For this work, the data used are provided by the Agency National Meteorology (ANM) and include climatological series related to daily, monthly and annual precipitation from 1994 to 2023. The choice of this period is justified by the need to consider a maximum number of precipitations in the observation networks and to have a long series to better analyze precipitation variability in the study area.

The methodological approach begins with a search for general and specific scientific papers and other documents that have helped to better define the aspects of the subject. An analysis of the literature reviewed enabled us to identify the methods best suited to the study. Software and formulas were used to produce average seasonal rainfall cycles and maps, and to calculate certain values and statistical tests.

This is the ratio between the standard deviation and the mean. It shows whether or not there is a high degree of variation in the data. When it is less than 100, it indicates a low level of data dispersion. On the other hand, a $C_v$ value > 100 means that there is a high degree of variation in the data. In this case, we can assume the presence of heterogeneity. Its mathematical expression is given by the equation:

$C_v=\frac{\sigma }{X}\times 100$(1)

This is an index that measures the deviation from an average established over a long period based on station data. The annual rainfall index is defined as a reduced-centered variable [18] [19] . It is obtained by the equation:

$Ip=\frac{X_i-X_m}{\sigma }$ (2)

where $X_i$: rainfall in year $i$; $X_m$: mean interannual rainfall; $\sigma$: standard deviation of interannual rainfall; $Ip$: rainfall index. The rainfall index determines the number of surplus or deficit years and the corresponding significance in terms of the magnitude of the climatic phenomenon ( Table 1 ).

Source: [18] [19] .

Trends in maximum and minimum rainfall have been highlighted by a regression line of the type given by the equation:

$y=ax+b$ (3)

where y is the value of the variable whose trend is being investigated; a is obtained by calculating the slope, the directing coefficient of the regression whose positive (+) or negative (−) signs express respectively the increasing and decreasing trend over time x and b, a constant such that:

$a=\frac{\left({\displaystyle \sum y}\right)\left({\displaystyle \sum x^2}\right)-\left({\displaystyle \sum x}\right)\left({\displaystyle \sum xy}\right)}{N{\displaystyle \sum x^2}-{\left({\displaystyle \sum x}\right)}^2};b=\frac{N\left({\displaystyle \sum xy}\right)-\left({\displaystyle \sum x}\right)\left({\displaystyle \sum y}\right)}{N{\displaystyle \sum x^2}-\left({\displaystyle \sum x^2}\right)}$

where the coefficients a and b are respectively the y-intercept and the slope of the line

There are several methods for detecting stationary breaks in rainfall series. In this study, the Pettitt test was used to detect stationary breaks in rainfall data. The series studied is divided into two samples of size m and n respectively. The values of the two samples are grouped together and ranked in ascending order. The sum of the ranks of the elements of each sub-sample in the total sample is then calculated. A statistical study is defined on the basis of the two sums, which is thus determined and tested under the null hypothesis that both sub-samples belong to the same population. This test is based on the calculation of the variable $U\left(t,N\right)$ defined by equation [20]

$U_{t,N}=\underset{i=1}{\overset{t}{{\displaystyle \sum }}}\underset{j=t+1}{\overset{N}{{\displaystyle \sum }}}{D}_{ij}$(4)

with

$D_{ij}=\text{sign}\left(x_i-x_j\right)$ hence, so, in which $\text{sign}\left(x\right)=1\text{if}x>0\text{and}-1\text{if}x<0$

For rainfall variables with a break in the time series, it is interesting to calculate the mean variations on either side of the break by applying the equation [21] :

$D=\frac{X_j}{X_i}-1$ (5)

where $X_j$represents the average over the period after the break and $X_i$the average over the period before the break.

Analysis of mean annual rainfall in the forest region over recent years (1994-2023) shows a heterogeneous distribution of rainfall across the region, with a gradual increase in rainfall levels from northwest to southeast. Areas of high rainfall are mainly located in the Guéckédou and Macenta zones, where values reach 2357.7 mm and 2267.1 mm respectively. Medium rainfall areas, with levels between 1904.7 mm and 1995.3 mm, are found in Yomou and N’Zérékoré, while low rainfall areas with intervals between 1632.9 mm and 1723.5 mm are found in Lola and Beyla ( Figure 2 ).

In this study, the average annual rainfall in N’Zérékoré was 1981.5 mm, 2265.5 mm in Yomou, 2192.2 mm in Macenta and 2126 mm in Beyla. Maximum annual rainfall averaged 3063 mm in N’Zérékoré, 3369 mm in Yomou, 3474 mm in Macenta and 3229 mm in Beyla. Table 2 gives details of the statistical coefficients of rainfall at these stations. We note that the coefficients of variation ( $C_v$) range from 33.09% to 42.19% and are less than 100, which means that rainfall in our study area shows little variability.

Analysis of rainfall indices (Ip) in Figure 3 shows that positive (surplus) rainfall anomalies vary from 0 to 2, characterizing rainy periods, marked by indicators of extreme rainfall events. Those below 0, on the other hand, are negative (deficit) anomalies and indicators of rainfall regression. Looking at Figure 3(a) , the

N’Zérékoré synoptic station shows that the years with the most deficit are 1996 and 2000 to 2007. This deficit is marked by a long period of persistent drought at this synoptic station. The most surplus years are 1998, 2014 and 2018 to 2023. The other synoptic stations ( Figures 3(b)-(d) ), Yomou, Macenta and Beyla respectively, have the same rainfall characteristics as the N’Zérékoré station. In terms of positive and negative rainfall indices (Ip), the study area as a whole has experienced years of high humidity and years of strong drought sequences over recent decades.

Ip grids were used to characterize the level of rainfall deficits observed and to assess the extent of drought sequences or humidity ( Table 3 ). Analysis of this table shows that the study area recorded an average of 00% years with extreme humidity, 17.5% years with high humidity, 40.84% years with moderate humidity, 15% years with moderate drought, 26.67% years with high drought and 00% years with extreme drought. The types of rainy years recorded are heterogeneous from one synoptic station to another. It is worth noting that for the entire 1994-2023 series, there were eight years (1996, 2001, 2002, 2003, 2004, 2005, 2006 and 2007) of severe drought at all stations, indicating the existence of hydroclimatic hazards in Guinea-Forester.

For the four (4) stations, Figure 4 shows interannual rainfall variability in the forest region, with rainfall amounts varying between 600 mm and 1600 mm during the period 2000 to 2007. For all stations, rainfall increased slightly from 1994 to 2000. And fall from 2001 to 2007, with higher rainfall recorded from 2008 to 2023. The N’Zérékoré station ( Figure 4(a) ) shows that rainy years are numerous and varied, but 2008 and 2022 remain the wettest years, with rainfall totaling 3000 mm. We note that from 2007 to 2023, recorded rainfall is at its highest, varying between 1500 mm and 3000 mm, with peaks in 2010, 2013, 2019 and 2022. At the Yomou station ( Figure 4(b) ), we note that during our study period, rainfall peaks were identified in 1998 (3400 mm), 2009 (3000 mm), 2010 (3000 mm), 2014 to 2023, which constitute the wettest years. Figure 3(c) shows a very marked variability at the Macenta station, with low rainfall values recorded during the period from 2001 to 2007. Over our study period, the years 1998, 2009, 2014 to 2016 (3000 mm - 3010 mm) and 2018 to 2023 (2990 mm - 3000 mm) are identified as the wettest years. Considering the Beyla station ( Figure 4(d) ), we note a similar variability to that of Macenta during the period 2001 to 2007. At this station, 1998 (3300 mm) and 2014 (3350 mm) are the two (2) rainfall peaks of our study period. We have identified the following years 1998 (3300 mm), 2007 to 2010 (2700 mm to 2800 mm), 2014 to 2016 (3000 mm - 3010 mm) and 2018 to 2023 (2990 mm - 3000 mm) as being the most rainy years due to the quantities recorded.

Application of the Pettitt test ( Figure 5 ) to rainfall series from 1994 to 2023 in Guiana-Forester shows that there has been a break in stationary at all the stations studied. There was a break in the same year for Figure 5(b) and Figure 5(c) (2013) and a different date for Figure 5(a) and Figure 5(d) (2014 and 2007 respectively). Given that the calculated p values are below the alpha significance level equal to 0.05, we must reject the null hypothesis and retain the alternative hypothesis of a break in stationary in the rainfall series studied.

In order to verify the breaking levels of these graphs, Pettitt’s statistical test using Monte Carlo simulations with a 99% confidence interval presented the results below ( Table 4 ). Analysis of this table reveals that the rainfall series as a whole are homogeneous and show break-up times.

Table 5 shows the increase in mean rainfall heights at the various stations expressed by the increase in rainfall deficit. Analysis of the table reveals that the rainfall increase deficit is higher at the Macenta station than at the other three stations. However, for all four stations, the rainfall increase deficit recorded is around 0.04% for the stations (Yomou and N’Zérékoré) and 0.28% for Macenta and Beyla.

Climate change in the forest region is reflected in the increasing frequency of extreme climatic events. Indeed, rainfall indices ( Figure 3 ) range from −1.5 to 1.6 for the N’Zérékoré synoptic station, −1.6 to 1.4 for the Yomou synoptic station, -1.7 to 1.4 for the Macenta synoptic station and −1.6 to 1.6 for the Beyla synoptic station over the study period (1994-2023). HE risks associated with extreme rainfall events are discussed, as there is little research in Guinea that systematically reports the losses and effects due to these events, yet their impacts on agricultural production, rural livelihoods and urban and economic sectors are obvious and considerable. Analysis of Figure 4 shows that average annual rainfall is unevenly distributed and varies across Forest Guinea. They range from 986 mm to 3229 mm for the Beyla and Yomou stations, 646 mm to 3474 mm (Macenta station) and 900 mm to 3063 mm for the N’Zérékoré station. For all stations, rainfall rose slightly from 1994 to 2000, then fell from 2001 to 2007, with higher levels recorded from 2008 to 2023. Similarly, Figure 5 shows that annual maximum precipitation experienced a very remarkable increase in rainfall height between 2014-2023 with trend and determination coefficients of 42.03 and 0.282; 50.18 and 0.297; 54.01 and 0.264 and 29.43 and 0.134 for Figures 4(a)-(d) respectively. The rainfall increase deficits observed show that the periods after the break were rainier than those before. In addition, the rupture periods cited in ( Table 5 ) are not significant, but rather a spatio-temporal change in rainfall for the various localities in the study area. These changes in rainfall data show an increase in rainfall of 0.59% at the N’Zérékoré station, 0.64% at Yomou, 0.71% at Macenta and 0.43% at Beyla. These results show that the absence of a significant break does not guarantee the absence of a drop in rainfall. Indeed, the break levels defined do not take into account the abruptness with which the change occurs in the rainfall data. The classification of a break as significant or insignificant does not reflect the extent of the phenomenon, or at least does not allow us to judge the significance of the decrease or increase recorded within the precipitation series. This result is in line with those found by other African researchers [22] - [24] .