*3.2. Trend Analysis of Temperature and Rainfall*

In this study, the Mann–Kendall (MK) trend test was used. In regards to the temperature time series, a significant positive Z-scores value (9.3067) which is far greater than 1.96 was found, suggesting that the temperature trend is increasing. However, for rainfall, the MK trend test shows no-significant negative Z score (−0.17143) which above −1.96, suggesting weak decrease in rainfall variability.

Figure 6a shows the sequential statistic values of forward/progressive (Prog) u(t) (solid red line) and retrograde (Retr) u'(t) (black solid line) obtained by SQ-MK test for Conakry yearly mean temperature. In general, SQ-MK indicates and upwards trends of temperature in Conakry which is noticeable in both Prog. and Retr. SQ-MK statistic. The possibility is that the upwards trend started before the beginning of the time series (1960) because the change detection point, a point where Prog. and Retr. cross each other did not occur in the graph. What is noticeable in this figure is that, it is only from 1984 that this progressive SQ-MK statistic becomes positive and significant. At the same pace, it gradually increases until 1989 and then stands until 1998, the year from which the trend has increased significantly far above the confidence level (+3.866541) up to 2016 (+9.306717). There is a significant upward trend which seems to coincide with the 1970s 1980s droughts episodes and strongly the 1998 and 2014–2016 strong El Niño event. In a study that uses the similar non-parametric test method, Suhaila et al. [64] reported that the detection points captured by Pettitt and SQ–MK tests in Peninsular Malaysia temperature series during the years 1996, 1997 and 1998 are possibly related to climatic factors, such as El Niño and La Niña events. The retrograde statistic values are significant and negatives during the period from January 1960 to 1992 before it continues to be within the 95% confidence level limits (±1.96) except the year 1998 which the retrograde statistic value is significantly positive.

**Figure 6.** Sequential Mann-Kendal statistic values of progressive u(t) (solid redline) and retrograde u'(t) (black solid line), obtained by Sequential Mann-Kendall test for temperature (**a**) and rainfall (**b**) of Conakry airport: 1960–2016.

Figure 6b depicts the sequential statistic values of forward/progressive (Prog) u(t) (solid red line) and retrograde (Retr) u'(t) (black solid line) obtained by SQ-MK test for Conakry annual rainfall data for the period from 1960 to 2016. A strong significant upward trend was observed in late 1961, with the significant trend turning point observed in June 1962, which means that 1961 is the only year that is characterized by a positive and significant trend over the entire study period. But a careful analysis of the trend in progressive and retrograde which are non-significant (between ±1.96) and sometimes negative or positive shows two distinct periods that correspond to that found by the analysis of precipitation anomalies. For the first period (1970–1974) and the second period (1981–1985), the Retrograde curve is below the progressive curve in the negative band, which corresponds to

periods of deficit rainfall. For the rest of the study period, the two curves intersect each other or the retrograde curve is above the progressive curve, that corresponds to periods with variable or normal rainfall. The response of the West African drought of 1970s and 1980s is clearly identified by the reduction in the rainfall at Conakry. Statistically there is a no-significant downward trend in rainfall since the end of 1960s.

In summary, the SQ-MK test andMK model for Conakry yearly data shows that the temperature and rainfall are subject to a significant increasing trend and a no-significant decreasing trend, respectively, during the period from 1960 to 2016. Thus, these methods seem to be useful for explaining the variability and trends of both temperature and rainfall.

In order to investigate physical relationships between climate forcing, precipitation and streamflow in the Mid-Atlantic region, Schulte et al. [4] selected eight climate indices. In the present study, four climate indices (Niño3.4, AMM, TNA and AN) were used as explanatory variables for this model because of their well-known possible influence on temperature and rainfall variability over West African region. The relevant time series of these climate indices are shown in Figure 7. Zebiak [27] specified that the dominant signature of ENSO is clearly focused on the Equator and its temporal variability is strongly focused at 3–5-year time scales.

**Figure 7.** The standardized monthly Niño3.4 (**a**), AMM (**b**), TNA (**c**) and AN (**d**) time series for the period from 1960 to 2016. The vertical dashed lines indicate the year 1998.

There are two main forms of coupled ocean–atmosphere variability that exist in the tropical Atlantic Ocean, namely: the first one Atlantic Meridional Mode (AMM) [65] which is also called the interhemispheric mode [66]. It was originally identified by Servain [67]. This mode of variability is characterized by an interhemispheric gradient in sea surface temperatures and by oscillations in the strength of surface winds that cross the Equator, thereby reinforcing sea surface temperature anomalies [68]. The pronounced coupled ocean-atmosphere variability in the Tropical Atlantic is generated by fluctuations in the Atlantic Meridional Mode (AMM) [68]. The AMM is characterized by an anomalous meridional shift in the Intertropical Convergence Zone (ITCZ) that is caused by a warming (cooling) of SSTs and a weakening (strengthening) of the easterly trade winds in the northern (southern) tropical Atlantic [69]. And, the second one is the zonal mode, also called the Atlantic Niño [70]. Its seasonal evolution is due to surface wind variations associated with the northward migration of the ITCZ [71].

The tropical northern Atlantic (TNA) SST anomaly pattern is an important component of the tropical Atlantic SST variability, which is characterized by warm (or cold) SST anomalies in the TNA [72]. Sea surface temperatures in the tropical North Atlantic (TNA) affect the meridional movement of the

ITCZ and its band of heavy rainfall and cloud cover [73]. The Atlantic Niño (AN) is often regarded as something like the little brother of El Niño. During Pacific El Niño events, sea-surface temperatures (SSTs) in the central and eastern equatorial Pacific become warmer than average. Prevailing theories on the equatorial Atlantic Niño are based on the dynamical interaction between atmosphere and ocean [74]. In very much the same manner, SSTs in the central and eastern equatorial Atlantic become warmer than average (or anomalously warm) during Atlantic Niño events. The Atlantic Niño index used in this study is obtained by calculating the area average of SST in the cold tongue region, defined as 20◦ W to 0 and 3◦ S to 3◦ N [26]. While El Niño usually peaks in northern hemisphere winter, the Atlantic Niño peaks in summer [75]. Therefore, understanding of the Atlantic Niño (or lack thereof) has important implications for climate prediction in those regions. Although the Atlantic Niño is an intrinsic mode to the equatorial Atlantic [27].

There may be a tenuous causal relationship between climate parameters and the Atlantic Niño in some circumstances. Therefore, MLR and Wavelet analysis are used to identify the dependency and coherence between temperature, rainfall and climate forcings.

The correlations between the four indices used in our study are shown in Figure 8. There is strong correlation coefficient between AMM and TNA (0.76). As the two explanatory variables are strongly correlated, the MLR analysis may have difficulties to separate the contributions. For that purpose, the wavelet analysis was used by calculating the coherence between explanatory variables and the dependent variable separately.

**Figure 8.** Correlation between the standardized monthly Niño3.4, AMM, TNA and AN time series for the period from 1960 to 2016, "n" is data number. The values are the correlation coefficient between the 4 parameters and we found that TNA and AMM are significantly correlated. The "\*\*\*"; "\*\*"; "\*"; show that the correlation is significant to the 0.001; 0.01; 0.1 level and " " mean that there is no correlation.

For the MLR the explanatory variables (AMM + Niño3.4+ TNA + AN) were used. The output of the MLR statistical analysis of temperature and the independent variables is shown in Table 1. Statistically the results in Table 1 reveal a significant relationship between temperature and Niño3.4 AMM, and TNA, with *p*-values of 0.0138, 1.99 <sup>×</sup> 107, and less than 2 <sup>×</sup> 1016, respectively. The *p*-value for AN indicates a statistically insignificant association with the temperature because of *p*-value which is far greater than 0.05.


**Table 1.** The output of Multiple Linear Regression (MLR) model in which temperature is a dependent variable and AMM, Niño3.4, TNA and AN are independent variables.

Significant codes: 0 '\*\*\*' 0.001 '\*\*' 0.01 '\*' 0.05 '.' 0.1 ' ' 1.

A comprehensive summary of the MLR analysis statistics encompassing rainfall, Niño3.4, AMM, TNA and AN is shown in Table 2. The results in Table 2 reveal a statistically significant relationship between rainfall and Niño3.4, AMM, and TNA with *p*-values of 0.04374, 0.00441, 0.00301. The *p*-value for AN indicates a statistically insignificant association with the rainfall because of *p*-value (0.73691) which is far greater than 0.05. A strong dependence between the two meteorological parameters, AMM and TNA were found. And then, Niño3.4 has a moderate influence on the temperature and rainfall of Conakry. The low dependency between the AN and these two meteorological parameters would be due to the distance between the Conakry site and the NA focus (3◦ S–3◦ N), and this could be verified for another station closer to the equator. To analyse the two-component dependence of which one in the temporal environment and the other in the frequency environment between the temperature, the rainfall and the forcings used in the study, unlike the MLR, the wavelet model has been evaluated and model outputs are explained in the next section.

**Table 2.** The output of Multiple Linear Regression (MLR) model in which rainfall is a dependent variable and AMM, Niño3.4, TNA and AN are independent variables.


Significant codes: 0 '\*\*\*' 0.001 '\*\*' 0.01 '\*' 0.05 '.' 0.1 ' ' 1.

An influential variable for most African rainfall areas is the zonal wind over the tropical Atlantic, the north-south SST gradient in the tropical Atlantic modulates rainfall in West Africa as expected [76].

Figure 9a depicts the time evolution of mean monthly temperatures, with the warming trend line superimposed. From Theil-Sen function, in this study, 684 points were used for trend estimation. The trend estimate is: *p* < 0.001 = \*\*\*, *p* < 0.01 = \*\*, *p* < 0.05 = \* and *p* <0.1 = +. The temperature increases at 0.02 ◦C per year (0.2 ◦C/decade) at Conakry. The superimposed red line indicates the obtained linear trend. And the dashed red lines indicate the 95% confidence interval. The annual evolution of rainfall exhibits a negative slope, which corresponds to decreasing trend (Figure 9b) at −8.14 mm per year (−81,4 mm/decade). Compared to other sites in West Africa, our results are similar to that found by [32], for stations in downstream Kaduna River Basin during 1975–2014, in Nigeria. The fifth Intergovernmental Panel on Climate Change assessment stated Africa surface temperature already increased by 0.5 ◦C–2 ◦C over the past hundred years and an observed drop in average annual rainfall of approximate 25–50 mm each decade from 1951–2010 in some parts of West Africa [77]. Globally, according to the IPCC Special Report [78], it has been reported that the warming of anthropogenic origin has already exceeded the environment.

**Figure 9.** Shows the long-term trend of monthly temperature (**a**) and rainfall (**b**). The solid red line shows the trend estimate and the dashed red lines show the 95% confidence intervals for the trend based on resampling methods. The overall trend is shown at the top-left as 0.21 ◦C per year (**a**) and −8.14 mm per year (**b**), and the 95% confidence intervals in the slope from 0.2–0.23 ◦C/year (a) and −10.5–6.03 mm/year (**b**). On the figures, the sign "\*\*\*" shows that the trend is significant to the 0.001 level.

The seasonal distribution of temperature is shown in Figure 10a. the increase in temperature is more significant in winter (December–January–February) of 0.03 ◦C per year than spring (March–April–May), summer (June–July–August) and autumn (September–October–November) of 0.02 ◦C per year. The Figure 10b depicts the seasonal distribution of rainfall; no significant trend is observed in the winter months and rainfall values seem to be stables. Negative linear trend was found in spring (−0.34 mm/year), autumn (−1.23 mm/year) and for summer, the trend is positive (0.1 mm/year).
