Sedimentation assessment using hydrological simulation and bathymetry survey: The case of river Amissa drainage basin, Ghana

Abstract

Studies on freshwater bodies in Ghana for their sustainability are mostly concentrated on a few large rivers. However, other equally important rivers that do not attract research attention provide varied services and benefits to inhabitants and living organisms in their riparian communities. The basin of River Amissa as a major source of freshwater supply for several communities within its catchment has undergone various changes due to rapid socioeconomic and increasing anthropogenic disturbances. This consequently has led to increased sediment yield on the reservoir beside the river and the river basin itself. Hence the need to estimate the amount of sediment accumulated in the reservoir and in the river's catchment for their sustainability and management purposes. Sediment yield in the reservoir and the river basin were estimated using a bathymetric survey integrated with Geographic Information System (GIS) and the Soil and Water Assessment Tool (SWAT) hydrological model respectively. Results from the bathymetric survey estimated the current capacity of the reservoir to be 4,321,060 m³. Reservoir storage capacity lost to sedimentation was 1,438,940 m³ representing 25% of the reservoir volume. Additionally, the SWAT (Arc SWAT) performance was very good with Nash-Sutcliffe efficiency coefficient (NSE) for calibration and validation being 0.88 and 0.84 respectively. For the entire 28-year period, simulated sediment yield increased by 10,263 tons per kilometer square.

1. Introduction

Sediment yield is the quantity of sediment leaving a catchment per unit area annually [1]. A high sediment yield is mostly accompanied by increased soil loss within a watershed, which has effect on soil productivity, water quality and quantity, fishing activity, reservoir lifespan, and stream channel morphology [[2], [3], [4]]. It is therefore a vital component of water resource management and development to have accurate information on sediment yields within river basins [5,6]. Water supply and irrigation dams are threatened by sedimentation in reservoirs. Consequently, sedimentation in reservoirs is very critical to reservoir performance and storage capacity.

There are however two techniques for determining sediment load in rivers; field (direct) measurement and physical or empirical modeling. For field (direct) measurements, suspended sediment loads and discharges are estimated [4,7], as well as eroded sediments in small catchments [4], and estimation of the volume of sediments in lakes, ponds or reservoirs [[8], [9], [10]]. In spite of that, Ghana, like most less developed countries (LDCs), often lacks adequate resources to monitor and forecast human impacts on water resources [11]. As a result of the costs, remoteness of the site, the number of tests, and technical difficulties associated with the field method, hydrologists in LDCs often resort to the use of hydrological models for assessing suspended loads in rivers that do not involve direct measurements [5,12]. Researchers in present times also make use of other methods such as, a combinatorial method containing multi-criteria decision-making, likelihood ratio and fuzzy logic [41], Topographic LIDAR [42], a convolutional neural network (deep learning method) [43], and many others all for mapping out landslide susceptibility and hazard assessment. These methods, however, are not without their limitations. It follows that one's choice of method is influenced by a variety of factors, including equipment availability, cost, convenience, and expected outcomes [13]. These and many other factors led the study to combine the bathymetry survey, GIS, and SWAT hydrological modeling. In spite of the various limitations of these hydrological models, the SWAT is preferred in this study due to its ability to model ungauged catchments, prediction of the relative impacts of scenarios (alternative input data) such as changes in management practices, climate, vegetation on water quality, quantity, or other variables [14]. With SWAT, each hydrologic response unit (HRU) is assumed to be spatially uniform in terms of slope, land use, soil type, and climate. SWAT is also attractive for its computational efficiency as it offers some compromise between the constraints imposed by the other model types such as lumped, conceptual, or fully distributed, physically based models [15].

Literature shows that only the major popular rivers in Ghana receive attention when it comes to hydrological studies especially on sedimentation [5,8,[10], [11], [12], [13],[16], [17], [18]]. However, these minor rivers that are mostly ignored, provide a wide range of services and benefits to the inhabitants and living organisms of their riparian ecosystems and at the same time, very vulnerable to environmental changes [19]. A typical example is the river Amissa which is considered a minor river in Ghana. Nonetheless, several communities rely on its freshwater supply for survival [20]. However, the river's basin has experienced rapid socioeconomic growth in recent years. The floodplain of the river has also attracted agricultural and settlement expansion due to the low-lying nature of the basin and the surrounding fertile soil. This has resulted in increased sediment yields in the river basin and in the reservoir due to soil erosion from agricultural fields, hill slopes, and settlement expansion. Field observations by the research team and reports from the officials of Ghana Water Company Limited (GWCL) on river Amissa revealed a drastic reduction in the river volume and this compelled the management of Ghana Water Company Limited to desilt some portions of the channel in April 2020 [20]. Furthermore, portions of the river's wider channel with relatively high discharge that attracts swimming has now become very narrow and shallow as a result of increased sediment yield and transport. Moreover, with the reservoir serving as both a water treatment and irrigation system, only few studies have been conducted on sediment assessment; hence, a geospatial assessment of sediment in the basin is required in the face of current climatic changes. In addition to filling an important gap in neglected river studies, this is also important for natural resources management, planning, and sustainable development.

The main objective of the study is to examine the sedimentation assessment within the river Amissa basin. Documenting the trend in sedimentation of the river basin and the reservoir will help ensure their sustainable management. Due to the large nature of the river basin, covering the full extent of the basin using direct field-based sediment monitoring would be very tedious and expensive, hence the use of a hydrological model to simulate the spatial distribution of sediment within the river basin.

The objective of the study was achieved through a field-based bathymetric survey integrated with GIS to estimate the amount of sediment accumulated in the reservoir. This was augmented with a hydrological simulation of sediment within the drainage basin of the river that feeds the reservoir. These two approaches were adopted based on their level of efficiency in estimating sediment yield. The use of hydrological models in estimating sediment may underestimate or overestimate sediment yield within a given basin. Therefore, using the field-based bathymetric survey for sedimentation assessment in a reservoir will help avoid these uncertainties.

The innovation point in this study is the application of a data-driven hydrological model in a limited data environment and a field-based bathymetric study to estimate the spatial distribution of sediment in the drainage basin. This is because most of the rivers in Ghana lack sediment monitoring stations and even those with flow monitoring points are outdated and poorly managed. This is basically as a result of the expensive and time-consuming laboratory analysis that follows the physical sample collection of suspended sediment concentration. However, the use of the SWAT hydrological model in recent times is widely acknowledged as one of the best tools for estimating sediment yield without necessarily visiting the site. And also, the application of bathymetric survey integrated with GIS is able to avoid the under and over-estimations sometimes experienced in the use of hydrological models. As a result, the adoption of these two techniques in a study of any geographic region is capable of providing reliable results. This combination of the direct measurement on a small water body (reservoir) and indirect measurement on a relatively larger river basin seeks to contribute to the evaluation of the quantity of sediments washed from a given watershed (River Amissa), which basically is responsible for filling an artificial basin (reservoir) used simultaneously for water treatment (drinking water) and irrigation.

2. Research methodology

2.1. Site description

The study was conducted on the river Amissa Basin (Fig. 1) and the reservoir facility on the river. This river's profile coincides with Latitudes 5° 44′N and 5° 11′N, and Longitudes 1° 20′W and 0° 56′W. Located in the Central Region of Ghana, the river begins to flow from Assin Fosu southwards for about 57 km long before discharging into the Gulf of Guinea at a point known locally as Ekumfi Asaafa all in the Central Region. It covers an area of 1413.3 km² which comprises of five administrative districts and municipalities in the central region, namely, Ekumfi, Mfantseman, Ajumako-Enyan-Esiam, Assin South and Assin Central [34]. A gentle undulating terrain characterizes the basin, with the northern segment of the basin having the highest elevation [10]. The vegetation of the basin comprises of Tropical rainforest, moist semi-deciduous forest type, coastal thicket and grassland and few mangroves along the coast. It has become an agricultural hub because majority of the basin lies in an agriculturally productive zone. Subtropical wet climate dominates the basin, with double maxima rainfall seasons (May–July and September–November). Meteorological stations that were used for the SWAT hydrological modeling includes Saltpond, Asuansi and Assin Fosu.

Fig. 1.

The river and reservoir in the study area.

There is also a dam (locally known as Mankessim Dam) constructed on the river to store water in the reservoir for domestic use. The dam is an earth fill type with a length of 195 m and the height of 10.4 m [21]. The dam is located at 5° 18′ 52.08″ N and 1⁰ 01′ 45.08″ W. The reservoir's surface area was 1.92 km² at the time of construction but now stands at a reduced area of 1.3 km². It was primarily constructed for irrigation purposes between 1974 and 1977 but now serves a dual purpose of both domestic water supply to households and irrigation. The total capacity of the reservoir after construction was 5,670,000 m³. The irrigated area for the reservoir is 17 ha. The major encroachment within the dam catchment is by estate developers (settlement expansion). There are two irrigation schemes located in the basin under the management of Ghana Irrigation Development Authority [20]. The reservoir also serves as a source of water for the water treatment plant at Baifikrom which provides pipe borne water for the entire Mfantseman Municipality, parts of Ekumfi district, and parts of Ajumako Enyan Essiam district [20]. It also serves as a standby water supply to the Cape Coast Metropolis if there is shortage of water supply from the Brimsu Head Works in the Cape Coast metropolitan assembly [20].

2.2. Datasets

2.2.1. Meteorological and hydrological data

Weather data were collected for Saltpond, Asuansi, and Assin Fosu weather stations for the study. Daily precipitation and temperature values from 1990 to 2020 were obtained from the Ghana Meteorological Agency (GMet), stream flow data was also obtained from the Ghana Hydrological Survey Department and Climate Forecast System Reanalysis (CFSR) data from National Center for Environmental Prediction (NCEP) to augment the GMet data.

2.2.2. SWAT model spatial datasets

ASTER GDEM, land use, and soil maps (Fig. 2a, b, and c) were the principal spatial input datasets used in ArcSWAT. For catchment delineation, ASTER GDEM was acquired at a resolution of 30 m. Through supervised classification of Landsat imagery, five LULC classes were recognized: water, built-up, closed forest, open forest, and farmlands. Soil data were collected based on the FAO DSMW (Digital Soil Map of the World) with a scale of 1:5,000,000. Rasterized shapefiles of the spatial data with their lookup tables were uploaded to ArcSWAT for classification to show the area of interest.

Fig. 2.

(a) Soil, (b) 2020 land use land cover map and (c) DEM.

2.3. Application of the SWAT

This model estimates water, chemical yields, and sediment yields from large complex watersheds using continuous, long-term, physically based data [22]. Hydrologic Response Units (HRUs) are the model's basic operating units that represent homogeneous land uses, land management, and soil characteristics. The HRUs are nested within the sub-basins, and simulations are aggregated at the HRUs before being transferred to the sub-basins [23]. Furthermore, the SWAT model predicts sediment fluxes using the Modified Universal Soil Loss Equation (MUSLE) expressed in equation (1) below [24]:

$$\mathrm{S}\mathrm{e}\mathrm{d}=11.8{\left(Q_{surf}*{\mathrm{q}}_{\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}*{\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}}_{\mathrm{h}\mathrm{r}\mathrm{u}}\right)}^{0.56}{\mathrm{K}}_{\mathrm{U}\mathrm{S}\mathrm{L}\mathrm{E}}{\mathrm{C}}_{\mathrm{U}\mathrm{S}\mathrm{L}\mathrm{E}}{\mathrm{P}}_{\mathrm{U}\mathrm{S}\mathrm{L}\mathrm{E}}{\mathrm{L}\mathrm{S}}_{\mathrm{U}\mathrm{S}\mathrm{L}\mathrm{E}}\mathrm{C}\mathrm{F}\mathrm{R}\mathrm{G}$$ (1)

where Sed is the yield of sediment (ton day⁻¹), 11.8 is the unit conversion factor, Q_surf is the volume of surface runoff (mm ha⁻¹), q_peak is the peak surface runoff rate (m⁻³ s⁻¹), area_hru is hydrologic response unit area (ha), K_USLE is USLE soil erodibility factor, C_USLE is USLE cover factor, LS_USLE is USLE topography factor and P_USLE is USLE soil protection factor, and CFRG accounts for stoniness [24].

Additionally, sediment fluxes are predicted based on two principal processes after peak runoff is generated: degradation and deposition. In the watershed, these processes determine how much sediment is entrained or transported out of the watershed [23]. For the current study, ArcSWAT was used to simulate sediment. Spatial and weather data were used in ArcSWAT to simulate flow and sediment.

2.3.1. SWAT model setup

SWAT delineated 13 sub-basins for the Amissa River basin. Land use, soil type, and slope characteristics were used to divide the catchment into 184 HRUs. Fig. 3 shows the general workflow as it was executed in the SWAT operation. For the purposes of zooming into relatively larger streams in the Amissa River basin, a threshold area of 1500 Ha was used when delineating the river networks. Land surface slopes were defined by multiple land surface slope definition based on three land surface slope classes (0–8, 8–30, and 30–9999%). Unlike using a single approach, the multiple slope approach considered the various slopes within the area. Using the SWAT2012. mdb database, land use and soil data were reclassified into forest, crop, and urban with their usersoil tables. In the definition of the HRU, land use, soil class, and land surface slope classes were set at 0%, 5%, and 2%, respectively choosing multiple criteria. In creating the HRUs, these limit levels were set to guarantee that all dominant land use and soil class definitions were considered [25]. Finally, the simulations covered a period of 30 years (1990–2020), with the first two years serving as warm-up period. A monthly time step was used for the output.

Fig. 3.

Flow chart for SWAT

2.4. Model calibration and validation

Many parameters affect sediment fluxes; therefore, sensitivity analysis became vital to identifying the most sensitive parameters. Calibration of the model was achieved both manually (with SWAT's Calibration Helper v1.0) and automatically with SWAT's Calibration Uncertainty Program (SWAT-CUP). To automatically calibrate the model, the Sequential Uncertainty Fitting (SUFI-2) algorithm was utilized. Calibration and validation were performed using stream flow data. A sensitivity analysis was conducted to calibrate the model (1992–2010), followed by a validation (2017–2019). A monthly time step was used for the calibration and validation.

3. Bathymetry

The second part of the sedimentation assessment was done on the reservoir within the drainage basin. For sedimentation assessment in the reservoir, the current volume of the reservoir as measured through bathymetry study was deducted from the volume of the reservoir as it was designed for. The designed volume of the reservoir was acquired from the Ghana Irrigation Development Authority (GIDA) while the current volume of the reservoir was acquired by conducting bathymetric study on the reservoir. The present reservoir depth points were surveyed with a canoe, Garmin Eco sounder, a GPS device, and a life jacket. Using the Eco Sounder for depth measurement, a Hand-Held GPS for taking the corresponding geographic coordinates of the depths, a canoe for the reservoir navigation, ArcGIS 10.7 software, and proper safety clothing, the bathymetric survey was done between May and August 2020. Reservoir depth at full supply level was 10.4 m

The bathymetric data obtained from the reservoir navigation was upgraded to highest pool (spillway) height, which is the greatest permissible water surface rise. This was done in light of the fact that the reservoir was not at its maximum height at the period of navigation for the depth measurements. The upgraded bathymetric data was used to create bathymetric study map (Fig. 4) for the reservoir with the use of ArcGIS 10.7. The depth points and their corresponding GPS points were processed in Excel, which were then loaded into a geographic information system (GIS). The (X, Y, Z) data was informed into the ArcGIS 10.7, a shapefile for the points was created, and their coordinates defined. The depth points were interpolated using ordinary kriging technique in ArcGIS. Kriging enjoys the advantage of giving a proportion of error or uncertainty for the assessed surface additionally to the estimated surface [31]. The interpolated depths were then assigned colour with regard to the depth, after which an elevation (raster) map was created out of it, with contour lines made up of points of the same colour along the reservoir's longitudinal length using the contour spatial analyst tool. Bathymetric contour map at 0.5 m wide was eventually produced demonstrating the spatial distribution of the reservoir's depth at its maximum height (Fig. 6). The bathymetric survey and the reservoir sedimentation estimation followed the workflow in the figure below (Fig. 5).

Fig. 4.

Bathymetric survey points in the study reservoir.

Fig. 6.

Spatial distribution of reservoir depth points at maximum height.

Fig. 5.

Bathymetric survey and Reservoir Volume Estimation.

Accordingly, the adjusted depth points and the corresponding coordinates were used to generate 2D and 3D bathymetric maps. The 2D map was created by interpolating from the reservoir depth points mapped. The interpolation method used was the ordinary kriging.

The 3D bathymetry map was additionally made using the Triangulated Irregular Network (TIN) model. The TIN uses the original sample points to constitute many non-overlapping triangles that cover the whole region. TINs are a type of vector-based computerized geographic data and are developed by triangulating a bunch of vertices (points). The vertices are linked with a series of edges to form a network of triangles. The triangles were made using the Delaunay triangulation method, so that all points are related using their two nearest neighbors to form triangles [26]. Along these lines, a bathymetric TIN map was made for the reservoir. The current volume of the reservoir and reservoir surface area at full supply level was then processed using the TIN model and the surface volume tool in ArcGIS. Using the bathymetry map, the volume of the reservoir lost due to sedimentation were calculated based on the preceding equations:

$$\mathrm{R}\mathrm{V}\mathrm{L}=\mathrm{O}\mathrm{V}\mathrm{R}-\mathrm{P}\mathrm{V}\mathrm{R}$$ (2)

where, RVL is the Reservoir Volume Lost through sedimentation (m³), OVR represent the Original Volume of the Reservoir as engineered (m³), and PVR relates to the Present Volume of the Reservoir (m³) [27].

$$\mathrm{A}\mathrm{A}\mathrm{R}\mathrm{S}=\mathrm{R}\mathrm{V}\mathrm{L}/\mathrm{A}\mathrm{g}\mathrm{R}$$ (3)

where AARS represents Average Annual Rate of Sedimentation (m³/yr), RVL relates to Reservoir Volume Lost (m³), and AgR represents Age of the Reservoir (yr) [28];

Average Annual Sedimentation percentage loss estimation was based on the expression:

$$\mathrm{A}\mathrm{A}\mathrm{R}\mathrm{S}=\mathrm{R}\mathrm{V}\mathrm{L}/\mathrm{A}\mathrm{g}\mathrm{R}\mathrm{x}100$$ (4)

where, AARS represents Average Annual Rate of Sedimentation (%/yr), RVL relates to Reservoir Volume Lost (m³), and AgR represents Age of the Reservoir (yr) [29].

The study additionally assessed the useful life of the Amissa Reservoir with the expression:

$$\mathrm{U}\mathrm{L}=0.5(\mathrm{O}\mathrm{V}\mathrm{R}-\mathrm{R}\mathrm{V}\mathrm{L}/\mathrm{A}\mathrm{A}\mathrm{R}\mathrm{S})$$ (5)

where UL represents Useful Life of the Reservoir in years when the original volume as engineered will lessen to half (50%), OVR is the Original Volume of the Reservoir (m³), RVL is the Reservoir Volume Lost (m³), and AARS is the Average Annual Rate of Sedimentation (m³/yr) [30].

3.1. Reservoir Volume Estimation and bathymetric TIN map generation

The processed bathymetric study vector data (X, Y, Z) was informed into the ArcGIS 10.7 in ESRI shape format. Through GIS kriging technique, the depth points were interpolated where raster map was created out of it. In addition, using the 3D Analyst tool, a TIN was created out of the data informed in the GIS database. These data points serve as the corners of TIN, which is a collection of triangles. The Delaunay triangulation technique was used to produce the triangles and subsequently, the output file was a bathymetric TIN map produced for the reservoir.

The present volume of the reservoir and its surface area at its maximum height were modeled using the Surface volume tool in 3D Analyst Tool in ArcGIS. The present reservoir volume and its surface area as obtained from the bathymetry and the original reservoir volume as engineered was obtained from GIDA and the present surface area of the reservoir were used to compute the reservoir volume lost using the expressions as explained in the methods above (equations (2), (3), (4), (5))). The reservoir volume lost represents the volume of sediments accumulated beneath.

4. Results and discussions

4.1. Performance of the SWAT model after calibration, and validation

Model calibration and validation was performed for 1992–2010 and 2017–2019 respectively. For the calibration period, R² and NSE were 0.89 and 0.88, respectively (Fig. 7a). Simulated and observed stream flows showed a very good correlation. Similarly, for the validation period, R² and NSE values were respectively 0.84 and 0.89 (Fig. 7b), also indicating a good correlation between simulated and observed stream flows. These calibration and validation correlations are graphically represented in the hydrograph in Fig. 8 and the performance statistics of the model tabled in Table 1. In general, most parameters used to assess the performance statistics of the model were within the recommended ranges (Table 1). They were in concurrence with those reported by Gyamfi et al. [23] and Abbaspour et al. [32]. A PBIAS of calibration and validation values of 8.66% and 15.82% were obtained. As indicated by the PBIAS results, the calibrations were strongly agreed upon, while the simulations were satisfactory. This however is in sharp contradiction with the findings of Arthur et al. [33] in a study on river Pra in Ghana.

Fig. 7.

a: Calibration

Fig. 7b: Validation.

Fig. 8.

Calibration and Validation hydrograph.

Table 1.

SWAT model performance.

Criteria	Calibration	Validation	Performance range	Performance rating
NSE	0.88	0.84	0.75 >NSE ≤1.00	Very Good
R2	0.89	0.89	0.75 ≤ R2 ≤ 1.00	Very Good
RSR	0.35	0.40	0.00 ≤ RSR ≤0.50	Very Good
PBIAS	−8.66	15.82	PBIAS <10, ±15 ≤ PBIAS< ± 25	Very Good, Satisfactory

4.2. Fitted parameter values and sensitivity analysis

The Table 2 gives global sensitivity parameters with their corresponding p-values and t-stats. Global sensitivity analysis emphasizes the importance of high absolute values of t-stats and small p-values [32,33]. At 5% significance level, GWQMN (t-stat = 11.77; p = 0), TMPMN (t-stat = −2.90, p = 0), and TMPMX (t-stat = −3.60, p = 0) exhibited the highest levels of sensitivity for calibration.

Table 2.

SWAT-CUP global sensitivity after final calibration.

Parameter	t-stat	p-value
4:V__GWQMN.gw	11.77	0
3:R__TMPMN(..).wgn	−2.90	0.004
2:R__TMPMX (..).wgn	−3.60	0.0004

Most sensitive parameters with their t-stats and p-values

4.3. Impacts of land use land cover variations on sediment yield in the river basin

Table 3 contains LULC variation and annual sediment yields based on simulated sediment yields from 1990 to 2020. Land use classes across board experienced changes. However, among the five land use classes, open forests and farmlands experienced the greatest changes. In all the years under consideration, built-up areas and farmlands continuously increased, while open forests decreased. Closed forest decreased by 2.52% from 1991 to 2002, it however increased subsequently from 2002 to 2015, 2015 to 2020 and 1991 to 2020 by 1.79%, 0.91%, and 0.18% respectively. Similarly, open forest increased by 1.06% from 1991 to 2002, but later declined by 19.17%, 21.33% and 39.44% from 2002 to 2015, 2015–2020 and 1991–2020 respectively for the entire study period. Conversely, farmlands expanded by 0.19%, 8.80%, 25.93% and 34.82% between the periods 1991–2002, 2002–2015, 2015–2020, and 1991–2020 respectively. Generally, open forest experienced the greatest decline in land use whiles farmland significantly increased throughout the entire study period.

Table 3.

Annual land-use and sediment yield dynamics in river Amissa basin.

TIME PERIOD	LULC Change (%)					Sediment Yield (t/km2.a)
	FRSE	FRSD	URBN	WATR	AGRL
1991	11.49	76.55	8.58	0.49	2.90	983.44
2002	8.97	77.60	10.03	0.32	3.09	5737.65
2015	10.76	58.43	17.77	1.16	11.88	7123.03
2020	11.67	37.11	12.48	0.94	37.81	11246.13
1991–2002	−2.52	1.06	1.45	−0.17	0.19	+4754.21
2002–2015	1.79	−19.17	7.74	0.84	8.80	+1385.38
2015–2020	0.91	−21.33	−5.29	−0.22	25.93	+4123.11
1991–2020	0.18	−39.44	3.90	0.45	34.92	+10262.69

(−/+ sign respectively shows decrease and increase in land use type/sediment yield).

* FRSE-Closed Forest; FRSD-Open Forest; URBN-Built-Up; WATR-Water Body; AGRL-Farmlands.

Simulations of sediment yield for the years 1991, 2002, 2015, and 2020 were performed using the SWAT hydrological model (Table 3). There were noticeable differences in the spatial distribution of sediment yield between the 13 sub-basins under similar climatic circumstances however with various land use scenarios. Different sub-basins had significantly different sediment yields (Fig. 9). The northern and southern sections of the catchment had higher sediment yields than the central section. These changes were generally because of the variations in predominant land uses (i.e., open forest, built-up areas, and farmland). In the north, for instance, there was a very pronounced conversion of open forest to farmland. For 1991, 2002, 2015, and 2020, sub-basins 1, 2, and 8 produced the highest sediment yield on a sub-basin basis.

Fig. 9.

Spatial distribution of total sediment for 1991, 2002, 2015 and 2020.

Similarly, high sediment yields in 1991, 2002, 2015 and 2020 were located in sub-basins 8, 2, 1; 2, 1, 8; 2, 1, 6; and 8, 2 and 13 respectively. According to the study, land use patterns had the greatest impact on sediment yield during 1991, 2002, 2015 and 2020. The sediment yield within the drainage basin was 983.44 tons per kilometer square for 1991, 5737.65 tons per kilometer square for 2002, and 7123.03 tons per kilometer square for 2015, and 11246.13 tons per kilometer square for 2020. Overall, the basin produced 10262.69 tons per kilometer square of sediment from 1991 to 2020 of the entire study periods.

4.4. Bathymetry results and discussions

After the bathymetry and ArcGIS assessments, the following outcomes were realized from the calculations using equations (2), (3), (4), (5)).

The reservoir volume (storage capacity) loss due to sedimentation was 1,438,940 m³. This signifies 25% loss of its volume to sediment in the reservoir. The Average Annual Rate of Sedimentation was 31,281.3 m³/yr and the percentage Average Annual Sedimentation Rate was 0.54%. These discoveries affirm other comparative investigations by Adwubi et al. [9] where 4 more modest reservoirs in the Upper East Region of Ghana lost volume quickly at an average rate of 1.75% annually and Abubakari [36] on the Tono at 1.74% annually. Similarly, Ceylan & Ekizogllu [37] findings at a rate of 0.51% annually in Turkey and Gomez-Fragoso [38] on the Lago La Plata Reservoir in Puerto Rico was likewise reported to have lost its volume by 0.56%.

It was also established from the study that the maximum depth of the reservoir lost to sediments was 1.4 m. Assuming the sediment distribution at the bottom of the reservoir is uniform, the reservoir loses its depth at a rate of 3.04 cm/yr. The reservoir depth lost in the current study is in sharp contrast with the findings of Rakhmatullaev et al. [39] who recorded a higher depth loss of 25.25 cm/yr in the Akdarya reservoir in Uzbekistan in 21 years of its existence and furthermore far lower than the 7.57 m (37.85 cm/yr) depth loss of the Jibia reservoir in Nigeria within a span of 20 years of operation [35]. The estimated useful life of the reservoir was 69 years.

4.5. Limitations

Generally, the applicability and limits of the direct and indirect methodologies for the evaluation of erosion are myriad; particularly, there is no approach or method that can be considered to be totally reliable if used individually; direct methods such as bathymetric analyses inside reservoirs can be strongly influenced by the presence of dredging/desilting operations carried out over time and almost never quantified; and the hydrological model (SWAT) method can certainly provide fascinating results only if calibrated and validated with direct measurements (flow data) [40]. Up until when the reservoir was constructed from 1974 to 1977 that the hydraulic parameters of the reservoir were taken, there hasn't been any major study carried on the reservoir to ascertain the authenticity of the volume of the reservoir. For that matter, if there is any error in the information provided by the Ghana Irrigation Development Authority (GIDA) for the current study, it will directly affect the outcome of the study. However, since GIDA is an institutionalized body, the data provided can be trusted.

Reservoir designers have also expressed worry about the lack of reliable data sources on sediment yield. There is no monitoring of sediment yield on the reservoir to serve as the baseline data. This in essence makes the results of the bathymetric survey the only existing survey on sedimentation on the reservoir. This inadequacy of data on the reservoir is largely because of the costs related to bathymetric studies.

Similarly, SWAT is a data driven hydrological model. Hence, poor data quality and quantity influences its quality of output. The quantity of flow data used for the calibration (1992–2010) was enough. However due to the poor data management, the flow data used for the validation was relatively not as large as that of the calibration (2017–2019). This however was overcomed by the use of the bathymetry study as an alternative technique in sedimentation assessment.

5. Conclusion

Water is life and there is no direct replacement for water after it has been depleted. Again, one natural resource which was seen previously as a free good with unlimited quantity is water resource. However, the availability of water is now a thing of the past due to rapid population increase, urbanization and diversified use such as fisheries, irrigation, industrial processes, hydropower generation and ecosystem support services. One of Ghana's water-related problems is reservoir sedimentation which reduces water quantity. It is therefore of much concern to ensure sustainable management of water resources. The SWAT hydrological model was used to simulate the spatial distribution of sediment within the river Amissa basin and a bathymetry study was conducted on the reservoir to estimate the quantity of sediment accumulated beneath the reservoir.

After a successful bathymetric survey on the reservoir and hydrological simulation within the entire river Amissa basin the following findings were realized:

The bathymetric survey estimated the current volume of the reservoir to be 4,321,060 m³ (4.3 M m³). Also, the reservoir's volume (storage capacity) lost due to sedimentation was 1,438,940 m³ (1.4 M m³) representing 25% of the reservoir volume, whiles the Average Annual Rate of Sedimentation was 31,281 m³/yr and reservoir's useful life was calculated to be 69 years.

On the other hand, the SWAT hydrological model performance was very good with NSE for calibration and validation as 0.88 and 0.84 respectively. Generally, for the entire 29-years period under study, simulated sediment yield increased by 10,263 tons per kilometer square.

The fact that the reservoir lost 25% of its volume to sediment is an indication of high rate of soil erosion caused by anthropogenic (agricultural) activities within the river basin. Most of the watershed is cultivated and the buffer zone around the reservoir is not adhered to.

Decerning from the results of the bathymetry study and the performance of the SWAT hydrological model, it can be concluded that the methods adopted by the study was successful and its applicable in similar studies.

Author contribution statement

Essel-Yorke K. A.: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Anim M.: Performed the experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Nyarko B. K.: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability statement

Data will be made available on request.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.