Changes in land use/ land cover and water balance components before and after dam construction in the Mono River Basin, West Africa

The intergovernmental Panel on Climate Change has predicted sub-tropical region to be more vulnerable to the effects of Climate Change(CC). Additionally, to CC, land use and land cover (LULC) changes and dam contruction often neglected, plays and important role on the spatial and temporal distribution of water balance components(WBC) for agriculture production and socio-ecological equilibrium. The aim of this study wa sto analyze and compare the changes in LULC and WBC for the periode before Nangbeto dam construction (1964-1986) and the period after its construction (1988-2010) in the Mono River Basin(MRB). To this end, the study used mainly WBC extracted from the validated Soil and Water Assessment Tool and LULC data of the years 1975 and 2000 in the MRB to explore their temporal distributions and the link in their changes. The results showed that the mean monthly actual evapotranspiration, percolation, water yield, surface runoff, groundwater and lateral ow represent 51.05%, 17.53%, 15.93%, 9.43%, 5.67% and 0.42%, respectively of total water balance between 1964 and 1986. The same components represented 51.02%, 9.17%, 20.43%, 6.30%, 10.56% and 2.59%, respectively between 1988 and 2010. The contribution of these WBC in mean-annual (1964-1986) period were for actual evapotranspiration (31.33%), water yield (25.95%), percolation(17.67%), groundwater(14.71%), surface runoff (9.94%) and lateral ow (0.40%). Meanwhile, between 1988 and 2010, the contribution of actual evapotranspiration, water yield, percolation, groundwater, surface runoff and lateral ow are 49.85%, 19.97%, 11.17%, 10.34%, 6.15% and 2.52%, resepctively. The results showed that the peaks of the evapotranspiration, surface runoff, percolation and water yield appeared in September corresponding to a month after the maximum of rainfall in August. However, our more detailed analysis showed that a signicant decrease of forest and savanna and increase of cropland let to a decrease in actual evapotranspiration and lateral ow over the second period of simulation compared to the rst period of simulation over the MRB scale. These ndings showed that sustainable management and conservation of natural vegetation is crucial for integrated water resource management and conservation in MRB.


Introduction
Water is a source of sustainable economic development because it gurantees supply of basic resources to society and ecosystems. Water resource is fundamental to various sectors of activities such as agriculture, industry, domestic water use and sanitation, hydropower generation, health and environmental security (Hanjra and Qureshi, 2010). Agriculture, hydropower dam and agricultural land irrigation are the sectors with large water consumption. Actually, water resource management is becoming more pressing issue due to climate change (CC) impacts (Mango et al., 2011;Yang, 2008). Most of researches in water resource management are dealing with complex system determined by several interactions between natural, socio-economic, political issues and climate change implications (PCCP, 2008). As demonstrated by the Intergovernmental Panel on Climate Change (IPCC) in its 2014 report, global climate change, demographic and economic changes will be felt more in tropic and sub-tropical regions (Paeth et al., 2009;Stanzel et al., 2018). In most cases, climate variability and human activities are the two major driving factors of hydrological processes and spatial temporal distribution of water balance components (WBC) in any river basin.
In other hand, land use and cover changes (LULCC) usually affect hydrologic cycle through their direct impacts on land surface processes like the amount of evaporation, groundwater in ltration and surface runoff that occur during and after precipitation events (Bronstert et al., 2002;Setyorini et al., 2017). These factors control the water yields of surface stream ow and groundwater aquifers and thus the amount of water available for ecosystem functions and human use (Anderson et al., 2011). For instance, dam construction on a river basin without appropriate management strategies and precautions can induce changes in stream ow with downstream ooding. These LULCC in addition to CC will have consequence on local and regional hydrological regimes. Consequently, the spatial and temporal water resource availability, or in general the water balance, will be signi cantly affected (Huntington, 2006). Therefore, more attention is needed in this sector.
Several studies have investigated the SWAT-simulated hydrological impact of land use change in Iran (Ghaffari et al., 2010). In China Zuo et al.(2016) assessed the effects of changes in land use and climate on runoff and sediment in China river basin ). Others studies assessed future land use changes impacts sed on hydrological process in Canada (Wijesekara et al., 2012) or performed a comparison of hydrological models for assessing the impact of land use and CC on discharge in a tropical catchment (Cornelissen et al., 2013a). In Togo, there are few studies on LULCC and water resource in the complex transboundary basin of Mono river (Badjana et al., 2017;Klassou and Komi, 2021).
Recently in the MRB, Koubodana et al. (2019) have shown that the watershed landscape is dominated by cropland, savanna, forest and oil palm plantation. LULCC in the MRB is characterized by losses of savanna and increase of cropland between 1975 and 2013, explained by demographic growth in Togo (Koglo et al., 2018;Koubodana et al., 2019). Over the past years, many projects like the Potential Con ict to Cooperation Potential (PCCP, 2008), the Integral Water Resource Management in West Africa (SAWES, 2011) and the Integrated Disaster and Land Management Project (PGICT) were implemented for promoting sustainable water resource management in the MRB. Previously, Kissi et al. (2015) and Tramblay et al. (2014) analyzed the social vulnerability of ood in the Bas-Mono prefecture embedded in the MRB. Ntajal et al. (2017) have investigated ood disaster risk mapping and analysis while Houngue (2018) has looked at the simulation of high stream ow using lumped hydrological and climate models in the small area of lower MRB. The authors concluded that the source of high stream ow is not only due to climate change but also to the regulation of the Nangbéto dam, land use and the social factors of the communities living in the catchment. Recently, Koubodana (2020a) have successfully run calibrated and validate SWAT semi-distributed model to assess stream ow change before dam installation  and for the period after dam installation . The authors suggested that land cover changes impacted on stream ow and probably on the others WBC which need further investigations.
In this study, outputs from already calibrated and validated SWAT model in Koubodana et al. (2020a) were used to analyze the temporal contribution of WBC for sustainable water resource management in the MRB. The speci c objectives of the study are to: (i) assess the temporal distribution of WBC for the periods before and after dam construction; (ii) compare the contribution and changes of WBC in the MRB before and after dam construction, and (iii) determine the link between LULCC and WBC changes in the same basin. The outputs of this study will allow to elaborate strategies for better planning and sustainable management of water and land resources in the Transboundary Mono River Basin.

Study Area
The MRB is drained by the Mono River and its tributaries. It is a transboundary basin shared by Togo and Benin Republics in the southern parts of the basin. The Mono River is located between 06°16' and 9°20'Northern latitude and 0° 42'and 1° 40' Eastern longitude (Figure 1). With a perimeter of 872, 092 km, the basin covers a surface area of 22,013.14 km², with 88% in Togo (PCCP, 2008). Flowing from its main source in Alédjo mountains in north Togo, to Atlantic Ocean in the South, the Mono River has a total length of 308.773 km. The elevation of the basin ranges from 12 to 948 m (http://srtm.csi.cgiar.org). The watershed shelters the biggest dam of Nangbéto that produce 20% of total hydroelectricity used by the two countries. To increase the electricity supply capacities, Togo and Benin have co-funded the construction project of a second dam on the same river at Adjarala.
The climate is a subequatorial climate from 0 to 8°N and with two rainy seasons and two dry seasons. It totals 1200 to 1500 mm/year in the mountainous area of the South-West and only, 800 to 1000 mm/ year on the coastal zone. From 8 to 10°N, the climate is tropical humid with one rainy season and one dry season (1000 to 1200mm/year). In the winter (December to March), there is an anti-cyclonic highpressure area centered over the Sahara. It drives the Harmattan, a desiccating, dusty wind that blows rather persistently from the northeast, drying out landscapes all the way to the coast (Arbonnier, 2000). However, the hydrograph has one peak that indicates that river discharge is mostly controlled by upstream tributaries. The mean annual temperature ranges from 22°C to 30°C and precipitation varies between 800 mm and 1300 mm/year (CILSS, 2016;Speth et al., 2010). Precipitation usually reaches the peaks in May-June and September-October.
Human activities in the MRB mainly include the management of the construction the hydroelectricity dam, irrigation activities in the downstream, water withdrawal for population needs, agricultural development and sheries. The rivers shelter the most important reservoir of Nangbéto Dam. The dam is built at 180 km from the mouth of the river and the surface area for water retention that feeds it is 15700 km 2 . The second dam under project of Adjarala will be built at 100 km downstream from Nangbéto and, between the two dams, the drained area is 11,000 km 2 (Rossi, 1996).

Water balance components datasets
The datasets used in this study are from the validated SWAT model outputs generated by Koubodana (2020a). The WBC considered were Precipitation (PCP), actual evapotranspiration (ET), percolation (PERC), surface runoff (SURQ), and groundwater ow (GW_Q), water yield (WYLD) and lateral ow (LAT_Q). These components were extracted from the calibrated SWAT model for the two periods. The values are provided on daily basis and for each sub-basin or reach point between 1964 and 1986 and from 1988 to 2010. The watershed was divided automatically into 24 sub-basins for the rst period of simulation  and 23 for the second period (1986-2010) (Gassman et al., 2007).

Land use and land cover change datasets
The LULC maps of 1975 and 2000 were used to re ect on land use/cover patterns for the period 1964-1986 (named as SIM1) and the period 1988-2010 (called SIM2), respectively. The Nangbéto Dam started operating in 1987, which is selected as the turning point of the climate data, because signi cant changes of land use may play an important role in local WBC. Land use and land cover datasets were initially analyzed (Koubodana et al., 2019). We de ne cropland and fallow with oil palm as a crop eld and fallow land, farms with crops and harvested croplands whereas agriculture represents cultivated areas with seasonal crops dependent on rainfall. Table 1 shows the areal proportions of the LULC units for the two years. and 111 HRUs for SIM1 and SIM2 respectively based on the same soil, land use, and slope (Arnold et al., 1998). More detail characteristics of the input data used for the SWAT model setup can be found in Koubodana (2020a). The surface runoff was estimated using the Soil Conservation Service (SCS) curve number method which is a function of land use, soil permeability and antecedent soil water conditions The Hargreaves's method, which requires only minimum and maximum temperature as input data was used for the evapotranspiration estimation in the model (Hargreaves and Samani, 1982;Koubodana, 2020a). A detailed description of the model setup, sensitivity analysis, calibration, and validation is presented by Koubodana (2020a). The SUFI-2 semi-automatic tools for calibrationvalidation-sensitivity & uncertainty analysis were then used to generated a validated and representative SWAT model over the catchment (Abbaspour et al., 2017).
The main WBC were extracted for both the periods before (SIM1) and after (SIM2) dam construction. These data were used to compute the average monthly and mean annual contribution over the whole catchment. Using SWAT Output Viewer (https://swatviewer.com/), it was possible to extract the contribution of each WBC at mean monthly and annual scales. Mean annual and mean monthly values were computed for each WBC contribution considered over SIM1 and SIM2. Therefore, the mean annual and monthly WBC contributions were used to show the percentage of each water balance components at annual and monthly scale.

Analysis of the temporal distribution of water balance before and after dam construction
The temporal distribution of WBC was assessed using Origin 2018 software. First, the SWAT models WBC contribution outputs between 1964 and 1986 or between 1988 and 2010 and distributions were averaged at sub-basin level. Next, the temporal contributions of a selected WBC value for the catchment were averaged using Origin 2018 software. Finally, the same software was used to compute the matrix where the component values of the water balance are listed according to month and year for the periods before and after dam construction.

Changes in land use/cover and water balance components before and after dam construction
The study has established the relationship between LULCC and hydrological components before and after dam construction. The SWAT model simulation was divided into two periods: SIM1 period  and SIM2 period (1988SIM2 period ( -2010. For SIM1, the land use map of 1975 was used as input and with climate variables extended between 1961 and 1986. Meanwhile, the input data for the SIM2 were the land use data of 2000 and climate variables extended between 1988-2010 and Nangbéto reservoir set in 1987.
More details about SWAT model setup model sensitivity analysis, calibration, validation and uncertainty analysis can be found in Koubodana, 2020a. Based on the validated model outputs after its performances and uncertainties analysis during SIM1 and/or SIM2, modelers could be able to deduce the impacts of LULCC on hydrological components before and after dam construction in 1987. Thus, the temporal intensity distribution and related statistics of WBC over MRB for SIM1 and SIM2 were respectively generated. Furthermore, the changes between the two periods of hydrological cycle components were computed and comparative methods allowed giving sustainable information for integral water resource management in MRB.

Results
3.1: Precipitation temporal distribution and changes before and after dam construction Figure 2 underscores the temporal distinction of wet and dry seasons and shows the monthly rainfall patterns for the whole basin between 1964 and 1986 (a) and between 1988 and 2010 (b). The onset of rainy season is May/June for both periods. The cessation of rains occurs in September/October and October/November for the period before and after dam construction, respectively. The peak of rainfall is reached in between June and September during SIM1 and August and October during SIM2. The dry season always starts from November/December to March/ April of the following year in the two simulations. There is an alteration of unimodal years (1966, 1968, 1979, 1980, 1985 & 1989, 1991, 1993, 1995, 1999, 2002, 2003, 2010) and bimodal years (1964, 1974, 1976, 1978 & 2007, 2009). Rainfall magnitude intensity between 1988-2010 has considerably decreased compared to the period  where there is the inverse situation.  1988, 1992, 1996, 1997, 2000, 2001, 2002, 2004 and 2005 which are known as drought years in the region.
3.4: Land use/cover and water balance component changes before and after dam construction in the Mono River basin Knowledge about land use and land cover (LULC) dynamics is of high importance for an integral water resource management in a given watershed. Therefore, LULCC were estimated between 1975 and 2000. The major land use changes are observed in savanna, forest, agriculture and cropland (   (Table 3). Consequently, the results showed that there were signi cant decreases of forest, savanna and increases of agricultural land involve a decrease of precipitation (PRECIPmm), actual evapotranspiration (ETmm) and lateral ow (LAT_Q_mm) over the second period of simulation compared to the rst period of simulation. The other component such as percoration (PERCmm), growndwater (GW_Qmm), surface runoff (SURQ_mm) and water yield (WYLDmm) show an increase in the second period of simulation.  1964-1986 1988-2010 1964-1986 1988 4. Discussion

4.1: Temporal analysis of water balance components
In water management strategy planning, the analysis of individual water balance component contribution is a requirement. Sathian and Symala (2009) indicated that precipitation, actual evapotranspiration, percolation, groundwater, surface runoff, water yield and lateral ow were the most important components of water balance in a watershed. Among these components, precipitation is an input in hydrological models such as SWAT while other inputs are predicted due to the paucity of observation data (Ghoraba, 2015). Actual evapotranspiration n percolation and water yield components contribution were the highest components over the two periods of average annual and seasonal timescales as displayed in Table 3.
Actual evapotranspiration is the highest amount of water loss by the watershed in annual and seasonal average scales. The high amount of actual evapotranspiration can be explained by the various type of vegetation and also by the global increase of temperature and particularly in the study area (Koubodana, 2020;Lawin et al., 2019). Meanwhile, it is important to note that actual evapotranspiration has increased from 31. 33% (1964-1986) to 49.85% (1988-2010) in inter-annual time scale and slightly from 51.02%  to 51.05% (1988-2010) for intra-annual period. This increase of water actual evapotranspiration from the period  to the period (1988-2010) is due to the increase of global land surface temperature since 1970, LULCC or decreasing wind speed (Koubodana et al., 2020b(Koubodana et al., , 2019.
The second major WBC is water yield which is net amount of water that leaves the sub-basin or the basin and contributes to stream ow in the reach during the time step. It is computed as WYLD = SURQ + LATQ + GWQ -TLOSS -pond abstractions. Therefore, an important amount of precipitation percentage received by the watershed is lost as stream ow. The percentage amount is ranging from 0.40%  to 2.52% . According to Figure 2b and Figure 4b, water yield decreases from 25.95% between 1964 and 1986 to 19.97% between 1988 and 2010 at -average annual timescale whereas Figure  3b and Figure 5b show on average seasonal timescale, it amounts has increased from 15. 93% (1964-1986) to 20.43% (1988-2010). Lateral ow is the lowest (1988-2010) for average annual time scale and from 0. 42% (1964-1986) to 2. 59% (1988-2010) for average seasonal timescale. This can be due to the low in ltration rate and also that lateral ow depends on the watershed local slope (Cornelissen et al., 2013b) which is not constant in the basin and ranges from 12 to 948m. The results on water cycle components contribution con rmed most analysis performed in West Africa (Akpovi et al., 2016;Begou et al., 2016;Hounkpè, 2016;Kumi M, 2015). For average annual timescale analysis, many years are associated with high and low contribution of surface runoff compared to the average over the period. For example, 1968For example, , 1979For example, , 1980For example, , 1995For example, , 1999For example, and 2003 runoff contribution is higher and with positive rainfall index. The years of 1977The years of , 1982The years of , 1983The years of , 1986The years of , 1990 and 2002 present the period with lowest surface runoff and associated with negative rainfall variability index and con rmed the years of drought in West Africa (Koubodana et al., 2020b;Oguntunde et al., 2006;Yabi and Afouda, 2012).
Rainfall matrix of Figure 2 shows the average seasonal precipitation from 1964 to 1986 and also from 1988 to 2010 sub-periods. These results con rmed that the climate in the MRB, is unimodal and bimodal according to the past studies (Koubodana et al., 2020b;Tramblay et al., 2014). Figure 3 and Figure 4 show the matrix illustration of water cycle components between 1964 and 1986 and from 1988 and 2010 sub-periods. These gures illustrate the dry and wet months and years over the entire basin. It displays also the nature of the season assigned in the basin. The results reported that most of WBC reach their peak between July and October which is exactly the period of rainy season in sub-tropical zone (Djaman et al., 2017;Giertz et al., 2006;Laux et al., 2009). The period of 1964, 1984,1982, 1981, 1977, 1976,1973, 1972, 1988, 1992, 1996, 1997, 2000, 2001, 2002 and 2004 characterized by a very low amount of surface runoff and precipitation are justi ed by the year where drought occurred in West Africa (Laux et al., 2009;Omotosho and Abiodun, 2007;Sylla et al., 2016). For example in 1984For example in , 1982For example in , 1981For example in , 1977For example in , 1976For example in ,1973For example in ,1972For example in and 1964 known in previous analysis as drought year with negative annual rainfall variability index (Descroix et al., 2009;Yabi and Afouda, 2012). and climate variability induce the increase of precipitation intensity, actual evapotranspiration and lateral ow whereas decrease in percolation, groundwater, surface runoff and water yield were found. One of the reason of this situation is that the conversion of forest and savanna in cropland caused the change in surface soil layer and vegetation canopy ). This con rms that LULCC play an important role in the changes in WBC, water in ltration, evaporation and water movement at local level (Hagemann et al., 2014). The results con rmed the analysis of Koubodana et al. (2019) over the basin. Figure 5 showed LULCC between 1975 and 2000 have affected water components annual mean for (1964-1986) and (1988-2010) periods respectively. Forest and savanna decreased and could be explained by agriculture expansion, bush re, timber extraction in response to population needs (Atsri et al., 2018;Koglo et al., 2018;Koubodana et al., 2019). Togo and Benin experience increase of population which involve more demands for agricultural lands and habit, energy wood consumption. According to Verstraeten et al. (2008), the actual evapotranspiration (ET) is the process from which water is transferred from the soil compartment and/or vegetation layer to the atmosphere. Therefore, any change in land cover (leaf index area) or land use will affect ET intensity. Soil characteristics and climate condition also impacted on water balance components variation (Sciuto and Diekkrüger, 2010). 4.3. Hydrological modeling and in uence of the sources of uncertainty in the analysis Indeed, there are many sources of uncertainty which could affect the results of this study. Some of these include: uncertainty associated with the hydrological modelling and input data quality. In many cases, the analysis of these predictive uncertainty helps in capturing the overall range of expected uncertainty propagated through modelling (Klein et al., 2016;Multsch et al., 2015). But this study was already subject of uncertainty analysis in Koubodana (2020a). Others studies prefer an ensemble hydrological modeling in order to reduce uncertainties (Gaba et al., 2015;Huisman et al., 2009). The ensemble of the hydrological models could therefore encompass the effects of model uncertainties, because the mean result is a more reliable estimation of hydrology characteristics and increases the confidence of the modeling. In fact, multi-model approach has been proven to be more robust and exhibits a better performance than individual models (Huisman et al., 2009). All these limitations will be considered in furthers analysis in order to minimaxes uncertainties for formulation of better policies strategies measures at local scale.

Conclusion
In this study, outputs on water balance components from a SWAT hydrological model and land use dataset of the years 1975 and 2000 were analyzed in the Mono River Basin over the period before Nangbeto dam construction  and the period after its construction . The results showed that mean monthly actual evapotranspiration, percolation and water yield represent 70% of total water balance in mean monthly and annual time scale. In details, actual evapotranspiration, surface runoff, percolation and water yield peaks appeared in September corresponding to one month after the maximum of rainfall in August. However, more detailed investigation showed that a signi cant decrease of forest, and savanna and increases of cropland involve an increase in precipitation amount, actual evapotranspiration and lateral ow over the second period of simulation compared to the rst period of simulation. Therefore, from this analysis it can be concluded that water balance component contribution, the runoff, evapotranspiration and water yield evolution depend strongly on different land-use type change, climate conditions and also on the presence or not of reservoir in the watershed. Finally, there is a strong need to develop sustainable adaptation measures in future studies including ensemble modeling to reduce uncertainties, particularly at local scale where the impact occurs, to mitigate the possible impacts of the projected change in climate.