2012 Null
2012 Null
2012 Null
ABSTRACT 1. INTRODUCTION
Climate change is described as the most univer- Climate has profound effects on the biophysical re-
sal and irreversible environmental problem fac- sources of the planet. It is the major factor controlling the
ing the planet Earth. While climate change is patterns of vegetation structure, productivity, and plant
already manifesting in Ethiopia through changes and animal species composition [1]. Many plants can
in temperature and rainfall, its magnitude is poorly successfully reproduce and grow only within a specific
studied at regional levels. The objective of this range of temperature and respond to specific amounts and
paper was to assess and quantify the magnitude seasonal patterns of rainfall, and fail to survive if climate
of future changes of climate parameters using changes. Thus, there is now a substantial concern over
Statistical Downscaling Mode (SDSM) version the global problem of climate change and its current and
4.2 in Amhara Regional State which is located be- future impacts. The impact of changing climate, such as
tween 8˚45N and 13˚45N latitude and 35˚46E rising global average temperatures and increases in fre-
and 40˚25E longitude. Daily climate data (1979- quency and severity of extreme events, droughts and
2008) of rainfall, maximum and minimum tem- floods, are already affecting human well-being, biodiver-
peratures were collected from 10 observed me- sity and ecosystems, economies and societies worldwide
teorological stations (predictand). The stations [2]. Seasons are shifting, temperatures are climbing and
were grouped and compared using clustering sea levels are rising around the globe. And meanwhile,
and Markov chain model, whereas the degree of our planet must still supply us and all living things with
climate change in the study area was estimated air, water, food and safe places to live [3].
using the coupled HadCM3 general circulation Africa contains one-fifth of all known species of
model (GCM) with A2a emission scenarios (Pre- plants, mammals, and birds, as well as one-sixth of am-
dictors). Both maximum and minimum tempera- phibians and reptiles. However, because of climate change,
tures showed an increasing trend; the increase in these ecosystems are threatened [4,5]. Similarly, the var-
mean maximum temperature ranges between ied ecology, edpahic and climatic conditions in Ethiopia
1.55˚C and 6.07˚C and that of the mean minimum which accounts for the wide diversity of biological re-
temperature ranges from 0.11°C and 2.81°C. While sources both in terms of flora and fauna is declining be-
the amount of annual rainfall and rainy days de- cause of climate variability [5,6]. The negative impacts
creased in the study Regions in the 2080s. The associated with climate change are also compounded by
negative changes in rainfall and temperature ob- many factors, including widespread poverty, human dis-
tained from the HadCM3 model in the current
eases, and high population density, which is estimated to
study are alarming and suggest the need for fur-
double the demand for food, water, and livestock forage
ther study with several GCM models to confirm
within the next 30 years [7]. For example; climate vari-
the current results and develop adaptation op-
ability and extreme weather events, such as high tem-
tions.
peratures and erratic rainfall, are critical factors in initi-
ating malaria epidemics and increase the spread of infec-
Keywords: Amhara Regional State; Climate
tious diseases especially in the highlands of Ethiopia
Change; Ethiopia; HadCM3; Statistical Downscaling
[3,8,9]. In additionally, [10,11] noted that rainfall vari-
#
Current address: CIMMYT, Addis Ababa, Ethiopia. ability and associated droughts have historically been
major causes of food shortages and famines in Ethiopia. commonly used GCM to develop climate change scenar-
Being able to predict seasonal climate can help to ios in area given and regional impact assessment studies
minimize the possibility of “climate surprises” in order [20,21]. Its good simulation of current climate without
to reduce impacts to society and ecosystems. With these using flux adjustments was a major advance at the time it
predictions, decision makers can be provided with reli- was developed and it still ranks highly compared to other
able scientific information about possible extreme cli- models [22]. Hence, HadCM3 was one of the major
mate events. Many sectors of society use these predic- models used in the IPCC Third and Fourth Assessments.
tions, including agriculture, fishing, forestry, energy, in- Further, (e.g. [23]) used HadCM3 models in order to de-
surance, public health, water resources, recreation, trans- velop climate change scenarios of rainfall, minimum and
portation, health, and construction. For example, with an maximum temperature in arid regions of Chile in order to
increased risk of drought, we know that some possible analyze the future potential impact of climate change on
impacts include water shortages, early onset of the wild- stream flow. In addition, in his study used HadCM3
fire season, and lower or no crop yields. This in turn GCM model outputs established on the A2 and B2 sce-
could lead to increased disaster assistance payments, narios was used to downscale climate change scenarios
higher food prices, and disrupted transportation on inter- at regional levels [24]. Further more, [25] also used the
nal waterways. With advance knowledge of drought, outputs of HadCM3 A2a and B2a scenarios for climate
water resource managers can adjust the timing of water change impact analysis using crop modeling.
releases from reservoirs, and farmers can alter the type of Accurate and timely climate information and predic-
crops they plant [12]. tions can help many sectors of society circumvent the
While most of us feel the effects of climate variation, impacts posed by climate variations. This reduces the
many businesses, services and activities depend on climate risk of economic setbacks and ecological damage. There-
prediction to prepare adequately, to manage risk, protect fore, the aim of this paper was to downscale and quantify
the environment, and to save lives [13]. A prediction proc- the magnitude of future changes of maximum and mini-
ess, begins with observing and accurately measuring the mum temperatures and rainfall in the Amhara Regional
most recent and current environmental conditions. The State using the HadCM3A2a GCM models.
tools of the predictor can therefore be based on statistics
from historical climate data which have been analyzed to 2. MATERIALS AND METHODS
show the relationships between the Earth’s surface condi- 2.1. Description of the Area
tions and climate [12].
Numerical models (General Circulation Models or Amhara region is located between 8˚45N and 13˚45N
GCMs), are currently most credible tools available for latitude and 35˚46E and 40˚25E longitude in North
simulating the response of the global climate system [14]. West Ethiopia. The total area of the region is estimated at
However, because of their coarse spatial resolution, GCMs 156,960 km2, which is divided into 11 administrative
to date are unable to provide reliable climatic informa- zones and 105 districts [26] (see Figure 1).
tion at regional and local scales [15]. As suggested in
different guidelines and documentation developed by the 2.2. Climate of the Study Area
IPCC [16,17], the climate change information required The climate of the Amhara region is affected sig-
for many impact studies is at a much finer spatial scale nifi-cantly by variation in altitude, its latitudinal position,
than that provided by GCMs. Because of this limitation, prevailing winds, air pressure and circulation and its
other alternative methods called down-scaling techniques proximity to the sea. Traditionally, the climate of region
using statistical downscaling model (SDSM) have been is divided in to Kola (hot zone) which represents and
developed in recent years to obtain fine resolution cli- cover 31% of the Region below 1500 meters above sea
mate change information at regional/local scales. The level (a.s.l.) and Woyina Dega (warm zone) encompasses
SDSM calculates statistical relationships, based on mul- areas between 1500 - 2500 m a.s.l and cover 44%, and
tiple linear regression techniques, between large-scale Dega (cold zone) which represents areas between 2500 -
(the predictors) and local (the predictand) climate [15]. 4620 m a.s.l which cover 25% of the Region [9]. The
These relationships are developed using observed cli- annual mean temperature of the Region is between 15˚C
mate data and, assuming that these relationships remain and 21˚C. But in valleys and marginal areas the tem-
valid in the future, and here can be used to obtain down- perature exceeds 27˚C [27].
scaled local information for some future time period by
driving the relationships with GCM-derived predictors 3. DATA SETS AND METHODOLOGY
[16,18,19].
3.1. Observed Data
Hadley Centre Coupled Model (HadCM3) GCM mod-
el outputs established on the A2 scenario is the most Thirty years observed rainfall data (1979-2008) was
Figure 1. Geographical location of Amhara region with reference to the whole country (Ethiopia) and Amhara.
collected from National Meteorological Agency (NMA) historic (observed) condition can be replicated or not
of Ethiopia for 10 meteorological stations in the Amhara [16]. It is therefore imperative that the methods used for
region. The meteorological stations were clustered by transforming the results of climate models to meteoro-
hierarchical clustering method with quantitative data set logical stations will generate precipitation rainfall and
and Squared Euclidean Distance with wards methods temperature series that should have the same statistical
using StatistiXL software. properties as observed meteorological data for use in
climate modeling [30]. The years from 1/1/1979 to 31/
3.2. General Circulation Model Data (GCM) 12/2008 were used as a baseline period for the study [5].
The GCM data were collected from the IPCC Data Thus, the HadCM3 data were downscaled for the base-
Distribution Center (DDC) which was established in line period A2 emission scenarios and the statistical
1998 by the World Metrological Organization and the properties (mean and variance) of the downscaled data
United Nations Environmental Program, following a were compared with observed data of the same period.
recommendation by the task group on Data and Scenario
Support for Impact and Climate Assessment [28]. How- 4. METHODS OF CLUSTERING
ever, because the daily fields are only available directly ANALYSIS AND HOMOGENITY TEST
from the respective modeling centers, the climatic data
4.1. Clustering Analysis
used for SDSM had been collected from the Canadian
Institute for climate studies website for model output of Clustering (or numerical classification) uses a data array,
HadCM3 [29]. The predictor variables were supplied on or matrix, to classify (or cluster) the cases (rows) of data
a grid basis of size 2.5˚ latitude × 3.75˚ longitude so that into groups according to values of their attributes (co-
the data were downloaded from the nearest grid box to lumns). Geographical location (latitude, longitude and
the study area. elevation) and climatologically mean monthly rainfall
As shown in Figure 2, Hence the nearest grid box for (1979-2008) of all stations were used to calculate their
the HadCM3 model, which represents the study area, is standardized standard deviation. Hence using Hierarchi-
the one at 10.95˚N latitude and 37.85˚E longitude (Y = cal Clustering method the stations were grouped; there
31 Latitude 10˚N and X = 12 Longitude 41.25˚E). was no prior knowledge of grouping (or any prior group-
ing was ignored). The data were separated data into
3.3. Baseline Period Data groups based on individual cases using clustering analy
One of the criteria used in evaluation of the perfor- sis [31]. Therefore, the data was summarized and stations
mance of any useful downscaling method is whether the were grouped using tree graph (dendrogram) (Figure 3).
Figure 2. The African continent window with 2.5˚ latitude × 3.75˚ longitude grid size from which the grid box for the
study area is selected (shown in red/or back if printed in black and white).
Motta
Gondar
Debre Markose
Bahir dar
Debre Tabore
Debark
Metema
Sirinka
Kombolcha
Mahil Meda
stations in each cluster have similar rainfall seasons or fourth root transformation was used as its data were
not. Moreover, first order Markov chain model clearly skewed and as its model was conditional. The range of
gave the probability of getting rain on a daily basis and variation of the downscaled daily weather parameters
the chance of getting rain in each season [32]. This was was controlled by fixing the variance inflation. The de-
used to identify the dry and rainy seasons of each cluster fault value, which is 12, was used for the daily tempera-
in the study area. ture values; where as for daily precipitation this value
The zero order Markov chain model was used for was set as 18, in order to magnify the variation. The bias
testing the homogeneity while first order was used to fill correction, which compensates for any tendency to over
missing daily rainfall values. The main reason to select or underestimate the mean of conditional processes by
the first order to fill the missing data rather than second the downscaling model, was set to be 0.8 for daily rain-
order was that first order didn’t exaggerate the resulting fall, and the default 1.0 for daily temperature (indicating
values and gave more accurate model to each cluster of no bias correction).
the study area as explained by the National Meteoro-
logical Agency of Ethiopia [33]. 4.6. Steps Used in SDSM Model Approach
4.4. Settings of the SDSM Data were checked for their quality; gross data errors,
missing data codes and outliers were identified prior to
According to [16], the concept of regional climate being model calibration. Predictors and/or the predictand were
conditioned by the large-scale state may be written as: also transformed. Then, empirical relationships between
R = F (L) gridded predictors and single site predictands (station
where, precipitation) were identified; the selection was done at
R represents the predicted (regional or local climate most care as the behavior of the climate scenario com-
variables), pletely depends on the type of the predictors selected.
L represents the predictor (a set of large-scale climate After this, models were calibrated and results for the
variables) and, different time slices were developed. Calibrated models
F is a deterministic/Stochastic function conditioned by (using independent data) and the synthesis of artificial
L and has to be found empirically from observed or mod- time series data representing current climate conditions
eled data sets. were verified. Then both derived SDSM scenarios and
The climate scenario for future period were developed observed climate data were interrogated with the analy-
from statistical downscaling using GCM HadCM3 pre- ses data screen of SDSM and monthly statistics produced
dictor variable for the emission scenarios for 100 years were plotted using the compare results screen. Then, rapid
based on the mean of 20 ensembles; and the analyses assessment of downscaled versus observed, and/or cur-
was done based on three 30-years periods centered on the
rent versus future climate scenarios were compared.
2020s (2011-2040), 2050s (2041-2070) and 2080s (2071-
Finally, the outputs of the Hadley Centre Coupled
2099).
Model (HadCM3) with A2a scenario, which is the most
commonly used GCM for downscaling future climate of
4.5. Settings Used for SDSM
an area (e.g. [23]), were used to develop climate change
For the observed and the National Centre for Envi- scenarios in the region.
ronmental Prediction (NCEP) data the year length was According to [16], the regression weights produced
set to be the default (366 days), which allows 29 days in during the calibration process were applied to the time
February in leap years. However, as HadCM3 have mod- series outputs of the GCM model. This is based on the
elled years that do only consist of 360 days, the default assumption that the predictor-predictand relationships
value was changed to 360 days. The base period used for under the current condition remain valid under future cli-
the model was from 1/1/1979 to 31/12/ 2008. mate conditions too. Twenty ensembles of synthetic daily
The event threshold value is important to treat trace time series data were produced for each of the two SRES
values during the calibration period. For the temperature scenarios for a period of 139 years (1961 to 2099). The
parameter, this value was set to be 0 while for daily rain- final product of the SDSM downscaling method was then
fall calibration purpose this parameter was fixed to be found by averaging the twenty independent stochastic
0.1 mm/day so that trace rain days below this thresh-old GCM ensembles of A2a and B2a emissions scenarios.
value were considered as a dry day. Missing data were
replaced by –999. 5. RESULTS AND DISCUSSIONS
For the daily temperature values, no transformation
5.1. Clustering Analysis
was used as it is normally distributed and the model was
non-conditional. However, for the daily rainfall, the A dendrogram (tree graph) is a common means of gra-
phically summarizing a clustering pattern using Statisti XL. resent climatologically homogeneous areas within the
The dendrogram usually starts with all individuals as study area (Figure 3). The geographical locations of the
separate clusters (“tips”) and shows the combination of stations grouped and the representative clusters used for
fusions back to a single “root”. The order of individu-als the study are given in Figure 4. Based on the clusters,
shown in the dendrogram was the order in which the climatologically the Amhara region has a mean maxi-
groups enter the clustering [31]. In this study, the maxi- mum temperature of 26.0˚C, 35.8˚C and 23.6˚C for Clus-
mum distance or similarity to classify as one group is 60 ter A, B and C, respectively, but inter-annual variability
Figure 3. of rainfall was observed in all clusters (Figures 5 and 6).
All available stations, which were 10 before clustering, Based on the maximum and minimum temperature
were clustered and classified in to three groups and each data, the mean monthly temperature over the Amhara
group was designated as Cluster/Region A (Kombolcha, region tend to increase until the beginning of June in all
Mahil-Meda and Sirinka), Cluster/Region B (Metema) regions and decreased during the main rainy season
and Cluster/Region C (Motta, Gondar, Debre Markose, ((June to September) (Figure 6)). Contrarily, Tmin had
Debre Tabore and Debark) randomly. greater value during the main rainy season, and the mean
The designation of each cluster was done just to rep- minimum temperature over Amhara region was 11.6˚C,
Figure 5. Variability of rainfall in clustered groups of Amhara Regional State for the period 1979-2008.
Figure 6. Monthly average temperature in clustered groups of Amhara region for the pe-
riod 1979-2008. Letters: A, B, C and a, b, c represent respective clusters for maximum
and minimum temperatures.
19.3˚C and 10.0˚C for Clusters A, B and C, respectively. 5.4. Predictor Variables Selected
The type and explanations of the predictors, which
5.2. Homogeneity of Stations
showed better correlation with the daily maximum tem-
Homogeneity test analysis showed that Cluster C and perature, daily minimum temperature, and daily rainfall
Cluster B had only one main rainy season. The rainfall in predictands at p < 0.05 significance level are shown in
Cluster C is long with early onset and late end date of the Table 2. Partial correlations indicate that the correspond-
rain, while in Cluster B rainfall is characterized by late ing predictor had the strongest association with the pre-
onset and early cessation of rain (Figure 7). dictand.
Cluster A has two rainy seasons which are locally
known as “belg” (small rainy season) and “kiremit” 5.5. Calibration and Validation
(main rainy season). The two rainy seasons have no dis-
The simulated maximum and minimum temperatures
tinct dry period on their translation; hence, the small
in Clusters A and C had better agreement with the ob-
rainy season is soon after followed by the main rainy
served results than rainfall/precipitation (Table 3). Though
season. In addition, the length of main rainy season is
simulated values of rainfall showed a lesser agreement as
similar to Cluster C.
compared to the maximum temperature and minimum
temperature in all clusters, the result was quite accept-
5.3. Selection Representative Stations
able due to the fact that precipitation is a conditional
Partial correlations of each station belonging to each process [14]. Conditional processes like precipitation are
group were analyzed. In each group, stations with better dependent on other intermediate processes like the oc-
correlation coefficient were selected (highlighted with currence of humidity, cloud cover, and/or wet-days. Un-
grey/black if it is printed in black and white in Table 1). conditional processes like temperature; however, are not
Kombolcha and Motta meteorological stations had good regulated by other intermediate processes.
correlation in their respective Cluster (Table 1). While, In addition, as indicated in the SDSM manual [14], lo-
Metema meteorological stations didn’t correlate with cal temperatures are largely determined by regional forc-
other observatory stations, hence it stood alone as Cluster ing whereas precipitation series display more “noise”
B. Henceforth, only these stations were used for further arising from local factors. Hence, larger differences can
analysis. be observed in precipitation ensemble members than that
These stations were taken as representative of the cli- of temperature.
mate stations with respect to average monthly rainfall On the other hand, the minimum and maximum tem-
(priority was given for rainfall than temperature). There- peratures over Cluster B simulation showed a very poor
fore, temperature data was taken from the selected rep- agreement with the observed one (Table 3). Even though
resentative station out of respective Cluster of a group. it is difficult to exactly point out the reason behind, this
Consequently, the climate scenarios were only developed could be due to low/inferior data quality used for the
for these selected stations. comparison.
Validation was done based on 15 year simulation from The correlation factors that were found during the
1993 to 2008. The validation statistics for maximum and calibration step are more or less maintained during the
minimum temperatures and rainfall in all clusters are validation period too, even better agreement was found
shown in Table 4. here.
Figure 7. Homogeneity test using Markov chain model (zero order) for three clusters in the Amhara Regional State, Ethiopia.
Table 1. Correlation matrix of stations average monthly rainfall in three clusters the Amhara Regional State, Ethiopia.
Cluster A 1 2 3
Cluster B 1 2 3
Cluster C 1 2 3 4 5 6
Table 2. List of predictor variables that gave better correlation results at p < 0.05.
Table 3. Calibration statistics of daily rainfall and maximum temperature, and minimum temperature for each region.
R2 Standard Error
Cluster Predictand
Unconditional Conditional Unconditional Conditional
Table 4. Validation statistics of daily rainfall and maximum temperature, and minimum temperature for each region.
R2 Standard Error
Cluster Predictand
Unconditional Conditional Unconditional Conditional
Table 5. Downscaled maximum and minimum temperatures and rainfall in three clustered regions of Amhara Regional State, Ethio-
pia.
Table 6. Summered statistics of maximum temperature in three clustered regions of Amhara regional State, Ethiopia.
Statistics
Cluster Year
Mean Std Maximum Minimum
A 1979-2008 26.44a 2.23 33.20 17.70
A 2020’s 26.71b 2.58 35.54 14.81
A 2050’s 27.57c 2.55 35.67 16.38
A 2080’s 28.50d 2.65 37.83 19.97
abcd; significant at 0.001 level
B 1979-2008 36.91a 2.66 47.85 27.81
B 2020’s 38.71b 2.74 49.46 28.23
B 2050’s 40.08c 2.85 51.22 31.97
B 2080’s 42.98d 3.01 55.06 35.95
abcd; significant at 0.001 level
C 1979-2008 23.78a 1.41 28.01 17.35
C 2020’s 24.00b 1.46 28.38 15.87
C 2050’s 24.60c 1.48 28.99 15.53
C 2080’s 25.33d 1.53 29.51 18.10
abcd; significant at 0.001 level
a, b, c, d; means with different superscripts within cluster groups and in a column are significant at * p < 0.05; ** p< 0.01; *** p < 0.001.
global temperature of up to 5.8˚C by the end of the 21st 5.7.2. Daily Minimum Temperature
century. The possible reason for rise in temperature ac- The present study revealed that the average daily
cording to [35] may be associated with the dramatic minimum temperature in Amhara region generally was
warming of oceans over the Atlantic and /or Indian increasing. For example, a rise in mean daily minimum
oceans which further warm the sea surface temperature temperature of about 2.80˚C, 0.11˚C and 0.64˚C are
in the two oceans and result in low pressure formation in noted in Clusters A, B and C by 2080s, respectively (Ta-
the area and decrease in moist air advection towards ble 7). This finding is in agreement with the historical
Ethiopia. climate of the country. According to [36], annual mini-
This might be the case for the increase in number of mum temperature has been increased by about 0.37 ˚C
days with Tmax > 28˚C, Tmax > 30˚C and Tmax > 32˚C every 10 years over the past 55 years. Although, the re-
in all clusters by 2080’s (Figure 8). Further, [5] reported sult in Clusters B and C showed less change in mean
that climate change may increase the frequency of ENSO minimum temperature than cluster A from the base pe-
warm phases in Africa. The present study also showed riod, it indicates a shift towards warmer conditions in the
the more likely occurrence of heat wave in all clusters future. In general, the projected climate in the three clus-
for the projected years than before. The projected climate ters indicated a decreas in the number of cold days (Fig-
in Cluster A showed that heat wave was more likely to ure 9)
occur starting from 2020’s, and it would be a common Days with minimum temperature less than 7˚C are
weather event for Cluster B. However, Cluster C will projected to become less common, in all clusters in the
stay safe and experiences wave free days even by 2080’s future climate scenario (Figure 9). This is in line with [5]
(Figure 8). and [37] who reported increasing trend in minimum
Table 7. Summered statistics for minimum temperature in three clusters regions in Amhara Regional State.
Statistics for Tmin over Amhara region
Cluster Year
Mean Std Maximum Minimum
A 1979-2008 12.32a 2.13 16.71 4.11
A 2020’S 12.84b 2.24 17.39 2.37
A 2050’S 13.92c 2.22 18.57 3.47
A 2080’S 15.13d 2.26 20.42 7.32
abcd; significant at 0.001 level
B 1979-2008 19.00 2.70 30.3 11.03
B 2020’S 19.00 1.42 23.84 12.68
B 2050’S 18.97 1.43 24.54 12.90
B 2080’S 19.11 1.45 24.59 12.07
NS
C 1979-2008 9.90a 2.36 14.27 3.64
C 2020’S 9.97b 2.41 14.54 4.21
C 2050’S 10.27c 2.40 14.49 4.25
C 2080’S 10.54d 2.38 14.96 4.42
abcd; significant at 0.001 level
Figure 8. Projected number of hot days per year in in three clusters re-
gions in Amhara Regional State.
Figure 9. Projected number of cold days per year in three clusters regions in Amhara Regional State.
temperature over Ethiopia with decreasing the number of This decrease in positive anomaly and increase in ne-
days with minimum temperature in the future. gative anomaly lead to an extended periods of below
average rainfall in the study areas. Rainfall variability in
5.7.3. Rainfall Variability and Extremesin general and decrease in positive anomalies in particular
Future Daily Pcpn is in accordance with the decrease in projected annual
rainfall by 27.2%, 42.3% and 12.2% in 2080s in cluster
Rainfall is not usually well approximated by normal A, B and C, respectively (Table 5). [36] also indicated
distributions [5].The projected scenario in the current that Ethiopia’s average annual rainfall has shown a very
study in all clusters in he study region showed that there high level of variability. Further, a study to analyze far-
was a decrease in the number of positive anomalies and mers’ perceptions of climate change in the Nile Basin of
an increased of negative anomalies compared with the Ethiopia by [38] indicated that temperature has increased
base period (1979-2008). In the base period climate, and the level of rainfall has declined. Furthermore, [5,
there were 13, 15 and 17 positive anomalies and 15, 15 39,40], add that the climate change induced warming of
and 13 negative anomalies in Clusters A, B and C, re- the Indian Ocean is likely to lead to persistent droughts
spectivelly As compared to 12, 10 and 13 “positive” in east Africa in the coming years; hence the monsoon
anomalies and 14, 18 and 17 negative anomalies respec- winds that bring seasonal rain to sub-Saharan African
tively in the 2080s (Figure 10). could be 10% - 20% drier than the 1950-2000 averages.
Figure 10. Annual rainfall anomaly for the period (1979-2008), 2020s, 2050s
and 2080s in three cluster regions.
Similarly, the finding in this study showed the more quency of drought occurrence in the past few dacades in
likely occurrence of persistent drought, (anomalies of ≥ Ethiopia.
−0.84. It means that the number of rainy days per year in Although the annual rainfall amount decreases (Figure
future climate will decreased dramatically (Figure 10); 12), the number of rainy days in the projected time peri-
which could shrink household farm production by up to ods at Cluster B did not decrease in equal magnitude as
90 percent of a normal year output [41]. Clusters A and C (Figure 11).
As shown in Table 5, there was high decrease in per-
centage change of rainfall in all clustered regions. This 6. CONCLUSIONS AND
could be due to a dramatic decrease in the number of RECOMMENDATIONS
rainy days in the 2080s as compared to the base period
6.1. Conclusions
(Figures 11 and 12). This finding is in agreement with
the finding of [42] who noted an increase in the fre- Downscaled temperature results for the Amhara region
Figure 11. Projected number of rainy days per year in three cluster re-
gions in the Amhara Regional State, Ethiopia.
1250
1000
750
500
y(A) = –5.7875x + 1217 y(B)= –5.2924x + 1121 y(C) = –4.3786x + 1431.8
250 R2 = 0.5547
R2 = 0.7114 R2 = 0.7776
0
1961 1971 1981 1991 2001 2011 2021 2031 2041 2051 2061 2071 2081 2091
Projected years
Figure 12. Future trends of rainfall in Clusters A, B and C in three cluster regions in the Am-
hara Regional State, Ethiopia.
using both A2a emission scenario showed increasing [2] UNEP (United Nations Environment Programme) and
trend for the three tri-decadal periods centered on the CMS (Convention on the Conservation of Migratory
Species of Wild Animals) (2006) Migratory species and
2020s, 2050s, and the 2080s. Maximum temperature
climate change: Impacts of a changing environment on
could rise up by 2.3˚C, 4.6˚C and 1.80˚C, while mini- wild animals UNEP/CMS Secretariat, Bonn, Germany.
mum temperature increased by 3.6˚C, 0.3˚C and 0.50˚C
[3] The nature conservancy website.
in Clusters A, B and C, respectively in 2080s. Unlike http://www.nature.org/climatechange/
temperature, rainfall results showed decreasing trend in
[4] World Wide Fund for Nature (2006) Climate change im-
all projected years. A percentage decrease in rainfall of pacts on East Africa: A review of the scientific literature,
about 27.2%, 42.3% and 12.2% was noted in Clusters A, Gland. WWF, Morges.
B and C, respectively in the 2080s. [5] IPCC Technical Summary (2001) Climate change 2001:
Generally, the effects of global warming due to green- The dcientific basis. In: Houghton, J.T., Ding, Y., Griggs,
house gas emissions and sea surface warming is reflected D.J., Noguer, M., van der Linden, P.J., Dai, X., Maskell,
in the study area with an increase in projected tempera- K. and Johnson C.A., Eds., Technical Summary of the
tures and a decrease in rainfall which resulted in in- Working Group I Report, Cambridge University Press,
creased number of projected hot days, fewer number of Cambridge, 94.
projected cold days and decreased number of projected [6] Biodiversity Strategy and Action plan (2005) Biodiversity
rainydays. These events may lead to increased occur- Strategy and Action plan. BSAP. Final document.
rence of drought and water shortage over the Amhara [7] Davidson, O., Halsnaes, K., Huq, S., Kok, M., Metz, B.,
Regional State. Sokona, Y. and Verhagen, J. (2003) The development and
climate nexus: The case of sub-Saharan Africa. Climate
Policy, 3, S97-S113. doi:10.1016/j.clipol.2003.10.007
6.2. Recommendations
[8] Zhou, G., Minakawa, N., Githeko, A.K. and Yan G. (2004)
Generally, based on the results of this study the fol- Association between climate variability and malaria epi-
lowing recommendations are forwarded: The projected demics in the East African highlands. Proceedings of the
changes in temperature and rainfall in the Amhara region National Academy of Sciences of the United States of
America, 101, 2375-2380.
urges decision makers to incorporate climate change
scenarios in devising sustainable regional coping strate- [9] Craig, M.H., Kleinschmidt, I., Nawn, J.B., Le Sueur D.
and Sharp, B.L. (2004) Exploring 30 years of malaria
gies, including: water harvesting technologies, supple-
case data in KwaZulu-Natal, South Africa: Part I. The
mentary irrigation, improved seeds which can tolerate impact of climatic factors. Tropical Medicine and Inter-
moisture and temperature stresses, afforestations and soil national Health, 9, 1247-1257.
conservation techniques, preventing free grazing and re- [10] Wood, A. (1977) A preliminary chronology of Ethiopian
lated strategies and measures. droughts. In: Dalby, D., Church, R.J.H. and Bezzaz, F.,
The result of any model depends on the quality of the Eds., Drought in Africa, Vol. 2, International African In-
input data; therefore, to make the evaluation of climate stitute, London, 68-73.
change studies more complete, it is recommended to con- [11] Pankhurst, R. and Johnson, D.H. (1988) The great drought
duct similar studies with large number of meteorological and famine of 1888-92 in northeast Africa. In: Johnson,
stations and GCM models. D.H. and Anderson, D.M., Eds., The Ecology of Survival:
Case Studies from Northeast African History, Lester
Crook Academic Publishing, London, 47-72.
7. ACKNOWLEDGEMENTS
[12] NOAA (The National Oceanic and Atmospheric Admini-
We express our sincere thanks to the Ministry of Education and stration) research website.
Haramaya University for their financial support to carry out this work. http://www.research.noaa.gov/climate/t_prediction.html
We also express our sincere thanks to Bahir Dar University for allow- [13] Australian Bureau of Meteorology.
ing the first author to pursue his PhD study. Equally important was the http://www.bom.gov.au/analclim.htm
support given by the Swiss NCCR North-South program and linking [14] Wilby, R.L. and Dawson, C.W. (2007) SDSM—A deci-
this particular research with the global climate implications.The Na- sion support tool for the assessment of regional climate
tional Meteorological Agency is also acknowledged for providing most change impacts. Environmental Modeling Software, 17,
of the daily rainfall and temperature data used for the study. 145-157. doi:10.1016/S1364-8152(01)00060-3
[15] Wilby, R.L., Dawson, C.W. and Barrow, E.M. (2002)
Statistical downscaling model SDSM, version 4.1. De-
REFERENCES partment of Geography, Lancaster University, Lancaster.
[1] Gitay, H., Suarez, A., Watson, R.T. and Dokken, D.J. [16] Wilby, R.L. and Dawson, C.W. (2004) Using SDSM ver-
(2002) Climate change and biodiversity. IPCC (Inter- sion 3.1—A decision support tool for the assessment of
governmental Panel on Climate Change) technical paper regional climate change impacts. User Manual, Leicester.
V, 86. http://www.ipcc.ch [17] Dibike, Y.B. and Coulibaly, P. (2005) Downscaling pre-
cipitation and temperature with temporal neural networks. IPCC Data Distribution Center.
Department of Civil Engineering, and School of Geogra- http://www.ipcc-data.org/
phy and Geology, McMaster University, Hamilton. [29] Canadian Institute for Climate Studies (CICS).
[18] Kilsby, C.G., Jones, P.D., et al. (2007) A daily weather http://www.cics.uvic.ca/scenarios/sdsm/select.cgi
generator for use in climate change studies. Environ- [30] Houghton, J.T., (2001) Climate change the scientific basis:
mental Modelling and Software, 22, 1705-1719. Contribution of working group I to the third assessment
doi:10.1016/j.envsoft.2007.02.005 report of the intergovernmental panel on climate change.
[19] Pryor, S.C., Schoof, J.T. and Barthelmie, R.J. (2005) Em- Cambridge University Press, Cambridge.
pirical downscaling of wind speed probability distribu- [31] Robert A. and Wither P. (2007) StatistiXL, version 1.8, a
tions. Journal of Geophysical Research, 110, Article ID: powerful statistics and statistical analysis add-in for Mi-
D19109. doi:10.1029/2005JD005899 crosoft Excel. Microsoft, Washington DC.
[20] Gordon, C., Cooper, C., Senior, C.A., Banks, H.T., Greg- [32] Stern, R., Knock, J., Rijks, D. and Dale, I. (2002) Instat+
ory, J.M., Johns, T.C., Mitchell, J.F.B. and Wood, A. (2000) (interactive statistics package). Statistics services center.
The simulation of SST, sea ice extents and ocean heat University of Reading, Reading.
transports in a version of the Hadley Centre coupled model
without flux adjustments. Climate Dynamics, 16, 147-168. [33] National Meteorological Services Agency (1985) Climate
doi:10.1007/s003820050010 and agro climatic resources of Ethiopia. Addis Ababa.
[21] Pope, V.D., Gallani, M.L., Rowntree, P.R. and Stratton, [34] Francis, D. and Hengeveld, H. (1998) Extreme Weather
R.A. (2000) The impact of new physical parameterizations and Climate Change.
in the Hadley Centre climate model—HadAM3. Climate http://www.mscsmc.ec.gc.ca/education/scienceofclimatec
Dynamics, 16, 123-146. doi:10.1007/s003820050009 hange/understanding/ccd/ccd_9801/sections/1_e.html
[22] Reichler, T. and Kim, J. (2008) How well do coupled [35] NCAR (National Centre for Atmospheric Research)
models simulate today’s climate? Bulletin of the Ameri- (2005) A continent split by climate change: New study
can Meteorological Society, 89, 303-311. projects drought in southern Africa. Boulder Co., Nairobi.
doi:10.1175/BAMS-89-3-303 [36] National Meteorological Services (2007) Climate change
[23] Souvignet, M. and Heinrich J. (2008) Future temperatures national adaptation program of action (NAPA) of Ethio-
and precipitations in the arid northern-central Chile: A pia. NMS, Addis Ababa.
multi-model downscaling approach. 6th Alexander von [37] National Meteorological Services Agency (2001) Initial
Humboldt International Conference, 24 June 2010. national communication of Ethiopia to the United Nations
[24] Zeray, L., Roehrig, J. and Alamirew, D. (2006) Climate framework convention on climate change (UNFCCC).
change impact on Lake Ziway watershed water availabil- NMSA, Addis Ababa.
ity, Ethiopia. Conference on International Agricultural [38] Temesgen, D. (2007) Measuring the economic impact of
Research for Development, Tropentag, 11-13 October climate change on Ethiopian agriculture: Ricardian ap-
2006. proach. World Bank policy research paper No. 4342.
[25] Tachieobeng, E., Gyasi, E., Abekoe, S.A.M. and Ziervo- World Bank, Washington DC.
gel, G. (2010) Farmers’ adaptation measures in scenarios [39] Hulme, M., Doherty, R., Ngara, T., New, M. and Lister, D.
of climate change for maize production in semi-arid (2001) Taking action against climate change in Ethiopia
zones of Ghana. 2nd International Conference on Climate, and South Africa.
Sustainability and Development in Semi-Arid Regions, [40] Funk, C., Senay, G., Asfaw, A., Verdin, J., Rowland, J.,
Fortaleza, 16-20 August 2010. Michaelson, J., Eilerts, G., Korecha, D. and Choularton, R.
[26] Central Statistics Authority (2005) The federal democ- (2005) Recent drought tendencies in Ethiopia and equato-
ratic Republic of Ethiopia statistical abstract. CSA, Addis rial-subtropical eastern Africa. FEWS-NET, Washington
Ababa. DC.
[27] Federal Democratic Republic of Ethiopia (1997) The con- [41] World Bank (2003) Risk and vulnerability assessment
servation strategy of Ethiopia: The resources base, its draft report. World Bank, Addis Ababa.
utilization and planning for sustainability, vol. I. Envi- [42] Ketema, T. (1999) Test of homogeneity, frequency analy-
ronmental protection authority in collaboration with the sis of rainfall data and estimate of drought probabilities in
ministry of economic development and cooperation. FDRE, dire Dawa, eastern Ethiopia. Ethiopian Journal of Natu-
Addis Ababa. ral Resources, 1, 125-136.
[28] Intergovernmental Panel on Climate Change (2011) The