Atmospheric Research 254 (2021) 105509

Contents lists available at ScienceDirect

Atmospheric Research
journal homepage: www.elsevier.com/locate/atmosres

Evaluation of CMIP6 models in simulating the statistics of extreme

precipitation over Eastern Africa
Akintomide Afolayan Akinsanola a, *, Victor Ongoma b, Gabriel J. Kooperman a
a
Department of Geography, University of Georgia, Athens, GA, USA
b
School of Geography, Earth Science and Environment, The University of the South Pacific, Laucala Campus Private Bag, Suva, Fiji

A R T I C L E I N F O A B S T R A C T

Keywords: The Eastern Africa region experiences frequent extreme precipitation events that can cause destruction of
ETCCDI property and environment, and loss of lives. Thus, there is a need to understand how these events may change in
CMIP6 the future and how well the global climate models that are used to make projections can simulate precipitation
Rainfall
extremes in this region before they can be used in downscaling or flood and drought impact assessment studies.
Extreme events
Climate change
In this work, we evaluated the ability of sixteen Coupled Model Intercomparison Project Phase 6 (CMIP6) models
to simulate present-day precipitation extremes over the Eastern Africa region during the two rainy seasons
(March–May and September–November). We used nine extreme precipitation indices (including seven (one)
indices of wet (dry) extremes) defined by the Expert Team on Climate Change Detection and Indices. The CMIP6
models were evaluated against two gridded observation datasets: Global Precipitation Climatology Project One-
Degree Daily Dataset and Tropical Rainfall Measuring Mission Multi-satellite Precipitation Analysis 3B42. Three
model performance metrics (percentage bias, normalized root-mean-square error, and pattern correlation coef­
ficient) were employed to further assess the strengths and weakness of the models. Our results show that the
multi-model ensemble mean generally provides a better representation of observed precipitation and related
extremes compared to individual models when considering all metrics and seasons. Several consistent biases are
evident across CMIP6 models, which tend to overestimate the total-wet day precipitation and consecutive wet
days, and underestimate very wet days and maximum 5-day precipitation in both seasons. Furthermore, no single
model consistently performs best, model performance varies with the season and index under consideration and
is generally independent of horizontal resolution.

1. Introduction starvation-related deaths (FEWS NET, 2011). A similar case was re­
ported during the 2015–2016 drought (Carty 2017). In addition to
Extreme weather and climate events can have significant impacts on drought, the region is also affected by wet extremes. Recently, the
the environment, livelihood, and economy around the world. Under­ United Nations Office for the Coordination of Humanitarian Affairs
standing the frequency, intensity, and spatial coverage of these extremes (OCHA 2019) reported that over 280 people died due to flooding caused
is a key step in averting the associated losses and thus calls for improving by the October – December 2019 heavy precipitation over Eastern Af­
the ability of climate models to realistically simulate these extremes. The rica. Many homes, infrastructure, and livelihoods were destroyed across
information is particularly crucial for developing countries due to their the region leaving more than 2.8 million people in need of humanitarian
low adaptive capacity as a result of limited access to information, assistance.
technology, finances, and capital assets (Sylla et al. 2016). In particular, To date, the region remains highly vulnerable to the effects of
the Eastern Africa region, where this study is focused, is prone to severe climate variability and change, largely due to its dependence on pre­
flooding and droughts (Nicholson 2017). These extreme events have cipitation for agricultural practices that sustains the economy (IPCC
resulted in loss of lives, destruction of property, and societal unrest. For 2007; Shimeles et al. 2018). Climate change is expected to increase the
instance, the severe drought event of 2011 led to crop failure over the frequency and intensity of extreme events on global-scales (Seneviratne
region and consequently resulted in significant food insecurity and et al., 2012). Given the socioeconomic impact of climate extremes,

* Corresponding author.
E-mail address: akinsanola@uga.edu (A.A. Akinsanola).

https://doi.org/10.1016/j.atmosres.2021.105509
Received 9 November 2020; Received in revised form 2 February 2021; Accepted 2 February 2021
Available online 5 February 2021
0169-8095/© 2021 Elsevier B.V. All rights reserved.
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

particularly wet and dry precipitation extremes, adequate planning to presented in Fig. 1. The region’s topography is highly heterogeneous,
mitigate these impacts is crucial. Although short-term weather forecasts and the weather and climate range from cool and wet highlands to semi-
have proved to be helpful in sub-seasonal planning (Coughlan de Perez arid lowlands (Ogwang et al. 2014). The climate of the region is affected
et al., 2019), there is a need for projections to provide long-term plan­ by many factors, including the migration of the Inter-Tropical Conver­
ning. Presently, the ongoing Severe Weather Forecasting Demonstration gence Zone (ITCZ), localized highlands and associated orographic cir­
Project (SWFDP – https://www.wmo.int/pages/prog/www/swfdp/), an culation, proximity to water bodies (Lake Victoria basin and Indian
initiative of the World Meteorological Organization (WMO), has helped Ocean), and large-scale atmospheric circulation (Nicholson 2017).
in forecasting climate extremes over the region. The project has shown In general, the mean annual precipitation over most of the region is
remarkable improvement in the lead-time and reliability for alerts and between 800 and 1200 mm, with higher precipitation observed during
warnings about high-impact events such as heavy precipitation, strong March–May (MAM, ‘long rain’) and September–November (SON, ‘short
winds, and high waves (Fitzpatrick 2015). However, there remains a gap rain’) seasons (Camberlin 2018). However, precipitation is highly var­
to be filled with regard to long-term predictions. iable in both time and space over the study area. On the western side
Global climate models have been utilized extensively to understand along the Equator, the region tends to experience precipitation all year
the past and future changes in climate events globally and regionally, round, which is partly influenced by moisture influx from the neigh­
and there is a growing use of these models beyond the scientific com­ boring Congo Basin (Yang et al. 2015). Contrary, there is a rain shadow
munity for climate impact applications and decision making (Dunning over the Eastern Rift Valley, which extends to Somalia and southeastern
et al. 2017; Wainwright et al., 2019; Rowell, 2019). Thus, a compre­ Kenya; these areas experience relatively drier conditions (about 700
hensive assessment of their performances is critical. Recently, analysis of mm) throughout the year. Furthermore, the Ethiopian Highlands into
experiments from the Coupled Model Intercomparison Project version 5 the north of the study area stands out as a relatively wet region,
(CMIP5) have been very impactful and numerous studies have evaluated recording most of its precipitation during June – September, locally
the performance of the CMIP5 models to reproduce precipitation char­ known as Kiremt. The Lake Victoria Basin, shared by Kenya, Uganda,
acteristics at global-scales (Seneviratne et al., 2012; Sillmann et al. and Tanzania records relatively high rainfall (1500–3000 mm) on an
2013; Kumar et al. 2013, 2014; Koutroulis et al. 2016; Nguyen et al. annual basis. This is mainly as a result of sea breezes originating from
2017) and some at regional- and local-scales (Wainwright et al. 2019; Lake Victoria. Further to the south, the Southern part of Tanzania re­
Rowell, 2019). A common result of many studies (Mehran et al. 2014; cords wet conditions as a result of the ITCZ presence during the austral
Sillmann et al. 2013; Samuels et al. 2018; Sonkoué et al., 2019) agrees summer period.
that the multimodel ensemble mean (MME) and median tend to repre­
sent precipitation characteristics better than individual models. How­ 2.2. Data
ever, despite some improvement of CMIP5 models as compared to the
previous version of CMIP3 at reproducing extreme precipitation, large We used the historical simulations of sixteen CMIP6 models (Eyring
uncertainties still exist, especially in tropical and subtropical regions et al. 2016), a detailed description is presented in Table 1. The choice of
(Kharin et al. 2013). In the Eastern Africa region in particular, CMIP5 sixteen models was based on data availability at the time of the analysis.
models have also shown generally good performance in capturing the
region’s precipitation characteristics based on the few available studies,
but considerable spatial and seasonal biases still exist (e.g.Sillmann et al.
2013; Yang et al. 2015; Zebaze et al., 2019).
The uncertainties in the models are partly due to their coarse hori­
zontal resolution (i.e., grid boxes are much larger than the scale of
convective events in the tropics/subtropics) and simplifications used in
the representation of physical processes (Taylor et al. 2012). Global
climate models that have contributed to the latest version of the Coupled
Model Inter-comparison Project (CMIP6; Eyring et al. 2016) include
updates to existing parameterizations, the addition of new physical
processes, and somewhat higher resolution relative to CMIP5. Assessing
how well CMIP6 models reproduce precipitation extremes over Eastern
Africa is crucial for determining whether these new model developments
have improved their ability to capture the essential physics controlling
convection and precipitation in the region. The goal of this study is to
evaluate the ability of CMIP6 models to represent the present-day
characteristics and spatial distribution of extreme precipitation over
the Eastern Africa region. We employ the standard set of extreme pre­
cipitation indices defined by the Expert Team on Climate Change
Detection and Indices (ETCCDI; Klein Tank et al. 2009; Zhang et al.
2011) and compare the model results with estimates from two gridded
daily observational datasets. Section 2 of this study describes the data
and methodology, the results and discussion are presented in Section 3,
and Section 4 gives a summary and conclusion of the study.

2. Data and methodology

2.1. Study area

The Eastern Africa region considered in this work covers nearly the Fig. 1. Map of Eastern Africa (Red rectangle). The color scale shows topog­
entire Great Horn of Africa (Kenya, Tanzania, Uganda, Rwanda, raphy in meters.
Burundi, Somalia, Djibouti, Eritrea, Ethiopia, Sudan, and South Sudan),
confined within latitude 10◦ S – 18◦ N and longitude 26◦ E – 54◦ E as

2
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Table 1
Information of the sixteen CMIP6 climate models used in this study.
S/N Model Institute Resolution (olon × olat) References

1 BCC-CSM2-MR Beijing Climate Center (BCC) and China Meteorological Administration (CMA) 1.13 × 1.13 Wu et al. (2018)
2 BCC-ESM1 Beijing Climate Center (BCC) and China Meteorological Administration (CMA) 2.81 × 2.81 Zhang et al. (2018)
3 CanESM5 Canadan Earth System Model 2.81 × 2.81 Swart et al. (2019)
4 CESM2 National Center for Atmospheric Research 1.25 × 0.94 Danabasoglu et al.
(2019a)
5 CESM2-WACCM National Center for Atmospheric Research 1.25 × 0.94 Danabasoglu et al.
(2019b)
6 CNRM-CM6-1 Centre National de Recherches Météorologiques (CNRM); Centre Européen de 1.41 × 1.41 Voldoire (2018)
Recherches et de Formation Avancéeen Calcul Scientifique
7 E3SM US Department of Energy 1.25 × 0.94 Bader et al. (2019)
8 EC-EARTH3 EC-EARTH consortium 0.70 × 0.70 EC-Earth (2019a)
9 EC-EARTH3-Veg EC-EARTH consortium 0.70 × 0.70 EC-Earth (2019b)
10 GFDL-CM4 Geophysical Fluid Dynamics Laboratory (GFDL) 2.50 × 2.00 Guo et al. (2018)
11 GFDL-ESM4 Geophysical Fluid Dynamics Laboratory (GFDL) 1.25 × 1.00 Krasting et al. (2018)
12 HadGEM3-GC31-LL Met Office Hadley Centre 1.86 × 1.25 Ridley et al. (2019)
13 IPSL-CM6A-LR Institute Pierre-Simon Laplace (IPSL) 2.50 × 1.26 Boucher et al. (2018)
14 MRI-ESM2-0 Meteorological Research Institute (MRI) 1.13 × 1.13 Yukimoto et al.
(2019)
15 SAM0-UNICON Seoul National University Atmosphere Model Version 0 with a Unified Convection 1.25 × 0.94 Park and Shin (2019)
Scheme
16 UKESM1–0-LL Met Office Hadley Centre 1.88 × 1.25 Tang et al. (2019)

Given that the region faces challenges in spatiotemporal coverage of (ETCCDI). These are non-parametric indices that describe moderate
observed data (Omondi et al. 2014), and that the assessment of models’ extremes with a recurrence time of at most a year and are calculated
ability to reproduce extreme precipitation is dependent on the choice of from daily precipitation. The nine extreme precipitation indices
reference data (Sillmann et al. 2013), the models were evaluated against (including seven (one) indices of wet (dry) extremes) used in this study
two gridded daily precipitation datasets, namely: Global Precipitation are described in Table 2 and discussed in detail by Zhang et al. (2011).
Climatology Project One-Degree Daily Dataset (GPCP 1DD; Huffman and These indices have been used to indirectly assess the possible occurrence
Bolvin 2013) and Tropical Rainfall Measuring Mission Multi-satellite of drought and flood events in many regions including Africa (e.g., Ou
Precipitation Analysis 3B42 (TRMM 3B42 version 7; Huffman et al. et al. 2013; Ongoma et al. 2018; Sonkoué et al. 2019; Akinsanola et al.
2007). The GPCP 1DD is developed as a combination of both surface and 2020b). Sillmann et al. (2013) and Akinsanola et al. (2020a) used the
satellite measurements at 1 × 1◦ spatial resolution and daily frequency, Normalized Root Mean Square Error (RMSE), among other metrics, to
with full global coverage (Huffman et al. 2001). TRMM 3B42 is a evaluate these extreme statistics in CMIP5 across the globe and CMIP6
satellite-based precipitation dataset at 0.25 × 0.25◦ spatial resolution over the United States, respectively. Following the same approach, the
and 3-hourly frequency, which was designed to monitor and study mean values for the nine precipitation extreme indices were computed,
tropical and subtropical precipitation (approximately 50◦ S to 50◦ N). and the percentage bias (BIAS), NRMSE, and Pattern Correlation Coef­
These datasets have shown reliable performance over the region ficient (PCC) of the models relative to TRMM observation were using
compared to ground-based data, and have been useful for evaluating eqs. 1–3.
climate models (Le Coz and van de Giesen 2020). In fact, Cattani et al. ∑n
i=1 (Mi − Oi )
(2016) evaluated the capability of several gridded precipitation prod­ %BIAS = ∑n X 100 (1)
ucts to reproduce the precipitation characteristics of Eastern Africa and i=1 Oi

recommended TRMM as a viable alternative for evaluating climate √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅

∑n ̅
models, although potential uncertainties may still exist. TRMM is
1 2
i=1 (Mi − Oi )
(2)
n
NRMSE = ∑n
therefore used as the reference dataset in this study. 1
i=1 Oi
n
The datasets have different temporal and spatial resolutions, but for
consistency, our analysis was carried out at the daily timescale for a Cov(M, O)
PCC(M, O) = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (3)
common period from 1997 to 2014, focusing on the two major precip­ Var(M)Var(O)
itation seasons (MAM and SON). Furthermore, in order to provide a fair
comparison to observations and to compute the ensemble mean, the where M and O are model and reference datasets means, respectively,
datasets were regridded to a common 2.81◦ latitude × 2.81◦ longitude and cov denotes covariance while Var is variance. The n represents the
grid (the lowest model resolution) using a distance-weighted interpo­ number of observations.
lation method in the climate data operator (https://code.zmaw.de/pro All the indices were calculated each year and for different season
jects/cdo). This approach uses the first-order conservative regridding separately. Lastly, the student t-test was used to test for any statistical
method described in Jones (1999) and reflects the viewpoint that pre­ significance difference that may exist between the CMIP6 EnsMean and
cipitation output from climate models represents an area-average over reference dataset, and the confidence level was set at 95%.
the grid cell (Chen and Knutson 2008). The multimodel ensemble mean
of the sixteen CMIP6 models defined herein as ‘EnsMean’ is employed to 3. Results and discussion
reduce natural variability and systematic biases in the individual model
members (Akinsanola and Zhou 2019). 3.1. Spatial distribution of mean and extreme precipitation

2.3. Methodology The ability of the CMIP6 models to simulate the mean and extremes
of precipitation over Eastern Africa is shown in Figs. 2–7. Although the
The study adopted a subset of indices that were recommended by focus of our discussion is mainly on the CMIP6 EnsMean, the
WMO’s Expert Team on Climate Change Detection and Indices

3
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Table 2 datasets alongside the EnsMean broadly agree that the Congo Basin and
List of precipitation extreme indices used in this study. Lake Victoria Basin record the highest amount of precipitation with
S/ Extreme Name Definition Units maximum around the western equatorial region. Relative to TRMM
N indices observations, the EnsMean overestimates precipitation by about 88%
1 SDII Simple daily Let PRwj be the daily mm/ over the north-western part of the study area (i.e., Sudan). This region
intensity precipitation amount on wet day receives the least rainfall during MAM, so the bias is most evident as a
days, PR ≥ 1 mm in period j. If percentage rather than an absolute difference. The mean biases over the
W represents number of wet larger portion of the study area are relatively small (− 20% to +20%)
days in j, then: SDIIj =
∑ and statistically insignificant. The magnitude of SDII in the EnsMean is
( Ww=1 PRwj ) / W
closer to GPCP than to TRMM observations, but is under-simulated
2 CDD Consecutive dry Let PRij be the daily days
days precipitation amount on day i
relative to the two datasets. The EnsMean produces the SDII pattern
in period j. Count the largest slightly to the west and north-central relative to observations, leading to
number of consecutive days a noticeable statistically significant dry bias that ranges from − 5% to
where PRij < 1 mm − 45% across most of the grid points especially to the east. Most of the
3 CWD Consecutive wet Let PRij be the daily days
ensemble members exhibit this dry bias (Fig. S1) although with some
days precipitation amount on day i
in period j. Count the largest spatial variations. For instance, the spatial biases in BCC-ESM1 are weak
number of consecutive days and statistically insignificant at the 95% level for most grid points, while
where PRij > 1 mm CNRM-CM6–1, EC-Earth3, and HadGEM3-GC31-LL exhibit the highest
4 PRCPTOT Total wet-day Let PRij be the daily mm
significant difference over most of the region. It is also worth noting that
precipitation precipitation amount on day i
in period j. If I represent the
the most statistically significant difference occurs over the eastern part
number of days in j, then: of the study area, covering the semi-arid lands of Kenya and Somalia
∑
PRCPTOTj = In=1 PRij where intense precipitation often results in flash flood, causing property
5 R10mm Heavy Let PRij be the daily days damage and losses of lives. Overall, the CMIP6 models reproduce SDII
precipitation precipitation amount on day i better in the highland areas of Mt. Elgon on the Kenya – Uganda border.
days in period j. Count the number of The CWD in the two observation datasets in the MAM season range from
days where PRij > 10 mm
6 R20mm Very heavy Let PRij be the daily days
0 to 20 days/year (Fig. 2), but is much higher in the models. The
precipitation precipitation amount on day i EnsMean, alongside all the ensemble members except CNRM-CM6–1,
days in period j. Count the number of exhibits significant overestimations of CWD (bias >90%) over most
days where PRij > 20 mm parts of the study areas (Fig. S2). The EnsMean CWD spatial pattern is
7 RX5day Max 5-day Let PRkj be the precipitation mm
similar to observations and resembles the mean pattern, but with higher
precipitation amount for the 5-day interval
ending k, period j. Then values throughout the region – a common problem in global models,
maximum 5 day values for which tend to rain too frequently. On the other hand, the spatial dis­
period j are: RX5dayj = max tribution of the CDD is more consistent with the two observations, but
(PRkj) with a noticeable EnsMean bias over central Ethiopia, extending to
8 R95pTOT Very wet days Let PRwj be the daily mm
precipitation amount on a wet
South Sudan. The grid points with relatively low mean precipitation
day w (PR ≥ 1 mm) in period i experience higher consecutive dry days (Figs. 2 and Fig. S3), although
and let PRwn95 be the 95th the magnitude of the biases varies significantly across the ensemble
percentile of precipitation on members with CNRM-CM6–1, EC-Earth3, and EC-Earth3-Veg exhibiting
wet days in the 1961–1990
the largest positive bias, while GFDL-CM4 performs relatively well in
period. If W represents the
number of wet days in the reproducing CDD with low bias.
period, then: R95pj = The spatial distribution of the maximum 5-day precipitation
∑w
w=1 PRwj , where PRwj > (RX5day), very heavy precipitation days (R20mm), very wet days
PRwn95 (R95pTOT), and total wet-day precipitation (PRCPTOT) over Eastern
9 R99pTOT Extremely wet Let PRwj be the daily mm Africa are presented in Fig. 3 for the MAM season from the two gridded
days precipitation amount on a wet
observations, EnsMean, and the bias of EnsMean relative to TRMM. Also,
day w (PR ≥ 1 mm) in period i
and let PRwn99 be the 95th the spatial bias of GPCP observations alongside the CMIP6 ensemble
percentile of precipitation on members are presented in Figs. S4, S6, S7, and S9, respectively. The
wet days in the 1961–1990 EnsMean captures the spatial distributions of all the wet precipitation
period. If W represents the indices fairly well, as presented in Fig. 3, but with considerable bias in
number of wet days in the
the magnitudes, particularly along the eastern coast region (Somalia and
period, then: R99pj =
∑w Kenya). The EnsMean magnitudes tend to be more consistent with GPCP
w=1 PRwj , where PRwj >
PRwn99 observation than to TRMM. Nevertheless, relative to TRMM, the CMIP6
EnsMean exhibit biases in RX5day range between − 40 to +70%, with
negative (positive) bias occurring mostly over the eastern (north-west­
performance of individual ensemble members are presented in the ern, where precipitation amounts are small) part of the study area.
supplementary material (Figs. S1–S18). Significant variations exist among the ensemble members (Fig. S4) as
The spatial distribution of mean precipitation (Pr), simple daily in­ intense dry (wet) bias dominates most grid points in CNRM-CM6–1, EC-
tensity (SDII, a measure of the precipitation intensity during events), Earth3, EC-Earth3-Veg, IPSL-CM6A-LR, and SAM0-UNICON (BCC-CSM-
consecutive wet days (CWD, number of consecutive days with precipi­ MR, BCC-ESM1, and CanESM5) models. Overall, the EnsMean out­
tation greater than 1 mm), and consecutive dry days (CDD, number of performs most models, with GFDL-CM4 having the smallest bias of in­
consecutive days with precipitation less than 1 mm) for the two gridded dividual models. Biases in R20mm and R95pTOT are similar to RX5day
observations (TRMM and GPCP; referred as observations henceforth) as the eastern half of the study area exhibits statistically significant
and the EnsMean, as well as the EnsMean bias relative to TRMM, during negative bias in both the EnsMean and most of the ensemble members
the MAM season are shown in Fig. 2. The mean precipitation provides a (Figs. S6 and S7). Similar patterns were also found in heavy precipita­
measure of the climatology of the study area, which generally reduces tion days (R10mm) and extremely wet days (R99pTOT) (Figs. S5 and S8,
from the western part of Eastern Africa to the east. The two observation respectively), although intense wet bias dominates the western half of

4
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Fig. 2. (left)–(right) Mean precipitation and indices SDII, CWD, and CDD for MAM season from (top)–(bottom) TRMM, GPCP, EnsMean, and bias of EnsMean relative
to TRMM over the present-day period, 1997–2014. Areas with statistically significance differences at 95% level are marked with black stippling.

the study area in some models (e.g., BCC-CSM2-MR and BCC-ESM1). It day− 1 of rainfall, is worse than the bias in Pr along the eastern side of the
should be noted that some models contributing simulations to CMIP6 are domain, suggesting that very light days (less than 1 mm day− 1)
not independent, and models from the same parent institution (e.g., BCC- compensate for the biases in wet day statistics, such that Pr is better
CSM2-MR and BCC-ESM1 as well as CESM2 and CESM2-WACCM) tend represented.
to agree on the sign and magnitude of the biases. As a result of the The spatial distributions of mean precipitation, SDII, CWD, and CDD
extreme wet bias earlier reported in CWD, the total wet-day (PRCPTOT) are presented in Fig. 4 for the SON season. Similar to the MAM season
MAM precipitation exhibit significant positive bias (>75%) over most (Fig. 2), the Eastern Africa region receives the highest mean precipita­
grid points in the EnsMean (Fig. 3) alongside most of the ensemble tion in the western half of the study area and decreases eastwards.
members (Fig. S9), although considerable spatial variations exist. The Relative to TRMM, it is evident that the percentage bias in the EnsMean
bias in PRCPTOT, which only considers days with more than 1 mm is an overestimation of mean precipitation by over 60% during this

5
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Fig. 3. (left)–(right) Indices RX5day, R20mm, R95pTOT, and PRCPTOT for MAM season from (top)–(bottom) TRMM, GPCP, EnsMean, and bias of EnsMean relative
to TRMM over the present-day period, 1997–2014. Areas with statistically significance differences at 95% level are marked with black stippling.

season, especially over the eastern half of the study area (Fig. 4). region (i.e., Somalia, Kenya, and Tanzania), although spatial variations
However, this wet bias is statistically insignificant at 95% level, except exist among individual models. Generally, the EnsMean outperforms
for a few grid points over the northeast corner of the Democratic Re­ individual models in reproducing the SDII over the Eastern Africa re­
public of the Congo where absolute precipitation rates are higher. gion. The SON season CWD exhibit statistically significant positive bias
Overall, the overestimation of mean precipitation is greater in SON that is larger than 90% over most of the grid points in the study area.
when compared to MAM, but occurs over the eastern side rather than This bias is evident not only in the EnsMean but also in most of the in­
northwest of the domain. Similar to MAM, the CMIP6 EnsMean, along­ dividual models (Fig. S11). The CDD on the other hand has a bias with
side the ensemble members, underestimates SDII over much of the re­ much lower magnitude, which is slightly underestimated in the
gion, about 35% on average (Figs. 4 and S10). For SON, this EnsMean over most of the study area except eastern Tanzania that en­
underestimation includes the Sudan region as well as the eastern coast compasses Rwanda and Burundi, and southeastern part of Democratic

6
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Fig. 4. Same as Fig. 2 but for SON season.

Republic of the Congo where a positive bias was found. While the ma­ significant negative biases are evident over many grid points for
jority of the CMIP6 ensemble members also underestimate CDD, region RX5day, R20mm, and R95pTOT. The magnitude of these biases is larger
associated with overestimation of CDD was larger especially in E3SM, among individual models than in the EnsMean (see Figs. S13, S15, and
CESM2, CESM2-WACCM, CNRM-CM6–1, and MRI-ESM-0 models S16). Similar biases with spatial and intermodel variations are also
(Fig. S12). evident in R10mm and R99pTOT in Figs. S14 and S17, respectively.
The ability of the CMIP6 EnsMean to reproduce the SON season However, while precipitation intensity is too weak across the region, a
spatial distributions of RX5day, R20mm, R95pTOT, and PRCPTOT wet bias dominates most grid points in Eastern Africa region for total-
precipitation extreme indices is presented in Fig. 5 (individual ensemble wet day SON precipitation (PRCPTOT) both in the EnsMean (Fig. 5)
members are shown in Figs. S13, S15, S16, and S18 for these statistics, and all the CMIP6 models (Fig. S18). As reported in MAM, the over­
respectively). The EnsMean reasonably captures aspects of the spatial estimation of PRCPTOT in the SON season is largely contributed by the
pattern of these wet precipitation indices (Fig. 5), but statistically positive bias in CWD, indicating that days with precipitation occur too

7
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Fig. 5. Same as Fig. 3 but for SON season.

frequently in CMIP6 models, which is a persistent problem found in Akinsanola et al. 2020a). We present the portrait diagrams for the nine
earlier versions. extreme precipitation indices evaluated as the percentage bias,
normalized root mean square error, and the pattern correlation coeffi­
cient between TRMM observation and all datasets used in this study
3.2. Descriptive statistics (Fig. 6). Furthermore, some studies (e.g.Sillmann et al. 2013; Akinsanola
et al. 2020a) have documented that the performance of models in
Previous studies have shown that a portrait diagram can be a useful simulating precipitation characteristics largely depends upon the choice
tool in evaluating and summarizing the performance of individual of the reference dataset. Thus, we further evaluated the percentage bias,
models in representing precipitation characteristics across the entire normalized root mean square error, and the pattern correlation
region with regard to a reference dataset (Sillman et al., 2013a;

8
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Fig. 6. Portrait diagrams showing the percentage bias (a,b), normalized root mean square error (c,d), and pattern correlation coefficient between TRMM obser­
vations and all the datasets. The descriptive statistics are for MAM (a,c,e) and SON (b,d,f).

coefficient using GPCP as the reference dataset separately and the results and BCC-CSM2-MR) resolution models exhibit the lowest NRMSE in
presented in Fig. 7. both seasons and the EnsMean exhibit lower NRMSE in SON compared
In both seasons, CWD and PRCPTOT exhibit high positive biases to MAM season (Fig. 6c,d). For pattern correlation coefficients (PCC), all
(red) that range from +25% to +90% across all models, which is evident the models have values greater than 0.75 (Figs. 6e & f), except for
in comparison to TRMM (Figs. 6a and b) and GPCP (Figs. 7a and b). SDII, R20mm where values are lower than 0.65 in both seasons. Overall in
R20mm, RX5day, R95pTOT, and R99pTOT exhibit negative percentage comparison to GPCP (Fig. 7), the percentage bias and NRMSE (PCC) are
bias that range from − 5% to − 75% across most of the models and sea­ smaller (larger) across individual models when compared with the sta­
sons relative to TRMM but not GPCP. Across both seasons and statistics, tistics relative to TRMM (Fig. 6). However, most of the patterns of in­
the EnsMean slightly outperforms individual models by exhibiting dividual model statistics relative to TRMM (Fig. 6) are generally
generally lower bias. Also, some low and moderate resolution models, consistent with GPCP (Fig. 7). Overall, our results have shown that the
such as GFDL-CM4, CNRM-CM6–1, HadGEM3-GCM31-LL, and performance of CMIP6 over Eastern Africa is sensitive to the reference
UKESM1–0-LL show better performance than some high-resolution dataset as the two gridded observations exhibit considerable differences,
models. The NRMSE presented in Fig. 6c,d is a useful metric to com­ particularly for precipitation intensity. Also, the model’s ability to
plement the biases discussed above because it is less influenced by realistically reproduce the pattern of extreme precipitation indices over
spatially compensating errors. Across the two seasons, the NRMSE is the region is largely independent of horizontal resolution and the
uniquely high for R20mm, RX5day, R95pTOT, and R99pTOT compared EnsMean outperforms most individual models.
to TRMM. Interestingly, some low (BCC-ESM) and middle (GFDL-ESM4

9
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

Fig. 7. Portrait diagrams showing the percentage bias (a,b), normalized root mean square error (c,d), and pattern correlation coefficient between GPCP observations
and all the datasets. The descriptive statistics are for MAM (a,c,e) and SON (b,d,f).

4. Summary and conclusion both seasons over Eastern Africa, which is consistent with an over­
estimation of total-wet day precipitation (PRCPTOT). Also, we found
In this study, we present the first statistical evaluation of the most that no single model is best for both seasons in reproducing the extreme
recent global climate models (CMIP6) in representing extreme precipi­ precipitation indices over the region. Specifically, individual models
tation indices over Eastern Africa. We investigated nine extreme pre­ have particular strengths and weaknesses relative to others, but the
cipitation indices that were recommended by WMO’s Expert Team on EnsMean outperforms any individual model when all the indices and
Climate Change Detection and Indices (ETCCDI) and compared the seasons are considered together. Our findings are consistent with pre­
model results with two gridded observation datasets (TRMM, and GPCP) vious global findings based on CMIP5 models (Sillmann et al. 2013) and
for MAM and SON seasons. Our findings suggest that the CMIP6 models recent findings over the United States based on CMIP6 (Akinsanola et al.
alongside their ensemble mean, overestimate (underestimate) the total- 2020a).
wet day precipitation and consecutive wet days (very wet days and Furthermore, we found that the relative performance of the CMIP6
maximum 5-day precipitation) indices over the region in both seasons. models may depend on the choice of the reference dataset. Specifically,
The drizzling bias that has been reported in the CMIP5 models (see Flato there is a considerable statistically significant difference among the two
et al. 2013; Herold et al. 2016; Trenberth and Zhang 2018) has not been gridded precipitation datasets used in this study. We have thus focused
alleviated in the CMIP6. Specifically, CMIP6 models generally exhibit primarly on a reference that has been shown to perform well for the
very high consecutive wet days precipitation (CWD, Pr ≥ 1 mm day-1) in region (TRMM 3B42, Cattani et al. 2016). Overall, all the models exhibit

10
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

pattern correlation that is larger than 0.7 across all the nine indices UN Office for the Coordination of Humanitarian Affairs (OCHA), 2019. Eastern Africa
Region: Regional Flood Snapshot (November 2019). https://www.unocha.org/sites/
(except for R20mm which exhibits the lowest PCC), which indicates the
unocha/files/2019OCHAannualreport.pdf. (Accessed 10 April 2020).
biases are more associated with poorly represented magnitudes rather EC-Earth, 2019a. EC-Earth-Consortium EC-Earth3 Model Output Prepared for CMIP6
than patterns. The documented biases in this study may not be simply CMIP Historical. https://doi.org/10.22033/ESGF/CMIP6.4700.
alleviated by increasing the model’s horizontal resolution, in part EC-Earth, 2019b. EC-Earth-Consortium EC-Earth3-Veg Model Output Prepared for
CMIP6 CMIP Historical. https://doi.org/10.22033/ESGF/CMIP6.4706.
because even high-resolution global models cannot resolve processes of Seneviratne, S.I., Nicholls, N., Easterling, D., Goodess, C.M., Kanae, S., Kossin, J., Luo, Y.,
convection in the tropics. In fact, we found that some lower and mod­ Marengo, J., McInnes, K., Rahimi, M., Reichstein, M., Sorteberg, A., Vera, C.,
erate resolution models exhibit better performance than higher resolu­ Zhang, X, 2012. Changes in climate extremes and their impacts on the natural
physical environment. In: Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J.,
tion models for some indices. Ebi, K.L., Midgley, P.M. (Eds.), Managing the Risks of Extreme Events and Disasters
Our study provides a first-order evaluation of the performance to Advance Climate Change Adaptation, A Special Report of Working Groups I and II
CMIP6 models in simulating precipitation extremes over Eastern Africa of the Intergovernmental Panel on Climate Change (IPCC). Cambridge University
Press, Cambridge, UK, and New York, NY, USA, pp. 109–230.
and thus, lays the groundwork for analysis of future changes. These Akinsanola, A.A., Zhou, W., 2019. Projections of West African summer monsoon rainfall
results also inform the extent to which these models should be used for extremes from two CORDEX models. Clim. Dyn. 52, 2017–2028. https://doi.org/
impacts studies in the region. Understanding and removing the sources 10.1007/s00382-018-4238-8.
Akinsanola, A.A., Kooperman, G.J., Pendergrass, A.G., Hannah, W.M., Reed, K.A., 2020a.
of bias in the CMIP6 models used herein is a critical next step for Seasonal representation of extreme precipitation indices over the United States in
improving the accuracy of regional climate prediction and for the CMIP6 present-day simulations. Environ. Res. Lett. 15, 094003 https://doi.org/
application of these models in research using downscaling to study is­ 10.1088/1748-9326/ab92c1.
Akinsanola, A.A., Kooperman, G.J., Reed, K.A., Pendergrass, A.G., Hannah, W.M., 2020b.
sues related to water resources, flooding, and drought projection.
Projected changes in seasonal precipitation extremes over the United States in
CMIP6 simulations. Environ. Res. Lett. https://doi.org/10.1088/17489326/abb397.
CRediT authorship contribution statement Bader, D.C., Leung, R., Taylor, M., et al., 2019. E3SM-Project E3SM1.0 model output
prepared for CMIP6 CMIP. Earth System Grid Federation. https://doi.org/10.22033/
ESGF/CMIP6.2294.
Akintomide Afolayan Akinsanola: Conceptualization, Data cura­ Boucher, O., Denvil, S., Caubel, A., Foujols, M.A., 2018. IPSL IPSL-CM6A-LR Model
tion, Methodology, Formal analysis and visualization, Writing - review Output Prepared for CMIP6 CMIP Historical. https://doi.org/10.22033/ESGF/
CMIP6.5195.
& editing Victor Ongoma: Methodology, Writing - original Draft
Camberlin, P., 2018. Climate of Eastern Africa. Oxford Research Encyclopedia of Climate
Gabriel J. Kooperman: Methodology, Writing - review & editing, Re­ Science. https://oxfordre.com/climatescience/view/10.1093/acrefore/9780190228
sources, Supervision 620.001.0001/acrefore-9780190228620-e-512. (Accessed 9 April 2020).
Carty, T., 2017. A Climate in Crisis. How Climate Change is Making Drought and
Humanitarian Disaster Worse in East Africa. https://www-cdn.oxfam.org/s3fs-pub
Declaration of Competing Interest lic/mb-climate-crisis-east-africa-drought-270417-en.pdf. (Accessed 11 April 2020).
Cattani, E., Merino, A., Levizzani, V., 2016. Evaluation of Monthly Satellite-Derived
Precipitation Products over East Africa. J. Hydrometeorol. 17, 2555–2573. https://
The authors declare that they have no known competing financial doi.org/10.1175/JHM-D-15-0042.1.
interests or personal relationships that could have appeared to influence Chen, C., Knutson, T., 2008. On the Verification and Comparison of Extreme Rainfall
the work reported in this paper. Indices from climate Models. J. Clim. 21 (7), 1605–1621.
Coughlan de Perez, E., van Aalst, M., Choularton, R., van den Hurk, B., Mason, S.,
Nissan, H., Schwager, S., 2019. From rain to famine: assessing the utility of rainfall
Acknowledgements observations and seasonal forecasts to anticipate food insecurity in East Africa. Food
Secur. 11, 57–68. https://doi.org/10.1007/s12571-018-00885-9.
Danabasoglu, G., Lawrence, D., Lindsay, K., Lipscomb, W., Strand, G., 2019a. NCAR
We thank the modelling centers that provided the CMIP6 datasets CESM2 model output prepared for CMIP6 CMIP Historical. https://doi.org/
described in Table 1. The data is available from the ESGF archive of the 10.22033/ESGF/CMIP6.7627.
Program for Climate Model Diagnosis and Intercomparison (https://pc Danabasoglu, G., 2019b. Ncar cesm2-waccm model output prepared for CMIP6 CMIP
historical. https://doi.org/10.22033/ESGF/CMIP6.10071.
mdi.llnl.gov/CMIP6/) as a contribution to CMIP6 through the World Dunning, C.M., Allan, R.P., Black, E., 2017. Identification of deficiencies in seasonal
Climate Research Programme’s (WCRP) working group on coupled rainfall simulated by CMIP5 climate models. Environ. Res. Lett. 12, 114001. https://
modelling. We also acknowledge the centers that provided gridded doi.org/10.1088/1748-9326/aa869e.
Eyring, V., Bony, S., Meehl, G.A., Senior, C.A., Stevens, B., Stouffer, R.J., Taylor, K.E.,
precipitation datasets: TRMM TMPA 3B42 was provided by the NASA
2016. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6)
GSFC Mesoscale Atmospheric Processes Laboratory (https://science. experimental design and organization. Geoscientific Model Dev. 9, 1937–1958.
gsfc.nasa.gov/earth/mesoscale); and GPCP 1DD was downloaded from https://doi.org/10.5194/gmd-9-1937-2016.
Fitzpatrick, R.G.J., 2015. RMetS National Meeting - forecasting for Africa. Weather 70,
the National Oceanic and Atmospheric Administration /National Centre
176–177. https://doi.org/10.1002/wea.2487.
for Environmental Information (https://www.ncdc.noaa.gov/wd Flato, G., Marotzke, J., Abiodun, B., Braconnot, P., Chou, S.C., Collins, W., et al., 2013.
cmet/data-access-search-viewer-tools/global-precipitation-climatolo Evaluation of Climate models. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M.,
gy-project-gpcp-clearinghouse). AAA and GJK acknowledge support Allen, S.K., Boschung, J., Midgley, P.M. (Eds.), Climate Change 2013: The Physical
Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the
from the U.S. Department of Energy (DOE) Regional and Global Model Intergovernmental Panel on Climate Change. Cambridge University Press,
Analysis (RGMA) Program (DE-SC0019459 and DE-SC0021209), and Cambridge, United Kingdom and New York, NY, USA.
GJK acknowledges support from the University of Georgia ’s Office of Guo, H., John, J.G., Blanton, C., McHugh, C., Nikonov, S., Radhakrishnan, A., Zadeh, N.
T., Balaji, V., Durachta, J., Dupuis, C., Menzel, R., Robinson, T., Underwood, S.,
Research President’s Interdisciplinary Seed Grant Program. We also Vahlenkamp, H., Dunne, K.A., Gauthier, P.P., Ginoux, P., Griffies, S.M., Hallberg, R.,
thank the DOE’s RGMA program area, Data Management program, and Harrison, M., Hurlin, W., Malyshev, S., Naik, V., Paulot, F., Paynter, D.J., Ploshay, J.,
National Energy Research Scientific Computing Center for coordinating Schwarzkopf, D.M., Seman, C.J., Shao, A., Silvers, L., Wyman, B., Zeng, Y.,
Adcroft, A., Dunne, J.P., Held, I.M., Krasting, J.P., Horowitz, L.W., Milly, P.,
CMIP6 analysis activities. We thank the Editor and anonymous re­ Shevliakova, E., Winton, M., Zhao, M., 2018. NOAA-GFDL GFDL-CM4 Model Output
viewers for their constructive comments and suggestions. Prepared for CMIP6 CMIP Historical. https://doi.org/10.22033/ESGF/CMIP6.8594.
Herold, N., Alexander, L.V., Donat, M.G., Contractor, S., Becker, A., 2016. How much
does it rain over land? Geophys. Res. Lett. 43, 341–348. https://doi.org/10.1002/
Appendix A. Supplementary data 2015GL066615.
Huffman, G.J., Bolvin, D.T., 2013. Version 1.2 GPCP one-degree daily precipitation data
Supplementary data to this article can be found online at https://doi. set documentation. GPCP [Available online at. ftp://rsd.gsfc.nasa.gov/pub/1dd-v1.
2/1DD_v1.2_doc.pdf.
org/10.1016/j.atmosres.2021.105509. Huffman, G.J., Adler, R.F., Morrissey, M.M., Bolvin, D.T., Curtis, S., Joyce, R.,
McGavock, B., Susskind, J., 2001. Global precipitation at one-degree daily resolution
References from multisatellite observations. J. Hydrometeorol. 2, 36–50. https://doi.org/
10.1175/1525-7541(2001)002<0036:GPAODD>2.0.CO;2.
Huffman, G.J., Adler, R.F., Bolvin, D.T., Gu, G., Nelkin, E., Gu, G., Nelkin, E.J.,
Famine Early Warning Systems Network (FEWS NET), 2011. Special Report Kenya Food
Bowman, K.P., Hong, Y., Stocker, E.F., Wolff, D.B., 2007. The TRMM multisatellite
Security. USAID.

11
A.A. Akinsanola et al. Atmospheric Research 254 (2021) 105509

precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor Rowell, D.P., 2019. An observational constraint on CMIP5 projections of the East African
precipitation estimates at fine scales. J. Hydrometeorol. 8, 38–55. https://doi.org/ Long Rains and Southern Indian Ocean warming. Geophys. Res. Lett. 46, 6050–6058.
10.1175/JHM560.1. https://doi.org/10.1029/2019GL082847.
IPCC, 2007. IPCC Fourth Assessment Report, Climate Change 2007: Impacts, Adaptation Samuels, R., Hochman, A., Baharad, A., Givati, A., Levi, Y., Yosef, Y., Saaroni, H., Ziv, B.,
and Vulnerability. Working Group II Contribution to the 4th Assessment Report. In Harpaz, T., Alpert, P., 2018. Evaluation and projection of extreme precipitation
the Intergovernmental Panel on Climate Change. indices in the Eastern Mediterranean based on CMIP5 multi-model ensemble. Int. J.
Jones, P.W., 1999. First- and second-order conservative remapping schemes for grids in Climatol. 38, 2280–2297. https://doi.org/10.1002/joc.5334.
spherical coordinates. Mon. Wea. Rev. 127, 2204–2210. Shimeles, A., Verdier-Chouchane, A., Boly, A., 2018. Building a resilient and sustainable
Kharin, V.V., Zwiers, F.W., Zhang, X., Wehner, M., 2013. Changes in temperature and agriculture in sub-Saharan Africa. In: Building a Resilient and Sustainable
precipitation extremes in the CMIP5 ensemble. Clim. Chang. 119, 345–357. https:// Agriculture in Sub-Saharan Africa. https://doi.org/10.1007/978-3-319-76222-7.
doi.org/10.1007/s10584-013-0705-8. Sillmann, J., Kharin, V.V., Zhang, X., Zwiers, F.W., Bronaugh, D., 2013. Climate extremes
Klein Tank, A.M.G., Zwiers, F.W., Zhang, X., 2009. Guidelines on analysis of extremes in indices in the CMIP5 multimodel ensemble: part 1. Model evaluation in the present
a changing climate in support of informed decisions for adaptation, climate data and climate. J. Geophys. Res. Atmos. 118, 1716–1733. https://doi.org/10.1002/
monitoring WCDMP-No. 72, WMO-TD no. 1500 (56 pp). jgrd.50203.
Koutroulis, A.G., Grillakis, M.G., Tsanis, I.K, Papadimitriou, L., 2016. Evaluation of Sonkoué, D., Monkam, D., Fotso-Nguemo, T.C., Yepdo, Z.D., Vondou, D.A., 2019.
precipitation and temperature simulation performance of the CMIP3 and CMIP5 Evaluation and projected changes in daily rainfall characteristics over Central Africa
historical experiments. Clim. Dyn. 47, 1881–1898. based on a multi-model ensemble mean of CMIP5 simulations. Theor. Appl. Climatol.
Krasting, J.P., John, J.G., Blanton, C., McHugh, C., Nikonov, S., Radhakrishnan, A., 137, 2167–2186. https://doi.org/10.1007/s00704-018-2729-5.
Rand, K., Zadeh, N.T., Balaji, V., Durachta, J., Dupuis, C., Menzel, R., Robinson, T., Swart, N.C., Cole, J.N.S., Kharin, V.V., Lazare, M., Scinocca, J.F., Gillett, N.P., Anstey, J.,
Underwood, S., Vahlenkamp, H., Dunne, K.A., Gauthier, P.P., Ginoux, P., Griffies, S. Arora, V., Christian, J.R., Hanna, S., Jiao, Y., Lee, W.G., Majaess, F., Saenko, O.A.,
M., Hallberg, R., Harrison, M., Hurlin, W., Malyshev, S., Naik, V., Paulot, F., Seiler, C., Seinen, C., Shao, A., Sigmond, M., Solheim, L., von Salzen, K., Yang, D.,
Paynter, D.J., Ploshay, J., Schwarzkopf, D.M., Seman, C.J., Silvers, L., Wyman, B., Winter, B., 2019. The Canadian Earth System Model version 5 (CanESM5.0.3).
Zeng, Y., Adcroft, A., Dunne, J.P., Guo, H., Held, I.M., Horowitz, L.W., Milly, P., Geosci. Model Dev. 12, 4823–4873. https://doi.org/10.5194/gmd-12-4823-2019.
Shevliakova, E., Stock, C., Winton, M., Zhao, M., 2018. NOAA-GFDL GFDL-ESM4 Sylla, M.B., Nikiema, M., Gibba, P., Kebe, I., Klutse, N.A.B., 2016. Climate change in West
Model Output Prepared for CMIP6 CMIP Historical. https://doi.org/10.22033/ Africa: recent trends and future projections. In: Yaro, J.A., Hesselberg, J. (Eds.),
ESGF/CMIP6.8597. Adaptation to Climate Change and Variability in Rural West Africa. Springer.
Kumar, S., Merwade, V., Kinter, I.I.I.J.L., Niyogi, D., 2013. Evaluation of temperature https://doi.org/10.1007/978-3-319-31499-0_3. ISBN: 978-3-319-31497-6.
and precipitation trends and long-term persistence in CMIP5 twentieth-century Tang, Y., Rumbold, S., Ellis, R., Kelley, D., Mulcahy, J., Sellar, A., Walton, J., Jones, C.,
climate simulations. J. Clim. 26, 4168–4185. 2019. MOHC UKESM1.0-LL model output prepared for CMIP6 CMIP historical.
Kumar, D., Kodra, E., Ganguly, A.R., 2014. Regional and seasonal intercomparison of Version 20190902. Earth System Grid Federation. https://doi.org/10.22033/ESGF/
CMIP3 and CMIP5 climate model ensembles for temperature and precipitation. Clim. CMIP6.6113.
Dyn. 43, 2491–2518. Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2012. An overview of CMIP5 and the experiment
Le Coz, C., van de Giesen, N., 2020. Comparison of Rainfall Products over Sub-Saharan design. Bull. Amer. Meteorol. Soc. 93 (4), 485–498. https://doi.org/10.1175/BAMS-
Africa. J. Hydrometeorol. 21, 553–596. https://doi.org/10.1175/JHM-D-18-0256.1. D-11-00094.1.
Mehran, A., Aghakouchak, A., Phillips, T.J., 2014. Evaluation of CMIP5 continental Trenberth, K.E., Zhang, Y., 2018. How often does it really rain? Bull. Am. Meteorol. Soc.
precipitation simulations relative to satellite-based gauge-adjusted observations. 99, 289–298. https://doi.org/10.1175/BAMS-D-17-0107.1.
J. Geophys. Res. 119, 1695–1707. https://doi.org/10.1002/2013JD021152. Voldoire, A., 2018. CMIP6 Simulations of the CNRM-CERFACS based on CNRM-CM6-1
Nguyen, P., Thorstensen, A., Sorooshian, S., Zhu, Q., Tran, H., Ashouri, H., Miao, C.Y., Model for CMIP Experiment Historical. https://doi.org/10.22033/ESGF/
Hsu, K.L., Gao, X.G., 2017. Evaluation of CMIP5 model precipitation using CMIP6.4066.
PERSIANN-CDR. J. Hydrometeorol. 18, 2313–2330. Wainwright, C.M., Marsham, J.H., Keane, R.J., et al., 2019. ‘Eastern African Paradox’
Nicholson, S.E., 2017. Climate and climatic variability of rainfall over eastern Africa. rainfall decline due to shorter not less intense Long Rains. npj Clim. Atmos. Sci. 2
Rev. Geophys. 55, 590–635. https://doi.org/10.1002/2016RG000544. (34) https://doi.org/10.1038/s41612-019-0091-7.
Ogwang, B.A., Chen, H., Li, X., Gao, C., 2014. The influence of topography on East Wu, T., Chu, M., Dong, M., Fang, Y., Jie, W., Li, J., Li, W., Liu, Q., Shi, X., Xin, X., Yan, J.,
African October to December climate: Sensitivity experiments with RegCM4. Adv. Zhang, F., Zhang, J., Zhang, L., Zhang, Y., 2018. BCC BCC-CSM2MR Model Output
Meteorol. 2014, 143917. https://doi.org/10.1155/2014/143917. Prepared for CMIP6 CMIP Historical. https://doi.org/10.22033/ESGF/CMIP6.2948.
Omondi, P.A., Awange, J.L., Forootan, E., Ogallo, L.A., Barakiza, R., Girmaw, G.B., Yang, W., Seager, R., Cane, M.A., Lyon, B., 2015. The rainfall annual cycle bias over East
Fesseha, I., Kululetera, V., Kilembe, C., Mbati, M.M., Kilavi, M., King’uyu, S.M., Africa in CMIP5 coupled climate models. J. Clim. 28, 9789–9802. https://doi.org/
Omeny, P.A., Njogu, A., Badr, E.M., Musa, T.A., Muchiri, P., Bamanya, D., 10.1175/JCLI-D-15-0323.1.
Komutunga, E., 2014. Changes in temperature and precipitation extremes over the Yukimoto, S., Koshiro, T., Kawai, H., Oshima, N., Yoshida, K., Urakawa, S., Tsujino, H.,
Greater Horn of Africa region from 1961 to 2010. Int. J. Climatol. 34 (4), Deushi, M., Tanaka, T., Hosaka, M., Yoshimura, H., Shindo, E., Mizuta, R., Ishii, M.,
1262–1277. https://doi.org/10.1002/joc.3763. Obata, A., Adachi, Y., 2019. MRI MRI-ESM2.0 Model Output Prepared for CMIP6
Ongoma, V., Chen, H., Gao, C., Nyongesa, A.M., Polong, F., 2018. Future changes in CMIP Historical. https://doi.org/10.22033/ESGF/CMIP6.6842.
climate extremes over Equatorial East Africa based on CMIP5 multimodel ensemble. Zebaze, S., Jain, S., Salunke, P., Shafiq, S., Mishra, S.K., 2019. Assessment of CMIP5
Nat. Haz. 90, 901–920. https://doi.org/10.1007/s11069-017-3079-9. multimodel mean for the historical climate of Africa. Atmos. Sci. Lett. 20, e926
Ou, T., Chen, D., Linderholm, H.W., Jeong, J.H., 2013. Evaluation of global climate https://doi.org/10.1002/asl.926.
models in simulating extreme precipitation in China. Tellus, Series. Dyn. Meteorol. Zhang, X., Alexander, L., Hegerl, G.C., Jones, P., Tank, A.K., Peterson, T.C., Trewin, B.,
Ocean. 65, 19799. https://doi.org/10.3402/tellusa.v65i0.19799. Zwiers, F.W., 2011. Indices for monitoring changes in extremes based on daily
Park, S., Shin, J., 2019. SNU SAM0-UNICON Model Output Prepared for CMIP6 CMIP temperature and precipitation data. WIREs Clim. Chang. 2, 851–870. https://doi.
Historical. https://doi.org/10.22033/ESGF/CMIP6.7789. org/10.1002/wcc.147.
Ridley, J., Menary, M., Kuhlbrodt, T., Andrews, M., Andrews, T., 2019. MOHC Zhang, J., Wu, T., Shi, X., Zhang, F., Li, J., Chu, M., Liu, Q., Yan, J., Ma, Q., Wei, M.,
HadGEM3-GC31-LL Model Output Prepared for CMIP6 CMIP Historical. 10.22 2018. BCC BCC-ESM1 model output prepared for CMIP6 CMIP historical. Version
033/ESGF/CMIP6.6109. 20190902. Earth System Grid Federation. https://doi.org/10.22033/ESGF/
CMIP6.2949.

12