Climate change is becoming one of the most threatening issues for
the world today in terms of its global context and its response to
environmental and socioeconomic drivers. However, large
uncertainties between different general circulation models (GCMs) and coarse
spatial resolutions make it difficult to use the outputs of GCMs directly,
especially for sustainable water management at regional scale, which
introduces the need for downscaling techniques using a multimodel approach.
This study aims (i) to evaluate the comparative performance of two widely
used statistical downscaling techniques, namely the Long Ashton Research
Station Weather Generator (LARS-WG) and the Statistical Downscaling Model
(SDSM), and (ii) to downscale future climate scenarios of precipitation,
maximum temperature (
Six selected multimodel CMIP3 GCMs, namely HadCM3, GFDL-CM2.1, ECHAM5-OM,
CCSM3, MRI-CGCM2.3.2 and CSIRO-MK3 GCMs, were used for downscaling climate
scenarios by the LARS-WG model. The result from the ensemble mean of the six GCM
showed an increasing trend for precipitation,
The impacts of climate change on the hydrological cycle in general and on water resources in particular are of high significance due to the fact that all natural and socioeconomic systems critically depend on water. The direct impact of climate change can be variation and changing pattern of water resources availability and hydrological extreme events such as floods and droughts, with many indirect effects on agriculture, food and energy production and overall water infrastructure (Ebrahim et al., 2013). The impact may be worse on transboundary rivers like the Upper Blue Nile River where competition for water is becoming high from different economic, political and social interests of the riparian countries and when runoff variability of upstream countries can greatly affect the downstream countries (Kim, 2008; Semenov and Barrow, 1997).
According to IPCC (2007), between 75 and 250 million people are projected to be exposed to increased water stress due to climate change in Africa by 2020. The increasing water demand of upstream countries in the Nile Basin coupled with climate change impacts can affect the availability of water resources for downstream countries and in the basin, which could result in resource conflicts and regional insecurities. Moreover, climate variability (i.e., the way climate fluctuates yearly and seasonally above or below a long-term average value, caused by changes in forcing factors such as variation in seasonal extent of the intertropical convergence zone like El Niño and La Niña events) is already imposing a significant challenge to Ethiopia by affecting food security, water and energy supply, poverty reduction, and sustainable socioeconomic development efforts. To mitigate these challenges, the Ethiopian government therefore carried out a series of studies on Upper Blue Nile River basin (UBNRB), which has been identified as an economic “growth corridor”, focused on identifying irrigation and hydropower potential of the basin (BCEOM, 1998; USBR, 1964; WAPCOS, 1990). As a result, large-scale irrigation and hydropower projects including the Grand Ethiopian Renaissance Dam (GERD), the largest hydroelectric power plant in Africa, have been identified and are being constructed as mitigation measures for the impacts of climate change. However, most studies have put less emphasis on climate change and its impact on the hydrology of the basin, and hence identifying local impacts of climate change at basin level is quite important. This is especially important in the UBNRB for the sustainability of large-scale water resource development projects, for proper water resource management leading to regional security and for finding possible mitigation measures to avoid catastrophic consequences.
To this end, several individual studies have been done to study the impacts
of climate change on the water resources of Upper Blue Nile River basin. Taye
et al. (2011) reviewed some of the research outputs and concluded that clear
discrepancies were observed, particularly on the projection of precipitation.
For instance, as the results obtained from Bewket and Conway (2007), Conway
(2000) and Gebremicael et al. (2013) showed, there is no significant trend
observed in the amount of seasonal and annual rainfall, while Mengistu et al.
(2014) reported statistically nonsignificant increasing trends in annual and
seasonal rainfall. For the future projection, expected changes in
precipitation amount are unclear. For instance, Kim (2008) used the outputs
of six GCMs for the projection of future precipitations and temperature, and
the result suggested that the changes in mean annual precipitation from the
six GCMs range from
For the historical context, the discrepancies could be due to the period and length of data analyzed and the failure to consider stations which can represent the spatial variability of the basin and also errors induced from observed data. For the future context, besides the abovementioned reasons, discrepancies could be due to the differences in GCMs and scenarios used for downscaling, the downscaling techniques applied (can be dynamical and statistical), selection of representative predictors, the period of analysis, and spatial and temporal resolution of observed and predictor data sets.
To address uncertainty in projected climate changes, the IPCC (2014) thus recommends using a large ensemble of climate change scenarios produced from various combinations of atmospheric ocean general circulation models (AOGCMs) and forcing scenarios. However, it can become prohibitively time consuming to assess the climate change using many climate change scenarios and many statistical downscaling models simultaneously. As a result, researchers typically assess climate change and its impacts under only one or a few climate change scenarios, selected arbitrarily with no justification, for instance those that used only A1B and A2 scenarios. Yet there is no any hard rule to select an appropriate subset of climate change scenarios among the wide range of possibilities (Casajus et al., 2016).
Location map of the study area.
GCMs perform reasonably well at larger spatial scales but poorly at finer spatial and temporal scales, especially precipitation, which is of interest in hydrological impact analysis (Goly et al., 2014). Hence, the processes of downscaling that ensure that the scale discrepancy between the coarse-scale GCMs and the required local-scale climate variables for hydrological models will be narrowed down should be investigated for their contribution, which has often been overlooked in previous studies on climate change analysis in the UBNRB. Many researchers have tried to compare the comparative skill of downscaling methods in different study areas (Dibike and Coulibaly, 2005; Ebrahim et al., 2013; Fiseha et al., 2012; Goodarzi et al., 2015; Hashmi et al., 2011; Khan et al., 2006; Qian et al., 2004; Wilby et al., 2004; Wilby and Wigley, 1997; Xu, 1999). However, no single model has been found to perform well over all the regions and timescales. Thus, evaluations of different models is critical to understanding the applicability of the existing models.
Apart from the GCMs and downscaling techniques, most of the previous studies
(e.g., Beyene et al. (2010), Elshamy et al. (2009) and Kim (2008)) used CRU,
NFS and other gridded data sets constructed based on the interpolation of
a few stations in Ethiopia, which are relatively less accurate compared
with the station-based data (Worqlul et al., 2014). Therefore, the objective
of this study is (i) to evaluate the comparative performance of two widely
used statistical downscaling techniques, namely the Long Ashton Research Station
Weather Generator (LARS-WG) and the Statistical Downscaling Model (SDSM) over
the UBNRB, and (ii) to downscale future climate scenarios of precipitation, maximum
temperature (
Generally, downscaling methods are classified into dynamic and statistical downscaling (Fowler et al., 2007; Wilby et al., 2002). Dynamic downscaling nests higher-resolution regional climate models (RCMs) into coarse-resolution GCMs to produce complete set of meteorological variables which are consistent with each other. The outputs from this method are still not at the scale that the hydrological model requires. Statistical downscaling overcomes this challenge; moreover it is computationally undemanding, simple to apply and provides the possibility of uncertainty analysis (Dibike et al., 2005; Semenov et al., 1997; Wilby et al., 2002). Extensive details on the strength and weakness of the two methods can be found in Wilby et al. (2007, 1997). Among the different possibilities, two well-recognized statistical downscaling tools, a regression-based SDSM (Wilby et al., 2002) and a stochastic weather generator, LARS-WG (Semenov et al., 1997, 2002) were chosen for this study. They have been tested in various regions (e.g., Chen et al., 2013; Dibike et al., 2005; Dile et al., 2013; Elshamy et al., 2009; Fiseha et al., 2012; Hashmi et al., 2011; Hassan et al., 2014; Maurer and Hidalgo, 2008; Yimer et al., 2009) under different climatic conditions of the world.
The Upper Blue Nile River basin extends from 7
Selected global climate models from IPCC AR4 incorporated into LARS-WG.
B: baseline; T1: 2011–2030; T2: 2046–2065; T3: 2081–2100
The Upper Blue Nile River itself has an average annual runoff of about
49 billion m
The historical precipitation, maximum temperature and minimum temperature data for the study area were obtained from the Ethiopian Meteorological Agency (EMA), which were analyzed and checked for further quality control. A considerable length of time series data were missed in almost all available stations, and hence 15 rainfall and 25 temperature stations which have long time series and relatively short time missing records were selected. Filling missed or gap records was the first task for further meteorological data analysis. This task was done using the well-known methodology of the inverse distance weighting method. To check the quality of the data, the double mass curve (DMC) analysis was used. DMC analysis is a cross correlation between the accumulated totals of the gauge in question against the corresponding totals for a representative group of nearby gauges.
A new version of the LARS-WG5.5 was applied for this study that incorporates predictions from 15 GCMs which were used in the IPCC's Fourth Assessment Report (AR4) based on Special Emissions Scenarios SRES B1, A1B and A2 for three time windows as listed in Table 1. However, the fifth phase of Coupled Model Intercomparison Project (CMIP5) climate models, based on the new radiative forcing scenarios (Representative Concentration Pathway, RCP) which were used for IPCC Fifth Assessment Report (AR5), were not incorporated into LARS-WG at the time of the study.
As it is difficult to process all the incorporated 15 CMIP3 GCMs and large differences in predictions of climate variables among the
GCMs are expected, the performance of GCMs in simulating the current climate variables of
the study area (UBNRB) should be evaluated, and the best-performing GCMs were
selected. The MAGICC/SCEGEN computer program tool was used for the
performance evaluation of the 15 GCMs found in the LARS WG5.5 database, as it is
a standard method for selecting models on the basis of their ability to
accurately represent current climate, either for a particular region or
for the globe. In this study, we used a semiquantitative skill score that
rewards relatively good models and penalizes relatively bad models as
suggested by the Wigley (2008) user manual. The statistics used for model
selection are pattern correlation (
Name and description of all NCEP predictors on HadCM3 and canESM2 grid.
Moreover, atmospheric large-scale predictor variables used for representing the present condition were obtained from the National Centre for Environmental Prediction (NCEP) reanalysis data set. CanESM2, a second-generation Canadian earth system model (ESM) developed by Canadian Centre for Climate Modelling and Analysis (CCCma) of Environment Canada that represents CMIP5 and HadCM3 outputs from the Hadley Centre, United Kingdom (UK) representing CMIP3 were used in SDSM for the construction of daily local meteorological variables corresponding to their future climate scenario.
The reason for selecting these two GCMs was that they are models that made
daily predictor variables freely available to be directly fed into the SDSM,
covering the study area with a better resolution. Additionally, HadCM3 is the
most-used GCM in previous studies such as Dibike et al. (2005), Dile
et al. (2013), Hassan et al. (2014) and Yimer et al. (2009), and HadCM3
ranked first in performance evaluation done by MAGICC/SCEGEN
computer program tools and its downscaled results match with the ensemble
mean of the six GCMs used in the LARS-WG model. Furthermore, they can
represent two different scenario generations describing the amount of
greenhouse gases (GHGs) in the atmosphere in the future. HadCM3 GCM used
emission scenarios of A2 (separated world scenario), in which the
The NCEP data set was normalized over the complete 1961–1990 period data,
and interpolated to the same grid as HadCM3 (2.5
The canESM2 outputs for three different climate scenarios are RCP 2.6,
RCP 4.5 and RCP 8.5 for the period 2006–2100, while the outputs of HadCM3 for
A2a (medium–high) and B2a (medium–low) emission scenarios for the period
1961–2099 were obtained on a grid-by-grid-box basis for the study area from
the Environment Canada website
Schematic diagram of
LARS-WG is a stochastic weather generator which can be used for the
simulation of weather data at a single station under both current and future
climate conditions. These data are in the form of daily time series for
a group of climate variables, namely precipitation, maximum temperature and minimum
temperature, and solar radiation (Chen et al., 2013; Semenov et al., 1997).
LARS-WG uses a semiempirical distribution (SED) that is defined as the
cumulative probability distribution function (CDF) to approximate probability
distributions of dry and wet series, daily precipitation, minimum temperatures and maximum temperatures.
The inputs to LARS-WG are the series of daily observed data (precipitation, minimum temperature and maximum temperature) of the base period (1984–2011) and site
information (latitude, longitude and altitude). After the input data
preparation and quality control, the observed daily weather data at a given
site were used to determine a set of parameters for probability distributions
of weather variables. These parameters are used to generate a synthetic
weather time series of arbitrary length by randomly selecting values from the
appropriate distributions, with the same statistical characteristics as the
original observed data but differing on a day-to-day basis. The LARS-WG
distinguishes wet days from dry days based on whether the precipitation is
greater than zero. The occurrence of precipitation is modeled by alternating
wet and dry series approximated by semiempirical probability distributions.
The statistical characteristics of the observed and synthetic weather data
during calibration of the model are analyzed to determine if there are any
statistically significant differences using a chi-square goodness-of-fit test
(Kolmogorov–Smirnov, KS) and the means and standard deviation using
To generate climate scenarios at a site for a certain future period with a
selected emission scenario, the LARS-WG baseline parameters, which are
calculated from observed weather for a baseline period (1984–2011), are
adjusted by the
The SDSM is best described as a hybrid of the stochastic weather generator and regression based on the family of transfer function methods, due to the fact that a multiple linear regression model is developed between a few selected large-scale predictor variables (Table 2) and local-scale predictands such as temperature and precipitation in order to condition local-scale weather parameters from large-scale circulation patterns. The stochastic component of SDSM enables the generation of multiple simulations with slightly different time series attributes, but the same overall statistical properties (Wilby et al., 2002). It requires two types of daily data: the first type corresponds to local predictands of interest (e.g., temperature, precipitation) and the second type corresponds to the data of large-scale predictors (NCEP and GCM) of a grid box closest to the station.
The SDSM model categorizes the task of downscaling into a series of discrete
processes such as quality control and data transformation, screening of
predictor variables, model calibration, and weather and scenario generation as
shown in Fig. 2b. Detail procedures and steps can be found in Wilby et al. (2002) for further reading. Screening potentially useful predictor–predictand
relationships for model calibration is one of the most challenging but very
crucial stages in the development of any statistical downscaling model. It is
because of the fact that the selection of appropriate predictor variables
largely determines the success of SDSM and also the character of the
downscaled climate scenario (Wilby et al., 2007). After routine screening
procedures, the predictor variables that provide physically sensible meaning
in terms of their high explained variance, correlation coefficient (
The model calibration process in SDSM was used to construct downscaled data based on multiple regression equations given daily weather data (predictand) and the selected predictor variables at each station. The model was structured as a monthly model for both daily precipitation and temperature using the same set of the selected NCEP predictors for the calibration period. Hence, 12 regression equations were developed for 12 months. Bias correction and variance inflation factor were adjusted until the model replicated the observed data. Model validation was carried out by testing the model using an independent data set. To compare the observed and simulated data, SDSM has provided a summary statistics function that summarizes the result of both the observed and simulated data. Time series of station data and large-scale predictor variables (NCEP reanalysis data) were divided into two groups: for the periods 1984–1995 (1984–2000) and 1996–2001 (2001–2005) for model calibration and validation of HadCM3 (canESM2) GCMs.
The scenario generator operation produces ensembles of synthetic daily weather series given observed daily atmospheric predictor variables supplied by a GCM for either current or future climates (Wilby et al., 2002). The scenario generation produced 20 ensemble members of synthetic weather data for 139 years (1961–2099) from HadCM3 A2a and B2a scenarios and for 95 years (2006–2100) from canESM2 for RCP2.6, 4.5 and 8.5 scenarios, and the mean of the ensemble members was calculated and used for further climate change analysis. The generated scenario was divided into three time windows of 30 years of data: 2011–2040, 2041–2070 and 2071–2100.
A number of statistical tests were carried out to compare the skills of the
two downscaling models categorized into two main classes. First, quantitative
statistical tests using metrics, such as mean absolute error (MAE), root mean
square error and bias. These metrics are by far the most widely used and
accepted of the many possible numerical metrics (Amirabadizadeh et al., 2016;
Bennett et al., 2013) to evaluate the comparative performance of the models
to simulate the current climate variable of precipitation on the basis of
long-term monthly averages defined by using Eqs. (
Additionally, the varying weights technique was applied to the performance
metrics as given in Eq. (
Secondly, qualitative tests, comparing the skill of models in regard to
capturing the distribution of the observed data to the whole range and in
capturing the extreme precipitation events. For this purpose, statistical
metrics such as IRF, ABC, 99p, 95p, 1daymax, R1, R10, R20 and SDII and
graphical representations of box–whisker plots and KS cumulative
distribution test were applied. KS is used to compare the probability
distribution function (PDF) of the observations to the PDF of the simulated
precipitation (Simard and L'Ecuyer, 2011). These plots provide a convenient
visual summary of several statistical properties of the data set as they vary
over time. A scoring technique is applied to compare the accuracy of the
models. In this scoring technique, the bias of an evaluation metric for each
station is used: score 1 will be given to the model that has smaller bias,
score 3 to the one with a larger bias and 2 for the model with a value in
between. Afterwards, evaluation was carried out using an equally weighted
method only due to the assumption that the metrics have equal weights, as
discussed above for model ranking. For the Kolmogorov–Smirnov cumulative
distribution test, the observed and the simulated precipitation data from
each model were compared using a
Calibration results of the average statistical tests comparing the
observed data from 26 stations with synthetic data generated through LARS-WG.
The numbers in the table show the average numbers of tests gave a
Observed and simulated
IRF and ABC are recommended by Campozano et al. (2016), while 95p, 99p, 1day
max, R1, R10, R20 and SDII are recommended by Expert on Climate Change
Detection and Indices (ETCCDI). The interquartile relative fraction (IRF): to
evaluate the modeled variability representation relative to the observed is
defined by Eq. (
To verify the performance of LARS-WG, in addition to the graphic comparison,
some statistical tests were performed. The KS test is
performed to test equality of the seasonal distributions of wet and dry
series (WDSeries), distributions of daily rainfall (RainD), and distributions
of daily maximum (
For illustrative purposes, a graphical representation of monthly mean and
standard deviation of the simulated and observed precipitation,
Average partial correlation coefficient values of all stations for
precipitation and
Initially, offline correlation analysis was performed using SPSS software
between predictands and NCEP reanalysis predictors to identify an optimal
lag and physically sensible predictors for climate variables of
precipitation,
Calibration of observed and simulated of precipitation, maximum
temperature and minimum temperature for the Gondar station using SDSM from
canESM2 and HadCM3 from
Validation of observed and simulated of precipitation, maximum
temperature and minimum temperature for Gondar station using SDSM from
canESM2 and HadCM3 from
The predictor variables identified for each downscaling GCM and for the
corresponding local climate variables showed that different large-scale
atmospheric variables control different local variables. For instance, the
set of temp, mslp, s500, s850, p8_v, p500, shum comprises the most potential
or meaningful predictors for temperature, and the set of s500, s850, p8_u,
p_z, pzh, p500 performs best for predicting precipitation of the study area,
which is consistent with the result of offline correlation analysis. After carefully screening predictor variables, model
calibration and validation was carried out. The graphical comparison between
the observed and generated rainfall,
The statistical performance metrics of MAE and RMSE values for the monthly
precipitation modeled from canESM2 range from 3.5 to 14.8
Since the performance of LARS-WG during calibration and validation was very
good, downscaling of the climate scenario can be done from six selected multimodel CMIP3 GCMs under three scenarios (A1B, B1 and A2) for three time
periods. After downscaling the future climate scenarios at all stations from
the selected six GCMs, the projected precipitation analysis for the areal
UBNRB was calculated from the point rainfall stations using the Thiessen polygon
method. The result analysis (Fig. 4a) revealed that GCMs disagree on the
direction of precipitation change: two GCMs (CSMK3 and GFCM21) showed
decreasing trends, and a majority, or four, GCMs (NCCSM, Hadcm3, MPEH5 and
MIHR) showed increasing trends from the reference period in all three time
periods. By the 2030s, the relative change in mean annual precipitation is projected
in the range between
In a different way from precipitation, the projections of mean annual
Performance measure and ranking of models during the baseline period (1984–2011) for evaluation metric RMSE.
Statistical downscaling models ranking during the baseline period (1984–2011) for quantitative measures. The numbers in the table show the total ranking scores summed up from 15 stations.
Ranking of statistical downscaling models during the baseline period (1984–2011) for qualitative measures (distribution and extreme events of daily precipitation). The numbers in the table show the total ranking scores obtained from 15 stations.
Here, as it is difficult to process all the selected six CMIP3 GCM3 using SDSM, we choose the HadCM3 GCM as the best due to the fact that the downscaling result of HadCM3 using LARS-WG fits with the downscaling result of the ensemble mean model. Also, canESM2 from the CMIP5 GCMs was selected to test the improvements of CMIP5 over CMIP3. Results of downscaling future climate scenario of areal UBNRB using SDSM calculated from all stations using Thiessen polygon methods are summarized in Fig. 8. The overall analysis of the result indicates a general increase in mean annual precipitation for three time windows (2030s, 2050s and 2080s) for all five scenarios (A2a and B2a for HadCM3 and RCP2.6, RCP4.5 and RCP8.5 for canESM2) in the range of 2.1 to 43.8 %. The maximum (minimum) relative change of mean annual precipitation is projected to be 43.8 % (6.2 %), 29.5 % (3.5 %) and 19 % (2.1 %) in the 2080s, 2050s and 2030s under the RCP8.5 scenario of canESM2 (B2a) scenario of HadCM3. In general, the RCP8.5 scenario of canESM2GCM resulted in pronounced increases in all three time periods, whereas scenario B2a of the HadCM3 GCM reported minimum change over the study area.
Regarding temperature, the downscaling result of
Chen et al. (2013) argued that though major sources of uncertainty are linked to GCMs and emission scenarios, uncertainty related to the choice of downscaling methods give less attention to climate change analysis. Therefore, in this study, comparative performance evaluation of the downscaling methods has been given due emphasis and carried out in a number of statistical and graphical tests both quantitatively and qualitatively. The model skill was evaluated and ranked at each site for each metric as shown in Table 4 for metrics of RMSE. The overall rank obtained by summing up the score of each model for each metric is presented in Tables 5 and 6, for quantitative and qualitative measures respectively.
The result revealed that SDSM/canESM2 narrowly performed best in simulating
the long-term average values in both equally weighted and varying weights of
the quantitative metrics. However, LARS-WG performed best in qualitative
measures in reproducing the distribution and extreme events of daily
precipitation. For instance, absolute bias for the 95th percentile of daily
precipitation (95p) ranges from 4.35 to 12.4
Kolmogorov–Smirnov,
Kolmogorov–Smirnov test to compare the skill of the models for the observed precipitation distribution (upper three Alemketema stations, lower three Debre Markos stations).
Box plot showing the model performance for three stations on a monthly basis. Box boundaries indicate the 25th and 75th percentiles, the line within the box marks the median, whiskers below and above the box indicate the 10th and 90th percentiles, and dots indicate the extremes.
Furthermore, as the Kolmogorov–Smirnov test from Table 7 shows, LARS-WG captures
the distribution of the observed precipitation of 93.3 % from all stations
while SDSM captures only 20 % of the 15 stations equally both from
canESM2 and HadCM3 GCMs at 5 % significance level. The
Comparison of climate change scenarios
For future simulation, the HadCM3 GCM A2 scenario was used in common for two
(LARS-WG and SDSM) downscaling methods to test whether the downscaling
methods may affect the GCM result under the same forcing scenario. The
results obtained from the two downscaling models were found reasonably
comparable and both approaches showed increasing trends for precipitation,
The uncertainty related to climate change analysis can be due to climate models and downscaling methods among many other factors. In this study, we employed a multimodel approach to see that the uncertainties came from different GCMs. In total, 21 systematically selected future climate scenarios were produced for each time period, which we might think representative to understand fully and to project plausibly the future climate change in the study area and to retain information about the full variability of GCMs. Moreover, we applied two widely used statistical downscaling methods, namely the regression downscaling technique (SDSM) and the stochastic weather generation method (LARS WG) for this particular study.
The performance of the three models (HadCM3/SDSM, canESM2/SDSM and LARS-WG) were tested for the baseline period of 1984–2011 in representing the current situation, particularly for precipitation, as it is the most difficult climate variable to model. The result suggested that SDSM using canESM2 GCM captures the long-term monthly average very well at most of the stations and it ranked first from others. This could be attributed to the increasing performance of GCMs from time to time (i.e., CMIP5 GCMs performs better than CMIP3 GCMs) due to the fact that modeling was based on the new set of radiative forcing scenarios that replaced SRES emission scenarios, constructed for IPCC AR5 where the impacts of land use and land cover change on the environment and climate are explicitly included. However, LARS-WG performed best in qualitative measures in capturing the distribution and extreme events of the daily precipitation than SDSM. The better performance of LARS-WG in capturing the distribution and extreme events of the daily precipitation may be associated with the use of 23 interval histograms for the construction of semiempirical distribution, which offers a more accurate representation of the observed distribution compared with the 10 intervals used in the previous version (Semenov et al., 2010). The poor performance of SDSM would indicate the difficulty in finding climate variables from the NCEP data that could explain the variability of daily precipitation well. Therefore, LARS-WG would be more preferred in areas of the UBNRB where there is high climatic variability to correctly simulate the distribution and extreme events of the precipitation, which is crucial for a realistic assessment of flood events and agricultural production.
The downscaling result reported from the six GCMs used in LARS-WG showed large intermodel differences: two GCMs reported that precipitation may decrease while four GCMs reported that precipitation may increase in the future. The large intermodel differences in the GCMs showed the uncertainties of GCMs associated with their differences of resolution and assumptions of physical atmospheric processes to represent local-scale climate variables, which are typical characteristics for Africa and because of low convergence in climate model projections in the area of UBNRB (Gebre and Ludwig, 2015). These results further reinforce multimodel strategies for conducting climate change studies. The multimodel average result showed that in the future precipitation may generally increase over the basin in the range of 1 to 14.4 % which is in line with the result from HadCM3 GCM (0.8 to 16.6 %); this indicates that HadCM3 from CMIP3 GCMs has a better representation of local-scale climate variables in the study area, consistent with the previous study result by Kim and Kaluarachchi (2009) and Dile et al. (2013) in the same study area.
LARS-WG produces synthetic climate data of any length with the same
characteristics as the input record, and it simulates weather separately for
a single site. Therefore, the resulting weather series for different sites are
independent of each other, which can cause loss of a very strong spatial correlation
that exists in real weather data during simulation. However, a few
stochastic models have been developed to produce weather series
simultaneously at multiple sites, preserving the spatial correlation, mainly
for daily precipitation, such as space–time models, nonhomogeneous hidden
Markov models and nonparametric models that typically use a K nearest-neighbor
(K-NN) procedure (King et al., 2015). They are complicated in both
calibration and implementation and are unable to adequately reproduce the
observed correlations (Khalili et al., 2007). In this study, the simple
Pearson's correlation coefficient (
In conclusion, a multimodel average from LARS-WG and individual model result
from SDSM showed a general increasing trend for all three climatic variables
(precipitation,
In general, this study has shown that climate change will likely occur
that may affect the water resources and hydrology of the UBNRB. On the basis
of the results obtained in this study, both SDSM and LARS-WG models can be
adopted with reasonable confidence as downscaling tools to undertake climate
change impact assessment studies for the future. However, LARS-WG is more
suitable for extreme precipitation impact assessment studies, such as those dealing with floods and
droughts. Moreover, the paper provides substantial information that the
choice of downscaling methods has a contribution in the uncertainty of future
climate prediction. The authors would also like to suggest further
assessment of the study area using a large ensemble of CMIP5 GCMs. Further
relative performance of downscaling techniques for other climatic variables
such as
The data can be made available upon request to the corresponding author.
The authors declare that they have no conflict of interest.
We are grateful to the Ethiopian Meteorological Agency (EMA) for providing us the meteorological data for free. The first author received financial support for traveling money from the DAAD water–food–energy NeXus project. The authors are indebted to the two anonymous reviewers and the editor, Dimitri Solomatine, for their critique and constructive suggestions and comments on earlier versions of this paper, which were helpful in the improvement of the manuscript. This work was supported by the German Research Foundation (DFG) and the Technische Universität München within the funding programme Open Access Publishing. Edited by: Dimitri Solomatine Reviewed by: two anonymous referees