You are on page 1of 16

Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol.

13(2):252-267
Copyright Faculty of Engineering, University of Maiduguri, Maiduguri, Nigeria.
Print ISSN: 1596-2490, Electronic ISSN: 2545-5818, www.azojete.com.ng

PERFORMANCE OF SOME GENERAL CIRCULATION MODELS ON PREDICTING


TEMPERATURE AND RAINFALL IN THE SUDAN-SAHEL REGION OF NIGERIA

M. F. Amodu1, 2 and C. J. Ejieji1


(1Department of Agricultural and Biosystems Engineering, University of Ilorin, Nigeria
2
Department of Rural Development, Ministry of Agriculture and Rural Development, Musa
Usman Secretariat, Maiduguri, Borno State, Nigeria.)
Corresponding authors e-mail address: mfamodu@yahoo.com

Abstract
Twenty Global Circulation Models (GCMs) were evaluated by comparing their predictions of monthly average air
temperature and monthly rainfall amounts of the sudan-sahel region of Nigeria with observed data for the period
1981 to 2000. The GCMs were those in the Program for Climate Model Diagnosis and Intercomparison (PCMDI) of
the World Climate Research Program's (WCRP's) Coupled Model Intercomparison Project Phase 3 (CMIP3). The
observed data were from four representative stations namely Kano, Katsina, Maiduguri and Sokoto. Coefficient of
correlation (r), mean absolute error (MAE) and root mean square error (RMSE) were used as metrics for ranking the
GCMs. None of the GCMs was superior in performance across the four representative stations. The trends of
observed average monthly air temperature were generally better simulated with positive r-values in contrast with
their monthly rainfall predictions which correlated negatively with observed data. The r-values for temperature
averaged 0.211, 0.092, 0.373 and 0.212 for Kano, Katsina, Maiduguri and Sokoto respectively. For rainfall, the r-
values averaged -0.216, -0.280, -0.115 and -0.351 for Kano, Katsina, Maiduguri and Sokoto respectively. Simulated
rainy season commenced 4 to 6 months earlier than observed depending on the particular GCM. The average MAE
from the GCMs for the respective stations ranged from 5.23 to 14.31 K for temperature and 81.91 to 91.99 mm for
rainfall. For RMSE, the averages from the GCMs ranged from 5.84 to 14.42 K for temperature and 140.21 to 187.99
mm for rainfall. The results of the study have provided the basis for selection of appropriate GCMs for the locations
for further downscaling procedures that might be required for the application of a selected GCM to climate change
projection.

Keywords: Climate Change, Global Circulation Models, Sudan-Sahel, Nigeria


1. Introduction
In climate change studies General Circulation Models (GCMs) have been widely employed in
the simulation of past climates and the projection of that of the future under various greenhouse
gases emission scenarios (Nakicenovic et al., 2000). Their widespread use is based on their
recognition and adoption as currently the most credible tools for the purpose. General guidelines
on the use of climate projection data from the various GCMs for climate impact and adaptation
assessment under the various emission scenarios have been provided by the Task Group on
Scenarios for Climate Impact Assessment (TGCIA) of the Intergovernmental Panel on Climate
Change (IPCC) (IPCC-TGICA, 2007).
Christensen et al. (2007) reported that warming in Africa in the 21st Century was very likely to
be larger than the global annual mean warming throughout the continent and in all seasons with
drier subtropical regions warming more than the moister tropics. They also indicated that in
much of Mediterranean Africa and the Northern Sahara annual rainfall was likely to decrease and
that it was unclear how rainfall in the Sahel, the Guinean Coast and the Southern Sahara would
evolve. Temperature and rainfall however constitute major weather variables driving general
environmental, agro-hydrological and ecological processes. GCM-generated data on the
252
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

variables are therefore indispensable in the study of future climatic change impact on (i)
agriculture, food and fibre, ii) hydrology and water resources, (iii) terrestrial and aquatic
ecosystems (iv) human infrastructure and (v) human health. It has however been observed from
inter-comparison studies that generally the various variables are simulated with varying degrees
of success by different models and that no particular model was best for all variables and/or all
regions (Meehl et al., 2005). There is therefore no particular GCM that is globally best
performing. However, a GCM could project better in specific locations than others. Chiew et al.
(2009) therefore recommended that for more reliable climate impact assessment the better
performing GCMs should be preferred.
None of the existing GCMs (Table 1) originated from Africa hence the need to evaluate their
performance for informed local application. The objective of this paper was therefore to evaluate
the performance of the some GCMs in prediction of monthly rainfall and average monthly
temperatures for the Sudan Sahel region of Nigeria with Kano, Katsina, Sokoto and Maiduguri as
the representative stations. The GCMs (Table 1) involved in the study were those in the Program
for Climate Model Diagnosis and Intercomparison (PCMDI) of the World Climate Research
Program's (WCRP's) Coupled Model Intercomparison Project Phase 3 (CMIP3).

2. Materials and Methods


2.1 The Study Area
The study area comprise the Sudan Savanna and Sahel vegetative zones of Nigeria (Figure 1). It
o o o o
lies within latitude 11.50 N to 13.97 N, and longitude 2.63 E to 14.76 E. The area is semi-arid in
nature and boarders the southern part of the Sahara desert. The season wet starts in June and ends
in October, while the dry season is from November through May. Generally November to March
are Harmattan months with cool dry air dusty wind blowing North East from the Sahara desert,
while the hottest months are April and May. Most of the rains fall in the months of July, August
and September and are due to the south west monsoon winds from Atlantic Ocean. The annual
rainfall varies from 1050 mm in Kano to 550 mm in Maiduguri. The temperatures range between
o o
24 C and 44 C. The land areas are interspersed with woodland, shrubs, dense grasses with a
sprinkling of trees, riparian forests and seasonally flooded marshes and fadamas. Over-
cultivation, over-grazing, land degradation and deforestation are common features, which had
modified the vegetation of the basin. According to the Lake Chad Basin Commission, LCBC
(2006), the rate of desertification expansion became more pronounced after the prolonged
droughts of the 1970s and 1980s. The details of the representative stations of the study are
presented in Table 2.

2. 2. Data collection
Monthly average temperature and monthly rainfall for the 10-year period of 1981 to 2000 were
collected from Nigeria Meteorological Agency (NIMET) Abuja for the 4 representative stations
used in the study. The corresponding output data from the 20 GCMs outputs archived as
multi-model dataset for the WCRP's CMIP3 in Gridded Binary Edition 1 (GRIB1) format for
253
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and
rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

GCM grid cells appropriate for the respective stations were obtained from German Climate
Computing Centre i.e. Deutsches Klimarechenzentrum (DKRZ), Hamburg. A program was then
developed in Visual Studio 2013 platform for the extraction of the data to spreadsheet in order to
facilitate the analysis.

Table 1: The existing Coupled Atmospheric and Ocean Global Circulation Models
Model Code Description and Development Institution City Country
BCM2 BCCR-BCM2.0, Bjerknes Centre for Climate Research Bergen Norway
CGMR CGCM3.1(T47), Canadian Centre for Climate Modelling & Victoria Canada
Analysis
CGHR CGCM3.1(T63), Canadian Centre for Climate Modelling & Victoria Canada
Analysis,
CNCM3 CNRM-CM3, Mto-France, Toulouse France
CSMK3: CSIRO-Mk3.0, CSIRO Atmospheric Research, Victoria Australia
ECHOG: ECHO-G, Meteorological Institute of the University of Bonn, Miub Germany/
Meteorological Research Institute of KMA, and Model and Korea
Data group
GFCM21: GFDL-CM2.1, US Dept. of Commerce / NOAAA / Princeton USA
Geophysical Fluid Dynamics Laboratory
GIAOM: GISS-AOM, NASA / Goddard Institute for Space Shuttles New York USA
GIEH: GISS-EH, NASA / Goddard Institute for Space Shuttles New York USA
GIER: GISS-ER, NASA / Goddard Institute for Space Shuttles New York USA
INGV-SXG: INGV-SXG: Istituto Nazionale di Geofisica e Vulcanologia, Italy
HADCM3: UKMO-HadCM3, Hadley Centre for Climate Prediction and Exeter UK
Research / Met Office Devon
HADGEM UKMO-HadGEM1, Hadley Centre for Climate Prediction and Exeter UK
Research / Met Office Devon
INCM3 INM-CM3.0, Institute for Numerical Mathematics Moscow Russia
MIHR MIROC3.2(hires), Center for Climate System Research(The Tokyo Japan
University of Tokyo), National Institute
for Environmental Studies, and Frontier Research Center for
Global Change (JAMSTEC)
MIMR MIROC3.2(Medres), Center for Climate System Research(The Tokyo Japan
University of Tokyo), National Institute for Environmental
Studies, and Frontier Research Center for Global Change
(JAMSTEC)
MPEH5 ECHAM5/MPI-OM, Max Planck Institute for Meteorology, Hamburg Germany
MRCGCM MRI-CGCM2.3.2, Meteorological Research Institute, Japan Tsukuba Japan
NCCCSM CCSM3, National Center for Atmospheric Research Boulder USA
NCPCM PCM, National Center for Atmospheric Research Boulder USA
Source: Meehl et al. (2007)
254
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

Figure 1: Map of Nigeria: the delineation of vegetative zones adopted from Ejieji (2004). (The
locations of the representative stations for the study i.e. Kano, Katsina, Maiduguri, and Sokoto
are indicated with asterisks)

Table 2: Location details of the representative stations


Station Longitude Latitude Elevation
o o
( E) ( N) (m)
Kano 8.2000 12.0500 472.5
Katsina 7.6833 13.0167 517.6
Maiduguri 13.0833 11.8500 353.8
Sokoto 5.2500 13.0167 350.8

2.3 Evaluation of the GCMs


The GCMs were evaluated by comparing the predicted with the observed data using Pearsons
correlation coefficient (r), Root Mean Square Error (RMSE) and Mean Absolute Error of
prediction (MAE) as indices. The twentieth century coupled climate change model (20c3m)

255
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and
rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

scenario experiment outputs for most GCMs ended in 1999. Therefore the data used for the
evaluation were mostly for 1981 to 1999.
Correlation coefficient, r, was expressed as
( )( )
(1)
( ) ( ) ( ) ( )

where: xi and yi are respectively the observed and predicted data while n is the number of pairs
of data compared. Root mean square error was (RMSE) was obtained as follows

( )
( ) (2)

Mean absolute error (MAE) was estimated as follows

| |
(3)

In Equations (2) and (3), all the terms are as previously defined.

3. Results and Discussion


Some examples of graphical comparison of the observed and predicted data series are presented
in Figures 2 and 3 for monthly average air temperature and rainfall respectively. The plots show
for each station, with respect to observed and predicted data relationship, the three GCMs with
predictions having the best, median-ranked and worst correlations respectively. The relative
rankings of all the GCMs based r-values are presented in Tables 3 and 4 for air temperature and
rainfall respectively. The rankings relating to the MAE of predictions of the GCMs for monthly
average air temperature and rainfall are shown in Tables 5 and 6 respectively. Those for the
rankings based on RMSE are presented in Tables 7 and 8 for monthly average air temperatures
and monthly rainfalls respectively.
Generally, most of the GCMs simulated the trends of the monthly temperatures well (Figure2) in
contrast with their simulation of observed rainfall trends (Figure 3). The r-values were therefore
generally higher and mostly positive for temperature in contrast with the values for rainfall
which were mostly negative and generally lower. The r-values for temperature averaged 0.211,
0.092, 0.373 and 0.212 for Kano, Katsina, Maiduguri and Sokoto respectively. The corresponding
r-values for rainfall averaged -0.216, -0.280, -0.115 and -0.351 for Kano, Katsina, Maiduguri
and Sokoto respectively. The negative correlations observed for rainfall are due not only to
under-prediction of observed rainfall amounts but also to the predicted wet season
commencement generally having lead times of 4 to 6 months depending on the GCM. The

256
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

relatively poor performance of the GCMs in predicting rainfall is similar to finding of


Masanganise et al., (2013) that rainfall prediction was poorer than that of temperature in
Zimbabwe. Christensen et al., (2007) also reported that lack of consistency by the GCMs in
representing monsoon processes and the Intertropical Convergence Zone (ITCZ) resulted in
different projections by the models and increased the uncertainty in estimating of future
precipitation in West Africa.

257
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and
rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Figure 2: Observed and predicted monthly average air temperature data series for 1981 to 1999.
(The predicted series plotted for each station are for the 3 GCMs with data having, respectively,
the best, median and lowest-ranked correlation with observed data)

Figure 3: Observed and predicted monthly rainfall data series for 1981 to 1999. (The predicted
series plotted for each station are for the 3 GCMs with data having, respectively, the best,
median and lowest-ranked correlation with observed data)

258
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

Table 3: Ranking of the performance of the General Circulation Models (GCMs) based on the correlation between the predicted and
observed monthly average of air temperatures for the stations (Correlation coefficient (r) was used as the performance metric).
(a) Kano: (b) Katsina: (c) Maiduguri (d) Sokoto
GCM r-value Rank GCM r-value Rank GCM r-value Rank GCM r-value Rank
ECHOG 0.557 1 INGSX6 0.399 1 BCM2 0.593 1 INGSX6 0.574 1
BCM2 0.533 2 MIHR 0.387 2 NCPCM 0.583 2 MIHR 0.551 2
CNCM3 0.525 3 NCCCSM 0.328 3 CGMR 0.580 3 HADCM3 0.485 3
MIHR 0.497 4 HADCM3 0.317 4 CNCM3 0.507 4 HADGEM 0.466 4
INGSX6 0.497 5 GIEH 0.281 5 ECHOG 0.491 5 NCCCSM 0.441 5
NCPCM 0.445 6 GIAOM 0.272 6 MPEH5 0.457 6 GIEH 0.425 6
CGMR 0.324 7 HADGEM 0.251 7 MIMR 0.429 7 GIAOM 0.362 7
MIMR 0.268 8 CSMK3 0.120 8 CSMK3 0.427 8 GIER 0.320 8
INCM3 0.229 9 GIER 0.118 9 INCM3 0.407 9 CSMK3 0.286 9
HADCM3 0.215 10 GFCM20 0.089 10 MRCGCM 0.397 10 GFCM20 0.267 10
MRCGCM 0.190 11 NCPCM 0.084 11 MIHR 0.383 11 MIMR 0.140 11
NCCCSM 0.156 12 MIMR 0.083 12 INGSX6 0.376 12 NCPCM 0.067 12
GIAOM 0.105 13 BCM2 0.069 13 HADGEM 0.288 13 INCM3 0.063 13
GIEH -0.020 14 CGMR -0.018 14 GIEH 0.261 14 CGMR 0.061 14
GIER -0.038 15 MPEH5 -0.124 15 HADCM3 0.242 15 MPEH5 0.045 15
HADGEM -0.076 16 INCM3 -0.155 16 GIER 0.236 16 BCM2 -0.013 16
GFCM20 -0.097 17 MRCGCM -0.202 17 GFCM20 0.224 17 MRCGCM -0.078 17
CSMK3 -0.102 18 CNCM3 -0.250 18 GIAOM 0.158 18 CNCM3 -0.151 18
MPEH5 -0.203 19 ECHOG -0.304 19 NCCCSM 0.053 19 ECHOG -0.284 19

259
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Table 4: Ranking of the performance of the General Circulation Models (GCMs) based on the correlation between the predicted and
observed monthly rainfalls for the stations (Correlation coefficient (r) was used as the performance metric)
(a) Kano: (b) Katsina: (c) Maiduguri (d) Sokoto
GCM r-value Rank GCM r-value Rank GCM r-value Rank GCM r-value Rank
GFCM20 0.293 1 GFCM20 0.369 1 GFCM20 0.417 1 GFCM20 0.306 1
GIEH 0.183 2 GIER 0.225 2 GIER 0.402 2 GIER 0.200 2
GIER 0.137 3 HADGEM 0.108 3 HADGEM 0.382 3 GIEH -0.219 3
INCM3 -0.002 4 HADCM3 -0.040 4 GIEH 0.313 4 NCCCSM -0.242 4
HADCM3 -0.102 5 GIAOM -0.142 5 GIAOM 0.300 5 MRCGCM -0.242 5
HADGEM -0.161 6 GIEH -0.203 6 HADCM3 0.244 6 GIAOM -0.298 6
GIAOM -0.165 7 INCM3 -0.233 7 INCM3 -0.049 7 CNCM3 -0.340 7
MRCGCM -0.185 8 MRCGCM -0.238 8 MRCGCM -0.166 8 INCM3 -0.353 8
CSMK3 -0.283 9 NCCCSM -0.238 9 CSMK3 -0.190 9 CSMK3 -0.354 9
CNCM3 -0.294 10 CSMK3 -0.300 10 INGSX6 -0.301 10 HADGEM -0.355 10
INGSX6 -0.312 11 CNCM3 -0.346 11 NCPCM -0.303 11 HADCM3 -0.373 11
NCPCM -0.321 12 INGSX6 -0.405 12 MPEH5 -0.309 12 INGSX6 -0.376 12
NCCCSM -0.339 13 ECHOG -0.498 13 CNCM3 -0.330 13 ECHOG -0.524 13
BCM2 -0.360 14 MPEH5 -0.498 14 MIMR -0.364 14 MPEH5 -0.549 14
ECHOG -0.375 15 CGMR -0.539 15 CGMR -0.385 15 CGMR -0.561 15
MIMR -0.383 16 BCM2 -0.560 16 BCM2 -0.388 16 NCPCM -0.565 16
CGMR -0.432 17 MIMR -0.570 17 ECHOG -0.415 17 BCM2 -0.580 17
MPEH5 -0.469 18 NCPCM -0.574 18 NCCCSM -0.505 18 MIMR -0.593 18
MIHR -0.540 19 MIHR -0.641 19 MIHR -0.531 19 MIHR -0.652 19

260
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

Table 5: Ranking of the performance of the General Circulation Models (GCMs) based on their Mean Absolute Error (MAE) of
prediction of monthly average temperatures for the stations
(a) Kano: (b) Katsina: (c) Maiduguri (d) Sokoto
GCM MAE Rank GCM MAE Rank GCM MAE Rank GCM MAE Rank
(K) (K) (K) (K)
MIHR 2.50 1 NCCCSM 8.71 1 MPEH5 10.02 1 MIHR 9.60 1
CGMR 3.13 2 CSMK3 9.11 2 MIHR 10.21 2 NCCCSM 9.92 2
NCCCSM 3.23 3 MRCGCM 9.28 3 INGSX6 10.24 3 CSMK3 10.06 3
INGSX6 3.24 4 ECHOG 9.45 4 NCCCSM 10.65 4 INGSX6 10.28 4
BCM2 3.27 5 MPEH5 9.51 5 NCPCM 11.48 5 MRCGCM 10.48 5
MIMR 3.43 6 INGSX6 9.71 6 MIMR 11.53 6 MPEH5 10.53 6
ECHOG 3.57 7 MIHR 9.83 7 CGMR 11.89 7 ECHOG 10.67 7
NCPCM 3.63 8 GIER 10.45 8 BCM2 12.38 8 GIAOM 10.98 8
CSMK3 3.94 9 MIMR 10.93 9 CSMK3 12.71 9 GIER 11.66 9
MPEH5 4.10 10 NCPCM 11.28 10 MRCGCM 12.78 10 MIMR 12.15 10
MRCGCM 4.16 11 GIEH 11.33 11 ECHOG 13.15 11 NCPCM 12.51 11
CNCM3 4.77 12 GIAOM 11.51 12 CNCM3 14.64 12 GIEH 12.54 12
GIAOM 5.14 13 CNCM3 11.71 13 HADCM3 15.83 13 CNCM3 12.94 13
GIER 6.99 14 BCM2 12.97 14 HADGEM 18.20 14 GFCM20 13.43 14
HADCM3 7.33 15 CGMR 13.32 15 GIAOM 18.70 15 HADGEM 13.97 15
INCM3 8.96 16 HADCM3 14.13 16 GIEH 18.95 16 HADCM3 14.04 16
GIEH 9.01 17 HADGEM 14.48 17 GFCM20 19.34 17 BCM2 14.20 17
HADGEM 9.07 18 GFCM20 14.56 18 INCM3 19.34 18 CGMR 14.54 18
GFCM20 9.93 19 INCM3 16.44 19 GIER 19.80 19 INCM3 17.34 19

261
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Table 6: Ranking of the performance of the General Circulation Models (GCMs) based on their Mean Absolute Error (MAE) of
prediction of monthly rainfalls for the stations.
(a) Kano: (b) Katsina: (c) Maiduguri (d) Sokoto
GCM MAE Rank GCM MAE Rank GCM MAE Rank GCM MAE Rank
(mm) (mm) (mm) (mm)
GIAOM 71.21 1 GIER 71.97 1 GIER 78.85 1 GIER 99.96 1
CSMK3 71.28 2 GFCM20 72.30 2 HADGEM 79.62 2 GFCM20 100.61 2
CGMR 71.97 3 HADGEM 72.70 3 GFCM20 79.91 3 GIEH 101.51 3
NCPCM 71.98 4 HADCM3 73.20 4 GIEH 81.09 4 HADCM3 101.96 4
INGSX6 72.04 5 GIAOM 73.24 5 GIAOM 81.22 5 HADGEM 102.27 5
MRCGCM 72.07 6 INCM3 73.47 6 HADCM3 81.70 6 INCM3 102.97 6
CNCM3 72.08 7 GIEH 73.66 7 MRCGCM 83.02 7 GIAOM 104.74 7
GFCM20 73.50 8 CSMK3 81.45 8 INCM3 83.71 8 MRCGCM 110.24 8
INCM3 74.38 9 MRCGCM 82.36 9 CSMK3 84.78 9 CSMK3 113.07 9
HADGEM 74.72 10 INGSX6 88.38 10 CNCM3 85.45 10 CNCM3 115.07 10
NCCCSM 79.21 11 CNCM3 88.59 11 ECHOG 85.92 11 NCCCSM 115.89 11
HADCM3 79.85 12 NCCCSM 93.27 12 CGMR 87.11 12 INGSX6 115.93 12
GIER 87.04 13 ECHOG 93.70 13 MIMR 91.48 13 ECHOG 121.25 13
GIEH 88.16 14 MPEH5 105.20 14 INGSX6 91.57 14 MPEH5 138.13 14
ECHOG 88.74 15 NCPCM 113.86 15 BCM2 96.16 15 NCPCM 139.25 15
MPEH5 93.19 16 CGMR 119.16 16 MPEH5 99.42 16 BCM2 146.36 16
MIHR 98.61 17 BCM2 119.34 17 NCPCM 101.93 17 CGMR 147.08 17
MIMR 103.80 18 MIMR 122.21 18 NCCCSM 113.35 18 MIHR 147.71 18
BCM2 112.40 19 MIHR 129.67 19 MIHR 117.92 19 MIMR 149.69 19

262
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

Table 7: Ranking of the performance of the General Circulation Models (GCMs) based on their Root Mean Square Error (RMSE) of
prediction of monthly average temperatures for the stations.
(a) Kano: (b) Katsina: (c) Maiduguri (d) Sokoto
GCM RMSE Rank GCM RMSE Rank GCM RMSE Rank GCM RMSE Rank
(K) (K) (K) (K)
MIHR 3.02 1 NCCCSM 9.24 1 INGSX6 10.50 1 MIHR 10.01 1
CGMR 3.53 2 CSMK3 9.78 2 HADCM3 10.66 2 NCCCSM 10.32 2
BCM2 3.68 3 INGSX6 10.19 3 MIMR 10.71 3 CSMK3 10.51 3
INGSX6 3.72 4 MRCGCM 10.24 4 MIHR 11.18 4 INGSX6 10.60 4
NCCCSM 3.78 5 MIHR 10.31 5 NCCCSM 11.80 5 MPEH5 11.06 5
MIMR 3.86 6 MPEH5 10.38 6 ECHOG 11.92 6 MRCGCM 11.23 6
ECHOG 4.02 7 ECHOG 10.49 7 INCM3 12.20 7 GIAOM 11.39 7
NCPCM 4.12 8 GIER 11.13 8 NCPCM 12.67 8 ECHOG 11.54 8
CSMK3 4.48 9 MIMR 11.50 9 GIAOM 13.18 9 GIER 12.13 9
MRCGCM 4.60 10 NCPCM 11.76 10 CSMK3 13.19 10 MIMR 12.61 10
MPEH5 4.99 11 GIAOM 11.96 11 CNCM3 13.48 11 NCPCM 12.91 11
CNCM3 5.39 12 GIEH 11.98 12 HADGEM 14.94 12 GIEH 13.00 12
GIAOM 5.78 13 CNCM3 12.46 13 BCCM1 15.23 13 CNCM3 13.53 13
GIER 7.82 14 BCM2 13.44 14 GIEH 16.21 14 GFCM20 13.95 14
HADCM3 8.00 15 CGMR 13.92 15 GIER 18.52 15 HADGEM 14.26 15
INCM3 9.60 16 HADCM3 14.49 16 MPEH5 19.01 16 HADCM3 14.33 16
HADGEM 9.81 17 HADGEM 14.87 17 MRCGCM 19.36 17 BCM2 14.62 17
GIEH 10.05 18 GFCM20 15.05 18 GFCM20 19.61 18 CGMR 15.02 18
GFCM20 10.67 19 INCM3 18.00 19 BCM2 19.65 19 INCM3 18.62 19

263
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Table 8: Ranking of the performance of the General Circulation Models (GCMs) based on their Root Mean Square Error (RMSE) of
prediction of monthly rainfalls for the stations.
(a) Kano: (b) Katsina: (c) Maiduguri (d) Sokoto
GCM RMSE Rank GCM RMSE Rank GCM RMSE Rank GCM RMSE Rank
(mm) (mm) (mm) (mm)
GFCM20 133.75 1 GIER 133.90 1 CGMR 144.40 1 NCCCSM 177.65 1
INCM3 135.53 2 NCCCSM 134.24 2 MRCGCM 150.50 2 GIER 181.91 2
NCCCSM 136.74 3 GIEH 135.23 3 GIER 151.50 3 GIAOM 182.20 3
CNCM3 137.07 4 GIAOM 135.47 4 BCM2 152.20 4 GIEH 183.19 4
MRCGCM 137.55 5 HADGEM 136.03 5 MPEH5 152.80 5 HADGEM 184.10 5
INGSX6 137.61 6 GFCM20 136.24 6 GFCM20 152.87 6 HADCM3 184.66 6
HADGEM 137.92 7 HADCM3 136.32 7 GIEH 154.82 7 GFCM20 184.69 7
HADCM3 138.26 8 INCM3 137.00 8 CSMK3 155.92 8 INCM3 185.08 8
CGMR 138.61 9 CSMK3 138.50 9 CNCM3 156.16 9 INGSX6 185.21 9
MIHR 138.63 10 INGSX6 139.95 10 HADGEM 156.57 10 CSMK3 186.45 10
GIER 138.73 11 MRCGCM 139.97 11 INCM3 156.64 11 MRCGCM 187.11 11
MIMR 139.56 12 ECHOG 140.82 12 GIAOM 156.67 12 CNCM3 187.13 12
NCPCM 139.98 13 CNCM3 141.02 13 ECHOG 157.50 13 ECHOG 187.58 13
CSMK3 141.29 14 NCPCM 147.11 14 MIMR 158.04 14 NCPCM 188.95 14
ECHOG 142.25 15 MPEH5 147.43 15 NCPCM 158.72 15 MPEH5 194.47 15
GIAOM 142.58 16 BCM2 154.85 16 NCCCSM 160.93 16 MIHR 195.48 16
BCM2 146.99 17 CGMR 155.13 17 INGSX6 161.24 17 BCM2 197.78 17
MPEH5 148.74 18 MIMR 156.50 18 MIHR 163.73 18 CGMR 198.53 18
GIEH 152.18 19 MIHR 159.38 19 HADCM3 166.53 19 MIMR 199.72 19

264
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

The averages of MAE from all the GCMs for temperature and rainfall predictions respectively
were 5.23 K and 81.91 mm for Kano, 11.51 K and 91.99 mm for Katsina, 14.31 K and 89.70 mm
for Maiduguri and, 12.20 K and 119.67 mm for Sokoto. For RMSE the averages for temperature
and rainfall predictions respectively were 5.84 K and 140.21 mm for Kano, 12.17 K and
142.37 mm for Katsina, 14.42 K and 156.20 mm for Maiduguri and, 12.72 K and 187.99 mm for
Sokoto. The rankings presented in Table 3 to 8 however indicate that none of the GCMs was
generally top-performing across the four stations based on the stated ranking criteria. It should be
noted that the choice of metrics for evaluation usually depends on the purpose of a study. For
example, where good representation of trend is the major interest correlation coefficient should
be of main interest whereas mean absolute error or root square mean error should be considered
where quantitative accuracy is the major concern. It has however been recommended that more
than one metric be considered in order to achieve reasonable result when simulations of both
trend and quantitates are involved (Moriasi et al., 2007). The results obtained in this study
should be useful basis for the selection of appropriate GCMs for the locations for further
application of suitable downscaling procedures (Troin et al., 2015) before using the GCM output
data for climate change projection. In particular correction, forward time-shift of the predicted
series of rainfall amounts by 4 to 6 months, depending on the GCM, might be necessary in
further downscaling of the predicted rainfall data series (Amodu, 2016).

4. Conclusion
None of the 20 GCMs was superior in performance across the four stations in this study. The
models GCMs however generally simulated the trends of observed average monthly air
temperature better with positive correlation coefficients in contrast with their monthly rainfall
predictions which correlated negatively with observed data. Simulated rainy season also
commenced 4 to 6 months earlier than observed depending on the particular GCM. The results
obtained from the ranking of the GCMs would however be useful in the selection of appropriate
GCMs for the locations for appropriate downscaling procedures before their application to
climate change projection.

Acknowledgements
The authors are grateful for the technical assistance and data access granted by the Model and
Data Group, German Climate Computing Centre i.e. Deutsches Klimarechenzentrum (DKRZ),
Harmburg.

References
Amodu, MF. 2016. Evaluation of General Circulation Models and Application to Climate
Change in Water Resources of Hadejia-Jamaare-Komadugu-Yobe River Basin, Nigeria. Ph.D.
Thesis, Department of Agricultural and Biosystems Engineering, University of Ilorin, Ilorin,
Nigeria.

265
Amodu and Ejieji: Performance of some general circulation models on predicting temperature and
rainfall in the sudan-sahel region of Nigeria.
AZOJETE, 13(2):252-267. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Chiew, FHS. Kirono, DGC. Kent, D. and Vaze, J. 2009. Assessment of rainfall simulations from
global climate models and implications for climate change impact on runoff studies. 18th World
IMACS / MODSIM Congress, Cairns, Australia.
http://www.mssanz.org.au/modsim09/I13/chiew.pdf. (Accessed 04/08/2016).

Christensen, JH., Hewitson, B. Busuioc, A., Chen, A., Gao, X., Held, I., Jones, R., Kolli, RK.,
Kwon, WT., Laprise, R., Magaa RV., Mearns, L., Menndez, CG., Risnen, J., Rinke, A.,
Sarr, A. and Whetton, P. 2007. Regional Climate Projections. In: Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report
of the Intergovernmental Panel on Climate Change [Solomon, S., Qin, D., Manning, M., Chen,
Z., Marquis, M., Averyt, KB., Tignor, M. and Miller, HL. (eds.)]. Cambridge University Press,
Cambridge, United Kingdom and New York, NY, USA.

Ejieji, CJ. 2004. Evaluation of two models for the distribution of daily rainfall amounts in
Nigeria. Journal of Agricultural Research and Development, 3: 61-74.

IPCC-TGICA, 2007. General Guidelines on the Use of Scenario Data for Climate Impact and
Adaptation Assessment. Version 2. Prepared by T.R. Carter on behalf of the Intergovernmental
Panel on Climate Change, Task Group on Data and Scenario Support for Impact and Climate
Assessment, 66pp.

Lake Chad Basin Commission. 2006. Country National Action Programme for Integrated Water
Resources Management in Lake Chad Basin in Nigeria, by Technical Task Team for LCBC-GEF
Project on Reversal of Land and Water Degradation Trends in the Lake Chad Basin Ecosystem:
Establishment of Mechanisms for Land and Water Management, August 2006.

Masanganise, J., Chipindu, B., Mhizha, T., Mashonjowa, E. and Basira, K. 2013. An evaluation
of the performances of Global Climate Models (GCMs) for predicting temperature and rainfall in
Zimbabwe. International Journal of Scientific and Research Publications, 3(8): 1-11.

Moriasi, DN., Arnold, JG., Van Liew, MW., Bingner, RL., Harmel, RD. and Veith, TL. 2007.
Model Evaluation Guidelines for Systematic Quantification of Accuracy in Watershed
Simulations. Transactions of the American Society of Agricultural and Biological Engineers,
50(3): 885900.

Meehl, GA., Covey, C. McAvaney, B., Latif, M. and Stouffer, R. 2005. Overview of the
Coupled Model Intercomparison Project. Bulletin of the American Meteoreorological Society,
86: 89 93.

266
Arid Zone Journal of Engineering, Technology and Environment, April, 2017; Vol. 13(2):252-267
ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng

Meehl, GA., Covey, C., Delworth, T., Latif, M., McAvaney, B., Mitchell, JFB., Stouffer, RJ. and
Taylor, KE. 2007. The WCRP CMIP3 Multimodel Dataset: A new era in climate change
research. Bulletin of the American Meteorological Society, 88: 1383-1394, doi:10.1175/BAMS-
88-9-1383

Nakicenovic, N., Alcamo, J., Davis, G., de Vries, B., Fenhann, J., Gaffin, S., Gregory, K.,
Grbler, A., Jung, TY., Kram, T., La Rovere, EL., Michaelis, L., Mori, S., Morita, T., Pepper,
W., Pitcher, H., Price, L., Raihi, K., Roehrl, A., Rogner, HH., Sankovski, A., Schlesinger, M.,
Shukla, P., Smith, S., Swart, R., van Rooijen, S., Victor, N. and Dadi, Z. 2000. IPCC Special
Report on Emissions Scenarios, Cambridge University Press, Cambridge, United Kingdom
599pp.

Troin, M., Velzquez, JA., Caya, D. and Brissette, F. 2015. Comparing statistical post processing
of regional and global climate scenarios for hydrological impacts assessment: A case study of
two Canadian catchments. Journal of Hydrology 520: 268 288.

267

You might also like