Spatio-temporal trends of hydrological components: the case of the Tafna basin (northwestern Algeria)

The impacts of climate change have interested scientists in northern Africa since the last decades of the 20th century, when the decrease of precipitation and the increase of temperature were noticed (Arnell et al. 2003; Zettam et al. 2017; Mami et al. 2021). The decrease of precipitation detected in northwestern Algeria is estimated as −25% in the 1970s compared to the period of 1950–2000 (Arnell 1999; Knippertz et al. 2003), predicting the rise of the vulnerability of this semi-arid region as regards water stress. The decrease of precipitation affects water resources, in which a decline of flow is already noted as −60% in the 1970s compared to the period 1930–1998 (Meddi et al. 2010; Taibi et al. 2017).

One of the domains affected by climate change is water resources, induced by the decrease of precipitation and the increase in temperature. This may have an impact on drinking water, and industrial and agricultural activities, leading to serious droughts in some regions, such as the southern region of the Mediterranean basin (Solomon et al. 2007; Schilling et al. 2012; AR5 Climate Change 2013: 5; Jacob et al. 2014; Price 2017; Brouziyne et al. 2018; Zeroual et al. 2018). The level of water resource stress in Algeria, assessed by Arnell (1999), is medium (−20 to −40% of use/water resource ratio) in the last decade of the 20th century, and becomes high, exceeding 40%, for the future period (2025–2085). The Tafna basin, situated in northwestern Algeria, is one of the most important catchments ensuring the water self-sufficiency in this region (Zettam et al. 2017). The Tafna basin was impacted by the decline of the water resource during the 1970s (Meddi & Hubert 2003) and a decrease of this resource is estimated in the Tafna basin over the 21st century (Mami et al. 2021).

The northwestern region of Africa is characterized by a wet winter and a dry summer (Yebdri et al. 2007; Zettam et al. 2017; Gevaert et al. 2018). In the future period, Brouziyne et al. (2018) mention that the months mostly affected by climate change are January for precipitation and July for temperature in the R'dom watershed (Morocco), near to the Tafna basin and with similar climate characteristics.

The diminution of precipitation affects the water cycle, composed of runoff, soil water availability, and evapotranspiration. In northern Africa, the decrease of runoff, river discharge, and precipitation is underlined mainly in winter (Arnell 1999; Brouziyne et al. 2018; Grusson et al. 2018; Naumann et al. 2018; Zeroual et al. 2018; Burek et al. 2019). Groundwater is also impacted, but it is difficult to separate the role of climate change and local anthropogenic interferences (Kundzewicz & Doll 2009; Taylor et al. 2011; Green 2016; Grusson et al. 2018). All of these components are connected and change in one of them could interfere with another one (Grusson et al. 2018), as evapotranspiration and soil water availability are strongly linked (Manabe et al. 2004; Calanca et al. 2006; Jung et al. 2010; Verrot & Destouni 2016; Grusson et al. 2018).

Recent studies use regional climate models (RCMs) to investigate climate change, in northwest Africa (e.g., CNRM-CM5-RCA4, IPSL-IPSL-CM5A-MR-CCLM4-8-17, ICHEC-EC-EARTH-DMI, etc.) from the Coordinated Regional Climate Downscaling Experiment (CORDEX) project's Africa domain by Mami et al. (2021) and Zeroual et al. (2018) for the Algerian part, Europe by Brouziyne et al. (2018), and Seifennasr et al. (2016) for the Moroccan part. The project ENSEMBLE, European domain, is also used by Taibi et al. (2017). The decrease of precipitation indicated in these studies is around −10 to −40% under the Representative Concentration Pathways (RCPs) 4.5 and 8.5, respectively. The temperature is also affected by climate change, with an increase of 1.3 °C on average at an annual scale within 2050 (Brouziyne et al. 2018). Also, the flow is not spared from climate change and is expected to decrease by −40% in the Mediterranean region within 2099 (AR4 2007).

As suggested by Zettam et al. (2017) and Todini (2007), the hydrological models are useful for diverse objectives, such as for decision-makers, namely, water resource managers, urban planners, and administrations (Noori & Kalin 2016). Different distributed physically based models can simulate temporal and spatial variations of hydrological processes and work out the influence of land use impacts (Lin et al. 2015). Among these models are Hydrological Infiltration Capacity (VIC) (Liang et al. 1994), Institute of Hydrology Distributed Model (IHDM) (Beven et al. 1987), WATFLOOD (Kouwen et al. 1993), and Système hydrologique Européen (SHE) (Abbott et al. 1986).

The Soil and Water Assessment Tool (SWAT) (Arnold et al. 1998) is used for a supplementary function to assess: agricultural management practices (Moriasi et al. 2011), hydrology and sediment transfer (Arnold et al. 2012; Gassman et al. 2014; Zettam et al. 2017, 2020), and evaluating the impacts of climate change (Singh & Gosain 2011; Grusson et al. 2018). Numerous studies use SWAT in semi-arid watersheds, such as in Southern Australia (Shrestha et al. 2016), Spain (Molina-Navarro et al. 2014), Brazil (Santos et al. 2019), and North Africa (Bouraoui et al. 2005; Yebdri et al. 2007; Sellami et al. 2016; Zettam et al. 2017), and others for assessing climate change (Nerantzaki et al. 2015; Yanjie et al. 2015; Jin et al. 2016; Brouziyne et al. 2018; Grusson et al. 2018; Saddique et al. 2019; Yuan 2019; Mami et al. 2021).

Our study aims to detect the impact of climate change over the future period (2020–2099) on the hydrological components of the Tafna basin, using hydrological modeling forced with climate model outputs (precipitation and temperature maximum/minimum) of the CORDEX-Africa initiative. The approach developed in this paper is considered for future climate throughout 2020–2099. This application is proceeded by the SWAT model to analyze the spatial and temporal evolution of hydrological water balance over the Tafna basin, for hydrological SWAT components (precipitation, river discharge, runoff, subsurface flow, soil water content, evapotranspiration, etc.). The trends are presented in this study with two Representation Concentration Pathway (RCP) scenarios (4.5 and 8.5) on a monthly scale.

The transboundary Tafna watershed is situated in the north-west of Algeria, and 30% of its surface area is situated in the northeastern part of Morocco (Yebdri et al. 2007). It represents one of the most important basins of the Oranie-chott-chergui region according to the classification of the Agence Nationale des Ressources Hydriques (1981) (ANRH: National Water Resources Agency). The area of the Tafna basin is 7,245 km² and flows into the Mediterranean Sea near Beni-saf city (Figure 1). The elevation of the basin is from 0 to 1,800 m a.s.l., situated between the Mediterranean Sea and the Sahara desert.

The climate of the Tafna basin is semi-arid, characterized by wet winters, and hot and dry summers. The annual mean temperature varies between 11 °C in winter and 28 °C in summer, and annual precipitation ranges from 220 to 673 mm/yr during 1981–2000 (Agence Nationale des Ressources Hydriques 1981; Mami et al. 2021). The monthly precipitation is almost negligible in summer (from June to September), not exceeding 10 mm, and varies between 20 and 60 mm in winter (Figure 2). While the maximum monthly temperature reaches 33 °C in summer, the minimum monthly temperature in winter can drop to 2 °C (Figure 2).

The Tafna basin includes five dams: Beni Bahdel (built in 1952 across Oued Khmis and Oued Tafna, with a capacity of 63 Mm³), Mefrouche (1963, O. Meffrouche, 15 Mm³), Sidi Abdely (1988, O. Isser, 110 Mm³), Boughrara (1998, O. Mouilah, 177 Mm³), and Sikak (2005, O. Sikkak, 27 Mm³) with a total capacity of 392 Mm³ (Yebdri 2006; ANBT 2015; Mami et al. 2021) intended for the supply of drinking, industrial and irrigation water for neighboring cities. The flow recorded during the studied period (1981–2000) ranges on average between 0 and 10 m³/s for 170 km of river course (Agence Nationale des Ressources Hydriques 1981).

Ten climate model outputs of precipitation and air temperature from the CORDEX-Africa project (CORDEX Data Search | CORDEX | ESGF-CoG n.d.) are used in this work (Table 1). Two regional climate models (RCM) are employed and forced with eight different global climate models (GCM): the regional climate model CCLM4-8-17 is forced by EC-EARTH and CNRM-CM5, and RCA4 is forced by EC-EARTH, CNRM-CM5, IPSL-CM5A-MR, CCCma-CanESM2, MIROC, MPI-ESM-LR, NCC-NorESM1, and NOAA/GFDL-ESM2M. The grid spacing of RCMs used in this study is 0.44 degrees (50 km of resolution) under two RCPs, 4.5 and 8.5. The reference period is the period with available observations (1981–2000). These climate model (CM) outputs are extracted and bias-corrected by the CMhyd software (Rathjens et al. 2016), using the distribution mapping method for correcting both precipitation and air temperature. The performance of bias correction used in this study is detailed in a previous study (Mami et al. 2021).

The SWAT model is made for hydrological modeling of watersheds varying from a few hundred to multiple thousands of square kilometers (Zettam et al. 2017), which is the case with the Tafna basin (7,245 km²). The application is divided into two steps: (1) the conception, calibration, and validation of the SWAT model for the Tafna basin detailed in a previous study (Mami et al. 2021) during the period (1981–2000), and (2) the run of the model forced with a set of bias-corrected RCMs' outputs to evaluate the future climatic and hydrologic components' evolutions (2020–2099).

The following inputs are introduced into the SWAT model to produce the hydrological modeling and calibrate the Tafna project. The digital elevation model (DEM) is introduced into SWAT with a resolution of 30 × 30 m to define the basin, the river channel network, and delineate the basin into sub-basins (in this study, 33 sub-basins result from the delineation of the SWAT model for an area of 7,245 km²). The maps soil and land-use are inserted to characterize, joined to DEM, each hydrological response unit (HRU). The slope classes are derived from DEM (four classes: 5%, 15%, 30%, and higher than 30%). More details are shown in a previous paper (Mami et al. 2021).

The meteorological and hydrometric datasets (Figure 1 and Table 2), situated in the Tafna basin, are collected by the National Agency of the Meteorology of Algeria. The daily precipitation and minimum/maximum temperature datasets from five weather stations (Figure 1) are introduced into the SWAT model, on a daily scale, to run the modeling of the Tafna project for 1981–2010. However, for the calibration and validation of the model, the measured discharge from four hydrometric stations, obtained from the ANRH, are used on a monthly scale for the period 1981–2010 (three years for warm-up, from 1984 to 1995 for calibration, and from 1996 to 2010 for validation). Additional information about datasets can be found in Table 2.

The hydrological modeling is proceeded by the application of the SWAT hydrological model with version 2012 and ArcGIS interface (version 10.2). The Tafna hydrological model is calibrated and validated at the monthly scale for the past periods 1981–1995 and 1996–2010, respectively (including a three-year warm-up period in the calibration period) (Table 3). The calibration and validation are performed using SWAT-CUP (SWAT calibration and uncertainty procedures) (Yebdri et al. 2007; Abbaspour 2007; Zettam et al. 2017, 2020) with SUFI-2 (Sequential Uncertainty Fitting 2) algorithm (Abbaspour et al. 2004). The correlation between measured and simulated discharge observed and simulated is from satisfactory (station C with NSE = 0.54) to very good (station D with NSE = 0.84). The correlation at station A is only unsatisfactory during the validation step (NSE = 0.21). A more detailed description of the performance of calibration and validation of the SWAT model in the case of the Tafna basin can be found in Mami et al. (2021).

The difference between precipitation and evapotranspiration is the water available for human use and management.

The water cycle components used in this study, namely, surface runoff, subsurface runoff, soil water content, and evapotranspiration, are computed by the SWAT model as follows.

This following application is made for the hydrological SWAT components: precipitation, discharge, runoff, subsurface flow (lateral flow), soil water content, and evapotranspiration.

After calibrating the SWAT model on the Tafna basin, the ten climate model outputs (precipitation and air temperature) are introduced into SWAT to run the model. The SWAT components' outputs using the ten climate models presented in this study are compared to outputs obtained using meteorological observations during the reference period (1981–2000) (Figure 3). This first step allows us to validate the climate model outputs used in this study, and investigate the trends of water cycle components at the Tafna watershed-scale over the future period (2020–2099).

Figure 3 presents the comparison between the SWAT component outputs' simulations (precipitation, evapotranspiration, soil water content, surface flow, and subsurface flow) from the observations and the ten bias-corrected climate model outputs used in this study. This comparison is proceeded by distribution, in the form of boxplots, of each climate model output and observation for the monthly SWAT component outputs, during the reference period (1981–2000). An evaluation of the similarity between SWAT component outputs using ten climate model outputs of CORDEX-Africa shows the concordance of the SWAT component outputs' distribution for the majority of climate model outputs. The similarity between the outputs of the climate models is more suitable in January. That can lead to proceeding with the evaluation of trends at the Tafna watershed-scale over the future period (2020–2099).

Table 4 illustrates the evolution of monthly precipitation from the reference period (1981–2000) to the projected period (divided into four future periods: 2020–2039, 2040–2059, 2060–2079, and 2080–2099). A considerable difference is detectable between the months of January and July. During January, the tendency of precipitation is not clear between the climate models during the studied period, except for the periods 2080–2099, and 2060–2079 and 2080–2099, under RCP 4.5 and 8.5, respectively, that belongs to a significant decrease to 24 mm. The geographic presentation in Figure 4 gives more information about the tendency of precipitation of the ensemble of climate models over the Tafna basin. In January, the precipitation in the reference period attains 60 mm in the southern and eastern part of the basin, in the future period the decrease appears throughout the period, and seems more important in mountainous areas. However, in July, an increase, ranging between 0.1 and 7.5 mm, is distributed over the basin compared to the reference period.

The ratio increase/decrease is balanced while the decrease of discharge appears significant in the periods 2040–2059 and 2080–2099 under RCP 4.5 and the two last periods 2060–2079, 2080–2099 under RCP 8.5 according to the Mann–Kendall test (Table 5). This significant decrease is also marked in Figure 5.

Figure 6 shows the decrease in runoff to be high in January for the periods 2040–2059 and 2080–2099 under RCP 4.5 and 2020–2039, 2060–2079, and 2080–2099 under RCP 8.5. For July the positive delta marked in the basin is negligible, not exceeding 1 mm. For the rest of the future period, the ratio of increase/decrease is balanced between the models (Table 6), and some models give a poor significance of tendency because the p-values are high.