Water balance and hydropluviometric variability of dams in Eastern Algeria Open Access

Given the increase in water needs linked to population growth and the rapid development of economic activities, ambitious programs are being implemented to build reservoir dams. In fact, more than 250 new dams are built around the world each year (Nandalal 2007). The World Register of Dams, maintained by the International Commission on Large Dams, lists almost 60,000 dams worldwide (Adamo et al. 2020; Wang et al. 2022). In Africa, fewer than 3,558 dams were recorded in 2020, representing 9% of the world's continental dams (Mulligan et al. 2020). While most developed countries in North America, Europe, and Oceania have seen a decline in dam construction since the 1970s, developing countries in Africa, Asia, and South America have seen a steady increase in dam construction since that time (Zhang et al. 2018). In North Africa, many dams and reservoirs have been built to secure water availability (Tramblay et al. 2018). In the three Maghreb countries, a vast dam construction program has been launched since the 1960s, with a cumulative capacity of 28 billion m³ (Mahé 2020). In Morocco, 145 large dams with a total estimated capacity of 18 billion m³, 13 hydraulic water transfer structures, and over a hundred small dams and reservoirs have been built. This storage capacity will be increased in 2,030 to 27 billion m³ with the objective of building 2–3 large dams per year (Loudyi et al. 2022). At the end of 2018, Tunisia had 37 dams with a total capacity of 2.285 Mm³as well as 258 hill dams with a total capacity of 365 Mm³ (Langenberg et al. 2021). In Algeria, 80 dams are in operation (Langenberg et al. 2021). Most Algerian dams have a short lifespan of about 30 years due to losses from evaporation, high levels of sedimentation and leakage (Benfetta & Ouadja 2017).

What conclusions can be drawn about the hydrological functioning of these dams from the data used to calculate their water balances? Do climatic fluctuations predominantly control hydrological variability patterns? What is the impact of water losses and evacuated volumes (flooding and bottom discharge) on the regulated volumes allocated to the dams' sectors of use?

To answer these questions and in continuity with previous work (Remini et al. 2009; Boutouatou 2020; Ikhlef et al. 2023), our analysis focuses on a sample of ten dams, fairly well representative of the strong physico-climatic disparities in the Eastern region of Algeria.

Understanding dam's functioning with respect to the hydrological behavior of the catchment areas (itself influenced by the climate) and the management instructions issued by dam operators is based on:

• the analysis of the spatio-temporal variability of hydrological inputs and rainfalls as well as their relationship with climatic fluctuations, in particular, the North Atlantic Oscillation (NAO) and the Mediterranean Oscillation Index (MOI) (Hosseini et al. 2021; Rust et al. 2021; West et al. 2021);

• the study of the variability extent in water losses through evaporation and infiltration as well as water volumes evacuated by flood discharges and bottom outlets (Zhao & Gao 2019; Essamin et al. 2020; Guido Aldana et al. 2021);

• Finally, the study of the interannual variation by use sector (water supply, irrigation, and industrial water) of the regulated volumes (Assaba et al. 2013; Bakken et al. 2016).

The basins controlled by the 32 dams in Eastern Algeria cover an area of 29,011 km². They are distributed across large orographic units (arranged from North to South: the Tell Atlas, the High Plains and the Saharan Atlas) (Figure 1). The average altitude of the basins varies in large proportions (from 471.5 to 1,629 m). Geologically, these basins are characterized by the structural complexity and the diversity of the materials that make them up. The climatic context is also marked by strong contrasts, with a humid temperate climate of the Mediterranean type in the Tell, a semi-arid climate of the continental type in the High Plains and an arid dry climate in the piedmont of the Saharan Atlas. The precipitated layer of water varies between 150 mm on the southern piedmont of the Saharan Atlas and 1,400 mm in the region of Small Kabylie (Mebarki 2005). The average annual runoff in the basins at the dams varies from 15 mm in the South (Foum El Gherza) to 501 mm in the North (Erraguene).

Given the morphological and hydrological variety of the basins, the dams have very different capacities, varying between 3 hm³ (Foum El Gueiss) and 963 hm³(Béni Haroun); the same applies to their annual theoretical regulated volume (3.2–435 hm³). The Beni Haroun dam alone holds 26% of the theoretical regulated volume of all dams in the region. The capacity/theoretical regulated volume ratio exceeds one unit (1) for almost all the reservoirs and reaches more than 5 at the Foum El Khanga, the Safsaf, and the Soubella dams. This reflects the deliberate choice of large storage capacities in drought-prone areas.

This contribution was made on the basis of data on the water balance (inflows, outflows, and variations in water stock) corresponding to daily chronological series obtained from the National Agency of Dams and Transfers (ANBT). Ten dams representative of the physical variety of eastern Algeria were selected, with capacities ranging from 3 to 200 hm³ and theoretical regulated volumes ranging from 3.2 to 95 hm³/year (Table 1, values in bold). Their data cover a common hydrological period of 23 years (September 1990–August 2013). The same series concerns rainfall chronicles observed at dam sites.

For data processing, we used Excel, Xlstat, and the computer program R.

We describe in the following the working methodologies adopted in this study.

The reduced centered index applied to annual rainfall and dam input data is derived from Bertin's MGCTI (the chronological graphic method of information processing) matrix (Nouaceur & Murarescu 2019; Nouaceur 2020). This index is calculated in three steps:

The first step is to classify and sort the annual data of the studied dams over the entire study series according to the quintile limit values (Q1, Q2, Median, Q3, and Q4):

• Years where the totals of each parameter are below the first quintile are considered very dry.

• Years between Q1 and Q2 are said to be in dry.

• Years between Q2 and the median are considered normal.

• Years between the median and Q3 are qualified as normal with a wet tendency.

• Years between Q3 and Q4 are considered as excess.

• Finally, years with totals above the fourth quintile are classified as high surplus years.

The next step is to assign codes (varying from 1 to 5) according to the characteristics already determined and assigned to each year. The coded values are then replaced by frames of colors ranging from light to dark. In order to examine the relationships between lines (years) and columns (dams), we will invert in a third step the numerical table into a graphic table that makes it possible to visualize the evolution of the studied hydropluviometric parameter according to time and space (Laignel et al. 2014). For each year, the ratio of the deviation from the interannual mean (annual sums of the obtained values over all the dams – the mean of the series) to the standard deviation is centered-reduced. The latter, used in several fields and notably in geography, highlights hydroclimatic trends and characteristic periods (Derdous et al. 2021).

This section aims to analyze the annual hydrological input and precipitation data from the ten dams using the regional index and then to apply the continuous wavelet spectral method to the same data at a monthly time step. Finally, it aims to identify potential relationships between hydrological variability and the NAO and MOI climate indices.

From the variability of inputs and rainfall, we can notice that the major change, observed in inputs from the year 2002/2003, is also observed in rainfall. Three periods are generally identified with regard to the series of inputs (before 2001/2002, between 2002/3003 and 2006/2007, and after 2006/2007). The same periods have been identified in the rainfall series, with slight shifts in the years.

Wavelet coherence spectra of inputs and rainfall indicate variability of inputs as well as strong coherence with rainfall in all frequency bands from annual to multi-annual scales (Figure 4(c)).

The strong coherence observed between these two signals suggests an influence of climate variability on the hydrological response of the study region. The calculation of the average input/rainfall consistency over the studied period reveals a significant average correlation, of the order of 96% for the frequency band of the annual scale and 88% for the frequency band of the multi-annual scale of the 8–10 years period (Table 2).

Three main periods are highlighted: the first is identified before 2001/2002, characterized by a reduction in frequency bands. The second period begins around 2002 and ends around 2016, characterized by high energy levels across all fluctuations. Finally, the third period, from 2016 to 2022, is characterized by the presence of only the 1-year frequency band. It is characterized by a total loss of energy in the spectra of the inputs from Zardezas and Foum El Gherza.

In view of these results, the last decade has shown a notable variability in time and space, linked to different physiographic and climatic contexts. Significant fluctuations were observed between 2013 and around 2016 for the three dams (the highlight of this period was the exceptional hyperhumid year of 2014/2015) and around 2019 for both the Cheffia and Zardezas dams. Moreover, an overall loss of energy appears at the Foum El Gherza dam after the 2016/2017 wet year, in particular the four consecutive dry years, from 2018/2019 to 2021/2022, reflecting the aforementioned pattern of the annual rainfall and input curve.

The input/rainfall consistency calculation shows that the Cheffia dam is very strongly influenced by rainfall (90.64%), the Zardezas dam is strongly influenced (85.03%), and finally, the Foum El Gherza dam is fairly strongly influenced by rainfall (64.25%).

The hydropluviometric variability is itself linked to climate fluctuations. In climate, there are two types of influence. First, there is a global influence linked to large-scale movements of atmospheric air masses. This can be plotted by climate indices such as the NAO and the MOI. Second, there is a local climate influence linked to parameters such as distance from the sea, altitude.

The choice of these indices is based on their significant influence in the Maghreb region. For example, in Tunisia, Jemai et al. (2017) showed that the variability of the Gabes rainfall was influenced by the NAO and the MOI, with a total consistency ranging from 65 to 66% for the NAO and a total consistency ranging from 71 to 73% for the MOI. In Morocco, Zamrane et al. (2016) indicate a strong coherence between NAO/streamflow and NAO/precipitation identified on an interannual scale. Non-stationarity can be observed in the late 1980s, 1990s, and 2000s. The NAO contribution varies from one basin to another, ranging from 67 to 77%. In Algeria, Turki et al. (2016) analyzed the relationship between NAO climate patterns and the hydrological variability in the frequency domain, and they have shown a mean explained variance of 40%. In addition, Khedimallah et al. (2020) have shown that the variability of rainfall and runoff in the Cheliff and Medjerda basins is linked to the variability of the NAO, with correlations ranging from 60 to 84% for rainfall and 67–74% for runoff.

The consistency is essentially distributed from the interannual to the multi-annual scale. Overall, the consistency of inputs and rainfall with climatic indices appears to be more important on the frequency bands of 1, 1–2, and 2–4 years than for the interannual variability scales of 4–8 and 8–10 years. This is, however, partly linked to the duration of the chronicles, which is not long enough to be demonstrated consistent with multi-year variability patterns beyond 8 years.

Concerning the relationship with the NAO index, a total consistency ranging from a minimum of 49.6% (Foum El Gherza dam) to a maximum of 55.9% (Cheffia dam) was observed for inputs. A total consistency ranging from 53.9% (Foum El Gherza dam) to 59.6% (Cheffia dam) was recorded for rainfall.

Concerning the relationship with the MOI index, a total consistency ranging from a minimum of 52.7% (Foum El Gherza dam) to a maximum of 59.1% (Cheffia dam) was observed for inputs. A total consistency ranging from 46.6% (Foum El Gherza dam) to 58.5% (Zardezas dam) was observed for rainfall.

In detail, the local wavelet coherence spectra of the Cheffia and Zardezas dams reveal common characteristics. The influence of the two indices is more significant for the 1-year frequency band than for the interannual variability scales (1–2, 2–4, and 4–8 years).

Compared with the two Cheffia and Zardezas dams, the Foum El Gherza dam is characterized by a reduced influence of the NAO and the MOI climatic indices on the variability of inputs and rainfall. This can be explained by its geographical location in a sub-arid climatic zone (southern piedmont of the Saharan Atlas).

The aim of this section is to study the annual variation in water losses (evaporation and leakage), volumes of evacuated water (flood discharge and bottom outlet), as well as consumption over the same period and for the same sample of dams.

Figure 8(b) shows the cumulative annual values that record the volume lost due to leakage in the ten reservoirs. It is worth noting that there is no leakage in both the Cheffia and Ain Zada dams.

Leaks reached their peak in the years 1990/1991, 2002/2003, 2003/2004, and 2004/2005. The 1990/1991 peak can be explained by the large volume of observed leaks, which appeared in the Foum El Gherza reservoir, founded on the crystalline limestones of the Maestrichtian up to 80-m-deep, and on the right bank of the Ain Dalia dam about 200 m from the downstream foot. This phenomenon led to work in 1995 to stabilize the catchment area and drain the resurgences by means of 12 sub-horizontal drains about 30-m-long (ANBT 2014). Peaks observed in 2002/2003, 2003/2004, and 2004/2005 are justified by the remarkable losses observed in the Hammam Grouz dam, which is located on a calcareous and marly site, strongly tectonized and moderately karstic. These losses are underestimated (Mihoubi et al. 2013), which is confirmed by the appearance in April 2003, of a first vortex with a diameter of 1.5 m, which amounted to 7.4 hm³ in 2002/2003, 32.3 hm³ in 2003/2004, and 26.2 hm³ in 2004/2005.

The results of the interannual averages of the volumes evacuated by flood discharge and by bottom outlet show remarkable spatial and temporal nuances. The three dams of Cheffia, Hammam Debagh, and Zardezas, located in basins of the subhumid to humid Tellian zone, recorded the highest values (53.3 and 15 hm³). The Hammam Grouz and Ain Zada dams, located in a semi-arid zone, recorded the lowest values (2.2 and 0.7 hm³). The average flood discharge values are observed on the three dams of Foum El Gherza (19.6 hm³), K'sob (11.7 hm³), and Foum El Gueiss (10.0 hm³), despite their location in semi-arid to arid zones. This fact is explained by the brutal nature of floods as well as the advanced rate of siltation of these reservoirs.

The year 2011/2012 is characterized by an evacuated volume of 194.1 hm³, the highest values of which are observed in the months of February and March (112.5 and 80.9 hm³). These peaks are explained by two successive floods: the first was observed on February 2022, with a peak flow of 1,429 m³/s, causing a daily input of 73.2 hm³ and an evacuated volume of 83.8 hm³; the second flood arrived on 9 March, generating a maximum instantaneous flow of 985 m³/s.

The extreme values recorded over the years range from 92.3 hm³/year (77.7% of the Cheffia dam reserve at the beginning of the year, ensuring a satisfaction rate of 97.1%) to 0 hm³/year in some particularly dry years. This is to be linked to the low level of the dam reserve in certain years. As well, this is due to the incomplete supply and distribution networks for the benefit of the usage sectors, which leads to the non-exploitation of water reserves in dams.

For more detail, we will study the regulated volume evolution of the Hammam Debagh dam over the period from 1990/1991 to 2015/2016. This volume is distributed over three sectors of use, with the irrigation sector occupying first place (69.04%), followed by the drinking water sector (27.5%), and finally the industrial sector (3.4%).

The industrial water supply appeared in 1990/1991 then in 1995/1996, following the water shortage that affected the industrial zone of Annaba. The part allocated to the drinking water supply of the agglomeration of Guelma and the small centers close to the dam represents an annual volume of about 9 hm³ on average. The gradual increase in abstraction for drinking water supply peaked at 18.6 hm³ in 2015/2016 (Figure 10(b)).

The dam began to supply water for irrigation of the Guélma-Bouchegouf perimeter (total equipped area of 12,200 ha) in May 1996. Since then, the volumes intended for irrigation have been very variable from 1 year to another, the minimum being recorded in 1995/1996 (7.1 hm³) and the maximum in 2014/2015 (46.9 hm³).

The focus of the statistical analysis of water balances on a sample of ten dams (out of a total of 32 dams in operation) made it possible to study the hydrological behavior of these structures, the geographical distribution of which is representative of the very contrasting regional climatic context (from humid to arid climates).

The results of the reduced center of inputs and rainfall indicate a strong relationship between these variables, which is expressed graphically and statistically (the regional index). A cyclicity resulting in a clear increase in dry hydrological years from 1990/1991 to 2001/2002 and a return of wet years from the year 2002/2003 is highlighted. The annual to multi-annual temporal variability of inputs and rainfall by the wavelet transform indicates that this evolution is structured by four modes of variability in particular: 1, 2, 5–8, and 5–12 years. Three periods are globally identified for the inputs: before 2001/2002, between 2002/2003, 2006/2007, and after 2006/2007. The same periods have been identified in the rainfall series, with slight shifts in the years. The study of the input/rain relationship by wavelet coherence has shown that the main modes of variability of input and rainfall are highly similar; the average total coherence observed is 91%. The results showed that the variability of these parameters appears to be half related to the variability of the NAO and MOI indices.

The distribution of the water balance terms of the dams highlights the significant importance of the average annual evaporated volumes and their spatial variations linked to the climatic contexts as well as to the influence of the extent and depth of the reservoirs. In addition, the volume of water leaks varies from one dam to another depending on the geological and geotechnical conditions of the installation sites. The presence of fairly developed karst formations on the settlement sites (basin and banks of the dyke) largely explains these losses. Flood discharges reflect remarkable volumes subtracted from the region's water use balances; the spatio-temporal nuances that characterize them are linked to the importance of the floods in the year and to the values reached by the maximum instantaneous flows.

The variability of water balances severely influences the volumes allocated to the different uses. The degree of inflow regulation failure requires the implementation of dynamic management of dam water reserves. In order to compensate for the unequal distribution of dams across the territory and to reduce the effects of droughts, the insertion of interconnected planning systems, making the basins of Eastern Algeria solidary, is a necessity.

These results open up new avenues of research to improve and deepen our scientific knowledge of the hydrological problems of dams in Eastern Algeria:

• Specific understanding of the impact of physical factors in the catchment on hydrological variability.

• Impact of external factors (other climatic indices) on hydrological variability.

• Flood flow modeling on a set of dams to improve dam management.

• Study of specific erosion in catchment areas, with the aim of comparing these results with those calculated using bathymetric surveys.

This article benefited from the scientific support of the Algerian-French PROFAS B+ Program (2018/2019) and the PHC Maghreb scientific cooperation project (17 MAG 32), between our laboratory (LASTERNE, Constantine) and our partners at the University of Rouen (UMR CNRS 6143 M2C, and UMR IDEES CNRS 6226; Ms Valérie Mesnage, PHC Coordinator). May they all be thanked for this fruitful collaboration. In addition, the authors would like to thank Mr Khaled DJIR and Mr Mourad HOUGLAOUEN (ANBT, Algiers), Mr Yazid LADASI (ANB, Zardezas Dam), Messrs. Mounir MEGHZILI and Mohamed NEKAKA (ANB, Beni Haroun Dam).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.