Abstract: The rainfall and temperature variations would produce crucial impacts on water quantity and quality. The Intergovernmental Panel on Climate Change in its fourth report already predicted an eventual warming of 2°C in 2050 for so-called “optimistic” and “median” scenarios. The main objectives of this study were (1) To evaluate the climate variability and its impacts on the water resources in the Central Africa area, (2) To forecast the variability trend for the following years and (3) To heighten the local decision-makers on its impacts on water resources using data collected from the local weather stations. This study has established an inventory of available water resources and water needs. It also emphasized warming and drought expansion in this region thanks to analytical and geostatistical methods. The results show that, climate is changing in this region: the average temperature (ranged from 21-29°C) increases; the average annual precipitation (ranged from 800-1600 mm) decreases and the groundwater table increased up to 4 m in 15 years. Despite this general downward trend, the rate of variation in rainfall compared to the average over the study period remains positive in several localities ranging from -6 to +10%. The average Standard Precipitation Index (SPI) value also decreases and varies from -0.6 to +0.6. All these results and observations in the first approximation can explain the draining of wetlands and surface water, the dryness of wells and the rise of static and piezometric levels in boreholes drilled in certain localities in the region. The results from this article can be extended to better control climate change phenomenon and its impacts on water resources in Central Africa zone and in similar regions in the world.
INTRODUCTION
According to the fourth report of the Intergovernmental Panel on Climate Change (IPCC, 2007), a warming of 2°C is possible in 2050 for so-called “optimistic” and “median” scenarios. The temperature rise of this magnitude will have a significant impact on water resources. Depending on the morphological and hydrogeological realities of each region, the consequences can result in one or a combination of three basic problems: Declining water (drought), over abundance of surface water (flooding), degradation of the water quality (pollution). All these problems are reflected in the scarcity or shortage of good quality water for domestic, industrial, agro-pastoral and hydropower needs (Crisci et al., 2002; Singh et al., 2008; Zhao et al., 2013). This region is already experiencing atrociously the water stress. This situation could be exacerbated by climate change phenomenon since the global prediction tends to make dry regions dryer and wet ones wetter (Sahin et al., 2007; Russo and Sterl, 2012). The variability of either surface water and groundwater is closely related to the rainfall since the aquifer recharge occurs in two processes (Djeuda-Tchapnga et al., 1999). By direct supply from vertical infiltration of rainfall and runoff from superficial network of rivers, or by a side supply from the banks of rivers through weathering, cracks and joints of the underground basement. Indeed, groundwater resources are supplied by precipitation, whether liquid (water droplets of rain, mist), solid (crystals of ice or snow) or a mixture of both (freezing rain). Rainfall is the main entrance in the water balance. The analysis of the water balance of a region permits to determine the proportion of effective rainfall that would feed the useful groundwater reserves by infiltration (Panwar and Chakrapani, 2013). On the other hand, the deficiency of rainfall has a direct impact on the rivers flow. Another parameter used in the water balance is evapotranspiration. It represents the sum of water vapor back into the atmosphere by evaporation of surface water and plant transpiration. This component mainly increases with temperature and to a lesser extent with the wind and the type of land occupation (Kumar, 2012). It should therefore be noted that a maximum temperature rise will accelerate the evapotranspiration which in turn will affect the effective rainfall available for recharging aquifers. The hydrographical system in the region consists largely of temporary and seasonal waterways called mayos (closely dependent on rainfall events) which reinforces the “water-speed record” link (Nouvelot, 1972). In general, surface water that irrigates this area derives from basins of Lake Chad, Niger and Sanaga. However, knowledge about groundwater reservoir is approximate and ridiculous. This study is not intended to determine the origins of climate change (Avila et al., 2012; Chen et al., 2012; Yang et al., 2013) or to disprove the global climate circulation (GCM) predictions that local administrators do not seem to fit into their overall policy. Its main objective is to warn these managers that climate change and variability are a reality in this tropical region that is already experiencing water stress. And to achieve this goal, we conducted a literature review to establish the inventory of available water resources and needs. The analytical method was used to study the variation of temperature, precipitation and Standard Precipitation Index (SPI). The geostatistical method is also used to interpolate rainfall and SPI variation.
MATERIALS AND METHODS
Studied area: Central Africa is an ecologically strategic region considered as the second lung of the world, it has the second forest reserve in the world after the Amazon Forest (Collins et al., 1991; Alpert, 1993). Overall, this is an equatorial region with a mild climate, moderate temperature and abundant rainfall. Cameroon, commonly called Africa in miniature is at the heart of this region and also has one of the wettest places in the world, Debundsha, with an annual rainfall of more than 10,000 mm (Frankham et al., 2004). However, its northern part shows signs of concern for climate variability which would extend over the entire sub-region with damaging and diversified environmental and socio-economic consequences. This study is devoted primarily to hydrological consequences.
The area comprises a diversity of physical and natural conditions. Located between 6 and 13°N and 11 and 16°E, it includes the Eastern part of Nigeria, the Western part of Central African Republic, the Southern part of Chad and the Northern part of Cameroon (Fig. 1) for a population estimated in 2007 at more than 20 million 700,000 inhabitants (BUCREP, 2010). This region is under the influence of the African monsoon and is dominated by the seasonal translation of the Inter-Tropical Convergence Zone and by the Harmattan winds coming from Sahara.
| Fig. 1: | Location of the study area and main weather stations |
It can be divided into two sub-climatic areas: The sahelian zone stretching over the 8th parallel and the sudanian (or sudano-sahelian) zone located southward. Mean annual temperatures vary from 28°C in the Sahelian region to 24°C over the Adamaoua region while monthly maximums are between 34 and 28°C, respectively. The mean annual rainfall amounts range between 500-1500 mm and show strong interannual variability.
Three main topographic units can be distinguished: The Mandara and Alantika Mountains, from 900-1885 m, in the western part of the northern zone, between Cameroon and Nigeria; the Adamaoua plateau (~1000 m) in the southern part of the region which shelter mountains reaching up to 2500 m; The plains (altitudes below 300 m), drained by the permanent rivers of Benoue and Logone and practically flooded from July to November (Djoufack et al., 2012).
This area is covered by three headwater basins (Lake Chad, Niger and Sanaga) which engender several rivers. Its drainage network consists largely of temporary and seasonal streams and waterways called mayos. They are very dependent on rainfall events and surface water that irrigates this region emanates from these three watersheds.
Chari and Logone are the only permanent rivers of the upper basin of Lake Chad. The rest of the river system consists of seasonal mayos. Interannual modules Chari and Logone, respectively 564 (1954-2008) and 445.7 m3 sec-1 (1948-2008). The interannual module mayo tsanaga is also known for the period 1953-1979 due to Olivry (1986). It is equal to 21.1 m3 sec-1. The volume of water flowed into upper basin of Lake Chad is 32.52 km3. In this total volume, 10.79% are provided by the lower part of Lake Chad basin and 87.15% are from rainfall on Chadian territory.
The volume of water drowned in the lower part of Lake Chad basin was evaluated through hydrometric station of Logone Baibokoum. The volume of water flowed into this watershed is equal to 3.51 km3. Apart from the Cameroonian portion of Lake Chad, the only lakes that have been identified in the Lake Chad Basin are Lake Bini and Lake Dang. The volume of water stored in these lakes is not known. Some reservoirs have been built in the Lake Chad Basin. These are the Mokolo dam with 5.3x106 m3 of water, the Chidifi dam near Dourbeye with 5x106 m3 of water and the Maga dam which retains only 5x105 m3 of water (MINAGRI, 1986). For the entire Lake Chad basin, the volume of surface water is 32.52 km3.
Three main hydrometric stations were used to measure the volume of water drowned in the Niger basin: Garoua station on the Benoue River, Safaie station on the Faro River and Ngouri station on the Menchum River. The specific power modules of the Benoue and Faro basins, respectively 5.46, 12.27 and 10.25 l sec-1 km-2.
The volume of water flowed into the Sanaga basin was estimated from hydrometric data collected on the Sanaga station in Edea. The interannual unit is 1872.91 m3 sec-1 for the period 1945-2006. The specific power module is 14.24 l sec-1 km-2 and available surface water is summarized in Table 1, according to Global Water Partnership (GWP, 2009).
Water requirements can be sorted by sector and by headwater basin. Thus, needs are summarized in Table 2 (GWP, 2009).
The water balance provided in Table 1 and 2 above shows an abundance of water resources in the region. However, households and livestock remain largely unsatisfied because of the logistics failure. This is due to two reasons: not only the demographic disposition (distribution) moves populations away (far) from these resources but this surface water are of very poor quality for drinking and breeding. Given the lack of adequate treatment device afflicting the region, the exploitation of groundwater physico-chemical and bacteriological quality closer to the standards set by the World Health Organization (WHO) is suggested. Only groundwater sedimentary basin of Lake Chad was estimated at 3.2 km3. All in all, the deficit of good quality water is screaming.
Data description: Weather data are provided by national weather stations and parks, regional and international partner agencies like the Agency for the Safety of Air Navigation in Africa and Madagascar (ASECNA) and the World Meteorological Organization (WMO).
| Table 1: | Volume of surface water available in the area |
| Table 2: | Summary of water needs by sector and by basin |
For the sake of accuracy, we preferred these data, covering a half-century (1960-2010), to those of Climatic Research Unit (CRU). This interval constitutes our study period. Static level has been measured on wells and boreholes.
Standard precipitation index (SPI): To describe the interannual variability of rainfall, we will mainly use the analytical method and more precisely, descriptive statistics (mean values) and graphical representations. This method has been used by several scientists (Deka et al., 2013; Partal et al., 2006; Saidi et al., 2013). The variation in annual precipitation of each station is calculated relative to the reference period (study period). In addition, to better detect the rainfall deficit, the Standard Precipitation Index (SPI) is calculated for each station. It is expressed by the following equation (Servat et al., 1998):
| (1) |
Then for our study period, we calculate the average value of SPI with the Eq. 2:
| (2) |
where, Xi is the value of the annual rainfall of the year i;
Geostatistics: Geostatistical method is essentially based on the notions of variogram and kriging. In the context of developing countries where there is not enough weather stations this geostatistical approach is very important and allows interpolating study where measures have not been carried out effectively.
The variogram is a tool that is used to describe the spatial continuity of a phenomenon including weather behavior (De Carvalho Alves et al., 2013; Perez and Jury, 2013). The theoretical formulation of the variogram γ(r) uses the concept of variance (var) applied to the difference between two observations h(x) and h(x+r) separated by a distance r:
| (3) |
In practice, only the experimental variogram γe(r) is calculated from observations using the following equation:
| (4) |
where, γe(r) is the estimated value of the variogram for lag (r); N(r), the number of pairs of points separated by distance r; h(xi) and h(xi+r) are values of h at positions xi and xi+h, respectively.
Ideally, a point of the experimental variogram is considered representative if N(r)≥30. At these point values, a suitable theoretical variogram model is adjusted. The main current eligible models are nugget effect, linear, gravimetric, cubic, pentaspherical, spherical, exponential, power, Gaussian, Cauchy and logarithmic variograms (Chiles and Delfiner, 1999). A model is admissible if any variance calculated from the model is positive.
The description of a variogram model is based on the quantification of multiple parameters identified in Fig. 2. The range (length) a is the distance where the correlation between observations becomes zero. At this distance, the variogram reaches the sill (scale) σ2 which is the sum of the nugget variance C0 and the partial sill (variance) C. The nugget effect derives from various sources such as measurement errors, existence of a microstructure smaller than the size of the sample and/or the presence of a microstructure with a range less than the distance between the two closest observations. It may be impossible to quantify the contribution of each source.
When the minimum number of pairs of points is not reached, other methods such as maximum likelihood or cross-validation can be used (Marcotte, 1995).
| Fig. 2: | Experimental variogram |
As a last resort, an ad hoc variogram model inspired by the physical features, the scale of the study, measurement errors associated with the variables and methods used of the studied phenomenon can be used.
Kriging is a commonly used method of interpolation (prediction) for spatial data. The data are a set of observations of some variable(s) of interest, with some spatial correlation present. Usually, the result of kriging is the expected value and variance computed for every point within a region.
Thus, it is a direct approach with a unique solution to an estimation problem and can be used to estimate the unknown value h* of a variable at a point from the surrounding known values hi using the following equation:
| (5) |
where, λi represent the kriging weights.
Obtaining a minimum variance of estimation σ2e means to minimize the expression:
| (6) |
Substitution of the linear estimator can rewrite Eq. 6 as follows:
| (7) |
To ensure no bias for the linear estimator (Eq. 7), the constriction:
should be integrated into the model. This constraint means that the local average of the observations is constant throughout the field. The minimization of a quadratic function with the presence of an equality constraint (Eq. 8) is effected by the method of Lagrange which involves the Lagrange multiplier μ:
| (8) |
With the substitution of σ2e the Eq. 8 can be rewritten as:
| (9) |
Equation 9 provides ordinary kriging when cancel all the partial derivatives with respect to each λi and compared to μ. The ordinary kriging system becomes:
| (10) |
The minimum estimation variance of the system (kriging variance) σe2 is determined by the substitution of kriging equations in Eq. 10 to obtain the 11:
| (11) |
In practice, it is easier to use the matrix form of the kriging system Eq. 12:
| (12) |
where, Ks is the (nxn) matrix of covariance between observations, ks, the (nx1) matrix of covariance between the n observations and the point to be estimated, λ. The solution of this system is provided in matrix form as Eq. 13:
| (13) |
Where:
| (14) |
And finally:
| (15) |
Thus, Eq. 15 is used to calculate the kriging weights λi needed to estimate a point defined by the linear estimator with Eq. 5 (Oyoa et al., 2012; Diab et al., 2013; Meli’i et al., 2012, 2013; Nouck et al., 2013).
For each interpolation, several theoretical models of variogram are plotted for the visual analysis; Root Mean Square Errors (RMSE) are also evaluated for the best choice. Following equations express certain theoretical models that have been evaluated:
| (16) |
| (17) |
| (18) |
| (19) |
| (20) |
| (21) |
| (22) |
| (23) |
where, h, a, C0 and C, respectively represent lag, range (length), nugget and the covariance of the theoretical variogram.
RESULTS AND DISCUSSION
Variogram model for SPI: To determine the suitable variogram model to interpolate SPI data, Eq. 3-4 were used. The experimental variogram obtained according to these equations and the observed data in the study area, was plotted in Fig. 3. In the same figure, different theoretical models (Eq. 16-23) were evaluated and plotted for visualization. RMSE are also given in Table 3 for rational analysis.
Figure 3 shows that Gaussian model is the one that fits more with the experimental variogram. For a rational analysis, RMSE values of different models are given in Table 3.
This table presents the RMSE values of different variogram models evaluated to interpolate SPI. It reveals that among exponential, quadratic, spherical, pentaspherical, magnetic, gravimetric and cubic, the Gaussian model has the least RMSE.
| Fig. 3: | Gaussian variogram used for SPI interpolation C0 = 0.2, C = 0.8, a = 350 km |
| Fig. 4: | Magnetic variogram used for rainfall
interpolation C0 = 50 mm2, C = 200 mm2,
a = 250 km |
| Table 3: | RMSE values for different variogram models evaluated for SPI interpolation |
Variogram model for rainfall: Figure 4 shows the fitness of magnetic model with the experimental variogram. This model will be confirmed by Table 4 where RMSE of several models are evaluated.
This table reveals that the magnetic model has the least RMSE. According to the results from Fig. 4 and those from Table 4, the magnetic model was therefore used to interpolate rainfall data in the study area.
| Table 4: | RMSE values for different variogram models evaluated for rainfall interpolation |
| Table 5: | Variation of static level between 1995 and 2010. The average increase is 4 m in 15 years (1995-2010). |
Rise of groundwater table (static level): The static level of a well is the height of water in the well at equilibrium, attained when the pump has not been used for an extend period of time (approximately 6 h). Table 5 shows the rise of the static level from 1995-2010. This table reveals that the mean static level in this area, observed from several wells and wetlands has increased for about 4 m from the top to the bottom.
Variability of previous temperature: The analysis of temperature data shows an increasing trend in all stations except Tibati (Fig. 5). This increase is emphasized in some localities such as Maroua and Kaele. However, it is moderate in other areas like Tchollire and Poli. Kousseri station meanwhile experienced some malfunctions in the past. It is therefore difficult to assess climate change in this locality from these data.
Prediction of upcoming temperature: If the causes of climate change and variability remain unchanged, the temperature will increase until 2050 except in the localities of Poli and Tibati.
| Fig. 5(a-l): | Variation of average temperature. The
average temperature ranges from 21°C (Ngaoundere) at 29°C (Kaele)
(a) Poli, (b) Kousseri, (c) Mokolo, (d) Yagoua, (e) Tchollire, (f) Maroua,
(g) Tibati, (h) Garoua, (i) Ngaoundere, (j) Banyo, (k) Kaele and (l) Meiganga
station |
The largest increases are observed in Ngaoundere (+3.6°C) and Mokolo (+2.1°C). Overall, these estimates are uniform with those of the fourth report of the IPCC (2007). These results are shown in Table 6.
The value of the temperature is obtained by the equation of the linear regression. And the variation is calculated in relation to the value of the reference period. If current conditions remain, temperature will increase excepted in Poli and Tibati.
Standard Precipitation Index (SPI): The Standard Precipitation Index (SPI) decreases in all stations except Kaele, Banyo and Maroua (Fig. 6). This decline is significant in areas such as Meiganga and Ngaoundere. However, it is moderate in other areas such as Tibati. This trend supports the hypothesis of more severe drought in the future in this region which is already climatically vulnerable.
The spatial evolution of the SPI is given by the map of Fig. 7. The average SPI value varies from -0.6 to +0.6. The biggest downfall is observed in Meiganga with an amplitude of about 1.6.
Variability of previous rainfall: The analysis of precipitation shows a decreasing trend except Kaélé, Maroua and Banyo (Fig. 8).
| Table 6: | Equation predicting the upcoming temperature |
| Fig. 6(a-j): |
Variation of average SPI (a) Banyo, (b)
Garoua, (c) Guider, (d) Kaele, (e) Maroua, (f) Meiganga, (g) Ngaoundere,
(h) Poli, (i) Tibati and (j) Tchollire station |
The average annual precipitation ranges from 800 mm (Maroua) to 1600 mm (Meiganga and Tibati). The biggest breakdown is observed in Tchollire (from 1400-1100 mm). Despite this general downward trend, the rate of variation in rainfall compared to the average over the study period remains positive in several localities (Fig. 9). It ranges from -6 to +10 %.
Constant or even increasing trend doesn’t mean an abundant precipitation but the non-deterioration of the situation. The most comprehensive quantitative assessment of rainfall is shown on the map of Fig. 9. This map shows the percentage change in precipitation.
Prediction of upcoming rainfall: Under the same climatic conditions, only localities of Poli, Guider and Banyo will experience an increase in precipitation in 2050.
| Fig. 7: | Map of changes in average SPI |
| Fig. 8(a-j): |
Variation of yearly average rainfall
(a) Tchollire, (b) Tibati, (c) Poli, (d) Ngaoundere, (e) Meiganga, (f)
Maroua, (g) Kaele, (h) Guider, (i) Garoua and (j) Banyo station |
| Fig. 9: | Map of changes in precipitation |
| Table 7: | Prediction of the upcoming rainfall |
Table 7 gives the relative change of the variation with respect to the reference period.
Upcoming rainfall data are achieved thanks to the equation of linear regression. And the variation is computed in relation to the value of the reference period.
CONCLUSION
This study shows that, climate change phenomenon is evident in Central Africa. Even if, in some stations like Poli and Tibati, temperature and rainfall decrease, these two parameters increase in general in most of localities as it was predicted in the fourth report of IPCC. This warming will accelerate the process of evapotranspiration and impact the groundwater resources since the amount of effective rainfall available for recharging aquifers and streams is inversely proportional to evapotranspiration. This situation supports the hypothesis of the worst drought in this area that is already climatically fragile. First consequences are already visible: wetlands draining, wells dryness, rise of static and piezometric levels in boreholes drilled in certain localities. The results from this article can be extended to better control climate change phenomenon and its impacts on water resources in Central Africa zone and in similar regions in the world.
ACKNOWLEDGMENTS
We would like to thank Dr. Kouna Atangana Basile, the Cameroonian Minister of Water Resources and Energy and Mr. Nkili Robert, the Cameroonian Minister of Public Transports, who permitted us to access to the archives and records of all his national and regional services. They also want to thank Mr. Nsangou Soulemanou, Passah Mariatou, Mr. Njombissie Petcheu Igor Casimir, the late Mr. Jeannot Nouck and Mrs. Julienne Ngo Matip for the financial support, assistance, general comments and suggestions.