Hydrological variability and flood risk in a forest watershed undergoing accelerated urbanization: the case of Mefou (South Cameroon) Open Access

Changes in river flows generally result from interactions between climate variability/change and or anthropization (Aulenbach et al. 2017; Diem et al. 2018; Oudin et al. 2018; Ebodé 2022a; Ebodé et al. 2022b). It is, however, recognized that their sensitivity to these factors also depends on the natural predispositions (size, slope system, type of soil, etc.) of their watersheds (Gibson et al. 2005).

Most of the time, changes in hydrological regimes are only examined through rainfall-discharges relationships (Liénou et al. 2008; Ebodé et al. 2021). Few authors in sub-Saharan Africa for example (D'Orgeval & Polcher 2008; Amogu et al. 2010) have assessed the hydrological impacts of human-induced environmental changes. Likewise, studies that attempted to dissociate the latter from those arising from exclusively hydro-climatic fluctuations are quantitatively limited in Central Africa (Dzana et al. 2011; Ebodé et al. 2022a). However, there is no doubt that anthropogenic actions through urbanization and industrial agriculture, which have resulted in large-scale deforestation, have increased considerably over the past 30 years throughout Sub-Saharan Africa and even beyond.

The work devoted to investigating the impact of rainfall on discharge in sub-Saharan Africa has been ongoing since the 1980s (L'Hôte et al. 2003). Studies correlating precipitation and discharge are based on the detection of ruptures in the hydrometeorological series. Results from those studies confirm in the case of West Africa that the 1970s appear to be the main period of discontinuity marking the onset of hydroclimatic drought in the region (Nka et al. 2015; Bodian et al. 2020). In Central Africa, fluctuations in river discharges and rainfall have been observed at seasonal and annual (Ebodé 2022b; Kpoumié et al. 2022) time scales. Liénou et al. (2008) demonstrated in the case of three equatorial rivers (Ntem, Nyong and Kienke) that the most significant climatic variations leading to changes in river discharge result from variations in rainfall during the dry season. The authors explain the sensitivity of the studied basins to rainfall variability by the fact that their reduction induces a significant deficit in soil moisture and groundwater storage resulting in a decrease in river discharge. Conversely, an increase in soil moisture during the dry season, therefore, enhances discharge at the start of the rainy season. It appears that the variability of discharge regimes of equatorial rivers can be better appreciated when rainfall of these seasons is considered. Concerning the impact of land use change on river discharge, previous studies have used supervised classifications of satellite images using at least two dates, to assess the dynamics of land use and its impact on river discharge (Kouassi 2007; Ebodé et al. 2020; Ewane & Lee 2020). Their results confirm an increase in discharges following an increase in impervious surfaces (buildings, roads and cultivated areas).

Beyond the climate change that is currently raging throughout Central Africa (Ebodé et al. 2022a), the studied basin (Mefou) has experienced accelerated and uncontrolled urbanization through an excessive spatial extension of Yaounde city, whose population has been multiplied by 10 in about 30 years, from 50,000 in 1957 to 1,500,000 in 2008. Today it is estimated at 2,500,000 inhabitants. Projections predict that it will exceed 3,000,000 by 2030 (BUCREP 2011). This population growth which has as a corollary the increase in impervious areas has probably caused a change in the rainfall-runoff relationship, which requires a new study of hydrological variability. Another consequence of this accelerated population growth in the basin could be the increase in the occurrence of floods, which already pose serious problems (destruction of infrastructure, diseases, loss of human life, etc.) for populations in certain corners of the basin (Nsangou et al. 2022). Under these conditions, it is important to propose an appropriate tool (flooding risk areas map) that can enable political decision-makers and civil society to reduce the vulnerability of populations to this natural disaster.

The objectives of this study were (1) to investigate the impact of precipitation variability and land use change on the hydrological dynamics of the Mefou river and (2) to draw up a basin flooding risk area map. Considering that few studies have attempted to investigate these issues in Central Africa, this paper contributes to this debate by focusing our analysis on the Mefou river basin in South Cameroon, which is representative of Central African basins under the sub-equatorial Atlantic climate. It appears fundamental for this basin where many socio-environmental problems (floods for example) are observed. In addition, the data and new entrants would help in the long-term planning of water demand and use as well as improve future simulations of the river flows.

The rainfall data used in this work came from the Climate Research Unit (CRU) of the University of East Anglia in the United Kingdom. These data have been available since 1901, via the site https://climexp.knmi.nl/selectfield_obs2.cgi?id=2833fad3fef1bedc6761d5cba64775f0/ in NetCDF format, on a monthly time step and at a spatial resolution of 0.25° × 0.25°. Precipitation and temperature data from CRU have been used to validate CMIP models in the region (cases of Logone and Nyong basins, respectively) (Nkiaka et al. 2018; Ebodé et al. 2022a). Although these data have been available since 1901, the calculation of the basin's average rainfall only concerns the period 1950–2017.

The Mefou flow series comes from the CRH (Hydrological Research Center). It was incomplete during the 1980 and 1990s. Indeed, after 1987, due to budgetary constraints, the hydrological service no longer guaranteed the continuity of observations. We then note the abandonment of most of stations observed, including that of the basin studied.

The spatial data used to analyze changes in land use in the Mefou basin are Landsat 8 satellite images from January 2018, Landsat TM (from January 1987 and 1999) and Landsat MSS from March 1973. The spatial data used for the mapping of flooding risk areas include a 2020 Landsat 8 satellite image and a digital terrain model Shuttle Radar Topographic Mission (SRTM) of 30 m × 30 m spatial resolution. These data made it possible to obtain six (06) independent variables (land use mode, elevation, slope, distance from river, drainage density and topographic wetness index), which were then crossed to generate a flooding risk areas map in the basin investigated. These different variables are considered by some authors (Natarajan et al. 2021; Nsangou et al. 2022) to be the main factors in the occurrence of floods in a given region since they influence water runoff and infiltration. All spatial data used in this study are made available to the general public for free by the National Aeronautics and Space Administration (NASA), via the US Geological Survey site (https://earthexplorer.usgs.gov/), in GeoTIFF format. Uploaded images taken during the dry winter season (December to mid-March) were preferred over the rainy seasons because they are less affected by cloud disturbances.

In addition to spatial data, historic flood points were also used for mapping flood risk areas. A total of 47 historical flood points were used for the calibration (31) and validation (16) of the model retained for this study. These flood points were obtained from the urban community of Yaounde.

The analysis of average rainfall and flow was carried out using Pettitt (Pettitt 1979) and modified Mann Kendall (Hirsch & Slack 1984; Araghi et al. 2014) statistical tests at the 95% significance level. Following the application of autocorrelation (from the calculation of the R statistic) and seasonality (from the employment of the correlogram) tests to the rainfall and flow series used, it turned out that there is a seasonality in the latter. It is why we chose to use the modified Mann Kendall test of Hirsch & Slack (1984) to the detriment of the classic Mann Kendall test and other modified Mann Kendall tests.

If the null hypothesis is rejected, an estimate of the break date is given by the instant defining the maximum in absolute value of the variable Ut, N.

There is no significant trend in the series analyzed when the calculated p-Value is above the chosen significance level. More details on the modified version of the Mann Kendall test can be found in the relevant literature (Hirsch & Slack 1984).

To extract and to assess the behavior of extreme flows, the Indicators of Hydrologic Alteration (IHA) tool, version 7.1, developed by The Nature Conservancy was used. This tool offers the possibility of comparing the parameters characterizing the flow regimes under different conditions (Richter et al. 1998). It uses daily discharge values and produces several important statistics. Only four of them were considered essential for this study, among which are the average, the coefficient of variation (CV) of extreme discharge and the Julian date of the annual minimum and maximum. By dividing the series of values in the period before and after the discontinuity, the tool calculates the change that occurred in the evolution of each of these parameters after the discontinuity. We can thus analyze not only the sign of change between the two periods but also the magnitude of this difference.

Maps of elevation, slope, drainage density and distance from river were automatically generated using ArcGis 10.5 software.

The values of the FR indicate the types of correlation between the different factors and the floods. FR value greater than 1 indicates a strong correlation with flooding; on the other hand, a value less than 1 indicates a weak correlation with flooding.

To obtain the flood map, all variables are converted to raster format. The spatial resolution of each of the raster has been defined on cell size of 15 m × 15 m, and it is integrated into the ArcGIS database. This integrated database has been reclassified into five classes of flood sensitivity, namely very low, low, medium, high and very high.

The main trends in hydrological variability reported here are those identified for the mean (annual and seasonal) and extreme (maximum and minimum) flows.

The difference in evolution between the discharges of the Mefou and those of most other equatorial rivers require explanations which can be found by examining the hydrological modifications observed in the light of the climatic and environmental changes observed in the basin over the same period.

Changes of this type and such magnitude can only induce hydrological alterations. The most noticeable in the case studied are the increase in average and extreme discharges and the precocity in the observation of maximums. Concerning the increase in discharges, in a context where the precipitation of the rainy seasons that generate the maximum discharges decreased (Table 1 and Figure 2), the most logical thing would have been to see their drop, which is not the case. The urbanization of this basin seems to be the most relevant factor to justify this trend. In this case, the decrease in precipitation seems to have been compensated by the increase in runoff. The average annual runoff coefficients of this basin have indeed increased significantly since the 1980s (Table 1 and Figure 2), following a phase of accelerated urbanization in the western part of this basin (Yaounde region) from the 1980s, following the subdivision operations undertaken since the end of the 1970s by the municipal authorities (Dzana et al. 2004). In the case of the earliness of the maximums, the changes in surface conditions observed reduce the duration necessary for their appearance by accentuating the runoff they cause. Thresholds of impermeable surfaces beyond which urbanization is supposed to have, at the basin scale, a statistically significant influence on river discharge are proposed in literature, although the figures put forward by the various authors are somewhat dissimilar. Booth & Jackson (1997) thus place this threshold at 10%. Brun & Band (2000) at 20%, while for Yang et al. (2010), this threshold is between 3 and 5%. In all cases, it appears that, this threshold exceeded 40% in the studied basin. Under these conditions, it is logical that hydrological alterations such as those highlighted occur.

Summer rainfall are the only ones that increased significantly in the Mefou basin. The noted surplus after the break is +42.8% (Figure 2). Annual rainfall and that of other seasons decrease. Statistically significant decreases were noted for autumn (in 1972–73) and winter (in 1978–79) rainfall. The deficits recorded after the break in these two cases are −11% and −39% (Figure 2).

The different cases where a downward break was highlighted (autumn and winter) are also those for which the increases in flows are the least significant after the common rupture of 1985–86. On the other hand, the most significant increases in discharges were observed in cases where there is an upward break (summer) or no break (spring). Similarly, the 2010s, characterized by a significant drop in precipitation, is also the one for which the surpluses observed since the 1980s are generally the lowest, apart from the spring for which the excedent remains relatively high (Table 1).

The increase in runoff in a catchment area due to that of impervious areas or rainfall, as is the case in this study, is always accompanied by an increase in flooding. It is more likely insofar as floods generally occur not far from the rivers due to the overflow of the minor bed by the waters. It is that justifies this point being addressed in this study, following the presentation of the results relating to the evolution of flows. The idea, in the end, is to propose a flooding risk areas map of the basin.

As in the case of the land use modes, the basin elevations have been reclassified into four categories (657–742 m; 743–812 m; 813–921 and 922–1,210). Only the 657–742 m class (56.41% of the total area of the basin) is well correlated with flooding, with an FR of 1.54 (Table 4).

The basin slopes were also divided into four classes (0–5%; 6–12%; 13–22% and 23–69%). Only the slope classes of 0–5% (39.31% of the basin) and 6–12% (42.44% of the basin) are well correlated with flooding. Their respective FRs are 1.39 and 1.06 (Table 4).

The basin drainage densities are divided into four classes (3–8 km/km²; 9–14 km/km²; 14–20 km/km² and 21–25 km/km² (Table 4). Apart from the 21–25 km/km² class, no other class is well correlated with flooding. Its basin occupancy rate is 44.32% and its FR is 1.38 (Table 4).

Four classes (0–387 m; 388–774 m; 775–1,161 m and 1,162–1,548 m) also appear on the distances from the river map (Table 4). The 0–387 m class, whose occupancy rate is 53.62%, is the only one for which the FR is greater than 1 (Table 4).

The topographic wetness indices of the basin are also divided into four classes (3–6; 77; 8–9; 10–11). Classes 3–7 (29.83% of the basin) is the only one for which the FR is less than 1. The other classes are well correlated to flooding with FRs that vary between 1.03 (8–9) and 2.54 (10–11) (Table 4).

High and very high flooding risk areas are where flooding is most likely to occur. They are characterized by significant imperviousness, low elevations, low slopes, high proximity to the river, high drainage density and a high topographic wetness index (Table 4).

To validate the FR model, it is important to calculate the success rate and accuracy of the predictions. The success rate was calculated using 31 historical flood points. The accuracy prediction was calculated using 16 historical flood points. Future flooding is most likely to occur in high and very high flooding risk areas (Samanta et al. 2018). The prediction success and accuracy rates are 93.55% and 81.25% (Table 5). Such accuracy of the prediction validates the FR model in the watershed studied and proves at the same time that it is suitable for mapping the areas at risk of flooding in the region studied.

The objectives of this study were both to investigate the impact of precipitation variability and land use change on the hydrological dynamics of the Mefou river over the recent period (1963–2018), and draw up a basin flooding risk area map. Between 1973 and 2018, the Mefou watershed experienced changes in land use modes, mainly marked by a significant increase in impervious areas (+530%) to the detriment of forest areas (−52.9%) and water bodies (−80.4%). These changes caused a significant increase in the average and extreme flows of the Mefou river from 1985 to 1986. They are also at the origin of the precocity in the maximum appearance. The rainfall trends observed in this basin have contributed to amplifying this increase in some cases and reducing it in others. The summer and spring, whose rainfall respectively recorded an upward break and no break, are also the seasons for which the increases in runoff are the most significant. Conversely, autumn and winter, which recorded significant decreases in rainfall, experienced the weakest increases. Floods are often recurrent in basins with such hydrological behaviour. To enable policymakers to reduce the vulnerability of populations to this disaster, the proposed flood risk areas map shows that 2.67, 7.22, 22.5, 35.25 and 32.36% of the catchment area are respectively delineated into very low, low, medium, high and very high flood vulnerability classes.

Although this study provides useful information on the general behavior of flows in the Mefou watershed, it nevertheless leaves some uncertainties related to the gaps in the observed flow series. Other information that could have been obtained in the absence of these gaps has probably remained latent. Regular measurements of the flows of the Mefou river are therefore essential to solve this problem.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The author declares there is no conflict.