Analysis of rainfall dynamics in the three main cities of northern Cameroon

The northern zone of Cameroon, which depends mainly on agriculture, is considered one of the most vulnerable regions in the country to climate change. Few studies based on field data have examined the changes in climatic conditions that affect agriculture. This research focuses on fluctuations in precipitation that determine dry and wet seasons. From 1973 to 2020, data were collected from weather stations located in three major cities in northern Cameroon: Ngaoundere, Garoua and Maroua. Data were tested for homogeneity using the Pettitt and Buishand tests. Trends were analyzed using the Mann–Kendall test, Sen’s slope estimator and the regression line, while drought severity was assessed using the standardized rainfall index method. These data homogeneity tests were performed using two statistical tools, SPSS and XLSTA software. According to Pettitt’s test, rainfall increased by 29.6% in Ngaoundere from 1997 to 2020 compared to the previous years of 1973–1996; in Garoua, rainfall increased by 36.2% from 1988 to 2020 compared to the previous years of 1973–1987. However, from 1973 to 2020, the average rainfall in Maroua remained stable at approximately 716.5 mm, with a decreasing trend according to the Mann–Kendall test. In conclusion, this study shows that rainfall has increased significantly in the cities of Ngaoundere and Garoua, making these areas favorable for seasonal and market gardening. However, in Maroua, caution is advised, as rainfall is reportedly decreasing in this locality, increasing the risk of food insecurity. A credible climate warning system must be implemented on a large scale to guide farmers.


Introduction
The northern part of Cameroon includes three regions: Adamawa (Ngaoundere as its capital), the Far North (Maroua as its capital) and the North with Garoua as its capital. Bring and Moussa (2016) present this part of Cameroon as one of the most vulnerable areas of the country to climate change due to a severe lack of reforestation and the advancement of the desert, among other things. The Far North and Northern regions have experienced the most drastic degradation of 90% of their area in recent decades, and the Adamawa region is also affected by degradation and rapid drought, according to Bring et al. (2020). According to the Intergovernmental Panel on Climate Change, the manifestations of climate change at the global level might result in, among other things, a rise in rainfall and sea level and an increase in the frequency and intensity of extreme events: droughts, floods, cyclones, etc. (Parry et al., 2007). At the regional level, each continent, depending on its specific biophysical and human characteristics, is already undergoing these changes. According to Seneviratne et al. (2012), Africa in general would be one of the continents most exposed to these climatic risks, and the Sahel zone, where the northern part of Cameroon is located, stands out as one of the most affected by extreme events (rainfall instability, floods, water shortages, drought progression, etc. Accordingly, Challinor et al. (2007) also reported that Africa could indeed be considered one of the most vulnerable continents to the effects of climate variability and change due to its high dependence on climate-dependent agriculture, which plays a major role in supporting rural livelihoods and economic growth. The same argument of the continent's vulnerability is supported by Beucher and Bazin (2012); he estimated that Africa, although emitting less greenhouse gases (GHG) with a minimum emission rate of 4% and 2.5% of global CO 2 , would nevertheless be the most vulnerable continent to climate shocks and whose precipitation would decrease in the Mediterranean and Southern Africa. Camberlin (2007) specifies that the evaluation of the modalities of possible climate changes that affect or will affect an area assumes good knowledge of the functioning of its climate. Therefore, Niasse et al. (2004) define the concept of climate variability and change as the modification or significant variation of the climate, whether natural or due to factors of anthropogenic origin. Another definition is also given by Mitosek (1992), who states that climate variability would refer to the variability inherent in the stochastic stationary process that approximates the climate on the scale of a few decades and of which precipitation would be one of the four priority variables in assessing climate variation. Ouarda et al. (1996) reported that Kite and Harvey also identified other variables, such as evaporation, vegetation, soil moisture, management of hydraulic structures, timing of extreme events, water quality index, and glacier conditions, as factors to be considered in the analysis and understanding of climate change. According to the World Meteorological Organization and echoed by Mbog et al. (2020), climate is the average or typical weather condition observed over a long period of at least 30 years for a given geographical location; however, Nefzi (2012) recalls that experts have considered that the 30-year interval defining the period for an estimate of normal climate is not only too short to return the trend of the climate in a given region but also too long to decipher the anomalies and interannual variabilities.
Several organizations, e.g., The World Bank (2021), monitor the evolution of climatic conditions in Cameroon using reference data derived from default data classified according to climatic zones derived from the Köppen-Geiger climate classification system, which divides climates into five climatic groups, of which rainfall is one of the main parameters. According to MINEPDED (Ministry of the Environment, Nature Protection and Sustainable Development) et al. (2015), in Cameroon, observations of climate change indicate a decrease in rainfall in four of the five agro-ecological zones in the country (high Guinean savannahs, Sudano-Sahelian, high plateaus, and bimodal rainfall), with the exception of the monomodal rainfall zone, where rainfall has increased. Servat et al. (1999) mention that in Cameroon, during the period of disruption from 1969 to 1971, the average rainfall deficit observed was 16%, and this rainfall deficit was already being felt and would even seem to have increased over more than two decades from the 1970s to 1980. According to Gaymard et al. (2015), Cameroon is already experiencing the various effects of climate change, and by 2100, desert will dominate in the northern Cameroon zone. If nothing is done, the authors also note that in the Sudan-Sahelian zone, the projected changes in rainfall will range from −12 to +20 mm per month (−8 to 17%) by 2090. Bassirou and Bitondo (2022) reported that a decrease in rainfall would significantly affect agricultural productivity and that climate change is a reality in Cameroon and therefore poses a serious threat to Cameroonian agro-industries.
However, in Cameroon, scientific publications reporting on the analysis of the evolution of certain climatic parameters in localities remain insufficient, as noted by Menang (2017), who indicates that very few analyses of daily rainfall and temperature data have been carried out in Cameroon due, among other things, to the difficulty of accessing different meteorological field data. This argument is also supported by Sotamenou and Saleufeumeni (2013), who report the problem of unavailability of data during their study in the southern part of Cameroon, where the study had a small intertemporal dimension (10 years) due to the absence of available data. In the same vein, Saha (2019) notes that because of the problem of the availability of long-term data on both climatic and hydrological parameters, decision-making for preventive actions against natural hazards would not be sufficiently informed and that it would be difficult to establish the rainfall profile in the Sudano-Sahelian zone in Cameroon in particular.
In this context of difficult access to data, and contrary to studies where the data or part of the data would be derived from global or default data (Nicholson et al., 2018;Igri et al., 2022;Molua, 2006), to name but a few, the present study is based on updated field data available from the ASECNA (Agency for Aerial Navigation Safety in Africa and Madagascar) and the CCAA (Cameroon Civil Aviation Authority) weather stations. It covers the period from 1973 to 2020 with an intertemporal dimension of 48 years.
After the description of the study framework and methodology ("Location of the study areas and methodology" section), including the study framework ("Location of the study" section) and methodology ("Methodology" section), the results are presented in the "Results" section, which addresses the evolution and variations in annual and monthly rainfall, separately considering the locality of Ngaoundere in the Adamawa Region (the "Evolution and variation of rainfall in Ngaoundere" section); the locality of Garoua in the Northern Region (the "Evolution and variation of rainfall in Garoua" section); and the locality of Maroua in the Extreme North Region (the "Evolution and variation of rainfall in Maroua" section). A discussion of the observed trends and potential reasons for these variations in the various locales is presented (the "Discussion" section). The "Conclusion" section recalls the observed trends and suggests areas where further research is needed.
This research is based on observed precipitation fluctuations in the different study areas and provides benefits related to understanding and knowledge of how precipitation changes under agricultural conditions but does not focus on predicting future precipitation.

Location of the study areas and methodology
Location of the study The present study was conducted in the three main towns of the northern regions ( Fig. 1), representing two of Cameroon's five agroecological zones. These zones correspond administratively to the regions of Adamawa (high savannah zone located at latitude 7.35°N, longitude 13.56°E and an altitude of 1114 m) with a surface area of 17,196 km 2 ; the northern region located at latitude 9.33°N, longitude 13.38°E and an altitude of 242 m with an area of 13,614 km 2 ; and the far-north region located at latitude 10.45°N, longitude 14.25°E and an altitude of 423 m (Sudano-Sahelian zone) with an area of 4665 km 2 .

Data collection
The monthly and annual climate data for the weather stations studied come from the ASECNA operating in the Adamawa and North Regions, namely, weather stations 64,870 (FKKN) and 64,860 (FKKR) in Ngaoundere and Garoua, respectively; the data from the Far North Region come from weather station 64,851 (FKKA) managed by the CCAA in Maroua. This study is based on rainfall data; the observation period extends over forty-eight years (48 years), i.e., from 1973 to 2020 (Table 1), which gives the opportunity to conduct a set of statistical tests to provide useful information to be capitalized on. However, the observation shows a low availability of research data in Cameroon compared to West African countries (Benin, Senegal, and Burkina-Faso).
The monthly data collected at the different stations allowed us to obtain the total annual precipitation for each station based on the following equations: where: (1) and Na = Σ(Ny) 640 Page 4 of 18 Vol:. (1234567890) Pm = monthly or annual rainfall in mm Px = monthly or annual volume of rainfall collected at weather stations in mm Na = monthly or annual number of rainfall days Ny = monthly or annual number of days of rainfall recorded for the weather stations Data processing using statistical tools and tests

Statistical tools used
In this study, data processing was based on SPSS Statistics version 20, Excel 2016, and XLSTAT.

Statistical tests used
Pettitt test The Pettitt nonparametric test is commonly applied to detect a single point of change in hydrological or climatic series with continuous data (Pohlert, 2020). Pettitt considers a sequence of independent random variables X 1 , X 2 … X N . The sequence is assumed to contain a breakpoint at τ if the X t for t = 1… τ have a common distribution F1 (X), and the X t , for t = τ + 1… N have a common distribution F 2 (X), different from F 1 (X). The null hypothesis of "nonbreakage"  The alternative hypothesis of "breakage" A nonparametric statistical test is used to test this equation, and no particular condition is required for the functional forms of F1 and F2 except for continuity (Lubes-Niel, 1998).
Then, the variable The U t , N statistic is considered for values of t between 1 and N. To test H0 against H1, Pettitt proposes to use the variable: Using rank theory, Pettitt gives the approximate probability of exceeding a k value by: For a given first-species risk α, H0 is rejected if this probability is less than α. In this case, the series has a break at time t = τ defining K N . The test is most sensitive to a change in mean.

Buishand test
The Buishand test is a parametric test whose statistic is defined from the maximum of the cumulative sum of the deviations from the mean or the median. It is a test for detecting a temporal break in a data series. The alternative hypothesis of this test is an abrupt change in the mean, and the power function is estimated by generating series from independent normal variables of the same variance but with a break in the mean from a randomly chosen date (Doukpolo, 2014).
According to Lang et al. (2003), the Buishand statistic is derived from an original formulation given by Gardner in 1969; this Gardner statistic is used for a two-sided mean break test at an unknown time and is written as follows: P k denotes the a priori probability that the break occurs just after the k th observation.
This formulation assumes that the variance σ x 2 is known. If it is unknown, it can be replaced by the sample variance D 2 x , and if P k is chosen to be uniform, we finally obtain the statistic U defined by Lang et al. (2003):

Mann-Kendall test
The Mann-Kendall test, proposed by Mann (1945) and Kendall (1975), is a nonparametric test commonly used to detect monotonic trends in environmental, climatic or hydrological data series (Pohlert, 2020). The null hypothesis, H0, is that the data come from a population with independent realizations and are identically distributed. The alternative hypothesis (Ha) is that the data follow a monotonic trend. The analysis of the trends will only be considered representative when they are statistically significant at the 0.05 threshold, i.e. 5%, according to the nonparametric Mann-Kendall test (Yue et al., 2002). This test was applied to the time series of precipitation anomalies. The Mann-Kendall test is calculated as follows (Drouiche et al., 2019): Assuming that the data are independent and identically distributed, Kendall (1975) gives E(S) = 0, and the variance Var (S) is: where n is the number of data points in the series, p is the number of related groups, and tj is the number of data points in the group of order j. If the sample contains ten or more data points, the distribution of the test statistic Z below will be approximated by a center-reduced Gaussian.
The S statistic is closely related to Kendall's τ given by: Sen's slope Sen's slope is estimated by Sen's method, which is a nonparametric procedure developed by Sen in 1968(Sen, 1968; this slope is the median of all slopes computed between each pair of points (ti, Yi) and (tj, Yj); the set S of N distinct pairs (i, j) for which ti < tj, for which X ij is an estimate of the slope calculated as follows: According to Aswad et al. (2020), the Sen slope estimator proves to be a powerful tool for developing linear relationships, of which this Sen slope has the advantage over the regression slope in that coarse data series errors and outliers do not affect it much. A positive Sen slope indicates an upward trend, while a negative Sen slope suggests a downward trend.

Rainfall index analysis
The analysis of the standard deviation (indicator of climatic variability) allows us to evaluate the dispersion of the values around the mean. This standard deviation is determined by the following formula: where (x) and v represent the standard deviation and the variance, respectively.
From the standard deviation, the monthly and interannual reduced rainfall centered anomalies are calculated. The anomalies of the different parameters are calculated by the formula: where IP is the centered reduced anomaly for year i; X i is the value of the variable (precipitation); X is the series mean; and σ(x) is the series standard deviation. A drought occurs when the index continuously has a negative value of − 1.0 or less and ends when the index becomes positive (Svoboda et al., 2012).  Table 2 shows the values of the intensity of drought events as a function of the index value.

Synthesis of the methodological approach
In order to accomplish the study's goal, various steps and tools were used, as shown in Fig. 2. After data collection and preparation, the analysis of the homogeneity of the data was conducted using XLSTAT. Detection of breaks in the data series was done using the Pettitt test and the Buishand test. The Mann-Kendall test and Sen's slope estimator helped in the examination of trends in the data series. The rainfall index values enable to determine wet and dry periods using. Through this rainfall, variability was highlighted.

Results
The statistical analysis of the data reveals that the data arranged in the different study areas are remarkably homogeneous in the majority of cases. This allows validation of the quality of the data collected.
Evolution and variation of rainfall in Ngaoundere Figure 3 shows the evolution of rainfall in Ngaoundere from 1973 to 2020. It follows from this analysis that rainfall over these 48 years has fluctuated; the trends are positive, as the linear regression analysis shows an interannual increase in rainfall. It should be noted that the R 2 value = 0.2558 reflects significant variation depending on the year; therefore, the maximum  1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999  precipitation during this study was recorded in 1982, with a value of 3447 mm, followed by 2020, with a value of 2292 mm, and the lowest precipitation was recorded in 1975, with a value of 26 mm of rainfall.

Tests for breaking rainfall data in Ngaoundere
After conducting break tests of the rainfall series of the parent sample taken between 1973 and 2020, i.e. 48 years in Ngaoundere, it appears that the series shows a break at the date of 1996 with the Pettitt test and in 1981 with the Buishand test; because, given that the calculated p-value ((P-value of the Pettitt Test = 0.0043 < 0.05 (5%); P-value of Buishand's Test = 0.0046 < 0.05 (5%)) calculated is lower than the level of significance alpha = 0.05, the null hypothesis H0 must be rejected, and the alternative hypothesis Ha retained "There is a date from which there is a change in the data, and from which there is a break in the series"; see Fig. 4a, b. According to Pettitt's test, in Ngaoundere, rainfall increased by 448 mm over the period 1997 to 2020 compared to the previous period 1973 to 1996, when the average rainfall was 1068 mm between 1973 and 1996, and this value would have increased to 1568 mm between 1997 and 2020 (from 1997 to 2020, rainfall would have increased by 29.6% compared to the previous year's . However, the Buishand test indicates an increase in precipitation of 850.8 mm over the period 1982 to 2020 compared to the previous period 1973 to 1981, where the average precipitation was 601.2 mm between 1973 and 1981, and this value would have increased to 1452 mm from 1982 to 2020 (from 1982 to 2020, precipitation would have increased by 58.6% compared to the years prior to [1973][1974][1975][1976][1977][1978][1979][1980][1981].

Rainfall anomaly in Ngaoundere
The data analysis reveals that rainfall in Ngaoundere over the period 1973-2020 (48 years) shows 31 wet years and 17 dry years; see Fig. 5. The wet years have six periods from 1987 to 1989, 1991 to 2000, 2002 to 2006, 2008 to 2010, 2012 to 2014, and 2016 to 2020 and two dry periods from 1973 to 1981 and 1984 to 1986. Also noted is the presence of wet and dry years in the above periods: this is the case in 1982in , 1983in , 1990in , 2001in , 2007in , 2011in . From 1973 to 2020, the trend is positive. The linear regression analysis shows an interannual increase in precipitation from 1973 to 2020. It should be noted that the R 2 value = 0.2558 shows a significant variation in precipitation depending on the year. From this analysis, it appears that the extremely dry years are 1973 to 1976, 1985 and 1987 because their rainfall index is − 2; the years 1979, 1981 and 1986 were moderately dry because their rainfall index is − 1; however, the years 1983,1997,1999,2016,2018 and 2019 were moderately wet with a rainfall index of 1, and finally, the years 1982 and 2020 were extremely wet because their rainfall indexes are 4 and 2, respectively.

Trend of rainfall data in Ngaoundere
After conducting the trend test for the rainfall series in Ngaoundere from the parent sample taken between 1973 and 2020, i.e. 48 years, it is clear that there is a trend in the rainfall series, and therefore, significant variations are observed for this variable. Since the calculated p-value of the precipitation trend test (P-value of the precipitation test = 0.0000066) is below the significance level alpha = 0.05, the null hypothesis H0 must be rejected, and the alternative hypothesis Ha must be retained. "There are trends in the series" (Fig. 6). Figure 7 presents the monthly variability of rainfall in Ngaoundere over the period from 1973 to 2020 (48 years), which is necessary to distinguish between wet and dry months; from this analysis, it appears that the climate in Ngaoundere is a humid tropical climate with two seasons: a long dry season that runs from November to March (5 months) and a long wet season that runs from April to October (7 months); however, August is the wettest month, with an average rainfall of 239 mm, and January is the driest month, with 0 mm of rainfall. …

Monthly rainfall variation in Ngaoundere
Evolution and variation of rainfall in Garoua Figure 8 shows the evolution of rainfall in Garoua from 1973 to 2020. It follows from this analysis that rainfall over these 48 years has fluctuated; the trends are positive, as the linear regression analysis shows an interannual increase in rainfall. Notably, the R 2 value = 0.3018 reflects a significant variation depending on the year; therefore, we note that the maximum precipitation during this study was recorded in 2008, with a value of 1243 mm, and the lowest precipitation was recorded in 1974, with a value of 135 mm of rainfall.  1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

Tests for breaking rainfall data in Garoua
After performing break tests on the rainfall series of the parent sample taken between 1973 and 2020, i.e. 48 years in Garoua, it appears that the series shows a break at the date of 1987; since the calculated p-value ((P-value of Pettitt's test = P-value of Buishand's test = 0.0010 < 0.05 (5%)) is lower than the significance level alpha = 0.05, we must reject the null hypothesis H0 and retain the alternative hypothesis Ha "There is a date from which there is a change in the data, and from which there is a break in the series", see Fig. 9a 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999  Also note the presence of wet and dry years in the above periods: this is the case for the years 1978years , 1981years , 1982years , 1983years , 1984years , 1985years , 1986years , 1987years , 1988years , 1989years , 1990years , 1991years , 2007years , 2011years , 2012years , 2013years , 2014years , 2015years and 2020years . From 1973 to 2020, the trend is positive. The linear regression analysis interannually shows an increase in precipitation from 1973 to 2020. It should be noted that the R 2 value = 0.3018 reflects significant variation in precipitation over the years. From this analysis, it appears that the extremely dry years are 1973 to 1976 and 1979, as their rainfall index varies between −2 and −3; the years 1977,1983,1985 and 2011 were moderately dry, as their rainfall index is − 1; however, the years 1981, 1988, 1990, 1991, 1993 to 1997, 1999, 2001, 2007, 2008 to 2010, 2012, and 2017 to 2019 were moderately wet, with a rainfall index of 1.

Trend of rainfall data in Garoua
After conducting the trend test of the rainfall series in Garoua from the parent sample taken to 1973-2020,  1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006   i.e., 48 years, it appears that there is a trend in rainfall, and therefore, significant variations are observed for this variable. Given that the p-value of the rainfall trend tests (P-value of the rainfall test = 0.0019 < 0.05 (5%)) calculated is below the significance level alpha = 0.05, the null hypothesis, H0, must be rejected, and the alternative hypothesis, Ha, must be retained. "There are trends in the series", see Fig. 11. Figure 12 presents the monthly variability of rainfall in Garoua over the period from 1973 to 2020 (48 years), which is necessary to distinguish between wet and dry months; from this analysis, it appears that the climate in Garoua is a dry tropical climate with two seasons: a long dry season that runs from November to March (5 months) and a long wet season that runs from April to October (7 months); however, August is the wettest month, with an average rainfall of 199 mm; December, January and February are the driest months, with 0 mm of rainfall.

Monthly rainfall variation in Garoua
Evolution and variation of rainfall in Maroua Figure 13 shows the evolution of rainfall in Maroua from 1973 to 2020. It follows from this analysis that rainfall over these 48 years has fluctuated; the trend is negative, as the linear regression analysis shows a small interannual downward variation in rainfall. It should be noted that the R 2 value = 0.0086 reflects an absence of significant variation depending on the year; however, the maximum precipitation during this study was recorded in 1994, with a value of 1194 mm of precipitation, and the lowest precipitation was recorded in 2019, with a value of 251 mm of precipitation.

Tests for breaking rainfall data in Maroua
After performing break tests on the rainfall series of the parent sample taken to 1973-2020, i.e. 48 years in Maroua, it appears that the series does not show any break. Because, given that the p-value ((P-value of Pettitt's test = 0.8604 > 0.05 (5%); P-value of Buishand's test = 0.6684 > 0.05 (5%)) calculated is higher than the significance level alpha = 0.05, the null hypothesis, H0, cannot be rejected; "the data are homogeneous, and there is no break in the series", see Fig. 14. Furthermore, according to the Pettitt and Buishand tests, in Maroua, the average rainfall remained at approximately 716.8 mm from 1973 to 2020, so no significant variation was observed.

Rainfall anomaly in Maroua
This analysis shows that rainfall in Maroua over the period 1973-2020 (48 years) shows 31 wet years and 17 dry years (see Fig. 15). The wet years have 5 periods from 1975 to 1978, 1980 to 1982, 1988 to 1990, 1998 to 2010 and 2013 to 2015 and 2 dry periods from 1985 to 1987 and from 2016 to 2019. Additionally, note the presence of wet and dry years in the above periods: this is the case for 1973for , 1974for , 1979for , 1983for , 1984for , 1991for , 1992for , 1993for , 1994for , 1995for , 1996for , 1997for , 2011for . From 1973 to 2020, the trend was negative. The linear regression analysis shows no annual increase in precipitation from 1973 to 2020. It should be noted that the R 2 value = 0.0086 reflects a very small and insignificant variation in precipitation over the years. From this analysis, it appears that the extremely dry years are 1979,1987,2016 and 2019 because their rainfall index is between −2 and −3; the years 1975, 1983, 1985, 1986, 1997 and 2013 were moderately dry because their rainfall index is − 1; however, the years 1978, 1992, 1995, 1999, 2003 and 2010 were moderately wet with an index of 1, and finally, the years 1975, 1976, 1994 and 2020 were extremely wet because their rainfall indexes are between 2 and 3.

Trend of rainfall data in Maroua
After performing the trend test of the rainfall series in Maroua from the master sample taken between  1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006  1973 and 2020, i.e. 48 years, it appears that there is no trend in this series, and therefore no variation; ((P-value of the Rainfall Test = 0.7023 > 0.05 (5%)) calculated is greater than the threshold significance level alpha = 0.05. The null hypothesis H0 cannot be rejected, "There is no trend in the series", see Fig. 16. Figure 17 presents the monthly variability of rainfall in Maroua over the period from 1973 to 2020 (48 years), which is necessary to distinguish between wet and dry months; from this analysis, it appears that the climate in Maroua is a dry tropical Sahelian climate with two seasons: a long dry season that runs from November to March (5 months) and a long wet season that runs from April to October (7 months); however, August is the wettest month, with an average rainfall of 218 mm; November, December, January, February and March are the driest months, with 1 mm of rainfall for the month of March and 0 mm for the other months.

Discussion
Figures 3, 7 and 13 show the evolution of rainfall in Ngaoundere, Garoua and Maroua, respectively. The analysis of the results of the statistical tests illustrated in Figs. 4, 9 and 14 in Ngaoundere, Garoua and Maroua, respectively, shows that the rainfall dynamics in these localities are globally unstable, with a tendency to increase in rainfall in Ngaoundere and Garoua according to Figs. 6 and 11. These findings are consistent with those reported by Amougou et al. (2015), Vondou et al. (2021), and Cheo et al. (2013).
Ngaoundere's increasing rainfall is most likely due to its location, which is at an average altitude of 1114 m in a humid Sudanese tropical environment that encourages greater rainfall. Our findings are also in line with those of Amougou et al. (2015), who claim that the strong climatic variations observed in different parts of the country are partly explained by disturbances in the North Atlantic, whose regular movements of air masses easily reach the entire country, causing significant disturbances; they also claim that North Atlantic cooling causes an increase in rainfall in the Ngaoundere region.
According to Cheo et al. (2013), the increase in rainfall in Garoua is due to a rise in temperature, which increases the rate of evapotranspiration, resulting in an increase in water vapor in the atmosphere, which falls later as rain, and it is also due to the city's environment, as the Benoue River in Cameroon runs through it. However, among the three localities studied, it appears from these analyses that only in Maroua was the rainfall illustrated by an absence of break over the study period, represented in Fig. 14. This conclusion corroborates that reported by Bouba et al. (2017), who observed that between 1935 and 2011, the rainfall was illustrated by an absence of breaks at both the interannual and monthly scales. However, Fig. 16 indicates a downward trend in rainfall in Maroua, which is similar to the findings of Cheo et al. (2013) and Molua (2006), who indicated that rainfall in Maroua had already decreased between 1957 and 2006 and between 1960 and 2000. The lack of afforestation in Maroua, as well as the lack of rivers that may pass the towns, such as in Ngaoundere and Garoua, are probably to blame for the drop in reported rainfall. Figures 5, 10 and 15 show rainfall anomalies in Ngaoundere, Garoua and Maroua. The analyses of the rainfall index indicate that dry and wet periods alternate per year, a result that corroborates that reported by Vondou et al. (2021); however, there are years with long wet periods corresponding to six periods in Ngaoundere, three periods in Garoua, and five periods in the case of Maroua, and years with long dry periods also corresponding to two periods in Ngaoundere, one period in Garoua, and two periods in Maroua. Furthermore, these analyses of the rainfall index in the three localities show wet years interspersed with dry years, which is consistent with the conclusion of Nicholson et al. (2018) . Figures 7,  12 and 17 show the cumulative monthly rainfall in Ngaoundere, Garoua, and Maroua, respectively. These analyses reveal that cumulative rainfall of more than 50 mm occurs in Garoua and Maroua from May to September and in Ngaoundere from April to October, which is consistent with the results of Djoufack (2011). The 5 months with the greatest potential rainfall in these three research locations are May, June, July, August, and September, with significant rainfall occurring only in Ngaoundere from April onwards.
Rather than the findings of Igri et al. (2022), which demonstrate that the rainy season begins in June and ends in September, our findings reveal that the rainy season begins in April and concludes in October in the three cities analyzed, with an exceptionally early initiation of rainfall in March in Ngaoundere alone. November through March are generally dry months.

Conclusion
This paper provides a description of rainfall trends in the northern part of Cameroon for the period from 1973 to 2020. During this period, rainfall variability is clearly shown, with significant trends of increasing annual rainfall in Ngaoundere and Garoua and a decreasing trend in rainfall observed in Maroua (Figs. 3, 8 and 13). In the Ngaoundere area, a break in rainfall occurred in 1996 according to the Pettitt test and in 1981 according to the Buishand test, and in the Garoua area, a break in rainfall occurred in 1987 according to both tests (Pettitt and Buishand). On the other hand, no rainfall break is observed in the Maroua area. However, August is the month with the highest rainfall recorded in the three regions of Ngaoundere, Garoua and Maroua, with average rainfalls of 239 mm, 199 mm and 218 mm, respectively. This assessment shows that rainfall remains dynamic from 1973 to 2020 in this part of Cameroon, with an increasing trend in rainfall in Ngaoundere and Garoua and another decreasing trend in rainfall in Maroua. However, after the last wet period from 2016 to 2019 observed in Garoua, it is reported that a dry year will appear, particularly in 2020, and unlike Maroua, where during this same period (2016-2019), there was a dry period and where it is reported that a wet year will return in 2020; these different changes observed could be seen as a manifestation of a return to normal conditions in the case of Maroua and a usual alternation of dry years in the Garoua region. Contrary to similar studies conducted by Nicholson et al. (2018), where rainfall data were derived from global or default data on long series (from 1889 to 2014), Bouba et al. (2017), on long series (1935 to 2011) and Molua (2006) on long series (1960 to 2000), the present study dates from 1973 to 2020, with an intertemporal dimension of 48 years due to the updated field data available from ASECNA and CCAA weather stations. The results of this study could help in the predictions or simulations of crop models and allow the monitoring of climate change, particularly by integrating the rainfall parameter, so that farmers can better prepare their adaptation strategies to reduce the risks associated with climatic hazards, such as drought identified by dry periods that remain unfavorable to agricultural production, against wet periods that allow for better agricultural productivity. A credible climate warning system must be put in place on a large scale to guide farmers in defining their agricultural calendar.