The results obtained are demonstrated and discussed in the following subsections.
4.1 Selection of reference data
Based on the output from the Taylor diagram, the precipitation and temperature variables of NCEP-DOE reanalyses model has the highest correlation coefficient (CC) with values of 0.774 and 0.783 respectively compared to ERA5 and NCEP1 models (Table 2). The ration of the standard deviation (SD) of the reanalyses models to the observed dataset gave a good performance for NCEP-DOE reanalyses model. The ratio of the precipitation and the temperature is 0.686 and 0.72 respectively. The NCEP-DOE model performs better than the other models based on the performance indices as shown in Figure 3.
Table 2 Values of correlation coefficient of reanalysis models relative to observation
VARIABLE
|
MODEL
|
RATIO
|
CORRELATION OF COEFFICIENT
|
Precipitation
|
ERA-5
NCEP-DOE
NCEP1
|
0.491
0.686
0.699
|
0.773
0.774
0.701
|
Temperature
|
ERA-5
NCEP-DOE
NCEP1
|
0.619
0.72
0.697
|
0.584
0.783
0.623
|
Having a good performance, NCEP DOE reanalysis model has never been evaluated in the study region but has been applied around the world including sub-Saharan Africa where the study area lies (Zhan et al., 2016). Some reanalysis models have been used in Nigeria such as the Global Precipitation Climatology Centre (GPCC-V6) and Climatic Research Unit (CRU - version TS v. 3.23) (Shiru et al., 2021). However, the precipitation variable of the GPCC, the University of Delaware (UDEL) and NOAA’s Gridded Precipitation Reconstruction Over Land (PREC/L) have poor performance in the Northern part of Nigeria during the dry season but performed better during the wet season (Ogunjo et al., 2022).
4.2 Selection of representative GCM
4.2.1 Changes based on climate means
Changes in projected annual mean air temperature (∆T) and annual precipitation (∆P) for the short-term and long-term period with respect to the reference period are the bases for the initial selection process. We selected the five models closest to the 10th and 90th percentile value of temperature and precipitation representing cold and dry, hot and dry, and hot and wet corner (Figure 2). Some models show a close proximity and cluster around the vicinity of the respective corners, while others show a larger proximity and scatter around a corner. For example, the warm and dry corner for SSP4.5 and SSP8.5 during the short-term period, and the warm and dry, and warm and wet corner of SSP8.5 during the long-term period (Figure 2A, 2C and 2D) showed a close proximity. However, the cold and dry corner of SSP8.5 during the short-term and long-term period showed a scattered and a distant proximity to the corner. Interestingly, all CanESM5 models aligned towards the warm and wet corners for all study periods, all FGOALS-g3 and HadGEM3-GC31_LL ensembles aligned towards the warm and dry corner for all period except SSP8.5 during the long-term period where UKESM1-0 replaces FGOALs-g3. Majority of MPI-ESM1-2-LR ensembles occupies the cold and dry corner for all periods.
Values from the five models selected from each corner yielded a range of changes from the reference data set. The ∆T and ∆P range of projection for SSP8.5 are higher than SSP4.5 during the same period. The projection for ∆T and ∆P during the short-term ranges from 0.1-2.7⁰C and -2.6–4.4% respectively for SSP4.5 (Figure 2A), whereas for SSP8.5 (Figure 2C), the ranges are 0.1-3.5⁰C and -2.8-5.8% respectively. Likewise, the ranges of ∆T and ∆P during the long-term period are 0.3-2.8⁰C and -2.6-5.7% respectively for SSP4.5 (Figure 2B), whereas for SSP8.5 (Figure 2D), the ranges are 0.4-4.5⁰C and -1.3-7.7% respectively. Similarly, the ranges of projection for ∆T and ∆P is much higher for the long-term period compared to the short-term period during the same forcing pathway. Considering SSP4.5 (Figure 2A), the ∆T and ∆P ranges are 0.1-2.7⁰C and -2.6-4.4%, respectively, during the short-term period, whereas the ranges are 0.3-2.8⁰C and -2.6-5.7%, respectively, during the long-term period (Figure 2B). Likewise, considering SSP8.5, the ranges are 0.1-3.5⁰C and -2.8-5.8% during the short-term period (Figure 2C), whereas the ranges are 0.4-4.5⁰C and -1.3-7.7% during the long-term period (Figure 2D), respectively. This step leads the selection approach; this may result in reducing a substantial number of models for subsequent steps. The non-selected models might have good skills in simulating the climate extreme or the past climate. Another issue is the fact that the projected changes are averaged over the study area, which may result in dilution of spatial variation in projected changes. However most studies covering large areas exhibit similar methodology (Lutz et al., 2016).
4.2.2 Changes based on climate extremes
The models selected from the initial selection process (step 1) were subjected to changes in two ETCCDI indices for both temperature and precipitation for short-term period (2021-2050), long-term period (2051-2080) and the reference period (1981-2010). In this process, two models runs having highest combined scores from the combined temperature and precipitation indices were selected from each corner. In the event that multiple models runs have the same second highest combined score, more than two models are selected for that corner (Table 3). It is observed that, when computing combine score, some models with highest value change in one ETCCDI indices are not selected as a result of having a lower change in another ETCCDI indices. For example, in the warm and wet corner of SSP8.5, for the long-term period, the model CanESM5_r4i1p1f1 projects the highest change in R95pTOT, however, CanESM5_r6i1p1f1 and CanESM5_r9i1p1f1 were selected instead, due to its lower combined score. From (Table 3), models projecting large changes in mean air temperature results in projecting large changes in WSDI indices. Similarly, large changes in precipitation sum also project large changes in R95pTOT. For example, for SSP4.5, FGOALS-g3_r2i1p1f1 projected the largest increase in mean air temperature likewise show the largest increase in WSDI. Moreover, HadGEM3-GC31-LL_r1i1p1f2 shows a good combined score and present in all the warm and dry corner of the entire study period for both SSP4.5 and SSP8.5. Considering all the models relative to the corner and the season they appear, CDD is expected to decrease up to 112.7% in short-term, 113% in the long-term considering SSp4.5. Likewise, CDD is also expected to decrease up to 110.9% in short-term, 116.5% in long-term period considering SSP8.5. In addition, R95pTOT shows an increasing trend up to 137.8% in short-term, 139.1% in long-term considering SSP4.5. Similarly, R95pTOT increase up to 168.7% in short-term, 212.6% in long-term considering SSP8.5. Decrease in CDD and an Increase in R95pTOT predicts more wet days during rainfall seasons. The WSDI is projected to increase up to 3758.9% in short-term, 4159.0% in long-term considering SSP4.5. Likewise, WSDI is projected to increase by 4175.1% in short-term, 4692.3% in long-term considering SSP8.5. The CSDI is projected to decrease for all the study periods with the exception of EC-Earth-Veg_r1i1p1f1 which shows an increase of 9% with respect to the reference period. Increase in WSDI indicates an increase in maximum temperature, while a decrease in CSDI implies warm nights are expected in the future periods. Similar to step 1, area-averaged was applied to this step. This problem can be solved by dividing the area into homogenous entities.
Table 3 GCMs selected based on changes in ETCCDI indices. Model runs selected for step 3 are coloured with yellow
Seasonal projection
|
GCM Model Runs
|
∆CDD(%)
|
∆r95ptot(%)
|
∆WSDI(%)
|
∆CSDI(%)
|
∆T(°C)
|
∆P(%)
|
T Index rank
|
P index rank
|
Average Score
|
|
|
SSP4.5 (2021-2050)
|
|
|
|
|
|
|
|
FGOALS-g3_r1i1p1f1
|
-17.7
|
55.9
|
5149.8
|
-99.9
|
2.6
|
-2.3
|
3
|
4
|
3.5
|
|
FGOALS-g3_r2i1p1f1
|
-22.8
|
56.3
|
5228.9
|
-99.5
|
2.8
|
-2.2
|
5
|
1
|
3
|
warm and dry
|
FGOALS-g3_r3i1p1f1
|
-21.3
|
55.2
|
5191.2
|
-98.5
|
2.7
|
-2.3
|
4
|
2
|
3
|
|
FGOALS-g3_r4i1p1f1
|
-19.5
|
58.7
|
5121.5
|
-99.9
|
2.7
|
-2.7
|
1
|
3
|
2
|
|
HadGEM3-GC31-LL_r1i1p1f3
|
-14.4
|
53.2
|
5135.7
|
-13.8
|
2.1
|
-1.7
|
2
|
5
|
3.5
|
|
MIROC-ES2L_r1i1p1f2
|
-30.3
|
65.5
|
5284.6
|
-100.0
|
2.5
|
4.4
|
5
|
1
|
3
|
|
CanESM5_r5i1p1f1
|
-16.1
|
68.1
|
3751.4
|
-92.7
|
1.9
|
4.3
|
4
|
4
|
4
|
warm and wet
|
CanESM5_r7i1p1f1
|
-12.1
|
69.9
|
3711.4
|
-98.0
|
1.9
|
4.2
|
3
|
5
|
4
|
|
CanESM5_r10i1p1f1
|
-13.8
|
67.8
|
3554.7
|
-90.8
|
1.8
|
4.3
|
1
|
3
|
2
|
|
CanESM5_r22i1p1f1
|
-10.0
|
66.6
|
3673.6
|
-85.6
|
1.8
|
4.3
|
2
|
2
|
2
|
|
INM-CM4-8_r1i1p1f1
|
-24.0
|
-13.7
|
713.2
|
-91.3
|
0.1
|
-1.9
|
1
|
1
|
1
|
|
MPI-ESM1-2-LR_r6i1p1f1
|
0.8
|
-47.5
|
2460.2
|
-83.9
|
0.7
|
-1.9
|
3
|
4
|
3.5
|
cold and dry
|
CNRM-CM6-1_r1i1p1f2
|
1.7
|
30.4
|
2535.3
|
-65.4
|
0.4
|
-1.6
|
5
|
5
|
5
|
|
MPI-ESM1-2-LR_r3i1p1f1
|
-0.2
|
-44.1
|
2520.2
|
-89.5
|
0.7
|
-1.8
|
2
|
2
|
2
|
|
MPI-ESM1-2-LR_r5i1p1f1
|
0.6
|
-45.3
|
2398.5
|
-78.6
|
0.6
|
-1.7
|
4
|
3
|
3.5
|
SSP8.5 (2021-2050)
|
FGOALS-g3_r3i1p1f1
|
-19.5
|
58.6
|
5139.7
|
-100.0
|
3.2
|
-2.3
|
1
|
2
|
1.5
|
|
FGOALS-g3_r4i1p1f1
|
-17.9
|
60.4
|
5234.2
|
-99.2
|
3.3
|
-2.6
|
3
|
4
|
3.5
|
warm and dry
|
FGOALS-g3_r1i1p1f1
|
-18.0
|
60.9
|
5224.5
|
-99.8
|
3.2
|
-2.6
|
2
|
3
|
2.5
|
|
HadGEM3-GC31-LL_r1i1p1f3
|
-16.7
|
60.9
|
5792.8
|
-47.5
|
3.5
|
-1.9
|
5
|
5
|
5
|
|
FGOALS-g3_r2i1p1f1
|
-19.8
|
59.1
|
5236.4
|
-99.6
|
3.3
|
-2.8
|
4
|
1
|
2.5
|
|
CanESM5_r24i1p1f1
|
-10.7
|
69.0
|
3759.8
|
-90.6
|
2.9
|
5.4
|
2
|
3
|
2.5
|
|
CanESM5_r12i1p2f1
|
-13.0
|
65.9
|
3171.8
|
-97.8
|
2.9
|
5.4
|
1
|
2
|
1.5
|
warm and wet
|
CanESM5_r3i1p1f1
|
-9.7
|
71.3
|
3903.5
|
-92.7
|
2.8
|
5.6
|
5
|
4
|
4.5
|
|
CanESM5_r11i1p1f1
|
-11.9
|
65.7
|
3887.6
|
-86.1
|
2.9
|
5.8
|
4
|
1
|
2.5
|
|
CanESM5_r16i1p1f1
|
-14.6
|
79.0
|
3783.7
|
-90.5
|
2.8
|
5.5
|
3
|
5
|
4
|
|
CNRM-CM6-1_r1i1p1f2
|
-4.2
|
36.1
|
2499.6
|
-68.3
|
1.7
|
-2.3
|
4
|
2
|
3
|
|
EC-Earth3-Veg_r1i1p1f1
|
23.7
|
-23.2
|
2237.3
|
193.8
|
0.1
|
-1.9
|
5
|
5
|
5
|
cold and dry
|
MPI-ESM1-2-LR_r5i1p1f1
|
0.1
|
-39.1
|
2038.6
|
-85.9
|
1.4
|
-1.5
|
2
|
3
|
2.5
|
|
MPI-ESM1-2-LR_r9i1p1f1
|
1.4
|
-40.6
|
2088.9
|
-75.7
|
1.3
|
-1.5
|
3
|
4
|
3.5
|
|
INM-CM4-8_r1i1p1f1
|
-20.6
|
-19.5
|
1071.1
|
-95.2
|
0.6
|
-1.4
|
1
|
1
|
1
|
SSP4.5 (2051-2080)
|
FGOALS-g3_r3i1p1f1
|
-23.8
|
60.0
|
5598.1
|
-99.9
|
2.7
|
-2.4
|
1
|
1
|
1
|
|
FGOALS-g3_r4i1p1f1
|
-19.3
|
60.9
|
5688.0
|
-99.9
|
2.7
|
-2.4
|
2
|
2
|
2
|
warm and dry
|
FGOALS-g3_r2i1p1f1
|
-16.1
|
61.2
|
5714.0
|
-100.0
|
2.8
|
-2.6
|
4
|
4
|
4
|
|
FGOALS-g3_r1i1p1f1
|
-19.1
|
60.0
|
5688.7
|
-100.0
|
2.8
|
-2.6
|
3
|
3
|
3
|
|
HadGEM3-GC31-LL_r1i1p1f3
|
-14.6
|
52.3
|
5771.0
|
-80.6
|
2.3
|
-1.1
|
5
|
5
|
5
|
|
CanESM5_r3i1p1f1
|
-11.0
|
70.3
|
4957.3
|
-95.1
|
2.0
|
5.5
|
1
|
5
|
3
|
|
CanESM5_r6i1p2f1
|
-13.0
|
62.4
|
5086.6
|
-97.5
|
2.1
|
5.1
|
5
|
1
|
3
|
warm and wet
|
CanESM5_r8i1p1f1
|
-13.9
|
68.5
|
4966.2
|
-94.0
|
2.0
|
5.4
|
2
|
2
|
2
|
|
CanESM5_r9i1p1f1
|
-13.0
|
69.4
|
5065.1
|
-99.0
|
2.0
|
5.7
|
4
|
3
|
3.5
|
|
CanESM5_r1i1p1f1
|
-12.3
|
69.8
|
4978.1
|
-96.8
|
1.9
|
5.6
|
3
|
4
|
3.5
|
|
INM-CM4-8_r1i1p1f1
|
-24.6
|
-5.0
|
1383.0
|
-99.1
|
0.3
|
-2.0
|
1
|
1
|
1
|
|
CNRM-CM6-1_r1i1p1f2
|
-1.3
|
22.8
|
4396.9
|
-97.1
|
0.4
|
-1.4
|
4
|
3
|
3.5
|
cold and dry
|
MPI-ESM1-2-LR_r9i1p1f1
|
-2.2
|
-43.7
|
3678.9
|
-98.1
|
0.5
|
-1.4
|
3
|
2
|
2.5
|
|
MPI-ESM1-2-LR_r3i1p1f1
|
1.0
|
-42.9
|
3776.2
|
-95.3
|
0.5
|
-1.4
|
5
|
4
|
4.5
|
|
MPI-ESM1-2-LR_r10i1p1f1
|
2.7
|
-40.6
|
3724.9
|
-98.4
|
0.5
|
-1.3
|
2
|
5
|
3.5
|
SSP8.5 (2051-2080)
|
HadGEM3-GC31-LL_r1i1p1f3
|
-14.0
|
61.4
|
6493.5
|
-95.0
|
4.5
|
-1.3
|
5
|
3
|
4
|
|
UKESM1-0_r2i1p1f2
|
-19.9
|
63.6
|
6249.7
|
-84.8
|
4.1
|
-0.8
|
3
|
2
|
2.5
|
warm and dry
|
UKESM1-0_r3i1p1f2
|
-22.9
|
70.2
|
6328.3
|
-88.4
|
4.2
|
-0.7
|
4
|
1
|
2.5
|
|
UKESM1-0_r8i1p1f2
|
-9.1
|
64.6
|
6055.0
|
-85.8
|
3.9
|
-0.7
|
1
|
5
|
3
|
|
UKESM1-0_r1i1p1f2
|
-12.7
|
64.3
|
6223.0
|
-89.8
|
4.0
|
-0.6
|
2
|
4
|
3
|
|
CanESM5_r22i1p1f1
|
-13.4
|
87.2
|
6414.8
|
-100.0
|
4.2
|
7.4
|
4
|
1
|
2.5
|
|
CanESM5_r6i1p1f1
|
-11.4
|
89.2
|
6377.4
|
-97.2
|
4.1
|
7.5
|
3
|
4
|
3.5
|
warm and wet
|
CanESM5_r9i1p1f1
|
-10.4
|
87.5
|
6423.7
|
-100.0
|
4.2
|
7.2
|
5
|
3
|
4
|
|
CanESM5_r12i1pf1
|
-11.4
|
87.2
|
6372.5
|
-100.0
|
4.0
|
7.4
|
2
|
1
|
1.5
|
|
CanESM5_r4i1p1f1
|
-9.4
|
91.9
|
6315.1
|
-100.0
|
4.1
|
7.7
|
1
|
5
|
3
|
|
INM-CM4-8_r1i1p1f1
|
-27.3
|
-2.0
|
2924.7
|
-99.5
|
1.5
|
-0.7
|
2
|
2
|
2
|
|
INM-CM5-0_r1i1p1f1
|
-31.2
|
11.6
|
2868.8
|
-99.9
|
1.6
|
-0.7
|
1
|
1
|
1
|
cold and dry
|
MPI-ESM1-2-HR_r1i1p1f1
|
5.1
|
-13.4
|
5082.7
|
-97.8
|
2.2
|
-1.2
|
4
|
3
|
3.5
|
|
MPI-ESM1-2-HR_r2i1p1f1
|
5.2
|
-5.2
|
5103.1
|
-98.3
|
2.1
|
-0.8
|
3
|
4
|
3.5
|
|
EC-Earth3-Veg_r1i1p1f1
|
19.6
|
23.9
|
3542.9
|
-12.9
|
0.4
|
-0.4
|
5
|
5
|
5
|
4.2.3 Selection based on past performance skill
The models are subjected to a validation process based on their past performance skill in reproducing the NCEP-DOE reference dataset. Table 4 list the skill score for the GCM runs available after step 2. Table 5 list the combined score ranking for air temperature and precipitation resulting in the final ensemble selection (Table 6). From the precipitation and temperature skill score, it can be seen that the models that show good skill in simulating past climate for a particular subdomain, show equally a good skill in simulating the remaining subdomain. For example, HadGEM3-GC31-LL_r1i1p1f3, MPI-ESM1-2-LR_r9i1p1f1 and MPI-ESM1-2-LR_r10i1p1f1 exhibits good skill in all subdomain. Similarly, models with low skill for a particular subdomain show similar trend in another subdomain, for example CanESN5_r3i1p1f1.
CanESM5 model runs exhibit skill in capturing the warm and wet corner of the seasonal projection of the study area. The model however shows relatively good skill in simulating the past temperature for all subdomain but are poor in simulating past annual precipitation cycle in the entire subdomain. HadGem33-GC31_LL_r1i1p1f3, FGOALS, and UKESM1-0 models falls under the warm and dry corner. All the models show good skills in past performance for all subdomain. HadGEM3_GC3-1_LL_r1i1p1f3 show outstanding characteristics for both SSP4.5 and SSP8.5 and for all warm and dry corner by having a final skill score of 1. Similarly, MPI-ESM1-2-LR and MPI-ESM1-2-HR show good skill in representing both temperature and precipitation in the cold and dry corner.
Finally, Table 6 reveals that, the skilled GCM for a respective corner do not vary with radiative forcing scenarios (SSP4.5 and SSP8.5) as well as analysis period (short-term and long-term period), for example CanESM5 is selected in all warm and wet corner. However, the GCMs only exhibit different variant-ID in the CMIP6 metadata model runs, for example in the r1i1p1f1 nomenclature. The Realization (r) varies from r1-r10, with all Initialization (i) and Physics (p) as 1, and the Forcing (f) as 1 or 3 for all the model ensembles.
Table 4 Temperature and precipitation skill scores for the three subdomains
GCM runs
|
Precipitation Skill Score
|
Temperature Skill Score
|
|
ULN
|
ULB
|
SHJ
|
ULN
|
ULB
|
SHJ
|
CanESM5_r3i1p1f1
|
0.06
|
0.16
|
0.12
|
0.34
|
0.32
|
0.41
|
CanESM5_r5i1p1f1
|
0.07
|
0.27
|
0.15
|
0.47
|
0.44
|
0.49
|
CanEsm5_r6i1p1f1
|
0.05
|
0.16
|
0.09
|
0.39
|
0.33
|
0.28
|
CanESM5_r7i1p1f1
|
0.04
|
0.12
|
0.09
|
0.38
|
0.4
|
0.35
|
CanESM5_r9i1p1f1
|
0.08
|
0.23
|
0.14
|
0.42
|
0.49
|
0.45
|
CanESM5_r1i1p1f1
|
0.05
|
0.21
|
0.11
|
0.29
|
0.46
|
0.39
|
CanESM5_r16i1p1f1
|
0.06
|
0.15
|
0.11
|
0.27
|
0.35
|
0.38
|
CNRM-CM6-1_r1i1p1f2
|
0.27
|
0.27
|
0.25
|
0.43
|
0.44
|
0.42
|
HadGEM3-GC31-LL_r1i1p1f3
|
0.44
|
0.44
|
0.49
|
0.67
|
0.64
|
0.68
|
UKESM1-0-LL_r1i1p1f2
|
0.41
|
0.37
|
0.57
|
0.65
|
0.58
|
0.71
|
UKESM1-0-LL_r8i1p1f2
|
0.41
|
0.43
|
0.56
|
0.65
|
0.66
|
0.77
|
EC-Earth3-VEG_r1i1p1f1
|
0.11
|
0.1
|
0.09
|
0.36
|
0.33
|
0.29
|
FGOALS_g3_r1i1p1f1
|
0.36
|
0.13
|
0.25
|
0.34
|
0.27
|
0.41
|
FGOALS_g3_r2i1p1f1
|
0.28
|
0.08
|
0.22
|
0.29
|
0.22
|
0.33
|
FGOALS-g3_r4i1p1f1
|
0.3
|
0.05
|
0.16
|
0.2
|
0.19
|
0.29
|
MPI-ESM1-2-HR_r1i1p1f1
|
0.3
|
0.22
|
0.18
|
0.41
|
0.39
|
0.36
|
MPI-ESM1-2-HR_r2i1p1f1
|
0.3
|
0.22
|
0.17
|
0.44
|
0.41
|
0.4
|
MPI-ESM1-2-LR_r3i1p1f1
|
0.38
|
0.37
|
0.36
|
0.55
|
0.53
|
0.57
|
MPI_ESM1-LR_r5i1p1f1
|
0.23
|
0.22
|
0.18
|
0.56
|
0.51
|
0.47
|
MPI-ESM1-2-LR_r6i1p1f1
|
0.24
|
0.23
|
0.25
|
0.41
|
0.43
|
0.39
|
MPI-ESM1-LR_r10i1p1f1
|
0.41
|
0.41
|
0.33
|
0.67
|
0.69
|
0.59
|
MPI-ESM1-LR_r9i1p1f1
|
0.45
|
0.45
|
0.4
|
0.55
|
0.58
|
0.53
|
As stated earlier, some models with good skills in simulating the past climate might have been de-selected in the previous steps. Another limitation has to do with the final skill score calculation, which combines the precipitation and temperature skill scores.
The CMIP6 models have gained global applications for impact studies including Africa. However, the findings of the current study do not support a research conducted by Shiru et al. (2021) in Nigeria. They evaluated 13 historical GCMs each for minimum and maximum temperature and precipitation using only the first ensemble member (r1i1p1f1) of the models. Their ranking involve the application of compromise programming (CP) to rank the GCMs, while NRMSE, Pbias, NSE, R2 , and VE statistical performance measures were used for the evaluation. The concluded that, IPSL-CM6A-LR, NESM3, CMCC-CM2-SR5, and ACC-ESM1.5 have highest skills in simulating precipitation. For maximum temperature, the models INM-CM4-8, BCC-CSM2-MR, MRI-ESM2-0, and ACCESS-ESM1-5 performed better, while AWI-CM-1–1-MR, IPSL-CM6A-LR, INM-CM5-0, and CanESM5 are skilful in simulating the minimum temperature. Among these models, only CanESM5 was found to have good skills in this present study. A possible explanation for this result might be because of the different reference data sets used for both studies. In addition, in their study, minimum and maximum temperature were used separately, while ours considers the mean temperature. Another possibility is the presence of heterogeneous climate in Nigeria, which our study considers by focusing in the Northern part of the country having homogenous climate.
Table 5 Final ranking of GCM runs for the three subdomains
Period
|
SSP
|
Seasonal Projection
|
GCM Runs
|
Average score per subdomain
|
Rank per subdomain
|
Total
|
Final rank
|
|
|
|
|
ULN
|
ULB
|
SHJ
|
ULN
|
ULB
|
SHJ
|
|
|
2021-2051
|
4.5
|
Warm and Dry
|
FGOALS_g3_r1i1p1f1
|
0.35
|
0.20
|
0.33
|
13
|
26
|
12
|
51
|
18
|
HadGEM3-GC31-LL_r1i1p1f3
|
0.56
|
0.54
|
0.59
|
1
|
3
|
3
|
7
|
1
|
Warm and Wet
|
CanESM5_r5i1p1f1
|
0.27
|
0.36
|
0.32
|
18
|
13
|
14
|
45
|
13
|
CanESM5_r7i1p1f1
|
0.21
|
0.26
|
0.22
|
25
|
20
|
25
|
70
|
22
|
Cold and Dry
|
CNRM-CM6-1_r1i1p1f2
|
0.35
|
0.36
|
0.34
|
13
|
13
|
10
|
36
|
11
|
MPI_ESM1-2-LR_r5i1p1f1
|
0.40
|
0.37
|
0.33
|
10
|
10
|
13
|
33
|
10
|
MPI-ESM1-2-LR_r6i1p1f1
|
0.33
|
0.33
|
0.32
|
16
|
17
|
14
|
47
|
16
|
2051-2080
|
4.5
|
Warm and Dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
0.56
|
0.54
|
0.59
|
1
|
3
|
3
|
7
|
1
|
FGOALS_g3_r2i1p1f1
|
0.29
|
0.15
|
0.28
|
17
|
27
|
19
|
63
|
20
|
Warm and Wet
|
CanESM5_r9i1p1f1
|
0.25
|
0.36
|
0.30
|
19
|
11
|
16
|
46
|
14
|
CanESM5_r1i1p1f1
|
0.17
|
0.34
|
0.25
|
27
|
16
|
22
|
65
|
21
|
Cold and Dry
|
CNRM-CM6-1_r1i1p1f2
|
0.35
|
0.36
|
0.34
|
13
|
13
|
10
|
36
|
11
|
MPI-ESM1-2-LR_r3i1p1f1
|
0.47
|
0.45
|
0.47
|
9
|
9
|
8
|
26
|
9
|
MPI-ESM1-2-LR_r10i1p1f1
|
0.54
|
0.55
|
0.46
|
5
|
1
|
9
|
15
|
6
|
2021-2050
|
8.5
|
Warm and Dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
0.56
|
0.54
|
0.59
|
1
|
3
|
3
|
7
|
1
|
FGOALS-g3_r4i1p1f1
|
0.25
|
0.12
|
0.23
|
19
|
28
|
24
|
71
|
24
|
Warm and Wet
|
CanESM5_r3i1p1f1
|
0.20
|
0.24
|
0.27
|
26
|
23
|
21
|
70
|
22
|
CanESM5_r16i1p1f1
|
0.17
|
0.25
|
0.25
|
27
|
21
|
23
|
71
|
24
|
Cold and Dry
|
EC-Earth3-VEG_r1i1p1f1
|
0.24
|
0.22
|
0.19
|
22
|
24
|
26
|
72
|
26
|
MPI-ESM1-2-LR_r9i1p1f1
|
0.50
|
0.52
|
0.47
|
8
|
7
|
7
|
22
|
8
|
2051-2080
|
8.5
|
Warm and Dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
0.56
|
0.54
|
0.59
|
1
|
3
|
3
|
7
|
1
|
UKESM1-0-LL_r1i1p1f2
|
0.53
|
0.48
|
0.64
|
6
|
8
|
2
|
16
|
7
|
UKESM1-0-LL_r8i1p1f2
|
0.53
|
0.55
|
0.67
|
6
|
2
|
1
|
9
|
5
|
Warm and Wet
|
CanEsm5_r6i1p1f1
|
0.22
|
0.25
|
0.19
|
24
|
22
|
28
|
74
|
28
|
CanESM5_r9i1p1f1
|
0.25
|
0.36
|
0.30
|
19
|
11
|
16
|
46
|
14
|
Cold and Dry
|
EC-Earth3-VEG_r1i1p1f1
|
0.24
|
0.22
|
0.19
|
22
|
24
|
26
|
72
|
26
|
MPI-ESM1-2-HR_r1i1p1f1
|
0.36
|
0.31
|
0.27
|
12
|
19
|
20
|
51
|
18
|
MPI-ESM1-2-HR_r2i1p1f1
|
0.37
|
0.32
|
0.29
|
11
|
18
|
18
|
47
|
16
|
Among related studies in Africa, Gebresellase et al. (2022) adopted the advanced envelop-based approach in the Upper Awash basin (UAB) of Ethiopia and selected the skilled models. Similar to their study, this present study also found that MPI_ESM1-2-LR model is skilful in simulating the cold and dry seasons. Moreover, CanESM5 model was selected for having good performance in both studies. Despite having similarities in skilled models, the model differ in their ensemble members.
The selected models runs were downscaled from their coarser resolution to a finer resolution for impact studies.
Table 6 Final selected GCM runs with averaged projected changes in precipitation, mean annual temperature and ETCCDI indices over the study area for the two forcing scenarios (SSP4.5 and SSP8.5) between 2021-2050 and 2051-2080.
Period
|
SSP
|
Seasonal Projection
|
GCM Runs
|
∆CDD
(%)
|
∆r95ptot
(%)
|
∆WSDI
(%)
|
∆CSDI
(%)
|
∆T
(°C)
|
∆P
(%)
|
|
|
Warm and dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
-14.4
|
53.2
|
5135.7
|
-13.8
|
2.1
|
-1.7
|
2021-2050
|
4.5
|
Warm and Wet
|
CanESM5_r5i1p1f1
|
-16.1
|
68.1
|
3751.4
|
-92.7
|
1.9
|
4.3
|
|
|
Cold and Dry
|
MPI_ESM1-2-LR_r5i1p1f1
|
0.6
|
-45.3
|
2398.5
|
-78.6
|
0.6
|
-1.7
|
|
|
Warm and dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
-14.6
|
52.3
|
5771.0
|
-80.6
|
2.3
|
-1.1
|
2051-2080
|
4.5
|
Warm and Wet
|
CanESM5_r9i1p1f1
|
-13.0
|
69.4
|
5065.1
|
-99.0
|
2.0
|
5.7
|
|
|
Cold and Dry
|
MPI-ESM1-2-LR_r10i1p1f1
|
2.7
|
-40.6
|
3724.9
|
-98.4
|
0.5
|
-1.3
|
|
|
Warm and dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
-16.7
|
60.9
|
5792.8
|
-47.5
|
3.5
|
-1.9
|
2021-2050
|
8.5
|
Warm and Wet
|
CanESM5_r3i1p1f1
|
-9.7
|
71.3
|
3903.5
|
-92.7
|
2.8
|
5.6
|
|
|
Cold and Dry
|
MPI-ESM1-2-LR_r9i1p1f1
|
1.4
|
-40.6
|
2088.9
|
-75.7
|
1.3
|
-1.5
|
|
|
Warm and dry
|
HadGEM3-GC31-LL_r1i1p1f3
|
-14.0
|
61.4
|
6493.5
|
-95.0
|
4.5
|
-1.3
|
2051-2080
|
8.5
|
Warm and Wet
|
CanESM5_r9i1p1f1
|
-10.4
|
87.5
|
6423.7
|
-100.0
|
4.2
|
7.2
|
|
|
Cold and Dry
|
MPI-ESM1-2-HR_r2i1p1f1
|
5.2
|
-5.2
|
5103.1
|
-98.3
|
2.1
|
-0.8
|
4.3 GCM downscaling
Daily GCMs outputs were downscaled for short-term and long-term periods for each climate extreme corners on the 50*50km2 grid of NCEP-DOE dataset for the Northern Nigeria region (Figure 1). As described in section 3.3, the performance of the three downscaling techniques differs for temperature and precipitation variables. For all the precipitation variables, delta method performs better with higher index of agreement whereas six GCMs were downscaled using EQM and three using QM for the temperature variable.
Precipitation
The study area-raining season follows the West African rainfall characteristics, which prevails during the northern hemisphere summer (June through October). The precipitation is expected to increase in each of the summer months with higher intensities compared to the reference period (Figure 4C and D). For example, an increment of 16% and 23% in June for SSP4.5 and SSP8.5, respectively during the short-term period and a much higher increment of 101% and 102% in October rains is expected during the same period and scenario. Similarly, during the long-term period, the increment is expected in September and October months. An increment of 38% and 81% is expected during SSP4.5 and SSP8.5, respectively in September rains whereas 89% and 148% in October rains for the respective scenarios. Increase in precipitation predicts more rainfall is expected in the future leading to severe flooding events.
Average annual precipitation is also expected to increase for all the GCM models in the future. The average annual precipitation is expected to increase in the future: 8-17% during short-term (CanESM5_r5i1p1f1 predicts highest increase) and 11-35% during long-term (CanESM5_r9i1p1f1 predict highest increase), considering the SSP4.5 scenario. Higher precipitation is expected in the SSP8.5 scenario compared to SSP4.5 scenario because the annual precipitation is expected to increase by 5-38% in short-term (CanESM5_r3i1p1f1 predicts highest increase) and 29-60% in the long-term (CanESM5_r9i1p1f1 predict highest increase) period. Similarly, for the ensemble means (average of 3 GCMs) of the study period, the average annual precipitation is expected to increase in the future: In SSP4.5 during the short-term period, the expected average annual precipitation is 850mm/year yielding a 13% increase with respect to the reference dataset. Likewise, the average annual precipitation (% increase) of 906mm/year (20%), 926mm/year (23%) and 1061mm/year (41%) is expected during long-term (SSP4.5), short-term (SSP8.5) and long-term (SSP8.5) respectively.
The historic, present and future average annual precipitation values of the ensemble mean for SSP4.5 and SSP8.5 for the entire study period with their standard deviations is shown in Figure 5b. In Figure 5b, the dark orange and green lines represent the average values of the annual precipitation while the blue line represent the historic and present average precipitation. The light orange and green areas represent the standard deviation of the average precipitation for SSP4.5 and SSP8.5 respectively. The average annual precipitation values for the reference precipitation variable is 750mm/year.
4.3.2 Temperature
All the GCMs have predicted an increase in temperature with respect to the reference dataset (figure 4A and B). The seasonal increment during the harmattan season (December through February) with respect to the reference dataset varies from 0.2-1.0⁰C and 1.6-2.4⁰C for short-term and long-term periods, considering SSP4.5 scenario. The changes in SSP8.5 is much higher than SSP4.5 scenario with increment of 0.5-1.2⁰C and 2.0-2.7⁰C for short-term and long-term period respectively. Increase in harmattan temperature predicts warmer period during the cold season. In the same vein, during the hottest period (March through May), the average monthly mean temperature is expected to increase which varies from 0.2-1.0⁰C and 1.7-2.7⁰C for short-term and long-term respectively, considering SSP4.5 scenario. Similarly, and expected increment of 0.3-1.1⁰C and 2.0-3.0⁰C for short-term and long-term period, considering SSP8.5 scenario. The rainfall months (June to Spetember) is not an exception to this increment. An increase of 0.9-1.6⁰C and 2.2-2.8⁰C is expected for short-term and long-term period respectively, considering SSP4.5 scenario whereas 1.1-1.7⁰C and 2.3-3.2⁰C is expected to increase during the short-term and long-term period, considering SSP8.5 scenario. The temperature is expected to increase more compared to hot months and harmattan period.
The average annual mean temperature is expected to increase in the future: 0.26-1.6⁰C in short-term and 0.87-4.04⁰C in long-term, considering SSp4.5 scenario. The model CanESM5_r5i1p1f1 is expected to show the highest increment during the short-term while CanESM5_r9i1p1f1 show the highest increment during the long-term period. Higher temperature is expected to increase in SSP8.5 scenario compared to SSp4.5 scenario. The average annual temperature is expected to increase in the short-term by 0.01-1.78⁰C and 0.01-4.3⁰C in the long-term period. The model MPI-ESM1-2-LR_r9i1p1f1 and MPI-ESM1-2-HR_r2i1p1f1 show the least increment of 0.01⁰C whereas HadGEM3-GC31-LL_r1i1p1f3 show the highest increment of both 1.78⁰C and 4.3⁰C in the short-term and long-term period respectively.
Similarly, the ensemble mean (average of 3 GCMs) the mean annual temperature is also expected to increase for all the GCMs. In SSP4.5 during the short-term period, the expected increase in mean annual temperature is 1.1⁰C. Likewise, the mean annual temperature is expected to increase by 2.5⁰C, 1.2⁰C and 2.7⁰C during the long-term (SSP4.5), short-term (SSP8.5), and Long-term (SSP8.5) respectively. The ensemble mean predicts lower temperature increment compared to some of the prediction made by some models, example HadGEM3-GC31-LL_r1i1p1f3.
The historic, present and the future mean annual temperature values of the ensemble means with their standard deviations for SSP4.5 and SSP8.5 for the entire study period is shown in Figure 5a. The dark orange and green lines represent the average values of the mean annual temperature while the blue line represent the historic and the present value. The lighter orange and green areas represent the standard deviations of the mean annual temperature. The mean annual temperature of the reference dataset is 25.6⁰C.