1 Sediment.Cdr

Sediment Discharges from Ghanaian Rivers into the Sea

S. A. Akrasi CSIR-Water Research Institute, P. O. Box M.32, Accra, Ghana Corresponding author; E-mail: [email protected]

Abstract Information on sediment yield of a river basin is an important requirement for water resources development and management. In Ghana, data on suspended sediment yield are limited owing to lack of logistic support for systematic sediment sampling activities. The paper presents the results of a study, using measurements of suspended sediment transport for 21 monitoring stations in southern Ghana to develop simple predictive models for suspended sediment yields of catchments for which no sediment measurements had been undertaken. Regression analysis was used to establish a relationship between specific suspended sediment yield and both the mean annual runoff, and the drainage basin area. One of the prediction models was used to estimate the sediment loads of the southern Ghana rivers and total suspended sediment discharges into the sea, in addition to specific suspended sediment yields from the drainage basins. The specific suspended sediment yield for the south-western and coastal basin systems ranged between 11 and 50 t km–2 year–1. The annual sediment discharges into the sea by the rivers ranged between 15,000 and 1.2 × 106 t, and total yearly sediment discharge into the sea by Ghanaian rivers is estimated to be 2.4 × 106 t.

Introduction Sediment yield data are very limited in the Sediment particles are transported through river basin systems of Ghana owing river system as a result of runoff from rainfall primarily to lack of logistics and the high cost through the processes of sheet, rill and gully involved in sediment data collection. Simple erosion. The eroded sediment particles are sediment yield predictive models have been eventually deposited in a flood plain, developed, using the suspended sediment reservoir or sea. discharge measurements that has been Many activities related to the conserva- carried out in the basins, to be used to tion, development and utilization of land, estimate sediment yields of ungauged river mineral and water resources either increase basins in Ghana. Sediment yield predictive or decrease the rate of sediment movement, models have been developed for the south- thereby, causing a variety of sediment related western river basins (Amisigo & Akrasi, problems. A change in runoff regime from a 2000), the Volta basin system (Akrasi, 2005) drainage basin, for example, may and the Pra river basin (Akrasi & Ansa- concentrate or disperse sediments in the Asare, 2008) using runoff and catchment stream channel and, in turn, affect the flow area. capacity of the stream. Because of the Modelling is an important research tool. complex interrelationship that affects Planning and management of soil and water erosion, transport and deposition, resources of a nation require knowledge of knowledge of climate, physical attributes of factors that cause loss of soil and water and drainage basins, hydraulic and hydrologic those that tend to minimise such losses. characteristics of stream flow, and Erosion and sedimentation by water involve quantitative and qualitative aspect of the processes of detachment, transport and sediment are all required to solve sediment deposition of soil particles. Sediment yield problems. on a Ghanaian catchment includes soil

2 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 3 erosion from agricultural fields, hill slopes The coastal basin system (dry equatorial) LEGEND and settlements. Factors that influence fluvial is relatively dry, with annual rainfall values N International Boundary sediment loads would, therefore, include the ranging between 740 and 890 mm (Dickson Catchment Boundary various land uses and cultural practices that & Benneh, 1980). The vegetation within the River / Stream GHANA take place in these areas. Since it is not basin falls into two zones: semi-deciduous Sediment Station practicable to undertake field measurements forest in the northern part of the basin, and National Capital Regional Capital of the effects of these practices on sediment coastal savanna and grassland along the THE STUDY AREAS yield, predictive models are used. Good coast. Forest and coastal savanna ochrosols, 0 30 60 90Km. sediment yield models could provide good lateritic sandy soils, tropical black clays, estimates of the rates at which drainage sandy soils, tropical grey earth and coastal basins are contributing sediment to the rivers. sands constitute soils of the basin. The They could also be used to predict the effects underlying rock formation is also pre- E

of various combinations of land use activities Cambrian. Rivers draining the coastal basin R I and cultural practices on erosion and O system are Amisa, Nakwa, Ayensu, Densu T V I ’ SUNYANI O sediment yield on drainage basins. and Tordzie. D G

The study reported in this paper is a The study comprised three major O collation of sediment yield data in south- components. First, the available suspended western and coastal basin systems. The sediment measurements for 21 monitoring Hwidiem KUMASI objective is to develop simple models to stations in the south-western and coastal river Mfensi estimate suspended sediment yields of the systems were collated and used to estimate Konongo HO catchments within the south-western and the mean annual suspended sediment yields Bunso Kpetoe coastal basins, and to use the models to Anwia-Nkwanta of the sub-basins controlled by the stations. Ankwaso Tordzienu Oda estimate the total sediment loads being Second, the data were used to develop simple Assin Praso KOFORIDUA discharged into the sea. empirical prediction models for estimating Jomoro the suspended sediment yield of other Twifu Praso Oketsew

Materials and methods ungauged basins within the south-western E

T Prestea Ochiso Okyereko Study area and coastal river systems. Third, the derived O ACCRA

Materials and methods ungauged basins within the south-western E

T Prestea Ochiso Okyereko Study area and coastal river systems. Third, the derived O ACCRA

C Ekotsi Bonsoso The south-western basin system is humid sediment prediction models were used to Mankessim with annual rainfall in the range of 1500-2000 estimate the sediment yields of the major Beposo CAPE COAST mm, with vegetation consisting of tropical rivers entering the sea and the total sediment Ewusijo rain forest and moist semi-deciduous forest. It discharge into the sea. SEKONDI is located at the extreme south-western corner The available sediment data comprised of the country. Soils of the tropical rainforest measurements of the suspended sediment Fig. 1. The map of the south-western and coastal basin system of Ghana showing sediment sampling sites are forest oxysols, which is acidic, and that of concentrations of water samples, collected the semi-deciduous forest is forest ochrosols, without the use of depth-integrating suspended sediment sample collected from decreasing as one move toward the water which is alkaline and well drained. The basin suspended sediment sampler. Instead, the water surface is likely to underestimate surface. is underlain by pre-Cambrian formation, and improvised point sampler, similar to the the mean concentration in the vertical or the In the absence of specialist depth- is classified into Birrimain and Tarkwaian dipping of ordinary sample bottle, was used overall cross-section (Demmak, 1976; integrating or point-integrating samplers, rocks. Major rivers draining the south- to collect samples from the surface (Akrasi & Rooseboom & Annandale, 1981). This is which could be used to derive correction western basin system are Bia (enters the sea in Amisigo, 1993; Akrasi, 2008). It is well because the sediment concentrations are factors for taking into account the degree of La Cote d’Ivoire), Tano, Ankobra and Pra. known that the concentration obtained for a usually greatest near the river channel bed, underestimation associated with samples 4 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 5 collected at the surface by dipping, reference comprised 15–66 measurements, undertaken The existing information, however, shows significance of the resulting prediction was made to other studies that have attempted on infrequent samples collected at different the suspended sediment regimes of the study models were assessed using the coefficient of to establish such correction factors. times of the year over a representative range catchments, and are characterized by determination and by testing the significance Rooseboom & Annandale (1981) reported the of water discharges. The sampling frequency extended periods of high flow with elevated of the regression coefficients, using the F-test results of a study undertaken in South Africa of these data was insufficient to interpolate suspended sediment concentrations during (Yamane, 1970). aimed at comparing the magnitude of the the sediment concentration record for the the wet season, rather than a flashy response suspended sediment concentrations period of sampling. Thus, the rating curve characterized by short-lived flood events. Results and discussions The mean annual specific suspended associated with surface dip samples with the technique (Walling, 1977) was used to Table 1 gives the summary of rating sediment yields of the 21 stations were mean values for vertically derived samples estimate the mean annual suspended relationships developed from the suspended derived by dividing the mean annual load by sediment concentration data collected for the collected with a more sophisticated sediment yield at the sampling sites. For each the catchment area. 21 monitoring stations. Each relationship Turbisonde sampler. In the study, the site, the measured values of suspended concentrations associated with the surface The values of specific suspended sediment was statistically tested to be significant at 5% sediment concentration (corrected) and 2 samples were approximately 25% less than yield obtained for the 21 measuring stations, level. Coefficients of determination (r ), discharge were used to establish a rating the mean value for the vertical samples. using the method described above, were associated with the relationships, range relationship: subsequently used to develop simple between 0.81 and 0.91. The relatively low A similar study, undertaken by Demmak n Qs = kQ (1) magnitude of the rating exponents (1976) at the Sidi Bel Alter gauging station on w empirical prediction models for estimating ”1 where Qs = sediment discharge (t day ), Qw = the specific suspended sediment yields of (0.64–1.50) reflects the insensitivity of the Oued Cheift in Algeria, compared the 3 ”1 water discharge (m s ), n = exponent and k = suspended sediment concentration to instantaneous suspended sediment load, catchments within the south-western and constant. increases in discharges. The exponent values calculated using the concentration value coastal river system, for which no sediment As these rating relationships were measurements were available. These simple presented in Table 1 are substantially lower obtained for surface-suspended sediment than the typical values between 2.0 and 3.0, samples, collected near the edge of the established using least squares regression on empirical prediction models were derived by the log-transformed values of sediment using multiple regression analysis to relate cited by Gregory & Walling (1973). These channel with the load estimate derived, using low exponent values suggest that the rivers discharge and water discharge, their use in the available values of specific suspended detailed sampling in the cross-section with remain turbid over a wide range of flows, and sediment yields to a range of variables, the Neyrpic Turbisonde sampler method, estimating the sediment loads associated the suspended sediment concentrations representing catchment characteristics likely concluding that the former underestimated with individual discharge values is likely to remain relatively high during low flows. to influence the spatial variability of specific the load by approximately 25%. result in a bias towards underestimation The situation also could reflect the suspended sediment yield within the south- The precise magnitude of any correction (Ferguson, 1986). As a result, the bias relatively low gradient of the river systems western and coastal river system. factor can be expected to vary according to correction factor proposed by Ferguson and the importance of localized storms in In view of the small number of sediment both the grain size composition of the (1986) was applied to the resulting load yield values available to parameterize the generating local sediment inputs to the river estimates. These rating curves were used in sediment and the hydraulic conditions within model, it was important to keep the number of systems, which are essentially unrelated to conjunction with the longer-term flow the channel (e.g. velocity, roughness and independent variables as small as possible, in the overall water discharge regime. Further duration curves for the measuring stations in turbulence). In the absence of site-specific order to increase the statistical significance of work is required to explore the localized information, however, a correction factor of order to estimate the mean annual sediment the resulting regression equation. Both the storm, which is a key feature of the sediment 25% was used and the measured values of the yields for these stations (Miller, 1951; dependent and independent variables were response of the south-western and coastal suspended sediment concentrations obtained Walling, 1977). These flow duration curves log transformed, with the resulting prediction river systems in more detail. In the case of the for surface dip samples collected near the were based on mean daily flows. The use of equation: Ofin river at Dunkwa, the exponent of the channel edge were increased by 25% to daily mean flow data to estimate the sediment a b z sediment load discharge relationship is less Sy = K U1 U2 ……..Un (2) provide mean concentration values for the loads of the 21 sampling stations is unlikely where Sy = the mean annual specific than 1.0, indicating the concentrations tend to cross-section area. to introduce significant errors, as most of the suspended sediment yield (t km”2 year”1), K = decrease when the discharges increase. The The suspended sediment concentration catchments associated with the measuring the regression constant, and U1, U2...... Un = decrease in concentrations when discharges data for 21 sediment sampling stations stations are relatively large. the independent variables. The statistical increase is further emphasized by the 4 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 5 collected at the surface by dipping, reference comprised 15–66 measurements, undertaken The existing information, however, shows significance of the resulting prediction was made to other studies that have attempted on infrequent samples collected at different the suspended sediment regimes of the study models were assessed using the coefficient of to establish such correction factors. times of the year over a representative range catchments, and are characterized by determination and by testing the significance Rooseboom & Annandale (1981) reported the of water discharges. The sampling frequency extended periods of high flow with elevated of the regression coefficients, using the F-test results of a study undertaken in South Africa of these data was insufficient to interpolate suspended sediment concentrations during (Yamane, 1970). aimed at comparing the magnitude of the the sediment concentration record for the the wet season, rather than a flashy response suspended sediment concentrations period of sampling. Thus, the rating curve characterized by short-lived flood events. Results and discussions The mean annual specific suspended associated with surface dip samples with the technique (Walling, 1977) was used to Table 1 gives the summary of rating sediment yields of the 21 stations were mean values for vertically derived samples estimate the mean annual suspended relationships developed from the suspended derived by dividing the mean annual load by sediment concentration data collected for the collected with a more sophisticated sediment yield at the sampling sites. For each the catchment area. 21 monitoring stations. Each relationship Turbisonde sampler. In the study, the site, the measured values of suspended concentrations associated with the surface The values of specific suspended sediment was statistically tested to be significant at 5% sediment concentration (corrected) and 2 samples were approximately 25% less than yield obtained for the 21 measuring stations, level. Coefficients of determination (r ), discharge were used to establish a rating the mean value for the vertical samples. using the method described above, were associated with the relationships, range relationship: subsequently used to develop simple between 0.81 and 0.91. The relatively low A similar study, undertaken by Demmak n Qs = kQ (1) magnitude of the rating exponents (1976) at the Sidi Bel Alter gauging station on w empirical prediction models for estimating ”1 where Qs = sediment discharge (t day ), Qw = the specific suspended sediment yields of (0.64–1.50) reflects the insensitivity of the Oued Cheift in Algeria, compared the 3 ”1 water discharge (m s ), n = exponent and k = suspended sediment concentration to instantaneous suspended sediment load, catchments within the south-western and constant. increases in discharges. The exponent values calculated using the concentration value coastal river system, for which no sediment As these rating relationships were measurements were available. These simple presented in Table 1 are substantially lower obtained for surface-suspended sediment than the typical values between 2.0 and 3.0, samples, collected near the edge of the established using least squares regression on empirical prediction models were derived by the log-transformed values of sediment using multiple regression analysis to relate cited by Gregory & Walling (1973). These channel with the load estimate derived, using low exponent values suggest that the rivers discharge and water discharge, their use in the available values of specific suspended detailed sampling in the cross-section with remain turbid over a wide range of flows, and sediment yields to a range of variables, the Neyrpic Turbisonde sampler method, estimating the sediment loads associated the suspended sediment concentrations representing catchment characteristics likely concluding that the former underestimated with individual discharge values is likely to remain relatively high during low flows. to influence the spatial variability of specific the load by approximately 25%. result in a bias towards underestimation The situation also could reflect the suspended sediment yield within the south- The precise magnitude of any correction (Ferguson, 1986). As a result, the bias relatively low gradient of the river systems western and coastal river system. factor can be expected to vary according to correction factor proposed by Ferguson and the importance of localized storms in In view of the small number of sediment both the grain size composition of the (1986) was applied to the resulting load yield values available to parameterize the generating local sediment inputs to the river estimates. These rating curves were used in sediment and the hydraulic conditions within model, it was important to keep the number of systems, which are essentially unrelated to conjunction with the longer-term flow the channel (e.g. velocity, roughness and independent variables as small as possible, in the overall water discharge regime. Further duration curves for the measuring stations in turbulence). In the absence of site-specific order to increase the statistical significance of work is required to explore the localized information, however, a correction factor of order to estimate the mean annual sediment the resulting regression equation. Both the storm, which is a key feature of the sediment 25% was used and the measured values of the yields for these stations (Miller, 1951; dependent and independent variables were response of the south-western and coastal suspended sediment concentrations obtained Walling, 1977). These flow duration curves log transformed, with the resulting prediction river systems in more detail. In the case of the for surface dip samples collected near the were based on mean daily flows. The use of equation: Ofin river at Dunkwa, the exponent of the channel edge were increased by 25% to daily mean flow data to estimate the sediment a b z sediment load discharge relationship is less Sy = K U1 U2 ……..Un (2) provide mean concentration values for the loads of the 21 sampling stations is unlikely where Sy = the mean annual specific than 1.0, indicating the concentrations tend to cross-section area. to introduce significant errors, as most of the suspended sediment yield (t km”2 year”1), K = decrease when the discharges increase. The The suspended sediment concentration catchments associated with the measuring the regression constant, and U1, U2...... Un = decrease in concentrations when discharges data for 21 sediment sampling stations stations are relatively large. the independent variables. The statistical increase is further emphasized by the 6 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 7

TABLE 1 Parameters for Equation 2 mean annual runoff in the models indicates predicting the specific suspended sediment that specific suspended sediment yields yields of a river in the south-western and Station River k n r2 increase in areas with higher mean annual coastal basin system is further demonstrated with the plots of predicted versus observed Hwidiem Tano 2.540 1.500 0.87 runoffs, in turn, reflecting the increased Jomuro Tano 4.195 1.211 0.93 erosion and suspended sediment transport specific suspended sediment yield for the Ankwaso Ankobra 6.368 1.054 0.88 associated with the increased runoff. catchments above the 21 gauging stations, Prestea Ankobra 5.322 1.003 0.85 Although the relationship between the for which sediment yield data were Bonsaso Bonsa 2.805 1.143 0.92 specific suspended sediment yield and the measured, as presented in Fig. 2–5. All the Ewusijo Butre 1.072 1.197 0.85 catchment area has traditionally been viewed observed specific suspended sediment yield Mfensi Ofin 1.363 1.328 0.81 values are closely matched by the predicted Anwia-Nkwanta Ofin 5.514 1.135 0.86 as being negative (Walling & Webb 1996), Dunkwa Ofin 138.400 0.640 0.45 the positive relationship obtained in the study values obtained with the models (1–4). Thus, Bunso Birim 1.423 1.305 0.84 is consistent with a number of studies the models (1–4) can be used to provide an Oda Birim 4.086 1.105 0.90 (Dedkov, 2004) that have demonstrated that a estimate of the specific suspended sediment Konongo Anum 3.132 1.128 0.91 positive relationship can occur when channel yield for the respective river basins within the Assin-Praso Pra 1.083 1.331 0.90 south-western and coastal river basin Twifu-Praso Pra 4.473 1.137 0.86 erosion is an important sediment source. The systems of Ghana, for which suspended Beposo Pra 5.753 1.099 0.89 positive relationship reported by Dedkov sediment data are lacking, provided Mankesim Amisa 6.250 1.100 0.89 (2004) has been explained in terms of the Ekotsi Nakwa 6.606 1.110 0.86 information on the catchment area and the increased discharges and, therefore, the Ochiso Nakwa 2.620 1.468 0.85 mean annual runoff is available. Okyereko Ayensu 9.412 1.033 0.84 energy available for channel erosion as the Model 1 was developed using all the data Oketsew Ayensu 3.333 1.205 0.85 catchment area increases. Tordzienu Tordzie 6.635 1.101 0.79 in the south-western and coastal basin The negative exponent is commonly Kpetoe Tordzie 3.499 1.284 0.84 systems, including the Pra river basin, and it explained in terms of the increasing is the predictive model which was used to widespread evidence of dredging activities The corrected coefficients of multiple opportunity for sediment deposition and 2 determine the suspended sediment discharge for gold in the Ofin river at Dunkwa. determination, r ranges from 0.79 to 0.91. storage, both on the slopes of a catchment and into the sea. Though it is the model that The multiple regression analysis of the This is an indication that the variables, runoff within the channel system, as the catchment predicts the least, it is the only model that data sets gave the relations: and catchment areas accounted for a large area, and travel distances and times increase. covers a wide range of catchment areas. S =0.09Q 0.854 A0.228, r2 = 0.77 (3) y w proportion of the variance of the suspended The positive exponent for the catchment area Model 2 was developed to predict suspended Model 1 is for all the rivers in the basins in the sediment yield. The coefficients 0.858 and term determined in this study probably sediment yield in the south-western and south-western and coastal river systems, 0.228, 1.018 and 0.138, 1.438 and 0.757, and reflects the increased importance of channel coastal basin system, excluding the Pra river including Pra river basin: 1.386 and 0.141, were statistically tested to erosion, in the larger rivers in response to both basin (Bia, Tano, Ankobra, Butre, Amisa, S = 0.03Q 1.018A0.138,r2 = 0.86 (4) w be significant at 5% significance level. The the increased discharges and the more Nakwa, Ayensu, Densu and Tordzie). Model Model 2 is for the rivers in the south-western established models (1–4) can be used to extensive areas of alluvial deposition, which 3 was developed to predict suspended and coastal basin system, excluding Pra river predict specific suspended sediment yield at can be remobilized by channel erosion sediment yield in the coastal river basins basin: -5 1.438 0.757, 2 associated with larger drainage basins. The (Amisa, Nakwa, Ayensu, Densu and S = 3.9x10 Q A r = 0.91 (5) any cross-section within the drainage basins y w with no sediment data. The mean annual mean runoff is a measure of the sediment Tordzie), whereas model 4 was developed to Model 3 is for the rivers in the coastal basin 2 1.139 0.141, 2 transport capability of the runoff, ,and these predict suspended sediment yield in the S = 0.014Q A r = 0.91 (6) runoff (mm) and the catchment area (km ), y w results indicate that transport capacity is a south-western river basins (Bia, Tano, Model 4 is for the rivers in the south-western can also be used to estimate the total very important factor influencing specific Ankobra and Butre). basin, suspended sediment discharge of the river suspended sediment yield in the basin. where Q = mean annual run-off (mm), and A = into the sea. The simple predictive tools afford a means 2 The ability of the simple models (1–4) for catchment area (km ). The positive exponent associated with the of establishing the probable spatial variation 6 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 7

TABLE 1 Parameters for Equation 2 mean annual runoff in the models indicates predicting the specific suspended sediment that specific suspended sediment yields yields of a river in the south-western and Station River k n r2 increase in areas with higher mean annual coastal basin system is further demonstrated with the plots of predicted versus observed Hwidiem Tano 2.540 1.500 0.87 runoffs, in turn, reflecting the increased Jomuro Tano 4.195 1.211 0.93 erosion and suspended sediment transport specific suspended sediment yield for the Ankwaso Ankobra 6.368 1.054 0.88 associated with the increased runoff. catchments above the 21 gauging stations, Prestea Ankobra 5.322 1.003 0.85 Although the relationship between the for which sediment yield data were Bonsaso Bonsa 2.805 1.143 0.92 specific suspended sediment yield and the measured, as presented in Fig. 2–5. All the Ewusijo Butre 1.072 1.197 0.85 catchment area has traditionally been viewed observed specific suspended sediment yield Mfensi Ofin 1.363 1.328 0.81 values are closely matched by the predicted Anwia-Nkwanta Ofin 5.514 1.135 0.86 as being negative (Walling & Webb 1996), Dunkwa Ofin 138.400 0.640 0.45 the positive relationship obtained in the study values obtained with the models (1–4). Thus, Bunso Birim 1.423 1.305 0.84 is consistent with a number of studies the models (1–4) can be used to provide an Oda Birim 4.086 1.105 0.90 (Dedkov, 2004) that have demonstrated that a estimate of the specific suspended sediment Konongo Anum 3.132 1.128 0.91 positive relationship can occur when channel yield for the respective river basins within the Assin-Praso Pra 1.083 1.331 0.90 south-western and coastal river basin Twifu-Praso Pra 4.473 1.137 0.86 erosion is an important sediment source. The systems of Ghana, for which suspended Beposo Pra 5.753 1.099 0.89 positive relationship reported by Dedkov sediment data are lacking, provided Mankesim Amisa 6.250 1.100 0.89 (2004) has been explained in terms of the Ekotsi Nakwa 6.606 1.110 0.86 information on the catchment area and the increased discharges and, therefore, the Ochiso Nakwa 2.620 1.468 0.85 mean annual runoff is available. Okyereko Ayensu 9.412 1.033 0.84 energy available for channel erosion as the Model 1 was developed using all the data Oketsew Ayensu 3.333 1.205 0.85 catchment area increases. Tordzienu Tordzie 6.635 1.101 0.79 in the south-western and coastal basin The negative exponent is commonly Kpetoe Tordzie 3.499 1.284 0.84 systems, including the Pra river basin, and it explained in terms of the increasing is the predictive model which was used to widespread evidence of dredging activities The corrected coefficients of multiple opportunity for sediment deposition and 2 determine the suspended sediment discharge for gold in the Ofin river at Dunkwa. determination, r ranges from 0.79 to 0.91. storage, both on the slopes of a catchment and into the sea. Though it is the model that The multiple regression analysis of the This is an indication that the variables, runoff within the channel system, as the catchment predicts the least, it is the only model that data sets gave the relations: and catchment areas accounted for a large area, and travel distances and times increase. covers a wide range of catchment areas. S =0.09Q 0.854 A0.228, r2 = 0.77 (3) y w proportion of the variance of the suspended The positive exponent for the catchment area Model 2 was developed to predict suspended Model 1 is for all the rivers in the basins in the sediment yield. The coefficients 0.858 and term determined in this study probably sediment yield in the south-western and south-western and coastal river systems, 0.228, 1.018 and 0.138, 1.438 and 0.757, and reflects the increased importance of channel coastal basin system, excluding the Pra river including Pra river basin: 1.386 and 0.141, were statistically tested to erosion, in the larger rivers in response to both basin (Bia, Tano, Ankobra, Butre, Amisa, S = 0.03Q 1.018A0.138,r2 = 0.86 (4) w be significant at 5% significance level. The the increased discharges and the more Nakwa, Ayensu, Densu and Tordzie). Model Model 2 is for the rivers in the south-western established models (1–4) can be used to extensive areas of alluvial deposition, which 3 was developed to predict suspended and coastal basin system, excluding Pra river predict specific suspended sediment yield at can be remobilized by channel erosion sediment yield in the coastal river basins basin: -5 1.438 0.757, 2 associated with larger drainage basins. The (Amisa, Nakwa, Ayensu, Densu and S = 3.9x10 Q A r = 0.91 (5) any cross-section within the drainage basins y w with no sediment data. The mean annual mean runoff is a measure of the sediment Tordzie), whereas model 4 was developed to Model 3 is for the rivers in the coastal basin 2 1.139 0.141, 2 transport capability of the runoff, ,and these predict suspended sediment yield in the S = 0.014Q A r = 0.91 (6) runoff (mm) and the catchment area (km ), y w results indicate that transport capacity is a south-western river basins (Bia, Tano, Model 4 is for the rivers in the south-western can also be used to estimate the total very important factor influencing specific Ankobra and Butre). basin, suspended sediment discharge of the river suspended sediment yield in the basin. where Q = mean annual run-off (mm), and A = into the sea. The simple predictive tools afford a means 2 The ability of the simple models (1–4) for catchment area (km ). The positive exponent associated with the of establishing the probable spatial variation 8 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 9

)

-1

tear

-2

)

-1

year

-2

yiels (t km

sediment yield t (km

Oberved mean annual suspended

Observed mean annual suspended sediment

-2 -1 Predicted mean annual suspended sediment yield (t km year ) Predicted mean annual suspended sediment yield (t km-2 year-1)

Fig. 2. The plot of predicted and observed mean annual suspended sediment yields of Model 1 Fig. 4. The plot of predicted and observed mean annual suspended sediment yields of Model 3

) -1

)

-1

year -2

year

-2

sediment yield (t km (t yield sediment Observed mean annual suspended suspended annual mean Observed

sediment yield (t km

Observed mean annual suspended

Predicted mean annual suspended sediment yield (t km-2 year-1)

)

-1

tear

-2

)

-1

year

-2

yiels (t km

sediment yield t (km

Oberved mean annual suspended

Observed mean annual suspended sediment

-2 -1 Predicted mean annual suspended sediment yield (t km year ) Predicted mean annual suspended sediment yield (t km-2 year-1)

Fig. 2. The plot of predicted and observed mean annual suspended sediment yields of Model 1 Fig. 4. The plot of predicted and observed mean annual suspended sediment yields of Model 3

) -1

)

-1

year -2

year

-2

sediment yield (t km (t yield sediment Observed mean annual suspended suspended annual mean Observed

sediment yield (t km

Observed mean annual suspended

Predicted mean annual suspended sediment yield (t km-2 year-1)

Predicted mean annual suspended sediment yield (t km-2 year-1) Fig. 5. The plot of predicted and observed mean annual suspended sediment yields of Model 4 Fig. 3. The plot of predicted and observed mean annual suspended sediment yields of Model 2 10 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 11 of specific suspended sediment yields across and the associated low-energy conditions and TABLE 3 the south-western and coastal river basins, low efficiency of sediment delivery from the Mean annual suspended sediment load and specific suspended sediment yield for the major rivers in the south-western and coastal river basins and the Volta river basin system thereby, identifying the main areas that catchment surface to, and through, the supply sediment to the sea. The reliability of channel system. The low efficiency of River Drainage Area Mean annual suspended Mean annual suspended the predictor is similar to one developed by (km2) sediment load sediment yield sediment delivery is further emphasized by 6 Hydraulics Research for eastern and southern the widespread evidence of sediment (× 10 t year-1) (t km-2 year-1) Africa using small dams (Lawrence, 2003). production within the basins from the Ankobra 8,366 <1 48.15 The coefficient of determination of the surfaces of feeder roads, human settlements Tano 16,061 <1 24.14 observed versus predicted values was 95%. and illegal mining activities. Large amount of Bia 10,135 <1 25.53 Pra 23,168 1 49.17 Model 3 was used to estimate the mean sediment produced in the basins might be re- Volta † 400,000 17 52.25 annual specific suspended sediment yields of deposited in depressions and floodplains. Black Volta † 142,056 4 28.05 the rivers catchment area. Table 2 presents the The total sediment discharge from Ghanaian White Volta † 106,742 4 32.56 Oti † 75,110 5 63.26 total sediment loads of the major rivers in the rivers into the sea was estimated to be H” 2.4 south-western and coastal basins after a 10% 6 ”1 × 10 t year . a broader context, however, the specific –2 –1 bed load correction factor has been added to km year , obtained for some major rivers in The equivalent estimates of suspended sediment yield of the rivers in Ghana must be the suspended load. The correction factor was various parts of West Africa (Walling, 1984). sediment load and specific suspended interpreted as low when compared with that The total sediment loads carried by the rivers arrived after analysing the suspended sediment yield have been derived for the of the Ganges river (1352.23 t km-2 year-1), into the sea amounted to 2.4 × 106 t per year. sediment concentration and the type of major rivers in the south-western and coastal which is more than an order of magnitude The total sediment load does not include material forming the channel of the river river systems of Ghana (Table 3), togather higher. The estimates of sediment discharge sediment loads from the Densu and Volta (Akrasi, 2008; Tilrem, 1979). with estimates from the Volta river basin. into sea and the specific sediment yield of the rivers. The two rivers have dams across them All the specific sediment yield values are These values indicate that the specific catchment of the rivers in the south-western close to the sea and sediment loads after the low by world standards (Walling & Webb, suspended sediment yield of the Volta river ”2 ”1 and coastal basin systems of Ghana presented dams might be very low due to the significant 1983), with a maximum of 49017 t km year basin is similar to that of the Amazon basin reduction in flow rates. ”2 above relate specifically to the suspended for River Pra and a minimum of 11.01 t km (52.25 t km-2 year-1) but significantly greater Forest reserves, secondary forest, cocoa, ”1 sediment load. year for River Tordzie. These low values than that for the Congo (16.19 t km-2 year-1) coffee and oil palm plantations are the main reflect the low gradients of the river basins -2 -1 and Niger basins (t km year ). Considered in Conclusion land covers of the drainage basins. These Suspended sediment data within the south- vegetation types protect the soil from the western and coastal river basin systems of erosive activity of rainfall that is very high TABLE 2 within the basins. This might be the reason Sediment discharge into the sea Ghana have been analysed, and the yield and for the low sediment yield values. However, loads obtained have been used to develop River Drainage area Sediment yield Sediment discharge there is the need for comprehensive study to 2 -2 -1 simple empirical models predicting specific (km ) (t km year) t year ) confirm this conclusion because there is very suspended sediment yields within the basins. high density of feeder roads in the drainage Bia 10,135 25.53 258,747 All the statistics on the models showed that Tano 16,061 24.14 387,713 basins. This, together with other activities Ankobra 8,366 48.15 402,823 the models are reliable, and the parameters such as over exploitation of timber trees by Butre 422 35.34 14,913 used for the formulation of the models have timber firms and excessive illegal tree felling Pra 23,168 49.17 1,139,171 significant influence on the output functions. Amisa 1,298 27.49 35,682 for lumber, illegal gold and diamond The sediment yields in the south-western winning, and slash and burn by farmers, are Nakwa 1,409 35.85 50,513 -2 –1 Ayensu 1,709 16.75 28,626 and coastal basins (11–50 t km year ) may all major causes of forest degradation, which Tordzie 2,916 11.01 32,105 be considered to be low to moderate. They results in increased soil erosion and sediment are, however, within the range of 3.9–85.0 t production. 10 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 11 of specific suspended sediment yields across and the associated low-energy conditions and TABLE 3 the south-western and coastal river basins, low efficiency of sediment delivery from the Mean annual suspended sediment load and specific suspended sediment yield for the major rivers in the south-western and coastal river basins and the Volta river basin system thereby, identifying the main areas that catchment surface to, and through, the supply sediment to the sea. The reliability of channel system. The low efficiency of River Drainage Area Mean annual suspended Mean annual suspended the predictor is similar to one developed by (km2) sediment load sediment yield sediment delivery is further emphasized by 6 Hydraulics Research for eastern and southern the widespread evidence of sediment (× 10 t year-1) (t km-2 year-1) Africa using small dams (Lawrence, 2003). production within the basins from the Ankobra 8,366 <1 48.15 The coefficient of determination of the surfaces of feeder roads, human settlements Tano 16,061 <1 24.14 observed versus predicted values was 95%. and illegal mining activities. Large amount of Bia 10,135 <1 25.53 Pra 23,168 1 49.17 Model 3 was used to estimate the mean sediment produced in the basins might be re- Volta † 400,000 17 52.25 annual specific suspended sediment yields of deposited in depressions and floodplains. Black Volta † 142,056 4 28.05 the rivers catchment area. Table 2 presents the The total sediment discharge from Ghanaian White Volta † 106,742 4 32.56 Oti † 75,110 5 63.26 total sediment loads of the major rivers in the rivers into the sea was estimated to be H” 2.4 south-western and coastal basins after a 10% 6 ”1 × 10 t year . a broader context, however, the specific –2 –1 bed load correction factor has been added to km year , obtained for some major rivers in The equivalent estimates of suspended sediment yield of the rivers in Ghana must be the suspended load. The correction factor was various parts of West Africa (Walling, 1984). sediment load and specific suspended interpreted as low when compared with that The total sediment loads carried by the rivers arrived after analysing the suspended sediment yield have been derived for the of the Ganges river (1352.23 t km-2 year-1), into the sea amounted to 2.4 × 106 t per year. sediment concentration and the type of major rivers in the south-western and coastal which is more than an order of magnitude The total sediment load does not include material forming the channel of the river river systems of Ghana (Table 3), togather higher. The estimates of sediment discharge sediment loads from the Densu and Volta (Akrasi, 2008; Tilrem, 1979). with estimates from the Volta river basin. into sea and the specific sediment yield of the rivers. The two rivers have dams across them All the specific sediment yield values are These values indicate that the specific catchment of the rivers in the south-western close to the sea and sediment loads after the low by world standards (Walling & Webb, suspended sediment yield of the Volta river ”2 ”1 and coastal basin systems of Ghana presented dams might be very low due to the significant 1983), with a maximum of 49017 t km year basin is similar to that of the Amazon basin reduction in flow rates. ”2 above relate specifically to the suspended for River Pra and a minimum of 11.01 t km (52.25 t km-2 year-1) but significantly greater Forest reserves, secondary forest, cocoa, ”1 sediment load. year for River Tordzie. These low values than that for the Congo (16.19 t km-2 year-1) coffee and oil palm plantations are the main reflect the low gradients of the river basins -2 -1 and Niger basins (t km year ). Considered in Conclusion land covers of the drainage basins. These Suspended sediment data within the south- vegetation types protect the soil from the western and coastal river basin systems of erosive activity of rainfall that is very high TABLE 2 within the basins. This might be the reason Sediment discharge into the sea Ghana have been analysed, and the yield and for the low sediment yield values. However, loads obtained have been used to develop River Drainage area Sediment yield Sediment discharge there is the need for comprehensive study to 2 -2 -1 simple empirical models predicting specific (km ) (t km year) t year ) confirm this conclusion because there is very suspended sediment yields within the basins. high density of feeder roads in the drainage Bia 10,135 25.53 258,747 All the statistics on the models showed that Tano 16,061 24.14 387,713 basins. This, together with other activities Ankobra 8,366 48.15 402,823 the models are reliable, and the parameters such as over exploitation of timber trees by Butre 422 35.34 14,913 used for the formulation of the models have timber firms and excessive illegal tree felling Pra 23,168 49.17 1,139,171 significant influence on the output functions. Amisa 1,298 27.49 35,682 for lumber, illegal gold and diamond The sediment yields in the south-western winning, and slash and burn by farmers, are Nakwa 1,409 35.85 50,513 -2 –1 Ayensu 1,709 16.75 28,626 and coastal basins (11–50 t km year ) may all major causes of forest degradation, which Tordzie 2,916 11.01 32,105 be considered to be low to moderate. They results in increased soil erosion and sediment are, however, within the range of 3.9–85.0 t production. 12 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 13

Accurate estimates of suspended sediment Amisigo B. A. and Akrasi S. A. (2000). A suspended Walling D. E. (1984). The sediment yields of African Porto Alegre Symp., Dec. 1988) 337–350. IAHS load are essential for sediment budget sediment yield predictive equation for river basins rivers. In Challenges in African Hydrology and Publ. No. 174. investigations, since it may be necessary to in the south western river basin system of Ghana. J. Water Resources (Proc. Harare Symp., July 1984), Walling D. E. and Webb B. W. (1996). Erosion and appl Sci. Techonol. 5(1 & 2): 108–113. sediment yield: Global overviews. In Erosion and compare the sediment yields of sub-basins to 265–283. IAHS Publ. No.144. Carvalho N. D. O. (1988). Sediment yields in the Walling D. E. and Webb B. W. (1983). Patterns of Sediment Yield: Global and Regional Perspectives document the downstream changes in Velhas River Basin (Minas Gerais Brazil). In sediment yield. In Background to Palaeohydrology (D.E. Walling & B. W. Webb, ed.), pp. 3–19. IAHS sediment loads occasioned by conveyance Sediment budgets (M. P. Bordas and D. E. Walling, (K. J. Gregory, ed.), pp. 69–100. Wiley, Chichester. Press, Wallingford, UK. losses, or to investigate temporal trends in eds), pp. 369–375. Proceedings of the Porto Alegre Walling D. E. and Webb B. W. (1988). Reliability of Yamane T. (1970). Statistics: An Introductory sediment yield. Though, transformation bias (Brazil) Symposium on Sediment Budgets 11–15 rating curve estimates of suspended sediment yield: Analysis. International edn. Harper & Row, New corrections were applied in this study, the Dec. 1988, Porto Alegre, Brazil. IAHS Press, some further comments. In Sediment Budgets (Proc. York. Wallingford, UK. load correction estimators have been proved Dedkov A. P. (2004). The relationship between to be unreliable (Walling & Webb, 1988). The sediment yield and drainage basin area. In Sediment provision of reliable load estimates will, in Transport Through the Fluvial System (V. Golosov, many cases, necessitate the use of turbidity V. Belyaev and D. E. Walling, eds), pp. 197–204. monitoring equipment or sensors, frequent IAHS Press, Wallingford, UK. sampling or the use of sampling strategies Demmak A. (1976). Note sur l‘utilization de la “Turbisonade Neyrpic” pour l‘élaboration d’un specifically designed to produce accurate jaugeage de debit solide. Note Technique, No load estimators. Daily suspended sediment 114/SHYL, DEMRH, Algiers. sampling by surface dip by sampling Dickson K. B. and Benneh G. (1980).. A New assistants (Rooseboom & Annandale, 1981) Geography of Ghana. Longman Group Limited, should be employed in Ghana to obtain London. accurate sediment load estimates. Ferguson R. I. (1986). River loads underestimated by rating curves. Wat. Resour.Res. 22: 74–76. Lawrence P. (2003). Predicting sedimentation rates in Acknowledgment African small dams. Newsl. Wat. 16: 9. The author wishes to thank the technical staff Gregory K. J. and Walling D. E. (1973). Drainage of the Surface Water Division of the CSIR- Basin Form and Process. Edward Arnold, London. Water Research Institute, Accra, for their Miller C.R. (1951). Analysis of flow duration sediment cooperation and assistance in assembling the rating curve method of computing sediment yield. U S Bureau of Reclamation, Denver, Colorado, USA. data used in this paper. Rooseboom A. and Annandale G. W. (1981). Techniques applied in determining sedinent loads in References South African rivers. In Erosion and Sediment Akrasi S. A. (2005). The assessment of suspended Transport Measurement. (Proc. Florence Symp. sediment inputs to Volta Lake. Lakes & Reservoirs: June 1981). 219–224. IAHS Publ. No. 133. Research and Management 10: 179–186. Stocking M. (1984). Rates of erosion and sediment Akrasi S. A. (2008). Assessment of sediment transport yield in the African environment In: Challenges in in the south western and coastal river systems of African Hydrology and Water Resources (Proc. Ghana. Water Research Institute, Water Resources Harare Symp., July 1984), 265–283. IAHS Publ. Commission, Accra. No.144. Akrasi S. A. and Ansa-Asare O. D. (2008). Assessing Tilrem Q. A. (1979). Sediment transport in streams, sediment and nutrient transport in the Pra basin of sampling, analysis and computation. Vol. 5 of Ghana. West Afri. J. appl Ecol. 13: 61–73. Manual on Procedures in Operational Hydrology. Akrasi S. A. and Amisigo B. A. (1993). Sediment loads Walling D. E. (1977). Assessing the accuracy of of some major rivers in Ghana. Water Resources suspended sediment rating curves for a small basin. Research Institut, Technical Report. Accra. Wat. Resour. Res. 13: 531–538. 12 West African Journal of Applied Ecology, vol. 18, 2011 Akrasi, S. A.: Sediment discharges from Ghanaian rivers into the sea 13