1. Introduction
An increase in the occurrence and magnitude of extreme flooding events in the Niger River Basin has been registered in recent decades, especially during the last 10 years [
1,
2,
3,
4,
5,
6]. These events have stimulated researchers worldwide to shift their efforts to flooding comprehension and hazard evaluation. The first scientific approach to this issue was addressed by Tarhule [
7], who introduced extreme floods as a significant problem in addition to the water scarcity affecting African countries.
Focusing on the Sahelian regions, many studies have analyzed the rainfall time series from 1950. A moderate recovery of precipitation has been observed after the severe drought from 1970 to 1990 [
8,
9,
10]. The “return to wet conditions” showed a positive trend for annual rainfall and highlighted the increase in inter-annual variability and the changes in rainfall magnitude [
2]. This rainfall pattern might partially explain the increase of streamflow in Sahelian rivers but cannot justify the extreme floods of the past decade, since the amount of precipitation was smaller than that of the wetter period before 1970. This hydrological behavior was called the Sahelian Paradox [
11,
12]. In the last few years, two different causes were analyzed as possible drivers of the increase in flood magnitude: climate changes and land use/land cover changes. Descroix et al. [
4] argued that the uncontrolled deforestation and clarification of the savannah to create agricultural land is leading to soil crusting and to a consequent decrease in soil water holding capacity. These authors extensively observed this phenomenon on the right tributaries of the Middle Niger River (Sirba, Gorouol, and Dargol rivers). This change of land use and land cover enhances the runoff value and decreases the time of concentration of the basin. Therefore, Descroix et al. [
4] showed that the increase in river discharge is directly attributed to the increased runoff coefficient.
By contrast, Aich et al. [
13] considered that the cause of increasing flood intensity in the Niger River and its tributaries can be connected to climate changes. They found that the return to wet conditions was the driver key for the increased discharge, explained by a direct correlation of precipitation and streamflow trends in the last three decades.
More recent studies have claimed that it is not totally clear what the major driver is and have suggested that the mutual influence between these drivers strongly depends on the local climatic and territorial features [
5,
14,
15]. Thus, each basin or sub-basin should be analyzed carefully to understand which aspect leads to the intensification of the flood regime. In the latest studies, data-based and modeling approaches have been performed [
5]. However, reliable data and continuous datasets are needed to achieve robust results. Unfortunately, failures of gauging stations often cause loss of data or cause the rating curves to become obsolete and not representative of the cross-sections of the river reach.
The present research critically examines the existing river flow data of the Sirba River, the largest tributary of the Middle Niger River, and provides a revised dataset for the stream gauge station of Garbey Kourou. Working at the sub-basin scale, an uncertainty in river flow data was noticed due to a significant difference between the current rating curve and the discharge measured in the field. The recalibration of the set of rating curves was necessary to obtain a correct stage/discharge correlation and provide a revised and more reliable discharge time series. This revised time series represents a new milestone for future analysis and answers the demands for data quality, especially for the calibration in the modeling approach. Moreover, the revised time series allowed the recalculation of a number of hydraulic features, enhancing the knowledge of the Sirba River such as the flow duration curve and trends. Our results are a contribution to the scientific debate about whether the increase of river flooding in the Sahelian part of the Niger River can be attributed to climate changes and/or land use/land cover transformation [
5,
16,
17,
18,
19,
20].
The paper is organized as follows:
Section 2 gives a brief contextualization of the study area, the dataset sources, and the methods;
Section 3 presents the main outcomes of the study and contains the discussion of the results;
2. Materials and Methods
To simplify the reading, the acronyms adopted in the research are listed in
Table 1.
2.1. Study Area
The study was carried out in the Middle Niger River Basin (MNRB), focusing on one of its main tributaries, the Sirba River. The Sirba River basin covers a surface of approximately 39,000 km
2 in the central Sahel. The territory is spread over two different countries: 93% in Burkina Faso and the remaining sector in Niger. The flat topography of the basin is characterized by a slight height variation between the upper level of 444 m a.s.l. and the lower level of 181 m a.s.l., without steep slopes. The river flows towards the confluence with a mean bed slope of 0.02%. The natural Sahelian landscape principally consists of a patchwork of shrub bush, fallow savannah, and rain-fed millet fields. Today, the original environment is strongly affected by the fast population growth. In fact, some areas originally covered by the savannah vegetation have been replaced by cultivated and pastoral areas [
4,
6].
The Nigerien part of the river basin was chosen as the study area. New and detailed hydrological and statistical analyses were conducted. The analyzed reach of the Sirba River was about 100 km from the state border (downstream the confluence of its three main tributaries, Faga, Koulouko, and Yali) to the confluence into the Niger River. There is a manual gauging station on each tributary in Bassieri (Koulouko River), Liptougou (Faga River), and Sebba (Yali River).
The gauging station of Garbey Kourou on the Sirba River was installed in 1956. It is located 8 km upstream of the confluence into the Niger River. The gauging station is equipped with six staff gauges (0–600 cm) and two water pressure measuring devices controlled by the NBA (Niger Basin Authority) and the DGRE (General Water Resources Management) of the Republic of Niger.
A new automatic gauging station was installed in June 2018 as a part of the international cooperation project ANADIA 2.0 for the implementation of an early warning system. The location of the station is in correspondence of the village of Bossey Bangou, a few kilometres downstream the state border with Burkina Faso. This station is equipped with eight staff gauges (0–800 cm) and a water pressure measuring device, controlled by the DGRE.
The gauging stations within the Sirba River basin are summarized in
Table 2. This table also contains the most significant gauging stations present in the MNRB: Kandaji and Niamey on the Niger River and Kakassi on the Dargol River. The main watercourses of the investigated area and all the mentioned stations are illustrated in
Figure 1.
The study area is located in the Inter-Tropical Convergence Zone (ITCZ) between the isohyets 400 and 700 mm, characterized by a Sahelian semiarid climate. The climatic year is divided into two seasons: the dry season (October–May) and the wet season (June–September) [
21,
22].
The Sirba River is an intermittent river that is dry for about six months a year. Its hydrology is deeply related to the rainfall variability and the flood magnitude is influenced more by the superficial runoff than by the groundwater flow.
2.2. Dataset
The data used in this study comprise stage and discharge time series of the Garbey Kourou gauging station. The observation period starts from the gauging station installation in 1956 until 30 September 2018. The observed river stages and discharges at the daily resolution were obtained from the database of the NBA and DGRE. After the first examination in which completeness and reliability were analyzed, we decided to keep as reference the stage time series provided by the NBA. The data also comprise the discharge measurements realized on site covering the period 1956–2018. The whole set of 140 measures is constituted by 91 from the NBA, 39 from the Monographie hydrologique du fleuve Niger [
23], three from recent measurements of the DGRE, and seven from measurement campaigns that we realized in the international cooperation project ANADIA 2.0. These last seven measures were realized using an Acoustic Doppler Current Profiler (ADCP) and Global Positioning System (GPS) devices in the rainy season of 2018. The data finally include the three rating curves previously used by the DGRE (for the intervals of 1956–1976, 1977–1979, and 1980–current date).
The additional discharge time series for the gauging stations of Kandaji, Kakassi, and Niamey (2006–2017) were provided by the DGRE.
2.3. Methods
2.3.1. Power-Law Stage/Discharge Rating Curve
Recent discharge measurements at Garbey Kourou have shown how the application of the current rating curve (RC) was causing an underestimation of the streamflow. The incongruence with the on-site measures can essentially be explained by the significant changes in the geometry of the cross-sections over time. These changes were observed through the land surveys.
Three different rating curves have been used during the 63 years of measurements and the last RC dates back to 1980. An upgrade was hence necessary. The set of 140 measures allowed the revision of the existing rating curves and the obtention of a new set. The three new curves were calculated using a homogeneous set of data, identifying three intervals of validity: 1956–1978 (based on 84 measures), 1979–2003 (based on 39 measures), and 2004–2018 (based on 17 measures). The breakpoint of each interval was fixed in correspondence to gaps in the measurement set. The decreasing number of on-site measures over the time highlighted the lack of attention paid to river monitoring.
The stage/discharge rating curves were calculated for each interval, following a power-law with two parameters [
24].
where
a and
b are the coefficient and the exponent that define the shape of the curve, respectively;
h is the stage height (m); and
Q is the discharge (m
3/s). Taking the logarithm of each on-site measure, the coefficients were extracted by fitting a linear line to the logarithm plot which maximizes the R
2 (coefficient of determination).
Once the different equations for each interval were defined, the successive step was the application of these rating curves to convert the observed river stages into discharge, resulting in a new time series.
For clarity, the updated time series of discharge is called ANADIA time series and the updated rating curves are called ANADIA RCs.
2.3.2. Flow-Duration Curves
As described by Leboutillier and Waylen, “a flow-duration curve represents the annual flow-frequency characteristics of rivers by depicting the cumulative frequencies for average ranked flows in a river” [
25].
The shape of the flow-duration curve (FDC) provides information about the basin and the characteristic discharge (Qn) can be extracted for supplementary analyses. The value Qn is defined as the discharge that exceeds the number of days (n) during the year over the entire period of record on which the FDC is based. This physically means that, on average, the discharge should be present at least n days per year.
2.3.3. Statistics
Since the Sahelian area is clearly characterized by a decadal pattern of climate, changepoints in the time series of the discharge were sought in order to recognize whether there is a correspondence to the rainfall pattern. To detect these changepoints, the approach developed by Wang et al. [
26,
27] based on the PMF (Penalize Maximal Function) test was used.
A comparison between the variability of the annual maximum discharges (AMAX) and their anomalies was performed in order to provide a clear visualization of the time series and its tendencies. The AMAX anomaly was calculated through the use of the Standardized Anomaly Index (SAI) introduced by E.B. Kraus [
28,
29], which is an index frequently used for climate change studies [
30,
31].
The local regression-fitting technique LOESS (LOcally Estimated Scatterplot Smoothing) was used in order to identify and visualize the tendencies in data distribution [
32]. This is a widely used non-parametric regression method that combines multiple regression models in a k-nearest-neighbor-based meta-model. With this procedure, a smooth curve called the Loess curve is generated through the dataset. In order to better understand the time series pattern, a moving average line was plotted.
Monotonic linear trends of detection and estimation were performed using MAKESENS [
33]. This tool is based on the use of two non-parametric methods: the Mann–Kendall (MK) test and the Sen approach. The MK test detects the presence of a monotonic increasing or decreasing trend, rejecting the null hypothesis (
H0) with a certain level of significance α. The Sen method computes the slope of the existing linear trend, evaluating its magnitude.
4. Conclusions
Considering the great attention that flood hazard and flood risk management in Sahelian countries have attracted, the present study was designed to enhance the reliability of flood information by revising discharge data of the Sirba River, the main tributary of Niger River in the MNRB. The present work identified a lack of accuracy in the transformation from the hydraulic stage to streamflow. A set of reviewed rating curves were first calculated for the gauging station of Garbey Kourou. Secondly, a new discharge time series was provided, which allowed the update of the existing databases managed by the NBA and DGRE. The updated time series will be the starting point for future hydraulic and hydrological researches in the central Sahel. Reliable data are also the key point to improve hydrological modeling accuracy and provide significant results by the analyses performed at regional and sub-regional scales. A hydrological evaluation was carried out to analyze the water balance at Niamey, which became more consistent with the measurements of the upstream gauging stations.
Moreover, this study highlighted the importance of verifying the quality of datasets both for decision-makers and the scientific community. More effort should be made to strengthen the reliability of data, in order to increase on-site measurements and keep control systems updated (e.g., rating curves).
The FDC evaluation in the river flow regime analysis showed a greater amount of streamflow since 1990, mainly explained by the increasing surface runoff in the whole area. This result is associated with the annual rainfall recovery of the past decades and with the reduction of the infiltration capacity of the soil [
18]. Unfortunately, the higher magnitude of AMAX in recent years has led to many problems in terms of flood disasters as well as to an increase in the number of people affected by flooding events. We confirmed the positive trend in flood magnitude and flood occurrence, contributing to the existing literature with an updated time series analysis of discharges.
The results of the trend analysis emphasized how the annual maximum discharge trend is clearly positive and is likely to increase in frequency and intensity in coming years. This thesis is supported by the ongoing land cover transformation due to the high population growth rate and the resulting necessity to modify the territory to produce more food. Furthermore, since climate change is one of the drivers of this phenomenon, certain studies have found that climatological projections predict a slight increase in future precipitation [
21]. Thus, these results should be used to develop targeted interventions aimed at flood mitigation and prevention.