3.3.1. Rainfall
The impact of rainfall is perceptible in the evolution of flows. In general, the annual precipitation of the two watersheds studied does not decrease significantly, while the flows register downwards discontinuities during the 1970s. The stationarity of the hydrological regime and that of the precipitations of the watersheds at this time step thus departs from the general logic, which admits that hydrological discontinuities occur following rainfall discontinuities. However, for both watersheds, the analysis of the decadal deviations from the long-term average of the two compared variables shows a synchronous evolution between them over the last two decades. For the latter, there is a concomitant decrease in rainfall and flows (
Table 3).
During the rainy seasons, primarily in the spring, precipitation has changed moderately, so that no sudden mean inflexion point has been detected in their series. In return, the discharges and the runoff coefficients for this season have decreased significantly since the 1970s. Similarly, the examination of the deviations from the mean of these variables generally shows significant decreases for the discharges and the runoff coefficients, while the rainfall does not decrease or decreases very slightly (
Table 3). The decline in winter rainfall since the 1970s appears to be closely related in amplitude to that of spring flows. Although the correlations between them are not significant (r < 0.32 on both catchments), the rhythm of evolution of these two variables over the decades is fairly constant. This is observed during the 2010s, for which a small drop in winter rainfall led to a similar evolution in spring runoff coefficients and discharges (
Table 3). This influence of winter precipitation has already been suggested by Liénou et al. [
19]. These authors wrote that from the 1970s, the decrease in winter rainfall created a greater water deficit (evaporation and soil water reserve) in the watershed area at the start of the first rainy season in spring. This aridification of winter means that a greater part of the precipitation received during the spring first contributes to filling this water deficit, and therefore, the fraction that generates runoff is reduced. This then translates into lower runoff volumes for the same average depths of precipitation during spring, which explains the drop in the runoff coefficient (
Figure 4).
In autumn, rainfall and flows decrease very slightly over the entire study period. However, between the 1970s and 1980s, flows and runoff coefficients increased, while the rainfall did not vary much overall. This increase would be due to the increase in summer rainfall. The impact of Summer rainfall on autumn flows is more noticeable during the 2000s and 2010s, for which there is a joint decrease in summer rainfall, discharges and runoff coefficients of autumn, despite a sometimes-negligible variation in autumn rainfall in the different watersheds (
Table 3). In view of these developments, it appears that the increase in summer rainfall during the decades from 1970 to 1990 considerably reduces the deficit of evaporation and soil water reserves at the beginning of the autumn, favoring runoff [
19]. The part of the precipitation that actually participates in the runoff increases, hence, there is an increase in the autumn runoff coefficient during these decades (
Table 3). The decrease in summer rainfall during the 2000s and 2010s has the opposite effect, which is why there has been a reduction in runoff coefficients during these decades, responsible for a decrease in discharges. This is all the more probable since the analysis of the average monthly flows shows that the variations of the summer rainfall are generally felt more on the flows of the first month (September) of autumn. The model developed for this purpose is significant, with a fairly correct quality of fit (r > 0.6) on the two basins (
Figure 7). It could be used to predict fairly accurately the runoff in early autumn (September) based on the rainfall recorded during the summer.
In the summer, we notice that rainfall increases statistically, while flows evolve according to varying magnitudes downwards. These opposite trends could be explained by the fact that the summer flows are partly linked to the volumes precipitated during the spring, which occurs just before, even if the correlation between them is not significant (r < 0.35 on both basins). This is why we can note, in the evolution of summer flows, dry decades even when the rainfall was abundant. This was the case in the 1980s, when there was a surplus above average for the summer rainfall, but shortages for both the spring rainfall and the summer flows. However, the variations in summer precipitation themselves seem to also be related to the flows in the season. They participate in the event of a moderate evolution of the spring rainfall, in the maintenance of flows in some cases, or in their reduction in others. This last scenario is observed during the 2010s, where spring rainfall did not vary considerably, but where a decrease in summer rainfall is concomitant with a decrease in flows in the different watersheds (
Table 3).
In winter, precipitation and flows decrease according to different magnitudes. The stationarity of these two variables does not reveal an apparent link in the case of Nyong. Additionally, the correlations between them are weak (r < 0.39). However, the analysis of the deviations from the average shows clearer links between these variables, since the beginning of the drought started during the 1970s in these two watersheds (
Table 3). The succession of deficit rainfall sequences that were observed between the 1970s and 1990s also led to deficits in terms of flows. Similarly, the resumption of rainfall observed during the 2000s caused an upsurge in the flows, which was interrupted again in the two watersheds studied during the 2010s, following the relapse of the rainfall (
Table 3).
Several common and divergent points can be noted in the evolution of rainfall and its impact on discharges during the different seasons in the two studied watersheds. Spring discharges have decreased since the beginning of the 1970s, more markedly for the Ntem in the South of Cameroon close to the Equator, while rainfall has not decreased significantly, even though since 2005, several years of slightly weaker spring rainfall have been recorded. In summer there is a net increase in rainfall in the two watersheds from the early 1980s to the late 2000s, then a return to less abundant rainfall. Summer discharges do not seem to follow a variation related to this variability in rainfall. In autumn the rainfall does not drop on the Ntem, except for a few years of less abundance during the most recent decades, while for the Nyong the rainfall has decreased slightly and fairly regularly since the end of the 1980s. Diminution in these watersheds would result from their respective belonging to the equatorial (Ntem) and sub-equatorial (Nyong) domain, which is already at the origin of the significant differences on the duration of each season on these. The discharges seem to have decreased slightly for the Nyong, since the 1990s, in apparent link with the slight decrease of the rainfall, while the discharges of the Ntem seem to decrease also since the middle of the 1990s, without the rainfall having decreased significantly. It should be noted that the period of the possible change in discharges also corresponds to a period of significant gaps in discharge data series, which does not allow great significance for these analyses. In winter the series of rainfall of the two watersheds show a discontinuity in 1976–77, which separates the wet period from the recent drier period. Although during the 2000s the rainfall became close to the totals for the wet period, the discharges of the two rivers followed the same variation, with values generally lower after 1970—even in the 2000s, when the rainfall apparently increased.
The results, obtained in this study from recent data, corroborate the first observations of change in the hydrological regime in the region [
19]. These changes attest to the visible and lasting impact of a change in the rainfall regime which began in the late 1970s, and which have major repercussions on hydrological regimes and the seasonal availability of surface water resources.
3.3.2. Changes in Land Use Patterns
The changes in land use patterns seem to have more or less significant links with certain hydrological modifications observed. The slightly early arrival of the maximum noted at the outlets of the two watersheds (
Table 4), in a context where rainfall is decreasing slightly, seems to be related to the increase in IA (+771% for the Nyong and +451% for the Ntem) and the diminution in forests (−6% for the Nyong and −1% for the Ntem). These changes could reduce the time it takes for the maximums to appear by increasing the runoff they cause. The stability observed for the Nyong maximums after the discontinuity (
Table 4) could also be a consequence of these same developments. Indeed, in a context where the rainy seasons rainfall, which generate the maximum flows, decrease, the most logical would have been to see the maximums decrease, which is, however, not the case. The current modest urbanization rate in this watershed seems to be the most relevant factor to justify this lack of observed trend. In this case, the decrease in precipitation appears to have been offset by the increased runoff. The annual runoff coefficients of this watershed have increased significantly since the 1980s (
Table 3), following an accelerated urbanization phase in the western part of this watershed (Yaoundé region) over the 1980s and 1990s, following the subdivision operations undertaken since the end of the 1970s by the municipal authorities [
45]. The decrease in the water bodies noted could have contributed to that of the minima in the two watersheds. When there are in large number, water bodies contribute to the supply of rivers during low water periods and contribute to the maintenance of low flows [
46]. However, otherwise, their support for the flows decreases, which causes a decrease in minimum flows.
The impact of the increase in the IA on runoff has already been demonstrated in other forest environments (South America), particularly in the cases of certain watercourses such as the Amazon [
47], Upper Xingu [
48] and Parana [
49]. The discharges of these watercourses have indeed increased, following a reduction in the forest area of their watersheds, which seem relatively greater than that observed in the Nyong and Ntem watersheds. The rate of deforestation in the entire Amazon basin, for example, reaches 30% [
47].
Certainly, there is reason to wonder whether such small changes in land use could cause changes in extreme flows as visible as those observed in the watersheds investigated. Since the evolution of these extremes does not agree with other phenomena that have a proven impact on flows, in this case the decrease in precipitation highlighted in this study and the rise in temperatures postulated by Sighomnou [
50] and Amougou et al. [
51] in the region, which act rather in the direction of a reduction in flows and a late observation of the maximums, it seems appropriate to think that the changes in land use, although modest overall, might be the main cause of these evolutions. Their action on flows could have been considerably amplified by their location in the watersheds (to the west part, near the outlets of the watersheds studied), which would favor the rapid routing of rainwater bodies with the least possible losses to the outlets, thereby erasing the size and forest cover effect of these watersheds. If this situation persists, there is every reason to believe that there will be an increase in maximum flows in the Nyong within a few years, despite the decrease in precipitation.
To be complete, a further study should also analyze the evolution of daily rainfall and dry days within the rainy season, and even the possible changes in the frequency and intensity of rainfall events, which may have significant impact on runoff, as is the case in the Sahel, for instance [
52].