Ghana J. Sci. 54 (2014), 19 – 32

FLOOD PULSE ALTERATIONS OF SOME RIVER BASINS IN GHANA

F.Y. Logah*, K. Kankam-Yeboah, D. Ofori and P. Gyau-Boakye CSIR- Water Research Institute, P.O. Box M.32, Accra, Ghana. *Corresponding author’s email: [email protected]

Abstract Flood analysis is crucial for designing drainage, managing water quality, assessing the impact on lives and properties among others. The objective of the study was to determine the trends, occurrences and magnitude of floods in Ghana using hydrological data. The study looked at flood phenomenon in Ghana, because flooding has been observed recently as the most common natural hazard in the coun- try. Specifically, the rainfall amounts from selected gauging stations in Ghana and river discharges from the three main river systems namely, Volta (at Bamboi), south western (at Twifu Praso) and the coastal (at Okyereko) river systems in Ghana were considered. Peak streamflows from the se- lected river systems were also analysed. Flood frequencies were derived using the flow duration curve (FDC), and the high flow frequency (return period) determined to support the analysis. The threshold values above which streamflows are peak flow were estimated from the FDC at 90 per cent probabil- ity of non-exceedance for the selected rivers. The results of the analysis showed that increases in the number of occurrence of high streamflows with declining trends of monthly rainfall in the last decade (2000-2010) have been prevalent. This means that rainfall amount is not necessarily the major cause of recent high streamflows (or flood), though there are evidences of direct runoff during rainy season in Ghana. It is, therefore, recommended that as a country, there is the need to improve the drainage systems in the cities, and also provide adequate storage for floods during the rainy seasons. It is fur- ther recommended that flood pulse analysis be carried out on a continuous basis as and when new meteorological and hydrological data is available.

Introduction gion) are most vulnerable (Kankam-Yeboah Floods in Ghana have become a major envi- et al., 2011; Kagblor, unpublished). ronmental problem causing loss of lives and Despite the commitment to physical infra- property and displacement of people. Major structural development such as roads, runoff low-lying areas are subjected to severe per- channels and storm drains constructions to ennial flooding which is generally attributed assuage flood effects, the situation never to flaws in drainage design, improper waste gets better. Flooding and flood damage espe- management leading to siltation, and reduc- cially in Accra is a serious issue. (Ludlow, tion in the carrying capacity of drains. Some 2009). Flooding in Accra has become a per- of the major cities in Ghana which have been ennial rainfall ritual that does not receive the severely affected over the years include Ac- needed attention any longer, because various cra, Tamale, Bolgatanga, Sekondi-Takoradi interventions in the past had yielded little and Kumasi. In terms of coastal flooding, ar- results. These interventions may be ad-hoc eas like Ada, Dansoman-Glefe area (Greater measures, because the cause-effects had not Accra) and Agavedzi-Kedzi area (Volta Re- been studied enough to warrant pragmatic 20 GHANA JOURNAL OF SCIENCE Vol. 54 interventions. Between 1955 and 1997, of the total area of the country (Kankam- an estimated GH¢30,000,000.00 worth of Yeboah et al., 2011) were studied (Fig. 1). properties were destroyed, 100 lives were The Volta system consists of the Black and lost either during the flood period or after rivers and the systems. the floods and 10,000 people displaced from These three rivers run through Upper West, their homes (Adinku, 1994). Upper East, Northern Region, parts of the Due to this preventable loss, the Gov- Brong Ahafo, Ashanti, Eastern and Volta ernment of Ghana has implemented major Regions of Ghana. The south-western river projects (since 2000) to improve drainage system includes the Bia, Tano, Ankobra and systems and address the perennial flood- Pra rivers. These cut across parts of Ashanti, ing in the country. Most notable among Eastern, Central and Brong Ahafo Regions them include the Korle Lagoon Ecological of Ghana. Finally, the coastal river system Restoration Project (Afeku, 2005), and the made up of the Ochi-Amissa, Ochi-Nakwa, construction of Odaw river drains at Kwame Ayensu, Densu and Todzie Aka rivers span Nkrumah Circle. In addition, statutory agen- the Greater Accra, parts of the Central, East- cies such as the Ministry of Water Resourc- ern and the Volta Regions of Ghana. es, Works and Housing (MWRWH), City Engineers of Accra Metropolitan Assembly Drainage and Lands Department were established to The total annual runoff from all the three see to the reduction of the effects of flooding river systems in Ghana is estimated to be on life and properties (Nyarko, 2000). 56.4 billion m3 of which 41.6 billion m3 is The specific objective of the study was to accounted for by the . The mean determine the trends, occurrences and mag- annual runoff from Ghana alone is 38.7 bil- nitude of high streamflows in Ghana using lion m3 which is 68.6 per cent of the total hydro-meteorological data of the three river annual runoff. The Volta, south-western and systems in Ghana. The goals of the study coastal systems contribute 64.7 per cent, were to (i) use the outcome of the analysis 29.2 per cent and 6.1 per cent, respectively, of short-term existing dataset as a tool in of the annual runoff from Ghana (CSIR- predicting preventive, and mitigative ap- WRI, 2000). proaches that would provide better manage- ment options and, (ii) provide scientific data Hydro-meteorological data collection and information to engineers and city plan- Daily mean streamflow data for the period ners for proper design and decision making. 1980–2008 were obtained from the Hydro- logical Service Department (HSD) of the Experimental Ministry of Water Resources, Works and Study area Housing (MWRWH), Accra, Ghana. These The three main river systems in Ghana, i.e. were obtained from two different sources the Volta, south-western and coastal river namely, the CSIR-Water Research Institute systems that cover approximately 70 per cent, (CSIR-WRI) and the Ghana Meteorological 22 per cent, and eight per cent, respectively Agency (GMet) both in Accra. Specifically, Vol. 54 GHANA JOURNAL OF SCIENCE 21

Flood analysis

L E G E N D 11° The arithmetic mean method was used to International Boundary

Catchment Boundary Bolgatanga compute the monthly mean streamflow val- River Lake/Lagoon Regional Capital ues for the river stations for the period 1980

10° Wa – 2008 to determine the monthly flood situa- Tamale tion. On the other hand, the monthly rainfall totals for the period 1980 – 2008 were com- 9° puted for each synoptic station by summing V O L T A B A S I N rainfall values for each month. The mean

8° monthly rainfall for each river system was also computed by calculating the arithmetic Sunyani mean of the monthly rainfall totals from at 7° least three rainfall stations within the river Kumasi SOUTH - WESTERN BASIN system. Table 1 shows the selected river Ho flow and rainfall gauging stations used for 6° Koforidua the analysis.

Accra Cape CoastCOASTAL BASIN

5° Filling-in of missing gaps using correlation Sekondi 0 100 Km. coefficient 1°

3° 2° Fig. 1. Main river systems in 0°Ghana 1° Missing gaps in both the rainfall and river discharge data were filled-in using data from the monthly rainfall figures for Accra col- synoptic stations with good correlations. lected from CSIR-WRI and the others (Ta- Missing streamflow figures for Twifu-Praso, ble 1) from GMet were used. Generally, the Okyereko and Bamboi stations were filled- rainfall data were collected for the periods in with flow data from Assin Praso, Oketsew 1980 – 2008. Except for the south-western and Bui, respectively, because they gave river system, due to limited availability of good correlation coefficients. data, rainfall figures for the period 1995 –2008 were used. Extraction of high streamflows and corre- sponding rainfall amounts Two main methods are used in extracting

Table 1

Selected river flow and rainfall gauging stations of the river systems in Ghana

River systems Selected flow and rainfall gauging stations

River flow gauging stations Rainfall gauging stations

Volta Bamboi and Bui Tamale, Wa, Navrongo, Bole and Kete-krachi South-western Twifu Praso and Assin Praso Oda, Dunkwa-on-Offin and Kumasi Coastal Okyereko and Oketsew Accra, Asamankese and Agona Swedru 22 GHANA JOURNAL OF SCIENCE Vol. 54 high streamflows from the complete stream- Derivation of river discharge from flow du- flow series; the annual maximum series and ration curve the peak-over-threshold methods (Chow et One of the most informative methods of al., 1988). The former method extracts an- presenting the complete range of river dis- nual maximum streamflow from the com- charges from low flows to flood events is plete flow series, while the latter method de- by the use of FDC. The FDC defines the pends on a specified threshold above which relationship between any given discharge all streamflows are high flows (Willems, value and the percentage of time that this 2007). For long duration streamflow series discharge is equaled or exceeded or not (about 50 years), the annual maximum se- exceeded (Smakhtin, 2001; Mays, 2005). ries method is most widely used. However, Generally, FDCs may either be calculated for short duration streamflow series, the use as follows: (1) On the basis of the whole of the annual maximum flows method leads available record period (Vogel & Fennessey, to a lower number of extreme streamflows 1994), or long-term average annual (Sma- values, which can result in less accurate khtin et al., 1997), or (2) on the basis of all flood analyses. In that case, the peak-over- similar calendar months (long-term average threshold method is used (Willems, 2007). monthly FDC) or all similar seasons (long- The threshold selection can either be based term average seasonal FDC) from the whole on physical or statistical considerations. record period (Smakhtin et al., 1997). The peak-over-threshold method was The period of record for FDC represents adopted in extracting the high streamflows variability and exceedance probability of due to the short duration of streamflow data flow over the available period. Vogel & Fen- and missing gaps in the data available. For nessey (1994) considered FDCs for individ- each extracted high streamflow, the corre- ual years and treated those annual FDCs in a sponding mean monthly rainfall amount for way similar to a sequence of annual stream- the period of record was also extracted for flow maxima and annual streamflow mini- the analysis. The threshold selection was ma. This interpretation of the method allows based on statistical considerations, where mean and median FDCs to be estimated. the minimum threshold value above which Thus, the FDC was constructed in Mi- all flows were high flows, was estimated crosoft Excel spreadsheet by first arranging from the flow duration curve (FDC) using the streamflow values in ascending order of the probability of non-exceedance. The ex- magnitude. Then assigning rank numbers to tractions were done using the Microsoft Ex- each streamflow value with the lowest flow cel software (by selecting, copying and past- ranked as 1, and the largest by a value n, ing in new columns the high streamflows where, n is the total number of records. Af- that are equaled or not-exceeded 90% of the ter that, the probability of non-exceedance time; i.e., from 90% to 100% probability of for each flow value was calculated. The non-exceedance). percentage of time a given streamflow was ‘equaled’ or “not exceeded” (probability of non-exceedance) was computed using; Vol. 54 GHANA JOURNAL OF SCIENCE 23

r computed using; P = 100 × [1] n + 1 n Te = [2] where P is the percentage of time a given r flow is ‘equaled’ or “not-exceeded” and r where Te is the empirical return period (in is the rank of the flow magnitude. The FDC years) and r is the rank of the streamflow was finally developed by plotting all ranked magnitude. flows against their probability of -exceed A more simplified extreme value analysis ance (for low streamflow analysis), or non- was performed on the extracted peak stream- exceedance (for peak streamflow or flood flows by using the exponential extreme value analysis) (Smakhtin, 2001; Mays, 2005). distribution. For the probability distribution of the extremes above a threshold in years, Estimation of extreme streamflow and analy- the return period was calibrated to observa- sis of values tions using; The return period of high streamflow is an n 1 T = * [3] important parameter for studying flood, as it c t exp(–((x – xt)/b)) describes the probability of occurrence of extreme events. Flow frequency is normally where Tc is the calibrated return period constructed on the basis of a series of an- (years) based on exponential extreme value nual (daily or monthly) streamflow maxima, distribution (EVD), b is the calibrating pa- which can be extracted from the available rameter and t is the total number of extract- original continuous streamflow series (Sma- ed high streamflows used. The number of t b khtin, 2001). Since the available observed selected affects the values of xt and . streamflow records are normally insufficient The accuracy of the estimated distribu- for reliable frequency quantification of ex- tion parameter was then enhanced using the treme event, different types of theoretical equation (4): distribution functions are used to extrapolate n beyond the limits of ‘observed’ probabili- S (T – T )2 [4] RMSE = i=1 c c ties, and to improve the accuracy of high- n streamflow estimation (Smakhtin, 2001). The mean monthly river discharges for where RMSE is the root mean squared er- the period of record was sorted in descend- ror, n is the number of errors and Te–Tc is ing order of magnitude, and assigned rank the difference between the empirical and numbers with the largest flow ranked as 1 the calibrated return periods, respectively. and the lowest n, where n is the total number In calibrating the parameter of the extreme of records. The streamflow frequency curve value distribution, the distribution param- was drawn by plotting the river discharge eter was then optimised for the value which values against the estimated return periods. minimises the RMSE. This was achieved The recurrence interval of a flow with a by using the Solver Tool in Microsoft Excel certain magnitude (return period) was then spreadsheet. On the basis of linear regres- 24 GHANA JOURNAL OF SCIENCE Vol. 54 sions in the exponential quantile plots, the ance at a probability of 90 per cent, the design flow for a certain return period was threshold above which all streamflows are estimated by re-arranging equation (3) as: characterised as high flows for each system -1 n is noted as 995 m³ s (), 273 m³ x = x + b In (T) – In – [5] s-1 (Ayensu river) and 30 m³ s-1 (). T t ( t ) These threshold values represent the peak c where T is the estimated design flow at T- streamflows for the respective river systems. c years, t is the threshold value above which all streamflows are high flows, T is the re- Periodicity of high/peak streamflows of sam- turn period in years, n is the period of record pled river systems (in years), t is the total number of extracted Fig. 6 shows the periods (month and high streamflows used and b is the calibrat- corresponding year) in which high/peak ing parameter. streamflows were noticeable for the three river systems. Generally, over the 30-year Results period sampled, the occurrence and magni- Rainfall versus river flow regime tude of high/peak streamflows is noted to be The 28 years time-series plot showing the prevalent in the Volta system, followed by complete flow duration associated with rain- the south-western system compared to the fall (including missing data) for the three coastal system with the least. river basin system; (the Volta, the south- In the 1980’s streamflows were higher in western and coastal systems) showed vary- the south-western river system. However, ing mean streamflow (Figs 2, 3 and 4). This from 2000 to 2008, there were more fre- implies that at the different locations, (Bam- quent and much higher streamflows. From boi in the North, Twifu-Praso and Okyereko 2000 to 2004 the Volta system recorded gauging stations in the south), differences the highest streamflow (Fig. 6). The plots in monthly mean streamflow were observed of trends of high streamflows in the Volta, that ranged from 0 – 5, 270 m³ s-1 for the south-western and coastal river systems and Volta system, 0 – 124 m³ s-1 for the coastal their corresponding trends of monthly mean system and 0 – 704 m³ s-1 for the south-west- rainfall are shown in Fig. 7, 8, and 9, respec- ern system. tively. A decreases in rainfall is associated with increase in streamflow at Bamboi in Flow duration (FD) of the sampled river the Volta river system for the period 1986 - system 2003 (Fig. 7). Fig. 8 shows a fairly stable but Fig. 5 shows the FDs for the three river high streamflow associated with gradually basin systems sampled. The derived curves increasing high rainfall in the south-west- were based on the monthly mean streamflow ern river system at Twifu Praso for the pe- plots at the sampled river stations namely, riod 2000 – 2007. In terms of monthly high Bamboi (Black Volta), Okyereko (Ayensu flows, increases were noted for the coastal river) and Twifu-Praso (Pra river) over the river system. Very low monthly rainfall is 30 – year period (1980-2010). Non-exceed- associated with low streamflow at Okyereko Vol. 54 GHANA JOURNAL OF SCIENCE 25

8000 0 ) -1 s 3 6000 200

4000 400 Rainfall Streamflow 2000 600 Monthly rainfall (mm) Monthly Rainfall Monthly(mm) Rainfall Mean monthly flow (m Mean Monthly (m³/s) Flow Mean 0 800 Jan-80 Jan-84 Jan-88 Jan-92 Jan-96 Jan-00 Jan-04 Jan-08 Time

Fig. 2. Mean monthly rainfall and streamflow for the Volta basin system (Black Volta at Bamboi)

800 0

) Rainfall -1

s Streamflow 3 600 200

400 400

200 600 Monthly rainfall (mm) Mean monthly flow (m Monthly Rainfall Monthly(mm) Rainfall Mean Monthly (m³/s) Flow Mean 0 800 Jan-80 Jan-84 Jan-88 Jan-92 Jan-96 Jan-00 Jan-04 Jan-08 Time

Fig. 3. Mean monthly rainfall and streamflow for the south-western basin systems (Pra atTwifu Praso)

300 0 ) -1 100 s 3 200 200

Rainfall 300 Streamflow 100 400 Monthly rainfall (mm)

500 Monthly(mm) Rainfall Mean monthly flow (m Mean Monthly (m³/s) Flow Mean 0 600 Jan-80 Jan-84 Jan-88 Jan-92 Jan-96 Jan-00 Jan-04 Jan-08 Time

Fig. 4. Mean monthly rainfall and streamflow for the coastal basin systems (Ayensu at Okyereko) 26 GHANA JOURNAL OF SCIENCE Vol. 54

10000 South Western Coastal Volta (LOG)

-1 100 s 3

1 Streamflow (m³/s) (LOG) (m³/s) Streamflow Streamflow (m

0.01 0 10 20 30 40 50 60 70 80 90 100 Percentage of times a given streamflow is not exceeded

Fig. 5. FDC developed for the three river basin systems in Ghana

10000 South Western Coastal Volta (LOG)

-1 1000 s 3

100 Streamflow (m³/s) (LOG) (m³/s) Streamflow Streamflow (m

10 Jan-80 Jan-84 Jan-88 Jan-92 Jan-96 Jan-00 Jan-04 Jan-08 Time

Fig. 6. Magnitudes and occurrences of high streamflows in the river systems of Ghana for the period 1999 – 2007. Figs 10, 11 and 12 show the return period plots of extreme (peak) flows, which have Flow frequency (return period) curve been calibrated to t observations and extrap- Figs 10, 11 and 12 show the return peri- olated based on exponential extreme value od plots of streamflows for the Volta, south distribution (EVD). The value of the root western and the coastal river systems, re- mean square error (RMSE) shows the ac- spectively. Lack of data availability for cer- curacy of the estimated distribution parame- tain periods resulted in the differences in the ters. From the exponential EVD curves, the years of study for the three river systems. t chance for a flood with a certain magnitude is the total number of extracted high stream- to occur once in 1, 10 and 100 years have flows used in the calibration. The number of been estimated for the three river systems. t selected for the calibration affects the val- These values are shown in Table 2. For ex- b ues of xt and . ample, the return period flows of 650.6 m³ Vol. 54 GHANA JOURNAL OF SCIENCE 27

8000 400 High Streamflows High Rainfall )

-1 Trend of High Streamflow Trend of High Rainfall s 3 6000 300

4000 200

2000 100 Monthly rainfall (mm) Monthly (mm) Rainfall Monthly High (m³/s) Flow Monthly high flow (m 0 0 Jan-86 Jan-88 Jan-90 Jan-92 Jan-94 Jan-96 Jan-98 Jan-00 Jan-02 Time Time

Fig. 7. Trends of high streamflows and corresponding rainfall amount in the Volta river system (Bamboi)

) 800 400 -1 High Streamflows High Rainfall s

3 Trend of High Streamflow Trend of High Rainfall 600 300

400 200

200 100 Monthly (mm) Rainfall Monthly high (m³/s) Flow Monthly rainfall (mm)

Monthly high flow (m 0 0 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Time Time

Fig. 8. Trends of high streamflows and corresponding rainfall amount in the South–western river system (Twifu Praso)

200 400 ) High Flows Rainfall -1 Trend of High Streamflow Trend of High Rainfall s 3 150 300

100 200

50 100 Monthly Rainfall Monthly(mm) Rainfall Monthly rainfall (mm) Monthly high (m³/s) Flow Monthly high flow (m 0 0 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Time Time Fig. 9. Trends of high streamflows and corresponding rainfall amount in the coastal river system (Okyereko) 28 GHANA JOURNAL OF SCIENCE Vol. 54 s-1 and 6,407 m³ s-1 in the Volta river system long duration. On the other hand, rainfall at Bamboi is likely to occur averagely once in the northern part of the country is heavy every year and 100 years, respectively. Sim- with short duration. This results in a single ilarly, the return period flows of 296.4 m³ s-1 high peak flow in the year, especially at the and 36.5 m³ s-1 will occur on average once beginning of the season. These contrasting every year in the south-western river system scenarios are illustrated in Figs13, 14 and 15 at Twifu-Praso and the coastal river system with mean monthly rainfall totals and mean at Okyereko, respectively (Table 2). streamflows for the period 1995 – 2008 for the Volta, south-western and coastal river Discussion systems at Bamboi, Twifu Praso and Oky- Generally, rainfall is mainly bi-modal in the ereko, respectively. southern part of Ghana, where the bulk of Having the uni-modal rainfall pattern, the three river systems drain. The direct ef- peak monthly rainfall amount of 212 mm in fects on river discharges due to the two-peak August results in a peak streamflow amount flows are separated periods of low flows with of 1,068.6 m³ s-1 in September (Fig. 13).

8000 EmpiricalEmpirical return Return period Period EXP.EXP. Extreme extreme valueValue distribution Distribution Calibrated 6000 ) parameters -1

s n = 18 yr 3 t = 12 4000 3 -1 xt = 1,145.8 m³/sm s Flows (m³/s) Flows Flows (m β = 1,250 2000 RMSE = 6.50 mm³/s3 s-1 0 0.1 1 10 100 ReturnReturn Period period (years) (yr)

Fig. 10. Return period plots for the Volta river system using streamflow data from the Bamboi station

1000 EmpiricalEmpirical return Return period Period ) Calibrated -1 EXP.EXP. Extremeextreme valueValue distribution Distribution

s parameters 3 750 n = 24 yr t = 29 500 3 -1 xt = 272.6 m³/sm s β = 115

Peak Flows (m³/s) Flows Peak 250 Peak Flows (m RMSE = 2.08 mm³/s3 s-1

0 0.1 1 10 100 ReturnReturn Period period (years) (yr) Fig. 11. Return period plots for the south–western river system using streamflow data from the Twifu Praso station Vol. 54 GHANA JOURNAL OF SCIENCE 29

200 Empirical return Return period Period

) EXP. extreme value distribution Calibrated -1 EXP. Extreme Value Distribution

s 150 parameters 3 n = 26 yr t = 32 100 3 -1 xt = 30.3 mm³/s s β = 29.8 Peak Flows (m³/s) Flows Peak Peak flows (m 50 RMSE = 1.68 mm³/s3 s-1 0 0.1 1 10 100 Return Period (years) Return period (yr) Fig. 12. Return period plots for the coastal river system using streamflow data from the Okyereko station

In the south-western river system at Twifu tively (Fig. 5). The highest streamflow val- Praso, the monthly peak rainfall amounts of ues were observed in the Volta river system, 206.4 mm in June and 173.5 mm in Octo- whilst high streamflows in the coastal river ber resulted in two corresponding monthly system showed the lowest peak streamflow. peak streamflows of 138.0 m³ s-1 in June and In terms of the number of occurrences 215.0 m³ s-1 in October, respectively. of peak flows, there have been increases At Okyereko in the coastal river system, in the last decade (2000 – 2010), with the monthly peak rainfall amounts of 65.5 mm coastal and Volta river systems recording the in June and 32.3 mm and October resulted highest and lowest number of occurrences, in a direct peak streamflows of 26.7 m³ -1s in respectively (Fig. 6). In terms of the mag- June and 17.6 m³ s-1 in October, respectively. nitudes, the Volta and coastal river systems From the record, the monthly mean high recorded the highest and lowest high stream- streamflow for the Volta, south-western and flows, respectively, through-out the period coastal river systems ranged from 995 to of analysis. 5,270 m³ s-1, 273 to 704 m³ s-1 and 30 to 124 The Volta and coastal river systems at m³ s-1, respectively, with average values of Bamboi and Okyereko, respectively, have 1,684 m³ s-1, 374 m³ s-1 and 58 m³ s-1, respec- shown an increasing trend of high stream-

Table 2

Flood magnitudes with their return period

River systems (stations) T-years flood (m3 s-1)

Q1 Q2 Q5 Q10 Q25 Q50 Q100

Volta (Bamboi) 650.6 1,517.1 2,662.4 3,528.9 4,674.2 5,540.6 6,407.1 South-western (Twifu Praso) 296.4 376.1 481.5 561.2 666.6 746.3 826.0 Coastal (Okyereko) 36.5 57.2 84.5 105.1 132.4 153.1 173.7 30 GHANA JOURNAL OF SCIENCE Vol. 54

1200 250 ) -1

s Rainfall 200 3 900 Streamflow 150 600 100

300 Monthly rainfall (mm) Monthly Rainfall Monthly(mm) Rainfall

Mean monthly flow (m 50 Mean Monthly (m³/s) Flow Mean

0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec TimeTime Fig. 13. Mean monthly rainfall and streamflow for Volta river system at Bamboi

250 250 )

-1 Rainfall s 3 200 Streamflow 200

150 150

100 100

50 50 Monthly(mm) Rainfall Monthly rainfall (mm) Mean Monthly (m³/s) Flow Mean Mean monthly flow (m 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Time Time Fig. 14. Mean monthly rainfall and streamflow for south-western river system atTwifu Praso

40 80 ) -1 s

3 Rainfall 30 Streamflow 60

20 40

10 20 Monthly Rainfall Monthly(mm) Rainfall Mean Monthly (m³/s) Flow Mean Monthly rainfall (mm) Mean monthly flow (m 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec TimeTime Fig. 15. Mean monthly rainfall and streamflow for the coastal river system at Okyereko flow, with the Volta river system recording clining trends of monthly rainfall totals. The the highest rate of increase of high stream- south-western river system showed almost a flow over the period of records (Figs 8 and no-change in the trend of high streamflows 10, respectively). On the other hand, the at Twifu Praso (Fig. 9), although the basin Volta and coastal basin systems showed de- showed an increasing trend in high month- Vol. 54 GHANA JOURNAL OF SCIENCE 31 ly rainfall total for the period 2000 – 2005. uted the major cause of perennial floods in This, therefore, suggests that rainfall amount urban areas especially Accra, to drain clog- is not the major cause of the recent high ging as a result of siltation, design flaws streamflows or flooding in Ghana, though and improper waste management practices. there are evidences of direct runoff during Thus, the capacity of the drains is reduced, every rainy season, especially in the urban making it unable to hold the incoming high areas. streamflow. This high streamflow then over- With respect to the recurrence of high flows drains and finds its way into homes streamflow of a certain magnitude, it was es- and farms as flood, destroying lives and timated from the extreme value distribution properties worth millions of Ghana Cedis. that, there is a 100 per cent chance that, peak It is, therefore, recommended that drain- streamflows of magnitudes 650.6, 296.4 and age systems in towns and cities be improved 36.5 m³ s-1 will occur at least once every year to provide adequate storage volumes for in the Volta, south-western and the coastal run-off water during rainy seasons. City au- river systems, respectively, in Ghana. Simi- thorities should develop proper waste man- larly, streamflows of magnitudes 3,529, 561 agement and proper land use to reduce ero- and 105 m³ s-1 have the chance to occur at sion. The carrying capacity of drains should least once every 10 years in the Volta, south- be enhanced by de-silting to de-clog them western and coastal river systems, respec- regularly. Flood detection facilities should tively. be provided at vantage points along water courses to attenuate incoming flood waters. Conclusion and recommendations Also, studies on flood frequency analysis be The number of occurrences and trends of carried out on continuous basis as and when high stream flows confirm the increasing new meteorological and hydrological data number and magnitudes of flooding events are available. in the river systems recently. Even though there are evidences of direct runoff during the Acknowledgements rainy season, the study showed that rainfall The authors are most grateful to UNESCO, amount is not necessarily the major cause of through the Accra Cluster Office, Ghana, for the recent high streamflows or flooding. The funding the study under the framework of hypothesis that high streamflow does not Preventing the Promotion of Hydro-Hazard lead to flood is not entirely correct, because Disasters in Cluster Countries of West Af- high streamflows resulting from erratic rain- rica. They thank the Ghana Meteorologi- fall leads to flooding, especially in the urban cal Agency (GMet) and the Hydrological areas. Though heavy rainfall contributes to Service Department (HSD) of the Ministry flooding, the results show that rainfall is not of Water Resources, Works and Housing necessarily the major cause of the perennial (MWRWH) for providing the observed rain- flooding in the country. The study is also in fall and streamflow data, used in the study. conformity with findings from other studies They are also grateful to C.K. Asante-Sasu, (Kabglor, 2010, unpublished), which attrib- G. Appiah and F. Ofori (Technical Officers 32 GHANA JOURNAL OF SCIENCE Vol. 54 of the Surface Water Division, CSIR-Water poor region: A satellite data-supported flood Research Institute) for their assistance dur- model for Accra, Ghana, (MPhil Thesis).The ing data collection. Columbian College of Arts and Sciences, George Washington University. Reference Mays L. W. (2005) Water Resources Engineer- ing. Civil and Environmental Engineering Adinku, S. A. (1994) Disaster preparedness: Department, Arizona State University, Tempe, A sociological study of the flood problem in Arizona. the Odaw catchment in Accra. (MPhil The- Nyarko, B. K. (2000) Flood risk zoning of sis). Department of Sociology, University of Ghana: Accra experience, Department of Ghana, Legon. Geography & Tourism, University of Cape Afeku, K. (2005) Urbanization and flooding in Coast, Cape Coast, Ghana. URL: http:// Accra, Ghana. (MPhil Thesis). Department of www.isprs.org/proceedings/XXXIII/congress/ Geography, Miami University, Miami. part7/1039_XXXIII-part7.pdf Assessed on 7th Chow, T. V., Maidment, D. R. & Mays, L. W. December, 2010. (1988) Applied Hydrology. Civil Engineering Smakhtin, V. U. (2001) Low flow hydrology: A Series, McGraw Hill. review. Journal of hydrology, pp. 147-186. CSIR-WRI (2000) Climate change vulnerability URL: http://www.sciencedirect.com/science. and adaptation assessment on water resources Smakhtin, V. Y. & Watkins, D. A. (1997) Low- of Ghana. A UNFCC/EPA/WRI report, Accra. flow estimation in South Africa. Water Re- Kagblor, C. (2010) A spatio-temporal study of search Commission Report No. 494/1/97, Pre- urbanization and flooding in the Greater Ac- toria, South Africa. cra Metropolitan Area (GAMA) of Ghana, Vogel, R. M. & Fennessey, N. M. (1994) Flow (MPhil Thesis). Department of Environmen- duration curves, I. A new interpretation and tal Science, University of Ghana, Legon. confidence intervals, J. Water Resour.Plan. Kankam-Yeboah, K., Logah, F. Y., Gyau- Manag. 120 (4), 485 – 504. Boakye, P. & Ofori, D. (2011) Prevention of Willems, P. (2007) Statistic for water engineer- hydro-hazards disasters in Ghana. Update of ing. Lecture notes, September 2007, Katho- flood phenomenon in Ghana. CSIR-WRI/CAR lieke Universiteit Leuven (KULeuven), Leu- No. 359. ven, Belgium. Ludlow, D. C. (2009) Flood modeling in a data-

Received 11 May12; revised 16 Jul 12.