Spatiotemporal Changes in Temperature and Precipitation in West Africa. Part I: Analysis with the CMIP6 Historical Dataset

: Climate variability and change constitute major challenges for Africa, especially West Africa (WA), where an important increase in extreme climate events has been noticed. Therefore, it appears essential to analyze characteristics and trends of some key climatological parameters. Thus, this study addressed spatiotemporal variabilities and trends in regard to temperature and precipitation extremes by using 21 models of the Coupled Model Intercomparison Project version 6 (CMIP6) and 24 extreme indices from the Expert Team on Climate Change Detection and Indices (ETCCDI). First, the CMIP6 variables were evaluated with observations (CHIRPS, CHIRTS, and CRU) of the period 1983–2014; then, the extreme indices from 1950 to 2014 were computed. The innovative trend analysis (ITA), Sen’s slope, and Mann–Kendall tests were utilized to track down trends in the computed extreme climate indices. Increasing trends were observed for the maxima of daily maximum temperature (TXX) and daily minimum temperature (TXN) as well as the maximum and minimum of the minimum temperature (TNX and TNN). This upward trend of daily maximum temperature (Tmax) and daily minimum temperature (Tmin) was enhanced with a signiﬁcant increase in warm days/nights (TX90p/TN90p) and a signiﬁcantly decreasing trend in cool days/nights (TX10p/TN10p). The precipitation was widely variable over WA, with more than 85% of the total annual water in the study domain collected during the monsoon period. An upward trend in consecutive dry days (CDD) and a downward trend in consecutive wet days (CWD) inﬂuenced the annual total precipitation on wet days (PRCPTOT). The results also depicted an upward trend in SDII and R30mm, which, additionally to the trends of CDD and CWD, could be responsible for localized ﬂood-like situations along the coastal areas. The study identiﬁed the 1970s dryness as well as the slight recovery of the 1990s, which it indicated occurred in 1992 over West Africa.


Introduction
Climate change is a widely known global phenomenon with varying impacts across regions. Its indicator is significant changes in human and natural systems [1,2]. Ref. [3] assessed the global surface temperature and found an increase approximately 0.07 • C/decade during the period 1901-2010 and about 0.17 • C per decade from 1979 to 2010. According to [4,5], this increase, which is higher in developing countries, is definitely due to both natural changes and changes observed in local change inputs (industrialization, transport, etc.). Extreme weather events induce a large range of effects on the environment and society and raise important challenges about environment and resource management for developing countries [6]. perature and modification in rainfall (amount and spell duration) in WA have important implications on agriculture and water resource fields [42][43][44][45]. Regarding the projected change in the amounts and frequency of rainfall in WA [19,46,47], the uncertainty is considerably high, and several models do not agree on whether the change in rainfall will be negative or positive [48], especially for agriculture production. Ref. [49] explained that various studies on crops productivity over WA have shown that the impact of climate variability on crop yields is more frequently negative than positive. Thus, climate change impacts in the agriculture sector are a major challenge and need to be addressed. Some studies [20,50] have investigated adaptation strategies as a cornerstone to deal with sustainable agriculture [51] and climate change impacts [52]. It is important to understand whether the strategies are well designed and account for the best possible understanding of the trends of some parameters (e.g., precipitation and temperature). Using the CMIP6 dataset and referring to scenarios SSP1-2.6, SSP2-4.5, and SSP5-8.5, ref. [46] found that CMIP6 projected a continuous and significant increase in the mean annual temperature globally over all of Africa and subsequently over eight subregions of Africa during the 21st century. Meanwhile, ref. [53] adopted the same dataset to assess the representation of daily precipitation characteristics over West Africa and demonstrated how well CMIP6 reproduced observation patterns. However, these studies failed to investigate the trends in or frequency of precipitation and temperatures; they focused only on the variability of the parameters. The present study aimed to use the historical CMIP6 datasets (temperatures and precipitation) and analyze the trends of the changes in daily temperature and precipitation extremes over WA by calculating extreme climate index series, as well as using widely used and innovative statistical methods to assess the levels of changes. Section 2 of this paper, dedicated to the materials and methods, presents the study domain, the various datasets used, and the methodologies adopted, while the results and discussion are presented in Sections 3 and 4, respectively. Finally, the conclusion in Section 5 summarizes the main findings of the study.

Study Domain
The study area lay in West Africa, which is located between latitudes 0 • N and 25 • N and longitudes 25 • W and 25 • E ( Figure 1). The region is bordered in the South by the Gulf of Guinea and in the north by Mauritania, Mali, and Niger; the Cameroon highlands form the eastern boundary, while the Atlantic Ocean forms the western limit. Rainfall patterns over this region are mostly affected by ocean currents and local features such as topography. In terms of climatic zones, West Africa is divided into three different regions: the Sahel, which is characterized as a semiarid zone ranging from western Senegal to eastern Sudan, between 12 • N and 20 • N; the Sudano-Sahelian zone; and the Guinea coast, which is characterized by bimodality driven by the intertropical discontinuity (ITD).

Data
The model dataset utilized for the study was the Coupled Model Intercomparison Project 6th phase (CMIP6) database (precipitation and temperatures), and the observational datasets were CHIRPS, CHIRTS, and CRU. The CMIP6 historical dataset covers the period 1950-2014; CHIRPS covers 1981-present; CHIRTS, 1983-2016, and CRU, 1981-present.

Model CMIP6 Dataset
The datasets of the CMIP6 are divided into two phases (historical and projection) for all models used in the study. The datasets were published onto the ESGF data server and were freely accessed by searching the model name together with the reference ID (e.g., experiment name: "historical", source: "model name", variable: "pr", "tasmax", and "tasmin") at https://esgf-data.dkrz.de/search/cmip6-dkrz/ (accessed on 18 June 2021), or https://esgfnode.llnl.gov/projects/cmip6/ (accessed on 18 June 2021). Datasets are stored in various formats, but we opted for the NetCDF format, which could be explored with almost all climate dataset visualization software (NCL, http://www.ncl.ucar.edu/, accessed on 18 June 2021) or Python (https://www.python.org/, accessed on 18 June 2021), R (https://cran.r-project.org/, accessed on 18 June 2021). A historical experiment dataset was available that included data from 1850 to 2014. For climate projection experiments, we used a combination of the Shared Socioeconomic Pathway (SSP) and Representative Concentration Pathway (RCP), which provide data for the period 2006-2300. In this study, the study period of interest for the calculation of the extreme climate indices was 1981-2014. The rainy season in West Africa is mainly related to the West African Monsoon, which oscillates between May and September (MJJAS) each year. The remaining months of the year were counted as participating in the so-called dry season. The CMIP6 models of interest have variable spatial resolutions and are illustrated in Table 1. In total, we used 21 CMIP6 models that made available both precipitation and temperature.

Data
The model dataset utilized for the study was the Coupled Model Intercomparison Project 6th phase (CMIP6) database (precipitation and temperatures), and the observational datasets were CHIRPS, CHIRTS, and CRU. The CMIP6 historical dataset covers the period 1950-2014; CHIRPS covers 1981-present; CHIRTS, 1983-2016, and CRU, 1981-present.

Model CMIP6 Dataset
The datasets of the CMIP6 are divided into two phases (historical and projection) for all models used in the study. The datasets were published onto the ESGF data server and were freely accessed by searching the model name together with the reference ID (e.g., experiment name: "historical", source: "model name", variable: "pr", "tasmax", and "tasmin") at https://esgf-data.dkrz.de/search/cmip6-dkrz/ (accessed on 18 June 2021), or https://esgfnode.llnl.gov/projects/cmip6/ (accessed on 18 June 2021). Datasets are stored in various formats, but we opted for the NetCDF format, which could be explored with almost all climate dataset visualization software (NCL, http://www.ncl.ucar.edu/, accessed on 18 June 2021) or Python (https://www.python.org/, accessed on 18 June 2021), R (https://cran.r-project.org/, accessed on 18 June 2021). A historical experiment dataset was available that included data from 1850 to 2014. For climate projection experiments, we used a combination of the Shared Socioeconomic Pathway (SSP) and Representative Concentration Pathway (RCP), which provide data for the period 2006-2300. In this study, the study period of interest for the calculation of the extreme climate indices was 1981-2014. The rainy season in West Africa is mainly related to the West African Monsoon, which oscillates between May and September (MJJAS) each year. The remaining months of the year were counted as participating in the so-called dry season. The CMIP6 models of interest have variable spatial resolutions and are illustrated in Table 1. In total, we used 21 CMIP6 models that made available both precipitation and temperature.

Observed CHIRPS, CHIRTS, and CRU Datasets
Because of a lack of ground-based observation data, scientists refer to satellite datasets as an alternative to supply in situ observations. To address this limitation, in the present study, rainfall data from the Climate Hazards Group Infrared Stations (CHIRPS) version 2, developed by the Climate Hazards Group of the University of California, was used (precipitation: https://data.chc.ucsb.edu/products/CHIRPS-2.0/africa_daily/, accessed on 2 May 2021). In the same line, its corresponding temperature (maximum and minimum) dataset, the Climate Hazards Group Infrared Temperature with Station data (CHIRTS), was adopted (CHIRTS: [37]; Temperatures: http://data.chc.ucsb.edu/products/CHIRTSdaily/v1.0/, accessed on 2 May 2021) along with the Climatic Research Unit Time Series (CRU: https://crudata.uea.ac.uk/cru/data/hrg/, accessed on 2 May 2021; precipitation and temperatures). The dataset was used at the daily time step. The CHIRPS dataset is a quasiglobal rainfall dataset covering 50 • S-50 • N, while CHIRTS covers 60 • S-70 • N. To be consistent with the simulation model periods, we adopted a CHIRXS (where X is indicated for precipitation or temperature) period of 1983-2014 under 0.25 • × 0.25 • (~25 km × 25 km) horizontal resolution.
The CRU and CHIRPS datasets have been widely used in previous studies [19,20,54] and provided satisfactory responses in respect to observational datasets (in situ). In the present work, both observed datasets (CHIRPS/CHIRTS and CRU) were used to first investigate the recent release (CHIRTS [55]) in response to maximum and minimum temperatures and then to see whether, as CHIRPS fits well with CRU (precipitation) CHIRTS could be a response to CRU (temperature).

Methods
Several studies have been conducted to date by scientists in designing methods and applications for the computation and analysis of extreme climate indices for better monitoring and tracking of climate trends. The methodology adopted in this study consisted of: i. analysis of the precipitation and the temperature recorded in the observational dataset; ii. computation of the selected climate indices; iii. analysis of the spatial trends of the extreme indices computed with modified Mann-Kendall and Sen's slope tests; iv. assessment of the performance of a new method (innovative trend analysis: ITA) in respect to that of the methods in iii.; v. analysis of the temporal variability and trend of the computed indices.

Selected Extreme Climate Event
The climate indices agreed with the international committee of the Expert Team on Climate Change Detection and Indices (ETCCDI), which aims to provide a good mixture of daily statistics to assess changes in temperature and precipitation regimes in terms of duration, intensity, and occurrence [56,57]. Based on the combination of CDO commands and the Climpact (https://climpact-sci.org/indices/, accessed on 13 May 2021), the extreme indices listed in Tables 2 and 3 were calculated. The computed indices were classified into 5 sets of extremes climate indices: (i) absolute extreme indices-precipitation: RXnday (n = 1, n = 5, and n = 7), temperature : TXX, TXM, TMM, TNX, TNN, TNM, TXN; (ii) threshold exceedance indices such as R10mm, R20mm, and R30mm, which refer to the number of days on which a threshold was surpassed; (iii) indices that highlight the length of wet and dry spell duration, for instance, CDD and CWD; (iv) percentile-based precipitation and temperature indices (R95p, R99p, TN10p, TX10p, TN90p, TX90p, and WSDI); and (v) other indices-SDII and PRCPTOT.

Criteria of the Analysis of the Extremes
Some basic criteria were set to be accounted for in the analysis of the trend of each extreme climate event: (a) a minimum of 80% (17 models of 21) of the CMIP6 models needed to reflect the events or the trends; (b) a significant trend had to be demonstrated by at least 80% of the models.

Statistics Constraints to Evaluate the Performance of Models
The correlation analysis method was employed to evaluate the relationship between models and observations on one hand, and between indices on the other hand. Additionally, the nonparametric Mann-Kendall (MK) trend test [58,59], which is strongly recommended in this kind of analysis by the World Meteorological Organization (WMO), was adopted to evaluate the trends in extreme climate (both precipitation and temperature) indices. It was jointly associated with Sen's slope test [60] for better appreciation of the trend. The aforementioned tests were all based on hypotheses. The null hypothesis (Ho) assumed that the data were independent and randomly distributed, and the alternative hypothesis (H1) supposed that there was a monotonic trend in the dataset. The utility of these two tests was that the MK test (with its Z Kendall coefficient) provided an idea about the significance of the trend, while Sen's test (with the slope estimator) estimated the trend magnitude. The assessment was also based on a chosen 95% threshold of confidence level. An innovative trend analysis (ITA) was applied to detect monotonic trends and subtrends in the time-series dataset. This methodological approach was used to detect the annual and seasonal trends.
MK [58,59] is one of the common formulas used in hydrology and meteorology to identify trends in a time-series dataset. The test statistic (S) of series (x 1 , x 2 , x 3 , . . . and x n ) for which the trends are being checked can be expressed with the following formula: where n is the length of the dataset and x j and x k indicate the observations at times j and k. Sign A positive value for S indicates an increasing or upward trend, while a negative value shows a decreasing or downward trend in the time-series data.
The variance of S, VAR(S), can be calculated by the equation: where ρ denotes the tied group number of observations in group I, which is a set of sample data with similar values, and i indicates the extent of the ith tie number. The time-series data statistic (S) is identified with Kendall's τ (tau), which is given as follows: . Through the estimation of S and the variance VAR(S), the standardized test measurement Z is as follows when n > 10 [61]: The positive (+) values of Z indicate an increasing trend, and the negative (−) values depict a decreasing trend.
From [62], modified VAR(S) statistics can be estimated with the equation: Here, the correction factor n n * e is adjusted to the autocorrelated data as: ρ e ( f ) signifies the autocorrelation between ranks of observations and can be estimated as: • Sen's Slope Estimator The Sen's slope (Q) is the median of N values of Q i [63], where Thus if some zero values of Q i fall in between equal numbers of negative and positive values of Q i , the Sen's slope (Q) is zero. The greater the number of equal values there are in a time series, the higher the probability of a no-change trend for the series [64].

•
Innovative Trend Analysis (ITA) One recent method proposed for hydroclimatic variability assessment is innovative trend analysis (ITA), which was first elaborated in [65]. This new and robust technique is used in hydrometeorology for trend detection. It has been applied for different variables such as groundwater, rainfall, temperature, and evapotranspiration [66,67]. The methodology consists of dividing into two subseries of equal numbers in the observation variables of the study. For the next step, each subseries is reorganized in ascending order and plotted against the other in a Cartesian coordinate system [68]. The first half is plotted on the X-axis, and the second half on the Y-axis. After that, a straight line is fitted with the scatter plot that represents the "monotonic trend "or so-called "no trend". When the scatter is concentrated above the 1:1 line, the time series has an increasing trend, and if the scatter points concentrate below the line (1:1), a decreasing trend in the time series is indicated [69]. The trend indicator of ITA [65] is calculated from the following equation: where B represents the ITA slope, n denotes the extent of individual subseries, x j and x k represent the values of the consecutive subseries, and x represents the mean of the first subseries (x k ).   A positive slope of the B value indicates an increasing trend in the series, whereas a negative value of the slope signifies a decreasing tendency in the time series.

Results
This section focuses on presenting the main findings of the study. It depicts rainfall and temperature variabilities using selected climate indices over the study area and reports our analysis of their various trends with statistical tests. The reader is invited to see the supplementary documentation for additional figures and tables that are not shown here.

Scenario Models Validation
Validation of the CMIP6 models used was undertaken with observed rainfall and temperature (maximum and minimum) for the two databases in the period 1983-2014. Figure 2 shows the interannual distribution of rainfall averages for both the CMIP6 (21 models) and observed (CRU and CHIRPS) data as well as for the ensemble mean of the models. It can be seen that the total annual rainfall varied greatly across the study domain. For all the models and observations, there existed a northward gradient from higher to lower values, with the highest values recorded over the Guinea Highlands and Cameroon mountains (about 2500 mm/year for the observed and 2000 mm/year for the model ensemble mean). The Savannah region was generally wet, with a rainfall of about 700 mm/year. For the northern part, the rainfall recorded was less than 400 mm/year, and for the southern part (around the Guinea Coast), it was about 1100 mm to 1300 mm per year. The rainfall recorded during the monsoon period (MJJAS) over the study domain is depicted in supplementary material ( Figure S2). As in the case of the annual rainfall average, a northward gradient was noted, with the maximum and minimum of rainfall recorded over the same locations. A high correlation of models with observations was noted at latitude 5° N-20° N. The maximum cumulative rainfall registered during the monsoon period was about 2000 mm/year. Figure 3 presents the average percentage of the contribution of the monsoon to the annual total mean. The monsoon contributed significantly to the total rainfall in the Savannah region (even underestimating models confirmed this observation). Between the latitudes 5° N and 20° N, at least 85% of the rainfall amount was received during May-September. It was also observed that the West African Monsoon did not contribute to rainfall above 25° N or below the equator. The rainfall recorded during the monsoon period (MJJAS) over the study domain is depicted in supplementary material ( Figure S2). As in the case of the annual rainfall average, a northward gradient was noted, with the maximum and minimum of rainfall recorded over the same locations. A high correlation of models with observations was noted at latitude 5 • N-20 • N. The maximum cumulative rainfall registered during the monsoon period was about 2000 mm/year. Figure 3 presents the average percentage of the contribution of the monsoon to the annual total mean. The monsoon contributed significantly to the total rainfall in the Savannah region (even underestimating models confirmed this observation). Between the latitudes 5 • N and 20 • N, at least 85% of the rainfall amount was received during May-September. It was also observed that the West African Monsoon did not contribute to rainfall above 25 • N or below the equator.

Temperatures Evaluation
The assessment of the maximum and minimum temperatures (Tmax and Tmin respectively) revealed that, for observations as well as models, the temperatures varied widely. Models differently represented the temperature, both for Tmax ( Figure 4) and Tmin ( Figure S2). All the minima of Tmin were located in the northern part, especially in the northeastern area, while the maxima of the Tmin were captured in the western region; the other parts were, on average, 22 ± 3 • C. For the Tmax, some models underestimated the observations (NESM3, SAMO-UNICON, IPSL-CM6A-LR, etc.) while others overestimated (MIROC6, CanESM5, BCC-ESM1). The lowest values of Tmax were located in the northeastern regions (as in the case of the minimum of the Tmin) and scattered around Cameroon's mountains and Gabon's forests. Each model, whether over-or underestimating, agreed that the hottest area was the Savannah's band, located at the borders of Mali, Mauritania, and Senegal. The models' ensemble mean was closer to the annual average of CRU than CHIRPS and depicted the observations better than the individual models. This means that the models' ensemble provided a better estimate of the parameters than the individual models.

Temperatures Evaluation
The assessment of the maximum and minimum temperatures (Tmax and Tmin respectively) revealed that, for observations as well as models, the temperatures varied widely. Models differently represented the temperature, both for Tmax ( Figure 4) and

Interannual Rainfall and Temperatures Trend
To study the trends in the selected CMIP6 models, the work investigated the historical period 1950-2014, for which both precipitation and temperature data were available. To assess the annual rainfall and temperature trend, an advanced graphical method of trend detection was applied that included innovative trend analysis (ITA), the modified Mann-Kendall test (MMK), and Sen's slope. Figure 5, depicting the evaluation with the MMK test, shows, based on its corrected Z statistic (Zc) that the trend of the rainfall varied from one model to another. From Figure  5a-u, the red cross indicates where trends were significant at 99% confidence level (CL), and the black cross indicates where trends were significant at at 95% CL. Globally, all the models indicated significant (either 95% or 99% of CL) positive trends; nonsignificant negative trends of annual rainfall were noticed over the Guinea Highlands (reduction of about 4-5 mm/decade) and some other parts, such as Cameroon.

Interannual Rainfall and Temperatures Trend
To study the trends in the selected CMIP6 models, the work investigated the historical period 1950-2014, for which both precipitation and temperature data were available. To assess the annual rainfall and temperature trend, an advanced graphical method of trend detection was applied that included innovative trend analysis (ITA), the modified Mann-Kendall test (MMK), and Sen's slope. Figure 5, depicting the evaluation with the MMK test, shows, based on its corrected Z statistic (Zc) that the trend of the rainfall varied from one model to another. From Figure 5a-u, the red cross indicates where trends were significant at 99% confidence level (CL), and the black cross indicates where trends were significant at at 95% CL. Globally, all the models indicated significant (either 95% or 99% of CL) positive trends; nonsignificant negative trends of annual rainfall were noticed over the Guinea Highlands (reduction of about 4-5 mm/decade) and some other parts, such as Cameroon. The Savannah region experienced a significant increase in annual rainfall. In most locations in the study domain, a significant decreasing trend in annual rainfall was evident over Cameroon and Gabon, particularly in models such as AWI-ESM-1-1-LR, BCC-ESM1, NorCPM1; other models showed an increase in the yearly rainfall for these areas. In Figure 5v, the blue crosses indicate where at least 80% of the models depicted a trend with 99% CL, while in Figure 5w, the blue crosses express where at least 80% of models depicted a trend with 95% CL. The subplots Figure 5v,w indicated globally that, for the CMIP6 models studied, the Savannah showed a maximum rate of a significant increase in annual rainfall (about 4-5 mm/decade). A severe decline (nonsignificant) was identified around the Guinea Highlands but was revealed by less than 80% of the considered models.

Trend of Annual Rainfall
The same trends were observed from the results of Sen's slope test ( Figure 6) as from the MMK results ( Figure 5). The trends were located in the same areas and differed only The Savannah region experienced a significant increase in annual rainfall. In most locations in the study domain, a significant decreasing trend in annual rainfall was evident over Cameroon and Gabon, particularly in models such as AWI-ESM-1-1-LR, BCC-ESM1, NorCPM1; other models showed an increase in the yearly rainfall for these areas. In Figure 5v, the blue crosses indicate where at least 80% of the models depicted a trend with 99% CL, while in Figure 5w, the blue crosses express where at least 80% of models depicted a trend with 95% CL. The subplots Figure 5v,w indicated globally that, for the CMIP6 models studied, the Savannah showed a maximum rate of a significant increase in annual rainfall (about 4-5 mm/decade). A severe decline (nonsignificant) was identified around the Guinea Highlands but was revealed by less than 80% of the considered models.
The same trends were observed from the results of Sen's slope test ( Figure 6) as from the MMK results ( Figure 5). The trends were located in the same areas and differed only in magnitude. The yearly reduction in rainfall around the Guinea highlands here was about 0.2-0.5 mm, while the global increase noted reduced by 5 mm/decade. The yearly variability (average increasing) of the rainfall over West Africa was about 3 mm/decade, but around the Savannah, Sen's slope indicated a significant upward trend of about 4.5 mm/decade of at least at 95% CL under a minimum of 80% of the studied models.  The evaluation of the annual rainfall trend by applying the ITA is presented in Figure  7. In Figure 7a-u, the red crosses (+) express increasing trends, and the blue minuses (−) decreasing trends. The spatial distribution of the ITA trend indicator displayed the same variabilities and trends as did Sen's slope ( Figure 6) and MMK ( Figure 5) trend analysis. On one hand, declining trends circulated over the extreme western part of WA (Guinea highlands) and were shown by at least 80% of the models. On the other hand, about 60% of the models revealed that the Cameroon mountains and Gabon forests exhibited the same declining trends. Generally, the rest of the WA domain experienced a stable increasing trend. The trends highlighted were all significant, at least at 95% CL. Additionally, a few significant trends ignored by MMK and Sen's slope could be identified here, justifying the reliability of ITA to track unseen trends in time-series data. The evaluation of the annual rainfall trend by applying the ITA is presented in Figure 7. In Figure 7a-u, the red crosses (+) express increasing trends, and the blue minuses (−) decreasing trends. The spatial distribution of the ITA trend indicator displayed the same variabilities and trends as did Sen's slope ( Figure 6) and MMK ( Figure 5) trend analysis. On one hand, declining trends circulated over the extreme western part of WA (Guinea highlands) and were shown by at least 80% of the models. On the other hand, about 60% of the models revealed that the Cameroon mountains and Gabon forests exhibited the same declining trends. Generally, the rest of the WA domain experienced a stable increasing trend. The trends highlighted were all significant, at least at 95% CL. Additionally, a few significant trends ignored by MMK and Sen's slope could be identified here, justifying the reliability of ITA to track unseen trends in time-series data. Furthermore, the identified trends from Sen's slope and MMK were increased in magnitude under ITA. The maximum amount of the trend for ITA was about 10 mm/decade and was very high between the latitudes of 5 • N and 15 • N. In the northern part of WA, the trend indicator expressed a decrease of about +2 mm/decade.   The Z statistic score, the Sen's slope, and the ITA slope, as well as the latter's indicator, express how much the changes have impacted WA. For the Savannah regions, almost 90% of models showed change trends, but not with the same magnitude; only NESM3, MPI-ESM-1-2-HAM, and MPI-ESM1-2-HR behaved differently under the Sen's and MMK tests. For ITA, however, the trend was dynamically elaborated, showing, for instance, in the case of the aforementioned models, on one hand, a clear increasing trend (northward from Savannah: 0.6 mm/year), and on another hand, a southward decreasing trend for both Tmin and Tmax.
Generally, for either Tmax or Tmin, there was a significant increasing trend across WA with a northward gradient. The northern part grew hotter than the central and the southern parts. ITA revealed (Figure 9v) that all the models (red cross locations) identified a particular increasing trend in the northern part and the Guinea-coast. The increasing trends around the Guinea coast could be due to the gradual increase in sea surface temperature (SST) due to global warming [90]. The Z statistic score, the Sen's slope, and the ITA slope, as well as the latter's indicator, express how much the changes have impacted WA. For the Savannah regions, almost 90% of models showed change trends, but not with the same magnitude; only NESM3, MPI-ESM-1-2-HAM, and MPI-ESM1-2-HR behaved differently under the Sen's and MMK tests. For ITA, however, the trend was dynamically elaborated, showing, for instance, in the case of the aforementioned models, on one hand, a clear increasing trend (northward from Savannah: 0.6 mm/year), and on another hand, a southward decreasing trend for both Tmin and Tmax.
Generally, for either Tmax or Tmin, there was a significant increasing trend across WA with a northward gradient. The northern part grew hotter than the central and the southern parts. ITA revealed (Figure 9v) that all the models (red cross locations) identified a particular increasing trend in the northern part and the Guinea-coast. The increasing trends around the Guinea coast could be due to the gradual increase in sea surface temperature (SST) due to global warming [90].    However, the warm days (TX90p) and warm nights (TN90p), as well as the warm spell duration index (WSDI), revealed some very high-temperature events (warm extremes). The significant trend for the TN90p was well captured over the whole domain with all the statistical methods used. However, ITA displayed a dynamic and smooth southward gradient, with the peak of the increase noticed along the GC (more than +2 days/decade) and a lower increase in the northern part (about +1 day/decade). The increasing trend was slightly equally distributed across the study domain, with about +1.5 days/decade under Sen's slope test and almost double using the MMK test (3 days/decade), except some spikes over Lake Chad and the border junction between Burkina Faso and Cote d'Ivoire. For TX90p, the magnitude of the significant trends was regionally located, with the higher +1.4 days/decade under Sen's slope (+3.2 days/decade under MMK) in the north and the lower +0.5 day/decade under Sen's slope (+1.2 days/decade under MMK) in the south. Similar trends were noticed for the WSDI using MMK and ITA tests, but though the trends were well depicted by a minimum of 80% of the models, they were not significantly (at least 95% CL) indicated by a minimum of 80% of the models. This nonsignificant representation can be confirmed by the contrast shown with the spatial distribution of Sen's slope (Figure 10), which clearly displayed the changes in two trends, increasing over the north (+0.6 day/decade) and decreasing over the south (−0.3 day/decade).  However, the warm days (TX90p) and warm nights (TN90p), as well as the warm spell duration index (WSDI), revealed some very high-temperature events (warm extremes). The significant trend for the TN90p was well captured over the whole domain with all the statistical methods used. However, ITA displayed a dynamic and smooth southward gradient, with the peak of the increase noticed along the GC (more than +2 days/decade) and a lower increase in the northern part (about +1 day/decade). The increasing trend was slightly equally distributed across the study domain, with about +1.5 days/decade under Sen's slope test and almost double using the MMK test (3 days/decade), except some spikes over Lake Chad and the border junction between Burkina Faso and Cote d'Ivoire. For TX90p, the magnitude of the significant trends was regionally located, with the higher +1.4 days/decade under Sen's slope (+3.2 days/decade under MMK) in the north and the lower +0.5 day/decade under Sen's slope (+1.2 days/decade under MMK) in the south. Similar trends were noticed for the WSDI using MMK and ITA tests, but though the trends were well depicted by a minimum of 80% of the models, they were not significantly (at least 95% CL) indicated by a minimum of 80% of the models. This nonsignificant representation can be confirmed by the contrast shown with the spatial distribution of Sen's slope (Figure 10), which clearly displayed the changes in two trends, increasing over the north (+0.6 day/decade) and decreasing over the south (−0.3 day/decade). Temperature Indices (TXX, TXM, TMM, TNX, TNN, TNM, TXN) Figure 11 presents the spatial average of the maximum values of daily maximum temperature (TXX), mean daily maximum temperature (TXM), mean daily mean temperature (TMM), and maximum value of daily minimum temperature (TNX). As analyzed previously for the maximum temperature trend, the maximum TXX was located in the northern part, and specifically the northwest (45 • C). The same went for TXM (43 • C), TMM, and TNX. The lowest maxima were located in the southern area. The analysis was based on the criterion that a minimum of 80% of models had to show the trend; thus, the red crosses (+) indicate that at least 80% of models agreed with the increasing trend ( Figure 11). The green circles indicate where the trends were significant at 95% CL. analyzed previously for the maximum temperature trend, the maximum TXX was located in the northern part, and specifically the northwest (45 °C). The same went for TXM (43 °C), TMM, and TNX. The lowest maxima were located in the southern area. The analysis was based on the criterion that a minimum of 80% of models had to show the trend; thus, the red crosses (+) indicate that at least 80% of models agreed with the increasing trend ( Figure 11). The green circles indicate where the trends were significant at 95% CL. Regarding trends, all the statistical methods showed an increasing trend in maximum temperatures. The significant increase in maximum temperatures was higher in the north and lower in the south, with the appearance of a clear dynamic northward gradient. Sen's slope test indicated an increase of about 0.1 to 0.3 °C/decade; the ITA method, 0.1 to 0.25 °C/decade; and the MMK method, a very high 0.2 to 0.5 °C/decade. The maximum of Tmax was located in the same area (northern part) under the three methods used, with increases of 0.15 °C/decade, 0.25 °C/decade, and 0.35 °C/decade under ITA, Sen's, and MMK, respectively. The lowest value of the maximum of Tmax was across the GC. In contrast, the ITA overestimated the increasing trend of the maximum of Tmax over the Cameroon mountains and the Gabon forest compared to Sen's and MMK. This representation of ITA was due to the fact that it brought out the small trend changes or trends over each year. The ratio of the maximum Tmax to the minimum Tmax (TXN) was 2.1 (Figure 12), which means that the maximum of Tmax was about 2.1 times greater than the minimum of Tmax, and this was almost verified under all three of the analysis Figure 11. Spatial distribution of the maximum value of the daily maximum temperature (TXX), the average of the daily maximum temperature (TXM), the average of the daily mean temperature (TMM), and the maximum value of daily minimum temperature (TNX). The red crosses (+) indicate positive trends, the black minuses (−) negative trends, and the green circles significant trends with 95% CL.

Absolute Extreme
Regarding trends, all the statistical methods showed an increasing trend in maximum temperatures. The significant increase in maximum temperatures was higher in the north and lower in the south, with the appearance of a clear dynamic northward gradient. Sen's slope test indicated an increase of about 0.1 to 0.3 • C/decade; the ITA method, 0.1 to 0.25 • C/decade; and the MMK method, a very high 0.2 to 0.5 • C/decade. The maximum of Tmax was located in the same area (northern part) under the three methods used, with increases of 0.15 • C/decade, 0.25 • C/decade, and 0.35 • C/decade under ITA, Sen's, and MMK, respectively. The lowest value of the maximum of Tmax was across the GC. In contrast, the ITA overestimated the increasing trend of the maximum of Tmax over the Cameroon mountains and the Gabon forest compared to Sen's and MMK. This representation of ITA was due to the fact that it brought out the small trend changes or trends over each year. The ratio of the maximum Tmax to the minimum Tmax (TXN) was 2.1 ( Figure 12), which means that the maximum of Tmax was about 2.1 times greater than the minimum of Tmax, and this was almost verified under all three of the analysis methods adopted. The spatial average of Tmax, as well as the average of the mean temperature in Figure 12, expressed the same tendency of the increasing value from ITA to Sen's to the MMK. In the southern regions, increasing trends were around 0.09 • C/decade, 0.07 • C/decade, and 0.2 • C/decade for the mean of Tmax and 0.09 • C/decade, 0.1 • C/decade, and 0.22 • C/decade for the mean of the spatial average temperature under the ITA, Sen's, and MMK methods, respectively. The percentage of significant increase trend was 100% for an average mean temperature, greater than 70% for the maximum of the minimum temperature (TNx), and about 50% for the average of Tmax. Some areas were significantly increasing with 95% CL but were not shown by a minimum of 80% of models on which the authors agreed for the analysis in the present study.
Water 2021, 13, x FOR PEER REVIEW 20 of 34 0.1 °C/decade, and 0.22 °C/decade for the mean of the spatial average temperature under the ITA, Sen's, and MMK methods, respectively. The percentage of significant increase trend was 100% for an average mean temperature, greater than 70% for the maximum of the minimum temperature (TNx), and about 50% for the average of Tmax. Some areas were significantly increasing with 95% CL but were not shown by a minimum of 80% of models on which the authors agreed for the analysis in the present study.

Percentile-Based Precipitation Indices (R95p, R99p)
The analysis of Figure 13 indicated that the R95p and R99p variabilities were regionally distributed, with the maximum in the south and the minimum in the north. The figure marks an increasing trend of the percentile of the two indices, with scatterings of significance (at least at 95% CL) in the south. Less than 80% of the models used indicated a change in the northern part of the study domain with Sen's slope; it seems that the north did not have a clear change. However, both indices (R99p and R95p) recorded very low rainfall amounts (Figure 13a-h).

Percentile-Based Precipitation Indices (R95p, R99p)
The analysis of Figure 13 indicated that the R95p and R99p variabilities were regionally distributed, with the maximum in the south and the minimum in the north. The figure marks an increasing trend of the percentile of the two indices, with scatterings of significance (at least at 95% CL) in the south. Less than 80% of the models used indicated a change in the northern part of the study domain with Sen's slope; it seems that the north did not have a clear change. However, both indices (R99p and R95p) recorded very low rainfall amounts (Figure 13a-h).
For both indices (R95p and R99p), the ITA and MMK methods faintly depicted an increasing trend in the north of 0.3 mm/decade and less than 0.4 mm/decade, respectively. All three methods attested to the existence of change in the southern area and confirmed the maxima located around the Guinea highlands, the Cameroon mountains, and the Gabon forests for both indices. The maximum of the total annual rainfall from the heavy rain days (R95p) was about 320 mm/year (located around Guinea, Cameroon, and Gabon); the maximum of the total annual rainfall from the very heavy rain days (R99p) was about 110 mm/year and was located at the same area as R95p. The increasing trend in the south for both indices was quietly important (from 2 to 4 mm/decade) but of scattered significance. The increasing trend in the heavy rain days was significant (at least 95% CL) over countries such as Gabon, Cameroon, the southern part of Nigeria, the northern part of Benin, Burkina Ghana, and the southern part of Mali. However, for R99p, the marked increasing trend was not significant in the south. For both indices (R95p and R99p), the ITA and MMK methods faintly depicted an increasing trend in the north of 0.3 mm/decade and less than 0.4 mm/decade, respectively. All three methods attested to the existence of change in the southern area and confirmed the maxima located around the Guinea highlands, the Cameroon mountains, and the Gabon forests for both indices. The maximum of the total annual rainfall from the heavy rain days (R95p) was about 320 mm/year (located around Guinea, Cameroon, and Gabon); the maximum of the total annual rainfall from the very heavy rain days (R99p) was about 110 mm/year and was located at the same area as R95p. The increasing trend in the south for both indices was quietly important (from 2 to 4 mm/decade) but of scattered significance. The increasing trend in the heavy rain days was significant (at least 95% CL) over countries such as Gabon, Cameroon, the southern part of Nigeria, the northern part of Benin, Burkina Ghana, and the southern part of Mali. However, for R99p, the marked increasing trend was not significant in the south. (RX1day, RX5day, RX7day) The spatial display of the RX1day, RX5day, and RX7day varied widely and dynamically decreased from the south toward the north. From the data analysis, the average maximum RX1day of 65 mm was located between latitudes of 5° N and 12° N; that of RX5day, of about 121 mm, and that of RX7day, of about 169-187 mm, were located in the Guinea highlands, Cameroon, and Gabon. Figure 13. Spatial distribution of the extremely wet precipitation (R99p), the very wet precipitation (R95p), the annual total precipitation on wet days (PRCPTOT), and the intensity of the average precipitation on wet days (SDII). The red crosses (+) indicate positive trends, the black minuses (−) negative trends, and the green circles significant trends with 95% CL. (RX1day, RX5day, RX7day) The spatial display of the RX1day, RX5day, and RX7day varied widely and dynamically decreased from the south toward the north. From the data analysis, the average maximum RX1day of 65 mm was located between latitudes of 5 • N and 12 • N; that of RX5day, of about 121 mm, and that of RX7day, of about 169-187 mm, were located in the Guinea highlands, Cameroon, and Gabon.

Absolute Extreme Precipitation Indices
The indices had a positive trend across the whole of WA, and the extrema of the changes converged with the extrema of the indices. The strict criteria well captured some scattered significant increasing trends (5% of the whole domain). Figure 14 indicates for the three indices a rise of 1 to 1.7 mm/decade, 1 to 2.5 mm/decade, and 1.5 to 3.7 mm/decade under MMK, Sen's slope, and ITA, respectively, over the Savannah toward south. It was observed that the trends of these indices were quietly consistent but not significant everywhere. The more the number of cumulative days increased, the clearly southward the positive trend was oriented. In addition to the analysis of the R99p and R95p, this may impact the cumulative wet days in the southern region. scattered significant increasing trends (5% of the whole domain). Figure 14 indicates for the three indices a rise of 1 to 1.7 mm/decade, 1 to 2.5 mm/decade, and 1.5 to 3.7 mm/decade under MMK, Sen's slope, and ITA, respectively, over the Savannah toward south. It was observed that the trends of these indices were quietly consistent but not significant everywhere. The more the number of cumulative days increased, the clearly southward the positive trend was oriented. In addition to the analysis of the R99p and R95p, this may impact the cumulative wet days in the southern region. Figure 14. Spatial distribution of three absolute extreme precipitation indices (RX1day, RX5day, RX7day). The red crosses (+) indicates positive trends, the black minuses (−) negative trends, and the green circles significant trends with 95% CL. (R10mm, R20mm, R30mm, CDD, CWD) Figure 15 illustrates the spatially averaged numbers of heavy rain days (R10mm), very heavy rain days (R20mm), extremely heavy rain days (R30mm), consecutive dry days (CDD), and consecutive wet days (CWD). They exhibited a slight, increasing southward trend, as in the case of R99p and R95p. The CDD was very high in the north (298 days/year) and low in the south (61 days/year). The CWD was high in the south (112 days/year in the Guinea Highlands and Cameroon-Gabon and 48-62 days/year in other areas) and low in the north (12 days/year). Regarding the change trends, Sen's slope and MMK did not capture significant trends because of the high and strict criteria adopted in this study (the minimum of 80% of models that needed to express the trend). Focusing on the CDD, only a single decreasing trend was noticed, in the southeast of Niger, and an increase was observed in the southwest of the Congo. For the two analysis methods (Sen's and MMK) applied to CDD, globally, the study domain experienced a nonsignificant decreasing trend, except for in the Guinea highlands, where an increase up to 1 day/decade was observed, and a scatter of some increasing trends over the northeastern and northwestern parts as well as countries such as Benin, Togo, Ghana, and Cote d'Ivoire. The analysis of ITA revealed a clear and dynamic distribution of trends, with an increase in the CDD in the northern regions (up to 1.3 days/decade) diminishing toward the south to 0.2 days/decade. The assessment of the CWD revealed, indirectly, a contrast with the CDD; the locations of CDD maxima coincided with the minima of CWD and vice versa. The MMK and Sen's slope tests depicted, over the Guinea highlands and Cameroon, a decreasing trend in CWD. An increasing trend was noticed over the Figure 14. Spatial distribution of three absolute extreme precipitation indices (RX1day, RX5day, RX7day). The red crosses (+) indicates positive trends, the black minuses (−) negative trends, and the green circles significant trends with 95% CL. (R10mm, R20mm, R30mm, CDD, CWD) Figure 15 illustrates the spatially averaged numbers of heavy rain days (R10mm), very heavy rain days (R20mm), extremely heavy rain days (R30mm), consecutive dry days (CDD), and consecutive wet days (CWD). They exhibited a slight, increasing southward trend, as in the case of R99p and R95p. The CDD was very high in the north (298 days/year) and low in the south (61 days/year). The CWD was high in the south (112 days/year in the Guinea Highlands and Cameroon-Gabon and 48-62 days/year in other areas) and low in the north (12 days/year). Regarding the change trends, Sen's slope and MMK did not capture significant trends because of the high and strict criteria adopted in this study (the minimum of 80% of models that needed to express the trend). Focusing on the CDD, only a single decreasing trend was noticed, in the southeast of Niger, and an increase was observed in the southwest of the Congo. For the two analysis methods (Sen's and MMK) applied to CDD, globally, the study domain experienced a nonsignificant decreasing trend, except for in the Guinea highlands, where an increase up to 1 day/decade was observed, and a scatter of some increasing trends over the northeastern and northwestern parts as well as countries such as Benin, Togo, Ghana, and Cote d'Ivoire. The analysis of ITA revealed a clear and dynamic distribution of trends, with an increase in the CDD in the northern regions (up to 1.3 days/decade) diminishing toward the south to 0.2 days/decade. The assessment of the CWD revealed, indirectly, a contrast with the CDD; the locations of CDD maxima coincided with the minima of CWD and vice versa. The MMK and Sen's slope tests depicted, over the Guinea highlands and Cameroon, a decreasing trend in CWD. An increasing trend was noticed over the Savannah, and in the case of the MMK test, the change in some parts was captured based on the analysis criteria.

Threshold and Duration Extreme Precipitation Indices
The R10mm, R20mm, and R30mm showed increasing trends over the whole of WA, with values of 0.12 to 0.2 day/decade, 0.04 to 0.1 day/decade, and 0.04 to 0.06 day/decade, respectively, over the Savannah and the south area under the ITA test. The Sen's slope and MMK tests displayed the same trend dynamic, i.e., a southward increase in the number of days per decade. The increasing trend of the R10mm was significant in the southern regions, at about 1 day/decade under Sen's slope and 2 days/decade under MMK. There was almost no trend of R10mm over the northern regions. The R20mm and the R30mm increased from the Savannah to the south. MMK and ITA, based on our criteria, captured the change, while Sen's test could not. However, the Sen's slope results depicted on average an augmentation of about 0.2 to 0.4 day/decade in the very heavy rain days in the southern part. In the case of R30mm, the observations were quite similar to those for the R20mm when applying Sen's slope test. It was also noticed, based on the analysis criteria, that only the southern part of the WA experienced a nonsignificant increase trend, with picks around the Guinea highlands. Savannah, and in the case of the MMK test, the change in some parts was captured based on the analysis criteria. Figure 15. Spatial distribution of the numbers of heavy rain days (R10mm), very heavy rain days (R20mm), extremely heavy rain days (R30mm), consecutive dry days (CDD), and consecutive wet days (CWD). The red crosses (+) indicate positive trends, the black minuses (−) negative trends, and the green circle significant trends with 95% CL.
The R10mm, R20mm, and R30mm showed increasing trends over the whole of WA, with values of 0.12 to 0.2 day/decade, 0.04 to 0.1 day/decade, and 0.04 to 0.06 day/decade, respectively, over the Savannah and the south area under the ITA test. The Sen's slope and MMK tests displayed the same trend dynamic, i.e., a southward increase in the number of days per decade. The increasing trend of the R10mm was significant in the Figure 15. Spatial distribution of the numbers of heavy rain days (R10mm), very heavy rain days (R20mm), extremely heavy rain days (R30mm), consecutive dry days (CDD), and consecutive wet days (CWD). The red crosses (+) indicate positive trends, the black minuses (−) negative trends, and the green circle significant trends with 95% CL.

Other Indices (PRCPTOT and SDII)
The annual average of PRCPTOT ( Figure 13) had a similar spatial distribution and trend to those of R95p and R99p, wherein the values of indices were higher in the southern regions (2000 mm/year around Guinea, Cameroon, and Gabon). Its positive southward trend was about 2 mm/decade (from Savannah to GC) for all the statistical tests analyzed. As the case of previous indices, ITA revealed a very slight trend in the northern area, as did Sen's slope and MMK, of about 0.02 mm/decade.
The spatial distribution of the simple precipitation intensity indicator (SDII) presented in Figure 13, indicates a faithful correlation with PRCPTOT, as well as with R99p and R95p. The maxima and minima were located in corresponding areas from one index to another. SDII reached a maximum and minimum of 10 mm/day/year and less than 1 mm/day, respectively. According to the statistical methods and criteria applied in the present study, Sen's slope and MMK indicated a high positive trend in the SDII over the Savannah of about 2 mm/day/decade and a lower positive trend in the northeast and southwest of Cote d'Ivoire of about 0.01 mm/day/decade (almost no trend). ITA also displayed a positive trend in the SDII, but in contrast with the results obtained under MMK and Sen's slope, the maximum of the increase lay in the northern regions, with an increase up to 0.5 mm/day/decade. The trend in the central part of WA rose to 0.25 mm/day/decade.

Temporal Variability of Precipitation and Temperatures Indices
In this part, for better evaluation of the temporal variabilities [91], temporal analysis of trends was performed over five subregions [92]  The above analysis indicated that nighttime cooling was higher than daytime cooling in the period 1993-2014, and over the same period, it was observed that nighttime warming was higher than daytime warming. Figure 17 depicts the overall upward trend for indices including RX1day, RX5day, RX7day, R10mm, R20mm, and R30mm, as well as R99p, R95p, and PRCPTOT. All the indices showed both upward and downward trends (not clear trends) from 1950-1992. However, from 1993-2014, they slightly increased. For these indices, the WGC received the maximum rainfall amount, followed by the EGC and the whole domain. Very large differences between values were noted from one subregion to another. Obviously, the values of WSHL and CSHL were close, since they are located at the same latitude. This confirms that the rainfall experience over West Africa was due to the swaying of the West African Monsoon (WAM). The temporal representation of the change in CDD is opposite to that of the change in CWD for each subregion. There was an increase in CWD over WSHL, CSHL, and ESHL and a slight decrease over WGC and EGC, but the trends were mixed over the whole period. The overall trend (WA) was upward, with +0.012 day/decade. For the subregions, there was no obvious trend in CDD. However, the highest value of CDD was registered over the ESHL and seemed to then decrease from 1992-2014 by 0.009 day/decade. The CSHL trends were alternatively upward and downward in 1950-2014, while the upward and download trends over WSHL led to an overall increase of about +0.024 day/decade. The lowest values of CDD were located in WGC, which showed a decreasing trend in CDD over the whole region. On the other side, the lowest CWD was closed, located over WSHL, CSHL, and ESHL, and overall followed an upward trend of about 0.021 day/decade. The highest CWD (about 59 ± 6) was located in subregions such as WGC and EGC but showed a decreasing trend over time. to reach its peak in 2013. The above analysis indicated that nighttime cooling was higher than daytime cooling in the period 1993-2014, and over the same period, it was observed that nighttime warming was higher than daytime warming.   Figure 17 depicts the overall upward trend for indices including RX1day, RX5day, RX7day, R10mm, R20mm, and R30mm, as well as R99p, R95p, and PRCPTOT. All the indices showed both upward and downward trends (not clear trends) from 1950-1992. However, from 1993-2014, they slightly increased. For these indices, the WGC received the maximum rainfall amount, followed by the EGC and the whole domain. Very large differences between values were noted from one subregion to another. Obviously, the values of WSHL and CSHL were close, since they are located at the same latitude. This confirms that the rainfall experience over West Africa was due to the swaying of the West African Monsoon (WAM). The temporal representation of the change in CDD is opposite to that of the change in CWD for each subregion. There was an increase in CWD over WSHL, CSHL, and ESHL and a slight decrease over WGC and EGC, but the trends were mixed over the whole period. The overall trend (WA) was upward, with +0.012 day/decade. For the subregions, there was no obvious trend in CDD. However, the highest value of CDD was registered over the ESHL and seemed to then decrease from 1992-2014 by 0.009 day/decade. The CSHL trends were alternatively upward and downward in 1950-2014, while the upward and download trends over WSHL led to an overall increase of about +0.024 day/decade. The lowest values of CDD were located in WGC, which showed a decreasing trend in CDD over the whole region. On the other side, the lowest

Discussion
Some recent studies [46,47,93] investigated the performance of CMIP6 models and found that they had a better performance in regard to precipitation than CMIP5 models. In the present research, the spatial and temporal evolution of extreme climate events over the whole of West Africa and its five subregions during the period 1950-2014 was analyzed based on the CMIP6 dataset by selecting some indices of extreme temperature and precipitation.
The precipitation over WA was widely distributed when considering individual models. As aforementioned, monsoon period rain contributes more than 85% of the total annual precipitation over the latitudes 5 • N-20 • N. This seems to agree with [94], which found that in September, the rainfall was the highest (about 30.85% contribution of the total annual amount) and slightly increased from August (29.44%). The present study illustrated that the area with a decrease in total annual rainfall, as revealed by the models, was the Guinea Highlands, while Cameroon, Gabon, and the Savannah exhibited the highest increasing trends ( Figure 5). Regarding the extreme precipitation in WA, the study depicted strong links of its changes to climatic zones and showed its increase over the study period in Savannah. The identification and location of the trends were the same under all the statistical methodologies applied; the trends differed only in their magnitudes. The trends in precipitation were enhanced with the results on percentile-based metrics, which revealed the existence of a gradual north-south trend. The result converged with [95], which noted a drying trend over WA from 1951 to 2012. Additionally, ref. [96] noticed that during the period 1990-2010, both annual rainfall and the frequency of rainy days increased, leading to partial recovery from the severe dry period recorded in WA in the 1970s. This dryness was studied in [97,98], wherein it was illustrated that at the beginning of the 1970s, all climatic zones in tropical West Africa, from the arid Sahelian to the humid Guinea Coast climate, experienced a decade-long period of below-normal annual rainfall amounts. This finding converged with analysis of the temporal variability of trend, which revealed the existence of a breakpoint in 1992. Ref. [97] noted a decrease in the number of rainy events over the central Sahelian country of Niger in the two dry decades from 1970 to 1989; the results showed a quasinormal condition (or a slight increase) for temperature and precipitation from 1950-1992. A study led by [99] associated the "recovery" of WA in the 1990s to greenhouse gas (GHG) changes noticed between 1910 and 2008. This recovery in rainfall will influence the climate water balance (CWB, refs. [20,24]) and increase uncertainty in regard to hydrology, agriculture, and climate change. Although a general downward trend was noticed in the north, the ITA and MMK tests captured the existence of slight increases in R95p and R99p within the region and confirmed the maximum location as well as an upward trend in the amount of rainfall received around the Guinea highlands, Cameroon mounts, and Gabon forests. These results agreed with previous studies [47,100], which identified similar trends. Additionally, the present study indicated that the more the number of cumulative days increased, the more important the positive trend in the amount of rainfall received in the south was. R95p illustrated a significant positive trend in heavy rainfall near countries such as Gabon, Cameroon, the southern part of Nigeria, the northern part of Benin, Burkina Ghana, and the southern part of Mali. According to [101], it was expected that the number of consecutive dry days would be more pronounced in the northern regions than in the southern. The same analysis led to the conclusion in [102] that rainfall in the littoral zone of southern WA was more extreme than that inland.
The temperature has, over most regions across the world, an increasing trend. The annual means of daily maximum and minimum temperature exhibited a significant increase during the focus study period . Ref. [103] discovered the same warming trends over WA, while [104,105] found similar results in Eastern and Southern Africa, respectively. Ref. [106] reported that the temperature increase in high latitudes was greater than that in low latitudes. Regarding the temperature extreme changes in the whole of WA, TN90, TX90p, and WSDI indicated a general warming trend. That result was in line with the findings in [107] during the period 1979-2005, where an upward trend was noted in TX90p and TN90p associated with an increase in WDSI. The cold nights (days), TN10p (TX10p), showed a southward declining trend. This observation was confirmed by [101], which identified uniform declining trends in TN10p and TX10p over WA. Furthermore, during the period 1950-2014, the trends of cool/warm nights (TN10p/TX90p) were more significant than those of cool/warm days (TX10p/TX90p); this justifies the finding of [56], which reached the same conclusion over China. The absolute extreme temperature indices, such as TXX, TXM, TMM, TNX, TNN, TNM, and TXN, experienced an increasing trend over the whole study area. The Sahel area became warmer, with significantly high values noticed; a clear northward dynamic gradient was well represented. Overall, clear warming weather events were experienced in WA, with significant increases in all the extreme temperatures. This change in climate conditions revealed the manifest effect of climate change in the study domain. Studying the temporal variability and trends of temperature, the study revealed that the cold indices TX10p, TN10p (Figure 16a,b), and CSDI ( Figure 17k) expressed a general temporal negative trend. Those indices had breakpoints within their trends in 1962 in terms of temperature change because, prior to the clear declining trend, the indices observed a slight increase from 1950 to 1962 before adopting an upward trend until the end of the study period (2014).
The present study revealed that in WA, very few positive trends of CDD were generally observed, with the maxima converging in the northern part as well as over some other countries (Benin, Togo, Ghana, and Cote d'Ivoire). However, the increasing trends of CDD reduced southwardly. A contrast was noticed in that the area receiving a lot of rainfall (Guinea Highlands, Cameroon mounts, and Gabon forests) displayed a decreasing trend in CWD, while the savannah was more wet. The trends regarding CDD and CWD confirmed findings from [103]. The negative trend in CWD was directly correlated with a southward increase of R10mm, R20mm, and R30mm in WA, similarly as in the findings in [20], which detected increasing R10mm and R20mm over the orographic regions and the ocean boundary (Gulf of Guinea). Changes in CDD and CWD can lead to uneven temporal distributions of rainfall. It is manifest that the persistence in the upward trend in CDD and the simultaneous downward trend in CWD led to a negative impact on the PRCPTOT trend over the study domain and could have a negative influence on the water resource demand in WA in general and especially in the northern regions. This may lead to efforts to further irrigation plans to supply the needs of water for agriculture production in the northern area. Furthermore, the southward diminution in CWD might induce a reduction in the number of times WA is watered, and the concurrent upward noticed in SDII and the R30mm might induce localized floodlike situations over the southern regions. It is important to note that CDD and CWD are crucial in the magnitude of flooding (especially flash flooding) events because of their implications on the soil moisture state before the occurrence of flooding. Moreover, as [108] indicated, increasing soil moisture reduces the infiltration capacity of the study domain and then fosters flooding occurrence. A very strong correlation was noticed among R99p, R95p, and PRCPTOT. The increase in PRCPTOT over WA was illustrated in previous studies [103,107]. This attests that other rainfall events over the area did not contribute much to the total annual rainfall amount. Indirectly, extreme rainfall events increased in intensity in the southern area and might be the most accountable for the total annual rainfall.

Conclusions
In this study, the long-term spatial and temporal variabilities of and changes in rainfall and temperature were analyzed. Based on climate indices suggested by ETCCDMI, the study applied three statistical tests to assess the trends in the two variables. The following conclusions from the study can be made:

1.
The total annual rainfall was found to decrease around the coastal area, especially over the Guinea highlands, the Cameroon mountains, as the Gabon forests, but increase over the Savannah and Sahel regions. Furthermore, rainfall during the monsoon months contributed more than 85% of the total annual rainfall in the study domain.

2.
The interannual Tmax and Tmin both followed the same trends as the total annual rainfall, with a northward gradient. The warmest region was the Savannah-Sahel, while the coastal part was the coolest area. Using ITA, particular increasing trends were identified from the models in the northern part and the Guinea coast. The increasing trends around the Guinea coast may be due to the gradual increase in sea surface temperature (SST) due to global warming (GW).

3.
Extreme high-temperature indices (warm extremes) significantly increased, while the cold extremes indicated a significant upward trend. Both indices showed a breakpoint (abrupt changing point) in 1992, after which the trend increased more in power.

4.
The study revealed that the more the number of cumulative days increases, the more important the positive trend in the amount of rainfall received in the south was. Based on analysis of indices such as TN90, TX90p, and WSDI, the study also indicated a general warming trend over the whole of WA. However, over the study period, the trends in cool/warm nights (TN10p/TX90p) are more significant than those in cool/warm days (TX10p/TX90p).

5.
The upward trend in CDD and simultaneous downward trend in CWD led to a negative impact on the PRCPTOT trend over WA. This may affect water resource demand in general in WA and especially in the northern areas. Moreover, the decline in CWD showed a reduction in the wet spells in WA, and the concurrent upward notice in SDII and the R30mm might induce localized floodlike situations over the southern regions. Thus, it is important to note that CDD and CWD are crucial in regard to the magnitude of flooding events because of their implications on the soil moisture state before the occurrence of floods. 6.
The innovative trend analysis (ITA) methodology applied in this work was able to capture the most minute trends existing in a time series, including some that could not be detected by the usual tests so far used, such as Mann-Kendall and Sen's slope. The reliability of ITA in tracking unseen trends in time-series data encourages us to recommend it to the reader as a reliable method to be used in time-series trend detection. 7.
Information gathered together from this study can contribute to producing sustainable water resource planning and management. It could also be useful for policy makers and scientists for exploring extreme climate event trends on regional and local scales to plan the circumstances in which potential floods and droughts might occur.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/w13243506/s1, Figure S1: Spatial distribution of the contribution of the West African Monsoon in the annual rainfall over the period 1981-2014 based on the observed data (CHIRPS and CRU) and the selected CMIP6 dataset. Figure S2: Spatial distribution of the minimum temperature (Tmin) over the period 1981-2014 based on the observed data (CHIRPS and CRU) and the selected CMIP6 dataset. Figure S3: Spatial distribution of the minimum temperature trend over the period 1950-2014 based on the CMIP6 dataset and Sen's slope test. Figure S4: Spatial distribution of the minimum temperature trend over the period 1950-2014 based on the CMIP6 dataset and the modified Mann-Kendall (MMK) test. Figure S5: Spatial distribution of the maximum temperature trend over the period 1950-2014 based on the CMIP6 dataset and the innovative trend analysis (ITA) test. Figure