*3.3. Wavelet Analysis*

The Figure 11 shows the normalised wavelet power spectrums calculated for the time series of temperature (a) and rainfall (b) for the period from year 1960 to 2016. In this figure, the "u" shaped solid lines represent the cone of influence (COI) which define the region of the spectrum which should be considered in the analyses. The COI actually indicates areas where edge effects occurs in the time series [44]. The thick black contours are the 95% significant regions of confidence level [42].

The main purpose of using the wavelet transform technique in our study is to identify any dominant variability mode that may be present within the two meteorological parameters (temperature and rainfall). In general, the wavelet transform of temperature shows two distinctive peaks (Figure 11a) corresponding at 6 and 12-month periods. The 12-month period shows strong power spectrums during the distinctive periods of 1961–1965, 1970, 1984 and 1992–2016. Its intensity increases with no-interruption from 1992 to 2016. These dominant wavelet peaks seem to be consistent with the results presented in Figure 3c. The wavelet power spectra for rainfall indicate a strong power spectral peak of 12-month cycle which starts from year 1960 to 2016 (Figure 11b). Moreover, there is a weak power that seems to appear at 6-month period for a few years and within the 95% significant regions of confidence level for distinctive years 1961, 1965, 1967, 1970, 1986–1992, 1998–2000 and 2002–2007.

**Figure 10.** The plot shows the Seasonal trend distribution of monthly mean temperature (**a**) and rainfall (**b**) obtained with Theil-Sen's estimate. 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 vertical red line indicate year 1998. There is no trend in rainfall during winter (0 mm per year).

**Figure 11.** Wavelet transform of temperature (**a**) and rainfall (**b**) variability from 1960 to 2016 at Conakry. The black solid line contour delimits the region (red) where the power is strong and significant and the cone of influence indicates the 95% confidence level.

It is important to note that the distinctive power periodicities found in temperature may be associated with the annual and semi-annual cycles, which are controlled by the alternation between dry and wet seasons. The strong and continuous power spectrums shown by rainfall confirms the annual cycle variability of rainfall year-round. Sylla et al. [7] pointed out that depending on a given year, the onset of WAM may be strong or quiet in the second half of June and the West African rainfall is highly variable on intra-seasonal, interannual, and interdecadal time scales. The wavelet coherence analysis between temperature, rainfall and the four climate indices used in this study is shown below.

Wavelet coherence is a method for analysing the coherence and phase lag between two time series as both a function of time and frequency [43]. This method has been shown to be the best possible method to indicate teleconnection between two independent time series. Thus, this section focuses on investigating the teleconnection between both Conakry temperature and rainfall, and selected climate indexes. Figure 12 shows the cross-wavelet power spectra for (a) Temp–Niño, (b) Temp-AMM, (c) Temp–NA, and (d) Temp–AN, respectively. The phase relationship is represented by arrows. The regions where two cross-wavelet parameters are in phase is shown by arrows point to the right, anti-phase if the arrows point to the left, and temperature or rainfall leading (or lagging) if the arrows point upwards (or downwards), respectively. The vectors were only plotted for areas where the squared coherence is greater or equal to 0.5. More details about wavelet coherence calculations can be find in studies such as Grinsted et al. [41] and Schulte et al. [4]. The solid black line indicates the cone of influence (COI) where the edge effects become significant at different frequencies (scales), and the solid black line delimit the 95% significant regions of confidence level.

**Figure 12.** Wavelet coherence between temperature and Niño 3.4 (**a**), AMM (**b**), TNA (**c**) and AN (**d**) 1960–2016, the phase relationship is represented by arrows. The black solid line contour delimits the region (red) where the power is strong and significant and the cone of influence (solid black line) indicates the 95% confidence level.

Having found that the wavelet transform shows strong forcings with 6 and 12–months periods in temperature variability, we have proceeded to identify the wavelet coherence signature between temperature and the four climate modes (Niño3.4, AMM, TNA and AN). Figure 12a shows the coherence calculated for the monthly mean temperature and Niño 3.4 in both time and frequency domain. At the period band of 32–64 months, significant relationship between the temperature and Niño 3.4 is clearly visible from 1960 to 2016. According to the arrows (phase) which are pointing upward and then turning to the right at the period band 32–64, the temperature seems to lead the Niño 3.4. Also, there seems to be an in-phase relationship which may indicate a strong teleconnection of the Conakry temperature to Niño variability. It is also important to note that there is a distinctive appearance of periods which are less than 13 months in the time series, with varying phase relationships between the parameters.

The wavelet coherence between temperature and AMM (Figure 12b) is observed to delineate some areas that have high significant power at periods between 4–12 months with significant peaks of distinctive periods 1960–1970, 1984–1990 and 2005–2016. A study by Foltz et al. [79] also reported that there was cooling of SSTs in the equatorial North Atlantic (ENA; 2◦–12◦ N) in 2009 in response to a strong Atlantic meridional mode event. It is also important to mention that there are significant peaks appearing in the period band of 32–48 months during 1969–1976 and 2005–2010, respectively. A key component of the AMM is a positive feedback between the ocean and the atmosphere. Surface air pressure responds to the SST anomalies, becoming higher than normal over the anomalously cold SSTs and lower than normal over anomalously warm SSTs.

In Figure 12c, there is a significant in-phase relationship between the temperature and TNA during the period band of 8–12 months with strong power during 1962–1970, 1975 and 1973–1984. In addition, it is noted that a significant in-phase relationship is also found during 1965–2004 and 2006 at the period band of 30–64 months. The Figure 12d shows an in-phase coherence between temperature and AN corresponding at the period band of 8–12 months during 1965–1970 and 1972–1984. At the period band of 32–64 months, the AN lead the temperature, so the significant power appears at the period from 1963 to 1968, 1982 to 1992 and 1994. From these results, it can be suggested that the four climate indices contribute to drier conditions across the Conakry region. Then, the wavelet coherence spectra show that the Niño 3.4, AMM, TNA and AMM are coherent with temperature at different time scales. We can summarise that temperature is subjected to climate indices forcing at Conakry. Anomalous surface wind flow from the cold to the warm hemisphere, strengthening the mean south-easterly trade winds in the South Atlantic and weakening the north-easterly trade winds in the North Atlantic. The surface wind anomalies thus provide a positive feedback onto the initial SST anomalies by forcing changes in wind-induced evaporative cooling of the ocean.

The wavelet coherences between four climate modes and rainfall are shown in Figure 13. Significant in-phase coherence was found with the Niño 3.4 at a band period of 8–12 months (Figure 13a). This coherence suggests that the negative phase of Niño 3.4 is in agreement with dry years and its positive phase with wet years at Conakry site. The secondary peak of significant coherence appears at the band period of 16–32 months from 1962 to 1967. The observed significant coherence at a period of 16–32 months seem to be partially linked to the response of wet condition of 1960s. An in-phase relationship between rainfall and Niño 3.4 is found too at the band period of 128–256 months from 1970 to 2005. Since ENSO events can have substantial influence on African rainfall [80], the equatorial region exhibits more rainfall during El-Niño years than La-Nina years [81]. In the Indian Ocean basin, Narasimha and Bhattacharyya [82] suggested that the stronger coherence between Homogeneous Indian Monsoon and Niño 3.4 is found in the 2–7-year band and that both rainfall and the Niño 3.4 index appear irregular and random. During 1950–99, there were seven most significant El Niño events (1957–58, 1965–66, 1972–73, 1982–83, 1986–87, 1991–92, and 1997–98) for which the SST anomalies in the Niño 3 region (5◦ S–5◦ N, 150◦ W–90◦ W) exceeded 1 ◦C [83].

**Figure 13.** Coherence between rainfall and Niño3.4 (**a**), rainfall and AMM (**b**), rainfall and TNA (**c**) and rainfall and AN (**d**) 1960–2016, the phase relationship is represented by arrows. The black solid line contour delimits the region (red) where the power is strong and significant and the cone of influence (solid black line) indicates the 95% confidence level.

The Figure 13b depicts the wavelet coherence between rainfall and AMM and indicates that the AMM response to the rainfall variability shows an in-phase relationship at the band period of 8–12 months during the periods 1960–1970, 1978–1984, 1985–1990 and 1995–2016. The wavelet coherence analysis detected at the band period of 16–32 months lagged (i.e., AMM leading) relationship with the wet conditions during 1965, 1998–2004 and 2005–2016. Another peak is shown around 64 months from 1970 to 1984, which seems to be in relationship with the 1970s and 1980s droughts. The AMM is the dominant source of coupled ocean-atmosphere variability in the Atlantic and it affects rainfall in tropical cyclone development in the North Atlantic. During a positive phase of the AMM, the ITCZ is displaced northward. Warmer than normal SSTs and weaker than normal vertical wind shear during positive phases of the AMM tend to enhance tropical cyclone development in the Atlantic. The conditions are opposite for the negative phase of the AMM. The AMM exhibits strong variability on interannual to decadal timescales.

Figure 13c shows the coherence analysis between rainfall and TNA. A significant coherence and out-of-phase between rainfall and TNA is found at the band period of 8–12 months during 1961–1970 and 1980. A second significant coherence of in-phase relationship is shown at the band period of 16–32 months from 1965 to 1970 and 2005 to 2012. Comparing to the results of the case study for the northern part of Brazil, Uvo et al. [84] reported that the variations of April–May averaged precipitation are closely connected to the changes in the TNA SST. And sea surface temperatures in the tropical North Atlantic affect the meridional movement of the ITCZ and its band of heavy rainfall and cloud cover [73]. Furthermore, the results found by Sun et al. [85] clearly demonstrate that the climate indices have the influential consistent correlation relationship with the precipitation variation in Korea.

The wavelet coherence between AN and Conakry rainfall data was also computed (Figure 13d). The wavelet analysis detected a statistically significant coherence and in-phase relationship at the band period of 4–12 months during the years 1965, 1970, 1980, 1986 and 2005. Significant out-of-phase coherence was found at a band period of 32–48 months during the periods 1965–1984 and 2005–2013, suggesting that the positive phase of the AN contributes to drier and cooler conditions in Conakry. A period of significant coherence between the AN and rainfall extending from 1976 to 1994 was also identified at the band period of 128–190 months (~11-year), which may be due to solar cycle. Using wavelet techniques to examine the association between Indian monsoon and solar activity,

Bhattacharyya and Narasimha [86] found the power in the 8–16 y band during the period of higher solar activity at confidence levels exceeding 99.88%. The teleconnection between AN and both the temperature and rainfall measured at Conakry seems to be in agreement with previous studies (e.g., Hastenrath and Polzin. [87]; Rodriguez-Fonseca et al. [88] and others). In their study on the role of the SST anomalies in the West Africa droughts, Rodriguez-Fonseca et al., 2015 reported that the tropical Atlantic SST variability influence the West Africa rainfall in different time scales: the variability in areas closer to the equator and those at the south.

To compare our result to those of other areas, a study by Mbata et al. [89] reported for the sector of Democratic Republic of Congo that the wavelet analysis of the rainfall time series indicates an important fluctuation between practically 1960 and 1970. And then, Giannini et al. [90] suggested that the atmospheric convection and circulation changes due to the Atlantic Niño can cause increased precipitation across the equatorial Atlantic and decreases over the Sahel. These climate modes appear to have contributed substantially to the 1970s and 1980s drought in a different way and scales. The widespread influence of El Niño Southern Oscillation (ENSO) events on regional climate can have considerable socio-economic impact Climatic effects of ENSO, which vary substantially with region and season [91].
