forest and vegetation cover south of 10°N in Sudan are replaced by either grass or, as an extreme case, desert. The model experiments were conducted for a time period of almost 21 years, from January 1989 to August 2009, and were preceded by a control experiment to ascertain the fidelity of the model simulations. The experiments indicate that the vegetation changes affect precipitation and surface temperature in both Southern and Central Sudan significantly although the land cover changes were imposed only in the south. The precipitation during the rainy season (June through September) was reduced in the perturbed region by about 0.1–2.1 mm d−1 for the desert scenario and by 0.1–0.9 mm d−1 for the grass scenario. The surface temperature increases by about 1.2 and 2.4 °C in the grass and desert scenario, respectively. The precipitation reduction is thus not only local but also extends to Central Sudan and neighbouring regions. The study demonstrates significant dependency for Southern and Central Sudan precipitation on the land use in Southern Sudan and indicates that the deforestation has both local and non-local regional climatic effects.

1. Introduction

The Republic of Sudan and the Republic of South Sudan together form the largest political territories in Africa. Until 9 July 2011 they used to be one country known as the Republic of Sudan, extending from Sahara in the north to the wet equatorial rain forests in the extreme south and hereafter we refer to it by Sudan. The main water resources are the river Nile, its tributaries, groundwater and the rainfall. The main economical activity consists of agriculture which contributes around 40% to the gross domestic production, equivalent to about 80% of non-oil export (UNEP, 2007). All agricultural activities out of the Nile basin depend on rainfall as the main source of irrigation. The central part of Sudan, encompassing the area between 10°N and 16°N, lies in the semi-arid transitional belt from the humid tropical Savannah in the south to the extreme dry desert in the north. The region represents the economical centre of the country where most of the population lives. This area has been struck by a period of severe drought during the 1970s and 1980s. The drought peaked during the mid-1980s and rainfall started partially to recover during the 1990s and 2000s, but is still lower than during the wet period between the 1950s and the end of the 1960s. This drought can be considered as a spatial extension to the drought of the whole Africa's Sahel during the same time.

The Sahelian drought is the most remarkable climatic event in North Africa during the second half of the twentieth century, causing economical difficulties and raising many scientific questions. Two main hypotheses are outlined explaining the Sahel drought. The first relates the drought to changes in the land–atmosphere interaction; an example is the pioneering work of Charney (1975), proposing that the change in the surface albedo due to the deforestation is a driving force for the southward shift in the Sahelian rain belt. Also, Zheng and Eltahir (1998) examined the sensitivity of the Sahel monsoon to the perturbation in vegetation distribution using zonally symmetric model coupled to a land surface scheme. Taylor et al. (2002) simulated the Sahelian land use for two periods: 1996 and a projection for 2015. Then they used the simulated land use to force a General circulation model. The study concluded that the land-use induced evaporation change results in a precipitation reduction of 4.6%. This is not enough to explain the 20% reduction of the Sahelian drought. The second hypothesis relates to changes in ocean–atmosphere interactions, due to a change in the sea surface temperature (e.g. Giannini et al., 2003; Lu and Delworth, 2005). Other mechanisms have also been suggested; e.g. the role of and teleconnection between waves in the African easterly jet and the African monsoon studied by Druyan and Hall (1996) and also see Cook (1999). Kawase et al. (2010) examined dynamic and thermodynamic changes in the Sahelian troposphere due to aerosol loading as possible causes for the observed drying trend. Along the same line, Ackerley et al. (2011) studied the effect CO2 and sulphate forcing on the Sahelian drought during the last 20 years of the 20th century. Although Central Sudan can be seen as extension to the Sahel, and it has been hit by the same drought wave, most studies of the Sahelian drought are limited spatially to the area between the Atlantic Ocean in the west and extending to 10°E at most, which is called the western Sahel. This distinction between western and eastern Sahel is likely due to the variation in the mechanism of the monsoon. The annual meridional march of the Inter-Tropical Convergence Zone (ITCZ) is a main factor governing the monsoon in North Africa, but also other circulation patterns differ from region to region. The western Sahel is strongly influenced by the African easterly jet, driven by temperature and moisture gradients between Sahara and the Guinean coast. In contrast, the eastern Sahel monsoon is influenced by the tropical easterly jet, and to some extent by the Asian monsoon. In general, the western Sahel drought was intensely studied, while less attention has been paid to the causes of the drought in the eastern part of the Africa's Sahel.

The rainy season in Sudan (June to September) depends on humid maritime air masses originating from south of the equator and humid air from the Congo River basin. The most significant feature of weather maps during this period is the location of the ITCZ, which is used as a benchmark for rainfall prediction; precipitation is expected about 50 km south of the ITCZ (Elsayem, 1986). The rainy season starts first in Southern Sudan, in an area of a complex ecosystem composed of dense vegetation cover, forests and wetland, which contribute to the moisture flux to the atmosphere. The total northward transport of moisture from all these sources plays a vital role for the rainy season in Central Sudan. There are several studies and observations pointing out deforestation and consequent degradation of the natural environment over Sudan, but fewer studies assessing the climatic consequences of such degradation (Elagib and Mansell, 2000). Recently, Elagib and Elhag (2011) evaluated the status of the drought in Central Sudan using observations from 14 weather stations. Using Pedj drought index the results suggested an intensification of drought condition. Most of the forests in Central Sudan are removed for the purposes of fuel supplement, mechanized agriculture, furniture and building. However, in Southern Sudan, where teak and mahogany forests were commercially harvested, other forests are also cleared to facilitate the emerging oil industry. The mechanised agriculture area expanded from 4.5 million hectare in 1961 to about 14 million hectare by 1996 (Ayoub 1999), and most of this expansion was likely made on expenses of woody or forested land. Elsayem (1986) suggested the deforestation and replacement of vegetation cover in South Sudan as a possible cause for the reduction in annual precipitation in Central Sudan due to lost canopy interception. Wang and Eltahir (2000) have concluded that there is a significant role for vegetation dynamics in the Sahel rainfall decadal variability. The loss of canopy-intercepted water was believed to be the reason for a 25% reduction in annual precipitation around Kilimanjaro Mountain (Boko et al., 2007). Thus, the drought induced by land use changes remains a reasonable hypothesis. Moreover, a large project of water harvest has been proposed, in Southern Sudan. The building of the Jonglei canal is intended to change the course of the White Nile to avoid passing over the swampy area of the Sudd, to reduce evaporation and thus increase the White Nile inflow by about 4.7 billion cubic meter per year (Ahmed, 2008). The Sudd is known to have very high evaporation, estimated between 1460 and 1935 mm year−1 (Mohamed et al. 2006). Although the current study is not specifically designed to examine the effect of the canal, it will still provide some understanding on potential climatic consequences of such project.

In this study, we examine a suggestion that relates the reduction of precipitation in Central Sudan to a combination of land use changes in Southern Sudan, and the dynamics of the ITCZ during the monsoon period. We use a dynamical technique, coupling the atmosphere and the underlying land surface in order to investigate the role of the vegetation cover in Southern Sudan for the northward moisture transport. The change in land use and vegetation cover will affect parameters like surface albedo, surface roughness, potential evapotranspiration, which, in turn, will alter temperature, soil moisture, and evapotranspiration and thus also change the energy transfer as well as large scale moisture and momentum transport. These potential effects in precipitation, evapotranspiration, and surface temperatures are factors of large socio-economic relevance.

This article is organized as follows. In the next section, we introduce theoretical background for surface energy and moisture budgets. In section '3. Model, methodology and data', a short introduction to the model, the domain and the experiment design is given. The mean climatology of the control experiment is shown and evaluated in Section '4. Climatology and evaluation of the model' while Section '5. Results' presents the results. Section '6. Summary and conclusion' includes a summary and the conclusions.

2. Theory

To understand the effects of changing vegetation for the surface temperature it is illustrating to consider the surface energy balance equation:

equation image(1)

where Ts is the surface temperature, equation image and equation image are the incoming short and longwave radiation, respectively, α is the surface albedo, σ is the Stefan–Boltzman constant, FH is the sensible heat flux and Lv is the latent heat of evaporation, E is the evapotranspiration and G is the soil heat flux. Except for the radiation, fluxes are positive when directed upward.

Assume now that the right-hand side of this equation is in a rough balance for a certain region, so that the soil surface temperature is quasi-constant, e.g. integrated over a season. Changing vegetation, either so that the surface leaf area decreases or by removing the vegetation all together, will to a first order reduce the evapotranspiration since the transpiration capacity of the vegetation goes down, and this will result in a positive tendency for the soil surface temperature. A concurrent change in surface albedo may partly counteract this tendency; for example replacing forest with desert will increase the albedo and therefore decrease the surface net shortwave radiation, which will induce a relative cooling. As the surface temperature increases, however, the outgoing longwave radiation, which is a function of the effective surface temperature, will increase. Also the sensible heat flux and the soil heat flux are sensitive to changes in surface temperature and these will also increase. Together these changes act to bring the situation back to a new balance at a new temperature. If the change in the evapotranspiration is larger (smaller) than the albedo-induced effect, the new surface temperature will be higher (lower). Thus for relatively small albedo changes, we expect that a decrease in the vegetation's capacity to evaporate and to transpirate water will lead to a decreased evapotranspiration and a higher soil-surface temperature.

Assume instead that the precipitation is somehow altered, everything else being the same; this will also affect the evapotranspiration. For increased precipitation the evapotranspiration will increase, since there will be more water available in the soil, and the surface temperature will thus decrease. If on the other hand precipitation decreases, the effect will be the opposite and surface temperatures will increase.

It is, however, not obvious how vegetation changes affect the precipitation. To illustrate this we examine the budget equation for the total moisture for the atmosphere (e.g. Zheng and Eltahir, 1998):

equation image(2)

where q is the specific humidity, u, v are the wind components, Ps is the surface pressure and Pu is the pressure level of the atmosphere above which the water flux is negligible. The variables E and P denote the evapotranspiration, and the precipitation, respectively. The first term on the left-hand side represents the local tendency in precipitable water while the second represents the vertically integrated divergence of the horizontal flux of moisture. This equation states that the total column's moisture content can be changed by both the divergence of the moisture transport and by the difference between evaporation and precipitation (E-P). Higher local evaporation can thus be balanced by more local precipitation and by increased moisture transport out of this region, i.e. larger divergence of the moisture transport.

In summary, changes in evapotranspiration, as discussed above, will also change the precipitation; a reduced evapotranspiration will reduce the available water for precipitation formation unless imported from outside the region. At the same time, as we saw above, reduced precipitation will reduce the evapotranspiration even further by limiting the supply of water in the soil. Thus, there is a positive feedback between the evapotranspiration from vegetation and precipitation. Regarding the surface temperature, evapotranspiration works as cooling agent, keeping the soil-surface temperature down, hence reducing the evaporative capacity might lead to both a temperature increase and reduction of precipitation. It is also clear that there is a connection to the horizontal transport of water in the atmosphere that may cause effects to propagate outside of the region immediately affected by changes in the vegetation cover.

3. Model, methodology and data

3.1. Model description

For the purpose of this study the third version of the regional climate model (RegCM3) of the Abdus Salam International Centre for Theoretical Physics (ITCP) is used. The model solves the hydrostatic so-called primitive equations using finite differencing on a terrain-following vertical sigma-pressure coordinate system. For more details the reader is referred to Giorgi et al. (1993a, 1993b) and Pal et al. (2007). This version of RegCM is based on the dynamical core of the hydrostatic version of Pennsylvania State University (PSU)-National Center of Atmospheric Research (NCAR) model MM5 (Grell et al., 1994). The radiative scheme of the Climate Community Model version 3 (CCM3, Kiehl et al., 1996) is employed. It has some extra features compared with CCM2, such as accounting for the effect of additional greenhouse gases (Nitrogen dioxide, Methane and Chlorofluorocarbon), atmospheric aerosols and cloud ice (Elguindi et al., 2007). The planetary boundary layer is represented by the scheme of Holtslag et al. (1990) while the fluxes for water bodies follow the scheme from Zeng et al. (1998). The RegCM3 system contains several parameterization schemes for convective precipitation; here we use the parameterization from Grell et al. (1994), with the Fritsch and Chappell (1980) closure assumption. The Biosphere Atmosphere Transfer Scheme (BATS1E), documented by Dickinson et al. (1993), represents the surface processes.

3.2. Domain and experiments

The domain of this study is chosen to encompass a wide area of the tropical Africa, specifically the region from 12.3° to 52.25°E and from 11.7°S to 27.2°N (Figure 1(a)). It contains 88 grid points in both zonal and meridional directions with a normal Mercator projection. In the vertical, 18 sigma levels are used with a model top at 5 hPa. The resulting horizontal resolution is 50 km while the highest vertical resolution near the surface is about 100 m, with the first level at about 80 m. This domain is relatively large compared with the targeted area of change in the land use, to allow the model more freedom against the specified lateral boundary conditions to respond to the prescribed changes.

Figure 1.

Panel (a) shows the domain encompassed by the study, with colours shading indicates the land use and vegetation types, the key of vegetations is given in Table 1). Data obtained by RegCM from the USGS (Loveland et al., 2000). Panel (b) and (c) show the area targeted by change in vegetation cover: (b) grass scenario and (c) desert scenario

The standard vegetation cover of the whole domain is shown in Figure 1(a). This is provided as a part of RegCM3 system referred to in the model documentation as provided by United State Geographical Survey (USGS) (e.g. Elguindi et al., 2007; Pal et al., 2007). For more and detailed information see Loveland et al. (2000). The vegetation in Southern Sudan is diverse, characterized as forest, fields, wetland, tall grass and evergreen trees; the Sudd is prescribed as a wetland. These settings are used for the control experiment, hereafter denoted by CONTROL (Table 1).

Table 1. Key of the vegetation types shown as shaded colours in Figure 1
KeyVegetation type
1Crop/mixed farming
2Short grass
5Deciduous broadleaf tree
6Evergreen broadleaf tree
7Tall grass
8Desert
11Semi-desert
13Bog or marsh
14Inland water
15Ocean
16Evergreen shrub
17Deciduous shrub
19Forest/Field mosaic

Two sensitivity experiments are conducted, where the control vegetation cover was perturbed south of 10°N in and slightly outside of Sudan southern territories, representing the area of strong deforestation and other changes in land use. In the first perturbed scenario we have set the vegetation of Southern Sudan to tall grass as the dominant vegetation type, beside other types like deciduous shrub and semi desert, denominated as experiment GRASS (Figure 1(b)). The second perturbed scenario, DESERT, represents an extreme change in the surface conditions, with the vegetation of Southern Sudan essentially all removed and the surface set to desert (Figure 1(c)). A few grid points around the swamp of Sudd were left inhabited by tall grass and evergreen broadleaf tree, to make the scenario extreme but still realistic. Hereafter, we focus on the region between 2°N and 10°N referred to South Sudan, and the region between 10°N and 15°N denoted as Central Sudan.

The timeline for all three scenarios is the same. The model was integrated for almost twenty years: from 1 January 1989 to 31 August 2009, discarding the first 6 months to allow the model to spin up the soil moisture. The model was forced at the lateral boundaries by the latest reanalysis of the European Centre for Medium Range Weather Forecasts (ECMWF), the so-called ERA-interim (Dee et al. 2011).

3.3. Data for the model evaluation

We are using the Climate Research Unit (CRU) dataset to evaluate the model. The CRU data comprises monthly mean gridded observations in 0.5 × 0.5 degree resolution, for five variables: temperature, diurnal temperature range, precipitation, vapour pressure and cloud cover (Mitchell et al., 2004). For more details the reader is referred to New et al. (2000). In this study we use the gridded observations of precipitation and 2 m temperature to evaluate the model performance. The data was obtained from the RegCM3 home page, provided by RegCNet group and only extends up to 2002.

4. Climatology and evaluation of the model

4.1. Precipitation

To evaluate the simulated climatology against observations we use the mean bias and the Root Mean Square Error (RMSE) for monthly mean of two variables: 2 m temperature and precipitation. The bias captures errors in the mean state, e.g. whether the model is on average too wet or dry, warm or cold, or neutral compared with the observations, while the RMSE includes errors of the time-varying component.

Figure 2 shows the spatial distribution of precipitation during JJAS for the control run (CONTROL) and for the gridded observations (upper panels Figure 2(a) and (b)). The lower panels (Figure 2(c) and (d)) show the model mean error (bias) and RMSE, respectively. The CONTROL simulation exhibits good agreement compared to the CRU observations for the monthly mean precipitation, between 0 mm d−1 in the extreme north of the country to about 8–10 mm d−1 in the extreme south, aligned with the border to Central African Republic. The highest precipitation, about 14 mm d−1, occurs in connection with complex high topography in Ethiopia. In Central Sudan precipitation varies from 6 mm d−1 around 10°N to 1 mm d−1 around 15°N in both CRU and CONTROL. Biases vary in general around ± 0.5 mm d−1, while the RMSE stays mostly below 2 mm d−1. The largest biases, up to 3 mm d−1, occur in connection with complex topography at the western border of Sudan and in the Ethiopian highlands. These areas also display the highest RMSE, up to 5 mm d−1. Problems around topography might result from either model deficiencies around complex topography or a lack of observations in such a complex environment. Furthermore, the model underestimates the amount of precipitation areas in North Sudan (Figure 2(b)). A belt of 1 mm d−1 extends to 16°N in CRU data, while the model simulation displays this extension only to south of 15°N. An unrealistic land surface prescription might be the reason behind this discrepancy, as the USGS dataset defines the whole area of Central Sudan as desert, with characteristics similar to the Sahara. However, in the reality there is a gradient in the desertification, increasing gradually from Central Sudan and northward. Such shift in precipitation would agree with the assumption of Charney (1975), attributing the drying in the Sahel belt to the extension of the Sahara southward. He assumed that the desert has higher surface albedo and temperature than it's surrounding. With also little clouds this situation leads to a radiative heat sink, which induces sinking air and reduces the relative humidity of the atmosphere. Thus, the desert conditions sustain and enhance the dryness of the Sahara.

Figure 2.

Averaged precipitation (mm d−1) for the rainy season JJAS, (a) in CRU and (b) in the model CONTROL simulation, (c) and (d) are the bias (mean error) and the RMES, respectively

Figure 3 shows the annual cycle of precipitation for South and Central Sudan for CRU and CONTROL, along with plus and minus one standard deviation. On average the model results show a slight underestimation of the precipitation for Southern Sudan. The underestimation is largest in April through July and in October and November, at about ∼1 mm d−1, while in the remaining months the differences are minor, except a slight overestimation in September. In Central Sudan the precipitation is smaller and overestimated in the model from March to August. For September and October the model and CRU data agree almost perfectly. In the remaining months precipitation is absent in both model and observations. For both regions, South and Central Sudan, the ± 1 standard deviation from CONTROL and CRU overlap, so the differences are not statistically significant. The largest interannual variability occurs during the rainy season, 0.5–1 mm d−1 for the model and 1–2 mm d−1 for CRU in South and Central Sudan. The model exhibits a slightly higher variance during April and May in Central Sudan, which might be explained by the interannual variability in the onset of the rainy season. In summary, the biases are mostly under 1 mm d−1 and RMSE are around 2 mm d−1, higher magnitudes of biases and RMSEs appear in connection with complex topography. It should be noted that the quality of the CRU data also varies due to lack of observations and complex topography. A comparison with Global Precipitation Climatology Project (GPCP, not shown) data, however, yields a similar comparison with RegCM data. Overall, the model simulates the rainfall climatology of the region realistically; both the spatial distributions and magnitudes of precipitation are well captured.

Figure 3.

The mean annual cycle of precipitation in mm d−1 (solid) with ( ± ) one standard deviation (dashed-dotted) for the period between 1990 and 2002; blue is the CONTROL and red is CRU. Panel (a) shows values for Central Sudan and (b) for South Sudan

4.2. Near-surface temperature

Figure 4 presents the spatial distribution of the annual mean 2 m temperature, with CRU in the upper left panel and the control simulation to the right, while lower panels show the bias (left) and RMSE (right). Average 2 m temperature is fairly homogeneous at about 30–32 °C, with only weak spatial variations mostly due to topography, e. g. the Marra Mountains at the western border to Chad. The general spatial pattern of monthly mean temperature in CRU and CONTROL has the same appearance with only minor difference. The model bias is generally less than ± 1 °C, while RMSE varies between 1 and 2 °C with a maximum value of 3–3.5 °C over the southern Ethiopian highland. The model overestimates temperatures in western and South Sudan by about 1 °C and over the rain forest of the Congo basin by about 2 °C. Sylla et al. (2009) noted the same positive bias around Congo and South Sudan.

Figure 4.

The mean annual 2 m temperature ( °C): (a) CRU and (b) the CONTROL simulation, (c) and (d) are the bias and RMSE, respectively

Figure 5 shows the annual cycle of the area-averaged monthly mean 2 m temperature for the Southern and the Central Sudan in CRU and CONTROL, together with plus and minus one standard deviation. The model generally reproduces the annual cycle well. However, it overestimates the amplitude of the annual cycle in Southern Sudan; the annual average is also slightly high in the model by about ∼0.5 °C. Especially, from November to March the mean model 2-m-temperatures exceeds the CRU data by 1–2 °C. Minimum temperatures occur during the rainy season and are about 0.5–1 °C too warm in the model. In Central Sudan, the modelled temperatures follow the observations closely with a minimum of 22.5 °C during January and maximum of 30 °C during May. During the rainy season, the simulated temperatures are 1 and 2 °C too low in Southern and Central Sudan, respectively. As with the precipitation, the plus-minus one standard deviation in CONTROL and CRU mostly overlaps, except for winter in Central Sudan and July and August in Southern Sudan. For both Central and Southern Sudan, and for the CONTROL and CRU, the standard deviation varies between 0.25 and 0.5 °C through most of the year, except for the CONTROL simulation during April in Southern Sudan, which shows about 1 °C standard deviation. Many factors are possible causes for these deficiencies in the simulated temperature, e. g. Sylla et al. (2009) observed that RegCM3 is sensitive to surface albedo, vegetation type, and stomatal resistance. Moreover, sensitivity in the modelled temperature for these mentioned factors is varying from region to region. Finally, all these biases in 2 m temperature are of the same magnitude as the RMSE. This fact indicates that the model captures the spatial climatology of the temperature in a good agreement with the CRU data.

Figure 5.

Annual cycle of monthly averaged 2 m temperature ( °C) for the period between 1990 and 2002, for (a) Central Sudan and (b) South Sudan. Red lines are CRU, blue lines are CONTROL, while solid line is the average and the dashed-dotted lines are plus and minus one standard deviation

5. Results

  • a.Anomalies in precipitation

Figure 6 characterizes the spatial distribution of the mean precipitation during the rainy season JJAS as given by GRASS (left) and DESERT (right) in the upper panels, with the corresponding anomalies for the two perturbed scenarios in the lower panels. In Central Sudan, the amount of monthly mean precipitation varies mostly between 1 and 5 mm d−1 in GRASS and between 1 and 4 mm d−1 in DESERT. However, in South Sudan the precipitation ranges between 4 and 6 mm d−1 in GRASS and 2–4 mm d−1 in DESERT (Figure 6(a) and (b)). The model has no precipitation along the Red Sea coast; this is perfectly consistent with the observed climatology of the region. The region has a winter precipitation and is dry during JJAS. Over the Ethiopian highland the precipitation seems to be somewhat perturbed in the southwest vicinity of the highland. A belt of 6 mm d−1 located over South of Sudan in the GRASS scenario has shrunk to the west along the border with Central African Republic in the DESERT scenario, and in the southeast a belt of 4 mm d−1 is reduced to 2–3 mm d−1. The anomalies (Figure 6, lower panels) vary between 0.1 and 0.9 mm d−1 in GRASS and 0.1 and 2.1 mm d−1 in DESERT. These negative anomalies represent 16–20% and 16–30% of the mean CONTROL precipitation during rainy season for the two scenarios, respectively. The disturbance in the Ethiopian highland precipitation appears as negative anomalies of 0.1–1.5 mm d−1 in DESERT, which is significant at the 95% confidence level. This result suggests a connection between Southern Sudan evapotranspiration and highland precipitation. The peak reductions in both scenarios occur over the perturbed region, exactly over the Sudd region.

Figure 6.

The averaged precipitation mm d−1 during JJAS in the perturbed runs (a) GRASS and (b) DESERT. The lower panels show the corresponding anomalies in mm d−1 during JJAS for (c) GRASS versus CONTROL and for (d) DESERT versus CONTROL, the grey contour shows the significant t-test at the 95% confidence level

Figure 7 shows the area-averaged, monthly mean precipitation (Figure 7(a) and (b)) and surface temperature (Figure 7(c) and (d)) for the two regions of Central and South Sudan. South Sudan shows a remarkable reduction in area average precipitation, equivalent to 25, 17, 17, 15, and 12.5% during June, July, August, September and October, respectively, in the DESERT scenario. The reduction is not limited to these months but can be seen also in March, April and May, with 32, 28 and 24%, respectively. March, April and May are relatively dry months, thus the reduction in precipitation appears as high percentages. In the GRASS experiment the anomalies are generally smaller, at a 3.3 and 8% reduction during August and September for Southern Sudan precipitation.

Figure 7.

The averaged precipitation mm d−1 during JJAS in CONTROL, GRASS and DESERT, panel (a) Central and (b) South. The lower panels (c) and (d) show the temperature for the same regions, respectively

In Central Sudan (Figure 7(a)), reductions amount to 30, 11 and 12.5% in the DESERT experiment during June, July and August, respectively, while the GRASS experiment shows reductions of 4.5 and 10.7% during July and August, respectively. Note that this region is outside of the area where vegetation was perturbed.

  • b.Anomalies in surface temperature

Figure 7(c) and (d) shows the seasonal variation in 2 m temperature, for Central (panel c) and South Sudan (panel d). The two perturbed simulations have the same temporal structure of seasonal temperature cycle. The sharp drop in the simulated temperature for July and August in Central Sudan is markedly less pronounced in the CRU data (Figure 5). Southern Sudan shows increase in the temperature throughout the year; however, the increase in temperature in Central Sudan appears only during the raining season. The temperature anomalies range between 0.25 and 1 °C in Southern and are smaller than 0.5 °C in Central Sudan. The seasonal cycle in Figure 7 is calculated using an area-averaged temperature for the whole region; this attenuates the temporal variance to some extent. For more contrast we examine the spatial distribution of the anomalies in summer and rainy season. Figure 8 shows the anomalies between CONTROL and the two perturbed scenarios, GRASS in the upper panels and DESERT in lower panels for MAM (right) and JJAS (left). A t-test indicates that almost all shaded temperature anomalies are significant at the 95% confidence level (grey contours in Figure 8). These two seasons are chosen because of their vulnerability to change in surface conditions as seen in Figure 7. In the DESERT scenario the temperature anomalies during MAM peak at about 2.1 °C over the swampy region (Figure 8(d)). Anomalies up 0.4 °C extend northward up to 12°N and to the southwest and over the Congo rain forests.

Figure 8.

Anomalies in 2 m temperature ( °C) during (a and c) JJAS and (b and d) MAM (upper row) for (a and b) GRASS and (c and d) DESERT, the grey contour shows significant t-test at 95% confidence level

The anomalies in the GRASS scenario are smaller. A band of 0.2 °C spreads from Southern Sudan up to the vicinity of the Red Sea Coast during JJAS (Figure 8(a)). During MAM a peak of up to 2.1 °C in 2 m temperature anomaly appears around the swamp of Sudd similarly as in DESERT, while this region appears to have a dampened change especially in GRASS during JJAS (Figure 9). Possibly, the large amount of water during the flood season acts as heat storage and thus prevents any pronounced temperature change. In the DESERT scenario, the local anomaly in surface temperature reaches values up to 2.4 °C during JJAS. It should be noted that the increase in surface temperature is not limited to the perturbed region only; similarly as the precipitation anomaly it stretches northward up to 13°N during MAM and up to 20°N during JJAS. Interestingly, the northward extension is far stronger during the rainy season than during the hot summer months suggesting a contribution of moisture to these changes. The role of the surface moisture in the energy balance will be investigated in Section '5. Results'd. The JJAS anomaly in near-surface humidity follows the same pattern as the temperature anomaly (not shown).

  • c.Anomalies in surface hydrology and relative humidity
Figure 9.

Anomalies during JJAS in the top layer soil moisture (mm) for (a) GRASS and (b) DESERT, and in evapotranspiration (mm d−1) for (c) GRASS and (d) DESERT (d)

Figure 10 shows the anomalies in soil moisture in the GRASS (left) and the DESERT (right) scenarios in the upper panel and evapotranspiration in the lower panel, both during JJAS. The negative anomalies in soil moisture are confined to the perturbed region; however, the anomalies in evapotranspiration extend to north of 10°N. The peaks of the negative anomalies are centred on the wetland region of the Sudd and gradually spread outward. The GRASS scenario shows a reduction in evapotranspiration between 0.6 and 1.6 mm d−1, while the DESERT scenario shows a reduction of 0.2 and 1.8 mm d−1, widely spread over the whole southern region. These anomalies represent 30–40% compared with the CONTROL evapotranspiration during JJAS. The top layer soil moisture displays a reduction in the perturbed scenarios by 2–20 mm and 5–35 mm for GRASS and DESERT, respectively. The evapotranspiration is, among other things, a function of the availability of moisture in the top-soil layer. Thus as the soil moisture goes down, the evapotranspiration is expected to follow.

Figure 10.

The moisture divergence in mm d−1 (solid line) and EP (dashed line). The (green) CONTROL, (blue) GRASS and (red) DESERT (a) Central Sudan and (b) South Sudan

The changes in surface hydrology also extend to the surface run-off, with anomalies ranging between 0.2 and 1.2 mm d−1 in GRASS and 0.2 and 1.8 mm d−1 in DESERT. Pursuing the corresponding changes in near-surface, relative humidity (not shown) shows widespread negative relative anomalies, varying between 2 and 12% for DESERT and between 2 and 8% for GRASS. This reduction in near-surface relative humidity is consistent with the reduced local evapotranspiration and also with the increase of surface temperature, as can be shown using the Clausius–Clapeyron relationship assuming that the specific humidity would remain constant. A Student's t-test (not shown) indicates that changes in evapotranspiration, soil moisture, surface run-off, and relative humidity over the perturbed region (South Sudan) during the rainy season are all significant at the 95% confidence level in both scenarios.

  • d.Anomalies in surface energy

So far, we have presented a series of results showing changes in the regional climate and surface hydrology. In this and the next subsection, we seek to understand the physical mechanisms that caused these changes. Table 2 presents the terms in Equation ((1)) from the model output. The table shows net short wave, net long wave, latent and sensible heat fluxes, for each experiment, as JJAS area average for Southern and Central Sudan in Wm−2. In each scenario, the net short wave radiation balances the long wave, sensible and latent heat fluxes with maximum residuals significantly below 1 Wm−2, indicating a good energy balance in the model. The net short wave radiation in South Sudan decreases with the perturbation of vegetations by 1.4 and 7.5 Wm−2 in GRASS and DESERT, respectively, due to the removal of vegetation (the effect of the albedo change). However, the latent heat flux simultaneously decreases by 4.8, 14.6, 1.6, and 4.2 Wm−2 for GRASS and DESERT, South and Central Sudan, respectively. The sensible heat flux increases in both regions. In the GRASS scenario in both regions, the long wave radiation is the leading term that restores residuals of the energy from the surface perturbation. In the DESERT scenario the sensible heat flux also plays pronounced role, in restoring the energy balance. It increases by 1.2 and 3.4 Wm−2 in South and Central Sudan, respectively. In Central Sudan the net short wave increases although the surface is not perturbed there. The only possible reason for this increase is a reduction of cloud albedo, which is also manifested in the reduction of precipitation (see Section 5.1). All these results are highly consistent with what we expected in theoretical discussion, see Section '2. Theory'. Although, the net short wave radiation decreases due to the change of albedo, the reduction of the evaporative capacity has left a considerable anomaly in energy to be balance in the surface energy. The new balance requires the surface temperature to increase in order to increase the outgoing longwave radiation and the sensible heat flux. The reduction of latent heat seen in Central Sudan is likely a result of the reduction in precipitation.

  • e.Anomalies in moisture budget
Table 2. The surface energy terms in Equation ((2)) parameterized from the model output for JJAS, unit Wm−2
TermSouthern SudanCentral Sudan
 ControlGrassDesertControlGrassDesert
Net short wave185.3183.9177.8206.7208.2209.6
Net long wave48.350.854.466.868.370.5
Latent heat flux83.979.169.365.563.961.3
Sensible heat flux52.853.854.074.475.977.8

Equation ((2)) shows that, in time-mean balance with the local tendencies approaching zero, the precipitation will be reduced either due to changes in the moisture transport or due to reductions in the local evapotranspiration. Figure 10 shows the terms of equation (2) from the model output: the vertically integrated divergence of precipitable water (second term in left-hand side) and EP, in mm d−1, which is equivalent to kg m−2 s−1. First thing to note is the good agreement in the model water budget, although E and P are in the model accumulated over each 3 h period, while the wind components and the specific humidity for the atmospheric flux calculations are only provided instantaneously 6 h. In Southern Sudan (Figure 10(b)) moisture convergence starts from early of March and continues through until October with peak during August. During the rest of the year, November to March the region acts as a source of moisture with a positive EP (dashed line), indicating a moisture flux out of the region. The EP is negative during the period moisture convergence, with peak at August, indicating that precipitation is larger than the local evaporation. The GRASS scenario shows behaviour close to the CONTROL, however, DESERT shows a reduction in both moisture convergence and EP. The reductions in moisture convergence during JJAS are about 0.84 and 11.8% in GRASS and DESERT, respectively. This reduction in moisture convergence in Southern Sudan is unexpected and interesting, and will be discussed below. In Central Sudan (Figure 10(a)) weak moisture convergence starts in May to strengthen continuously until reaching peak in August and then gradually weakens until ending by September. The perturbed scenarios follow a similar cycle to CONTROL, but differ in the magnitude. EP is negative during the rainy season and becomes slightly positive from September till December. During the rainy seasons the meridional component of the wind is the main moisture supplier in both regions. The moisture convergence is reduced by 11.5% in GRASS and 21.9% in DESERT during the rainy season.

If the moisture convergence in Central Sudan is reduced due to a reduction of the local evaporation in the southern vicinity, as it is the source of Central Sudan moisture according to dynamics of the monsoon, what causes the reduction in moisture convergence in Southern Sudan? To answer this question, we examined the vertically integrated atmospheric flux of moisture (arrows) and EP (shaded) (see Figure 11). The anomalies in moisture flux (Figure 11(b) and (c)) show an anticyclonic pattern in Central Sudan that strengthens the easterly flow over the southern parts of the country. This pattern is almost centred over the Central Sudan. During this period of the year (JJAS) the south-easterly flow of moisture is dominating, which pushes the ITCZ to the northward of Sudan; this will determine the intensity of the rainy season. The anticyclonic pattern in the anomalies indicates that the north-westward moisture flux declines and is deflected more westwards in the perturbed scenarios. This anomalous westward flow of relatively moist air explains the positive anomaly of 0.6 mm d−1 in precipitation that appears along the south-western border of the perturbed region. The anomalies in EP indicate in both scenarios that Southern and Central Sudan experience a drying owing to the vegetation change. Aligned to the boarder with Congo the DESERT scenario shows an increase of EP by 0.5 mm d−1. In general, such a change in the monsoon circulation was noted by Zheng and Eltahir (1998); they concluded that the Sahel monsoon is more sensitive to tropical deforestation than to the deforestation at the boundary to Sahara.

Figure 11.

Panel (a) is the moisture flux during JJAS in kg m−2 s−1 and EP in mm d−1 (shaded) (b) and (c) are the anomalies in moisture flux and E-P in GRASS and DESERT, respectively

6. Summary and conclusion

In this study, we examined the sensitivity of Sudan climate to changes in land use (deforestation and replacement of vegetation cover) in Southern Sudan. Three scenarios of surface condition were applied, a control run and two perturbed scenarios. The dense equatorial forest and vegetation in the Southern Sudan supply moisture to the atmosphere. Hence, it increases rainfall locally and moisture transport to neighbouring regions.

The control run is evaluated over the area of interest for the mean 2 m temperature and precipitation and their annual cycle. The biases in temperature stay between ± 1 °C while the RMSE varies between 1 and 2 °C over the plain surface of Sudan and with slightly higher values in the vicinity of complex topography. For precipitation biases also stay between ± 0.5 mm d−1, maximum bias of 3 mm d−1 appears with a connection of complex topography. The model produces the observed annual cycles of precipitation and surface temperature. The model underestimates the surface temperature in Southern Sudan and the area of the Congo forest during the rainy season, likely due to the complexity of wetland surface. The RMSE remains mainly close to the value of the bias, which indicates an acceptable downscaled climatology of the region.

The deforestation in Southern Sudan results in a significant reduction of precipitation, gradually distributed from South to Central Sudan. During the rainy season (JJAS) the monthly mean precipitation declines about 2.1 mm d−1 in Southern Sudan and about 0.5–1 mm d−1 in Central Sudan in the DESERT scenario, while in the GRASS scenario the reduction varies between 0.1 and 0.9 mm d−1. The anomalies in precipitation spread up to the Ethiopian highland, at about 0.3–1.5 mm d−1, although in theoretical climatology the precipitation over this region is assumed to be less dependent on the south-westerly flow of humid air. More work is needed to investigate this result, because the Ethiopian highland precipitation contributes about 60% to the annual inflow to the river Nile. The 2 m temperature is increased over the perturbed region by 2.4 °C at maximum in DESERT during the rainy season and summer. The temperature in the Sudd swamp region showed no response during JJAS, possibly owing to the large soil water amount during the flood season JJAS. The perturbation also affects the total hydrology of the surface, e.g. reduced precipitation causes reduced run-off and top layer soil moisture and consequently the evapotranspiration and atmospheric relative humidity are reduced as well. The increase in surface temperature and the reduction of the absolute humidity are together responsible for the reduction in relative humidity. The change in hydrology interrupts the surface energy balance due to the change in latent heat flux; as a consequence the surface over the perturbed area is heated.

The moisture influx into Central Sudan during JJAS is reduced by 11.5 and 21.9% GRASS and DESERT, respectively. Although the changes in land use are applied in Southern Sudan, they affect the precipitation in Central Sudan; a regional teleconnection. This result provides support for the hypothesis of this study: changes in land use are one of the factors that contribute to the changes in precipitation in Central Sudan. The reduction of moisture influx in South Sudan during JJAS was an intriguing result. The decline of the south-easterly flow, which appears as an anticyclonic anomaly pattern, weakens the moisture flux into South and Central Sudan. The reduced moisture from changes in both local evapotranspiration and moisture influx are of comparable magnitude as the reduction of precipitation. Hence the perturbation of the surface has affected the amount and dynamics of the moisture budget. The land-use in Central Sudan has undergone a noticeable change during the 20th century (Elagib and Mansell, 2000; Ayoub 1999 and UNEP, 2007), which is not simulated in our numerical experiments. It is possible, however, that this may have contributed to the observed drying trend as well. The high evaporation in South Sudan is usually considered as “lost water”, thus projects for water harvesting are planned for South Sudan, in Jonglei and Bahr Elghazal basin, to reduce this evaporation. This study highlights the general importance of this evaporation to the regional climate outside of South Sudan. Detailed studies to assess the climatic effect of these planned water harvest projects in South of Sudan are therefore needed, to justify the feasibility of such projects in comparison with local and non-local climatic and environmental risks.

Acknowledgements

The authors acknowledge the critical, detailed and valuable comments from two anonymous reviewers.

Ancillary