Comparison of groundwater recharge estimation methods for ... · rect recharge refers to water that percolates to the groundwater through the beds of surface watercourses. Localized

Comparison of groundwater recharge estimation methods

for the semi-arid Nyamandhlovu area, Zimbabwe

Tenant Sibanda & Johannes C. Nonner &

Stefan Uhlenbrook

Abstract The Nyamandhlovu aquifer is the main waterresource in the semi-arid Umguza district in MatebelelandNorth Province in Zimbabwe. The rapid increase in waterdemand in the city of Bulawayo has prompted the need toquantify the available groundwater resources for sustain-able utilization. Groundwater recharge estimation methodsand results were compared: chloride mass balance method(19–62 mm/year); water-table fluctuation method (2–50 mm/year); Darcian flownet computations (16–28 mm/year); 14C age dating (22–25 mm/year); and groundwatermodeling (11–26 mm/year). The flownet computationaland modeling methods provided better estimates for aerialrecharge than the other methods. Based on groundwatermodeling, a final estimate for recharge (from precipitation)on the order of 15–20 mm/year is believed to be realistic,assuming that part of the recharge water transpires fromthe water table by deep-rooted vegetation. This rechargeestimate (2.7–3.6% of the annual precipitation of 555 mm/year) compares well with the results of other researchers.The advantages/disadvantages of each recharge method interms of ease of application, accuracy, and costs arediscussed. The groundwater model was also used toquantify the total recharge of the Nyamandhlovu aquifersystem (20×106–25×106 m3/year). Groundwater abstrac-

tions exceeding 17×106 m3/year could cause ecologicaldamage, affecting, for instance, the deep-rooted vegetationin the area.

Keywords Groundwater recharge/water budget .Groundwater age . Groundwater flow . Nyamandhlovuarea . Zimbabwe

Introduction

The problem of estimating groundwater recharge in semi-arid areas is that recharge amounts are normally small incomparison with the resolution of the investigationmethods (e.g. Allison et al. 1984). The greater the aridityof the climate, the smaller and potentially more variable inspace and time is the recharge flux. Direct groundwaterrecharge from precipitation in semi-arid areas is generallysmall, usually less than about 5% of the average annualprecipitation, with a high temporal and spatial variability(Gieske 1992).

Lerner et al. (1990) concluded that determination ofgroundwater recharge in arid and semi-arid areas is neitherstraightforward nor easy. This is a consequence of thetemporal variability of precipitation and other hydro-meteorological variables in such climates, and the spatialvariability in soil characteristics, geology, topography,land cover characteristics and land use. Holman (2006)points out that climate change, also in arid and semi-aridareas, may affect recharge in the near future.

De Vries and Simmers (2002) classified naturalgroundwater recharge mechanisms according to the origininto three types: direct/diffuse recharge, localized re-charge, and indirect/non-diffuse recharge. Direct rechargerefers to water that is added to the groundwater reservoirfrom precipitation by direct percolation through theunsaturated zone in excess of soil moisture deficits,interception, surface runoff and evapotranspiration. Indi-rect recharge refers to water that percolates to thegroundwater through the beds of surface watercourses.Localized recharge is an intermediate form of groundwaterrecharge resulting from horizontal (near) surface concen-tration of water (ponding) in the absence of well-definedchannels.

Received: 25 July 2007 /Accepted: 18 February 2009Published online: 22 March 2009

* Springer-Verlag 2009

T. SibandaWorld Vision,P.O. Box 2420, Harare, Zimbabwe

T. Sibandae-mail: [email protected]

J. C. Nonner ()) : S. UhlenbrookDepartment of Water Engineering,UNESCO-IHE,P.O. Box 3015, 2601 DA, Delft, The Netherlandse-mail: [email protected]

S. Uhlenbrooke-mail: [email protected]

S. UhlenbrookDelft University of Technology,P.O. Box 5048, 2600 GA, Delft, The Netherlands

S. Uhlenbrook

Hydrogeology Journal (2009) 17: 1427–1441 DOI 10.1007/s10040-009-0445-z

Groundwater recharge studies in arid and semi-aridregions of Southern Africa have been carried out bydifferent researchers. Larsen et al. (2002), Gieske (1992),Beekman and Sunguro (2002) concluded that recharge insemi-arid regions in Southern Africa is a small componentof the water balance (usually <5% of the average annualrainfall). Nyagwambo (2006) demonstrated that as thepotential evapotranspiration is higher than the rainfall,the recharge is dependent on rainfall intensity and theexistence of fractures, fissures and cracks in the tropicalcrystalline basement aquifers of Zimbabwe.

The influence of vegetation on recharge is also veryimportant. Studies by De Vries et al. (2000) in theKalahari Desert have detected acacia species with deeptap-root systems exceeding 50 m. Capillary rise and upliftinduced by deep-rooted vegetation has been reported tocause upward movement of water from large depths(Obakeng 2007).

Within this framework, the main objectives of thisarticle are, first, to report on the application of differentmethods (tracer and physical methods) to estimaterecharge rates in the semi-arid Nyamandhlovu aquifer,Zimbabwe. Second, the recharge estimation results havebeen compared, along with the methods themselves withregard to the accuracy of the method, the costs involved,and the ease of application. The studied Nyamandhlovuaquifer is located near Bulawayo, the second city of

Zimbabwe, in the semi-arid western part of the country(see Fig. 1). The groundwater demand is growing in thisarea and the determination of the recharge rate is crucialfor the groundwater management in that region.

The Nyamandhlovu aquifer

Physiographic and geologic settingThe Nyamandhlovu aquifer is located in the Umguzadistrict of Matebeleland North Province of Zimbabwe,about 40 km northwest of the city of Bulawayo. TheNyamandhlovu aquifer is part of the much larger ForestSandstone aquifer system. The coordinates Xmin 598000,Xmax 6780000 and Ymin 7795000, Ymax 7833000 give theposition of the boundaries of the Nyamandhlovu aquifer.The boundaries partly coincide with the local boundariesbetween the Khami River and the Umguza River catch-ments, and other small catchments of rivers in the north,east and southwest, and with the Gwayi River in the west(Fig. 2). This region experiences a typical semi-aridsavannah climate characterized by low annual precipita-tion and high evapotranspiration. The long-term averageannual precipitation is about 555 mm/year and thepotential evapotranspiration can go up to 2,000 mm/year.

0 75 km

Harare

Sanyati River

Shangani River

Manyame RiverGwayi River

Kadoma

Gweru

Zimbabwe

Victoria falls

MIDLANDS

MASHONALANDWEST

MATEBELELANDNORTH

Zambia

Botswana

Munyati River

Bulawayo

Study area (See Fig. 2)

Provincial boundary

River

Town

International Boundary

AFRICA

Legend

Lake KaribaN

EW

S

Fig. 1 Location map of study area in Zimbabwe

1428


The land surface is flat with isolated ridges normallycapped with basalts. Some isolated hills with silicifiedsandstone occur at the southeastern part of the area. Thehigher areas, occupied by savannah bush forest, arecharacterized by a low surface drainage density and thetopographic gradient gently slopes downward in anorthwestern direction. The main rivers (all ephemeral)in the area, the Umguza River and the Khami River,generally flow down gradient towards the northwest.These rivers join the Gwayi River which flows northwardsinto the Zambezi River.

There are five main geological formations that havebeen mapped in the Nyamandhlovu aquifer area. Figure 2presents the geological map and Fig. 3 shows cross sectionsbased on borehole information. The geological formationsinclude the alluvial deposits, Kalahari Sands, KarooBasalts, Forest Sandstone, and the Basement Complex.The alluvial deposits are only mapped close to the rivers orold river channels. Thin patches of aeolian Kalahari Sandscover the Karoo Basalts in places. The basalts overlay theForest Sandstone which rests directly on the granitic-gneissic Basement Complex. The lithological stratigraphyof the five principal geological formations in the study areais summarized in Table 1 (Macgregor 1937).

Previous hydrogeological investigationsBeginning in the 1970s, different researchers havestudied the Nyamandhlovu aquifer. The studies are multi-disciplinary encompassing preliminary hydrogeologicalinvestigations. Beasley (1983) carried out a study on thegroundwater potential of the area. The main aquifer unitwas identified as the permeable Middle and Lowerdivision of the Forest Sandstone. The overlying KarooBasalt is permeable where the rock is weathered andfractured. Kalahari Sands locally occur as unconsolidated

sediments on top of the basalt. They have a highpermeability thus enhancing direct recharge throughporous material.

Beasley (1983) carried out further studies on thehydrogeology of the area. He found transmissivity valuesfor the Forest Sandstone aquifer in the range of 4–94 m2/day with an average value of 35 m2/day for theNyamandhlovu area. The aquifer thickness was between51 and 119 m resulting in a permeability range of 0.10–0.85 m/day with an average of 0.34 m/day. It was shownthat the groundwater flow system forms a regional systemflowing to the northwest. It was also noted that there aresome local flow systems influenced by changes in thestream discharges in the area. Direct groundwater–surfacewater interactions occur at the Umguza and the KhamiRivers. Furthermore, several hints indicated that rechargetakes place mainly at the high plateaus.

Banda et al. (1977) carried out a study in the Umguzasub-catchment focusing on the exploration of groundwaterthrough the drilling of exploratory boreholes. Within theframe of the study reported here, 25 boreholes were drilledinto the Forest Sandstone and pump tested, yielding aweighted average for the transmissivity of 40 m2/day anda storativity of 5×10–4.

Martinelli and Hubert (1996) carried out a desk studyand made an inventory of the available informationpertaining to groundwater in the Nyamandhlovu aquifer.They also carried out step-drawdown and constantdischarge tests and obtained values for the transmissivityin the range of 0.2–230 m2/day for the Forest Sandstone.They recommended executing further geophysical inves-tigations, drilling of production and monitoring boreholes,geophysical borehole logging, pumping tests, hydrochem-ical sampling and age dating of water, groundwater flowmodeling, re-designing the monitoring program andaquifer recharge studies.

(

(

Umguza River

NyamandhlovuNyamandhlovuCentreCentreNyamandhlovu

Epping Forest FarmEpping Forest FarmEpping Forest Farm

Centre

Khami River

Section 1

Section 2

Section 3Section 4

SawmillsS26

U40

S27

U59

U13

N1

U41 S21 N10

U60U19

U54

U20

S22

U28

U23

N5

S40

S24

S26S26

U40

S27S27

U59

U13U13

N1N1

U41U41 S21S21 N10N10

U60U60U19U19

U54U54

U20U20

S22S22

U28U28

U23U23

N5N5

S40S40

S24S24

S26

U40

S27

U59

U13N1

U41 S21 N10

U60U19

U54

U20

S22

U28

U23

N5

S40

S24

S40S23S23S23

U9 U9

20 0 20 Kilometers

N

EW

S

Dolerite and Epidiorite

RiverAlluvial and minor conglomerate

Basement Granite

Kalahari SandKaroo Basalt

Forest Sandstone

Railway

Section line

Exploration boreholes

Legend

Fig. 2 Geology of the area showing traverses of cross sections

1429


Siwadi (2001) carried out a study on re-designing thegroundwater level monitoring network for the Nyamandh-lovu aquifer. For his investigation period from 1989 to1999, it was shown that the groundwater levels aredeclining at a rate of about 0.2–1.5 m/year, experiencinga high decline in areas of intense groundwater exploitationaround the Bulawayo City well field.

Hydrogeological conceptualizationThe previous studies indicate that the main aquifer in theNyamandhlovu area is contained in the Middle and Lowerdivision of the Forest Sandstone Formation. The ForestSandstone only crops out in the eastern part of the studyarea and elsewhere this unit is overlain by Karoo Basaltand Kalahari Sands. The occurrence of artesian wells inthe Forest Sandstone at Sawmills indicates that the KarooBasalt can act as a confining aquiclude. However, theKaroo Basalt can also form a secondary aquifer where therock is weathered and fractured. These conditions were

found at Epping Forest Farm where geophysical inves-tigations indicated that the rock is also very patchilydeveloped. The unconsolidated Kalahari Sands, whichmainly developed on the higher plateaus in between thelocal rivers in the study area, have a high permeability andcan be considered as aquifers in the few places where thelocal water table is positioned within this unit (Fig. 3).

Recharge into the Forest Sandstone aquifer is believedto consist of contributions from precipitation, from thegroundwater–surface water interaction at parts of the

Table 1 Lithological stratigraphy of the area (after Macgregor 1937)

Lithological unit Geological age

Alluvial deposits Pleistocene and RecentKalahari Sand Early and middle

TertiaryKaroo Basalt and interbeddedsandstones

Early and middleTertiary

Forest Sandstone Upper TriassicBasement Complex Pre-Cambrian

1200

Section 1

1100 Sawmills

S24U40

S23 N1 N5 S22

1000

900

2000 8000 14000 20000 26000 32000 38000 44000

Legend

Groundwater level

Ele

vatio

n in

m a

bove

sea

leve

lE

leva

tion

in m

abo

ve s

ea le

vel

Ele

vatio

n in

m a

bove

sea

leve

lE

leva

tion

in m

abo

ve s

ea le

vel

Distance in m Distance in m

Distance in m Distance in m

1200

1100

S27U41

S21 U09

1000

900

2000 8000 14000 20000 26000 32000 38000 44000

Section 2

1200

1100

U59 U13

?

N1S40

N10

1000

900

2000 8000 14000 20000 26000 32000 38000 44000

Section 3

1200

1100

U23 U28 U20 U54 U60 U19

1000

900

2000 8000 14000 20000 26000 32000 38000 44000

Section 4

Basement Granite

Kalahari Sand

Karoo Basalt

Forest Sandstone

a b

dc

Fig. 3 Cross sections (see Fig. 2 for locations)

1430


Umgazi and Khami rivers and from return flows atirrigated fields. Precipitation amounting to a long-termaverage of 555 mm/year is the largest contributor ofrecharge and takes place in a direct way where the ForestSandstone is exposed (Fig. 2). In the areas where KarooBasalt, overlaying the Forest Sandstone crops out,recharge takes place where the basalt is weathered andfractured. Also recharge may take place through theKalahari Sand into underlying permeable Karoo Basaltand Forest Sandstone.

Where the sandstone and the basalt are exposed, thevegetation is rather sparse and the drainage density is notthat large with, in places, no connections of gullies to themajor streams. This confirms the occurrence of ground-water recharge. The Kalahari Sands on the other handshow a densely vegetated terrain with abundant vegeta-tion. When soil moisture limits are exceeded after heavyrain, percolation takes place through the sands anddownward to the Karoo Basalt.

The middle and eastern part of the study area couldwell receive larger amounts of recharge than the westernpart. One of the reasons is that, partially, the middle andeastern parts contain the exposed Forest Sandstone, buteven more important is that the Karoo Basalt seems tobecome thicker in a westerly direction (Fig. 3). The largethickness of the basalt in the west makes it unlikely thatthis unit is weathered from top to bottom. The resultingless permeable nature of the basalt in these areas asdemonstrated by the confining nature of the rock atSawmills substantially reduces the scope for rechargefrom precipitation.

The recharge contribution from surface water is believedto be small since the major streams in the area areintermittent, only flowing after heavy rainfall. The contactarea with the underlying Forest Sandstone or Karoo Basaltis relatively small and flood water carries an appreciablesediment load thereby clogging fractures in the hard rock.Return flows from fields are difficult to estimate but in viewof the small size of the irrigated area of less than 2,000 ha,the overall quantities are considered to be small.

Recharge into the Forest Sandstone aquifer results in agroundwater flow in a northwesterly direction as indicatedon groundwater level contour maps prepared on the basisof information from boreholes where groundwater levelsare monitored. Along a flow path the dissolution of calcitepresent in the sandstone is one of the main hydro-chemicalprocesses taking place. This yields a predominantlycalcium-bicarbonate type of groundwater which is alsofresh in most places.

Need for additional analysesPrevious studies in the Nyamandhlovu area have tackledto some extent the recharge issue, but so far thedistribution of recharge across the area and the relatedrecharge processes have only been described in generalterms. Beasley (1983) and Martinelli and Hubert (1996)tried to quantify recharge for their groundwater potentialand groundwater assessment studies and adopted recharge

rates on the order of 125 and 14–28 mm/year correlatingwith 22% and 2.5–5% of the long-term average precipi-tation, respectively. However, they did not carry outdetailed investigations. The study reported here, however,overcomes this lack and presents a number of methods toaddress the analysis of the recharge process and theestimation of recharge rates for the Nyamandhlovuaquifer.

Methodology and applicability

Four methods have been selected to estimate groundwaterrecharge in the study area: The chloride mass balance, thewater-table fluctuation method, the Darcian flownetcomputation, and the groundwater age dating method(e.g. Kinzelbach et al. 2002; Scanlon et al. 2002). Thesemethods have been selected based on the availability ofdata and the hydrogeological set up of the study area. Inaddition, groundwater flow modeling has been undertakento complement and confirm the findings of the mentionedrecharge estimation methods.

Chloride mass balance methodThis is a tracer balance method whereby the amount ofchloride entering into the root zone system is balanced bythe amount of chloride leaving the system (Nonner 2006).The chloride mass balance equation can be described asfollows:

ClgwQperc ¼ ClpP þ D� CleE � ClrR ð1Þ

where:

Qperc Rate of percolation at the lower boundary of theroot zone (m3/day)

Clgw Chloride concentration in percolating groundwater(g/m3)

P Precipitation rate (m3/day)Clp Chloride concentration in precipitation (g/m3)D Dry chloride deposition (g/day)E Total evapotranspiration from the root zone

(m3/day)Cle Chloride concentration in evapotranspiration water

(g/m3)R Surface runoff (m3/day)Clr Chloride concentration in runoff water (g/m3)

To work out a simplified relation for recharge estima-tion, it will be assumed that the chloride concentration inthe evapotranspiration water is virtually zero and thatareas are being studied with very little surface runoff. Inthe longer run, the percolation rate also equates to therecharge rate which can then easily be determined bycalculating the ratio of the chloride concentration inprecipitation and the chloride concentration in groundwa-ter (Gieske 1992). Working out the balance equation andassuming that the dry deposition is also determined in the

1431


collected precipitation water, the following formula forrecharge estimation can be established:

Qperc ¼ ClpP

Clgwð2Þ

The successful application of this method depends on thefollowing assumptions:

& Chloride is a conservative tracer in the system. Thismeans that its concentration is neither diminished norincreased through chemical reactions in the soil

& There is no other source of chloride in the soil orgroundwater (incl. dry deposition) other than precipitation

& Steady-state conditions of atmospheric solute input(constant chloride concentrations in space and time)and solute flux in the subsurface in the long termprevails

& No surface runoff is present, or its flux can be accounted for& In the long run, the percolation at the lower boundary

of the root zone equals groundwater recharge

The chloride mass balance method has successfullybeen used before in the semi-arid hard rock regions ofBotswana and Zimbabwe (Beekman et al. 1996; Beekmanand Sunguro 2002; Nyagwambo 2006). The assumptionsunderlying the application of the method are largelysatisfied for the study area. For example, another sourceof chloride could be fertilizer used on agricultural fields.However, commonly used fertilizers in Zimbabwe areCompound D and ammonium nitrate which do not containchloride and it is reasonable to assume that thesechemicals were also used in the Nyamandhlovu area.Surface runoff as discussed in the previous section isthought to be rather small and to neglect this amount inthe recharge computation following the chloride massbalance method is justified.

Water-table fluctuation methodGroundwater recharge from precipitation may be estimat-ed from an analysis of water-table fluctuations in anunconfined aquifer (Scanlon et al. 2002; Nonner 2006).Recharge causes a rise in the water table which conse-quently causes an increase in the amount of water stored.Recharge is equal to the amount stored plus an amountthat is added to discharge. To assess recharge with thewater-table fluctuation method, a hydrograph of the watertable is considered to determine the so-called modified risein water table elevation. The modified rise in the watertable is equal to the rise taken from the extended recessioncurve to the peak of the table. The modified rise accountsfor both the water stored and the amount drained laterallyduring the recharge event. The following equation canthen be used to compute the recharge rate:

Qprec ¼ SyD�mod

Dtð3Þ

where:

Qprec Rate of recharge from precipitation (m/day)Sy specific yield (-)Δ�mod Modified rise in water table (m)Δt Observation period (days)

In cases where more than one recharge event occurswithin the observation period, modified rises in watertable can be summed up for the estimation of an averagerecharge rate. This approach may be applicable for largertime steps ranging from months to seasons where naturaldischarge is quite substantial.

The water-table fluctuation method has been used forcrystalline rock in Zimbabwe where, in parts of the area, thehard rock was overlain by Kalahari Sands (Nyagwambo2006). The availability of long time series of groundwaterlevels justifies the application of the water-table fluctuationmethod for the Nyamandhlovu area, but there are alsodisadvantages. Perhaps the main drawback is the uncer-tainty in obtaining a representative value of the specificyield Sy which is difficult to establish in the fractured hardsandstone and basaltic rock present in the study area. As aresult of the well screen setting in the Forest Sandstoneaquifer for which the water levels may not exactly reflectwater table behavior, the correct water table rises may alsobe difficult to establish.

Flownet computationRegional groundwater flow computations based ongroundwater flow equations may be used to assessgroundwater recharge. The approach follows the Dupuit-Forchheimer assumption whereby groundwater flow isassumed to be horizontal in aquifers and vertical inaquitards. Flownets can be prepared by drawing a set offlow lines perpendicular to equipotential contour lines ongroundwater level contour maps. The flow lines will thenform boundaries of stream tubes whereby it is assumedthat groundwater flowing through an individual streamtube is neither loosing nor gaining water from theneighboring tube (Fetter 2001).

Darcy’s law can be used to compute the flow ratethrough a cross-sectional area of the aquifer by summingup flow rates through individual stream tubes. Thegroundwater flow rate through a stream tube at a selectedcross-section can be computed by taking into consider-ation the discrete difference in hydraulic head for twosuccessive contour lines, the width of the stream tube andthe thickness of the aquifer. Darcy’s law equation for theflow rate computation can be written as follows:

Qs ¼ �KHwsD�Ds

ð4Þ

where:

Qs Flow rate (horizontal) through the stream tube(m3/day)

1432


K Horizontal coefficient of permeability (m/day)H Saturated aquifer thickness (m)ws Width of the stream tube (m)Δ� Discrete difference in groundwater level (m)Δs Discrete distance between contour lines (m)

The flownet computation method has not been usedmuch in Zimbabwe. Presumably, the dominance of hardcrystalline rock with a very uneven distribution of trans-missivities and a lack of sufficient monitoring wells hasprevented scientists from performing this type of rechargeassessment. However, the application of the methodappears to be justified for the Nyamandhlovu aquifer.Although there is a reasonable variation in the transmis-sivity of the Forest Sandstone across the area, a reliableaverage can be determined from pumping tests carried outat a large number of wells. The large number ofgroundwater level monitoring wells with suitable dataand with a reasonable distribution over the study areaallows the preparation of groundwater level contour maps.

Groundwater datingGroundwater recharge to aquifers can be evaluated fromcarbon-14 ages according to the Bredenkamp and Vogel(1970) model. The equation for the recharge estimation is:

Qprec ¼ �eH

tlnH

hð5Þ

where:

Qprec Recharge rate (mm/year)�e Effective porosity of the saturated zone in the

aquifer (-)h Saturated thickness of the aquifer (m)H Total thickness of the aquifer (m)t Carbon-14 age of water (years)

The approach assumes only vertical water fluxes, thusthe developed relationship takes only the depth ofsampling into account. The approach can be applied forparticular boreholes in a recharge area. The model hasbeen applied successfully in several groundwater rechargestudies in Southern Africa by various researchers (e.g.Bredenkamp and Vogel 1970; Bredenkamp et al. 1995;Beekman et al. 1996).

Analyses and results

The recharge estimation following the chloride massbalance method requires data on precipitation, totalchloride deposition from the atmosphere and the chlorideconcentration in the percolating groundwater. A long-termaverage annual precipitation of 555 mm/year was obtainedfrom a record of 31 years duration at two stations in thestudy area, the Nyamandhlovu Rail and Umguzanistations. The total chloride deposition from the atmo-sphere, mainly from precipitation and measured over onlytwo years in the study area, ranged between 0.5 and 1.0mg/L. Larsen et al. (2002) proposed for the total chloridedeposition from the atmosphere an average value of 0.5mg/L for western Zimbabwe. For the computation of therecharges a value of 0.6 mg/L has been used. For thedetermination of the chloride concentration in percolatinggroundwater, water samples have been collected at wellscreens in the saturated Forest Sandstone aquifer. Chlorideconcentrations ranging from 2.5 to 76.5 mg/L, determinedin samples from a selection of 24 wells, have been usedfor recharge computations.

Figure 4 shows the aerial distribution of recharge ratescomputed with Eq. (2). The rates vary between a high of133 mm/year for an observation point in the middle of thearea to a low of 4.4 mm/year near Sawmills. Since watersamples for the determination of chloride rates ingroundwater have been taken at well screens not exactly

(

(

20 0

N

EW

S

Recharge rate in mm/year28.2 Sampled well

4.4

18.9

32.6 24.928.2

62.8

41.6

23.8

33.6

41.641.6

41.6

47.5

32

23.8

10

21.2

42.255.5

15.9

107.4

14.216.6

133.2

Khami River

20 kilometers

Umguza River



Basement Granite


Forest Sandstone

Railway

Legend

Sawmills

Fig. 4 Recharge determined with the chloride mass balance method

1433


tapping the water table, the recharge rates rather apply tosome point ‘upstream’ of the wells. This makes it difficultto correlate recharge rates determined at wells located inthe Karoo Basalt and Kalahari Sand areas with directrecharge into these formations, since recharge may alsohave taken place in ‘upstream’ areas where ForestSandstone is exposed. However, the general trend seemsto be that in the middle and eastern parts of the study area,the recharge rates tend to be higher than in the westernsection. Finally, it can be concluded that for the entirestudy area, a recharge range of 19–62 mm/year yieldingan average rate of 38 mm/year can be computed with thechloride mass balance method. This average rechargevalue represents 6.7% of the total annual precipitation.

For the computation of recharge using the water–tablefluctuation method, specific yields and groundwater leveltime series are required. The calculations have been basedon the lower and upper limits of specific yield values of0.002 and 0.008, which were interpreted from the twopumping tests in the area clearly showing water-tableconditions. The wells are located in the Forest Sandstoneand Karoo Basalt Formations. Groundwater levels at fourmonitoring wells are continuously recorded with pressuretransducers whereas in another 37 monitoring wellsmonthly measurements have been carried out since the1988–1989 rainy season.

Figure 5 shows time series of groundwater levels ofthree of these wells and Fig. 6 presents the locations ofthese wells. The time series based on monthly measure-ments show a similar pattern as the continuous registrationof levels in the same wells. The Cawston and Mpandeniwells are located in the Karoo Basalt, but screened in theForest Sandstone aquifer, and the Umguza River Bridgewell is located and screened in the Forest Sandstone.Without the presence of a thick dampening zone ofunconsolidated material, the rather rapid reaction toprecipitation events, within 1 or 2 months (Fig. 6),indicates preferential flow through fractured rock. Con-sidering the whole time series the reaction to rain events inthe rainy season is evident and it is clearly shown thatafter a number of dry years, with little recharge from 1990to 1999, groundwater levels were restored in the muchwetter period from 1999 to 2002. In particular thegroundwater level time series in the wells at inlandMpandeni and Umguza River Bridge show a similarbehavior which could indicate that the influence of theseasonal floods in the latter well (at about 20 m from theriver) is not that large. The well at Cawston is different inthe sense that the restoration of the groundwater levelsafter 1999 is more evident than in the other two wells. Themore pronounced restoration in the former well is a resultof a lack of direct recharge in the area around this well inthe preceding dry period when soil moisture deficits wereapparently not exceeded. Lack of groundwater level data,like in the 2002–2003 rainy season, can be a reason for apoor correlation between these groundwater levels andprecipitation.

Figure 6 also shows the locations of the other fourwells that have been used for the recharge computationsusing the water-table fluctuation method. These wells arelocated in the Karoo Basalt and screened in the ForestSandstone whereby it has been assumed that the measuredgroundwater levels reflect the water table. Some justifica-tion for this assumption is given by the statements madeon the hydrogeological conceptualization of the area,describing the Karoo Basalt as highly weathered andfractured in places. The recharge computations carried outwith Eq. (3) for the period 1989–2002 show that therecharge rates vary between 2 and 12 mm/year when aspecific yield of 0.002 is used. They range from 9 to 50mm/year for a specific yield of 0.008. The locations of themonitoring wells are in most places not that far away fromproduction wells with mutual distances ranging between

1140

1141

1142

1143

1144

1145

1146

1147

1148

1113

1114

1115

1116

1117

1118

1119

1120

1121

1122

Gro

un

dw

ater

Lev

el (

m)

Gro

un

dw

ater

Lev

el (

m)

Gro

un

dw

ater

Lev

el (

m)

1178

1179

1180

1181

1182

1183

1184

1185

1186

1187

1188

0102030405060708090100

0

10

20

30

40

50

60

70

80

90

100

Rai

nfa

ll (m

m)

Rai

nfa

ll (m

m)

Rai

nfa

ll (m

m)

0

10

20

30

40

50

60

70

80

90

100

Cawston (4-UMGU-126)

Time

Dec 88 Dec 92 Dec 96 Dec 00 Dec 04

Time


Time


Umguza River Bridge (4-UMGU-092)

Mpandeni (4-UMGU-118)

Groundwater level fluctuation

Rainfall (data missing from 1988 to 1999)

Legend

a

b

c

Fig. 5 Precipitation–groundwater level relationships

1434


50 and 1,000 m. However, during groundwater levelmeasurements production wells are usually switched offfor a period of at least 24 h and the recovered ‘static’levels allow a reasonable computation of recharge values.Nevertheless, the effects of incomplete recovery at somewells (Mpandeni, Dillkosh and Khami; see Fig. 6) werenoted and corrections were made to come up with moreaccurate recharge rates. In view of the above, the limitsthat can be set for the recharge estimation using thewater table fluctuation method are 2 and 50 mm/yearcorresponding to 0.4% and 9% of the long term annualprecipitation.

To apply the flownet computation method, transmis-sivity values and aerial distributions of groundwater levelsare required. Transmissivities adopted for the ForestSandstone aquifer range between 11 and 97 m2/day andwere determined from pumping tests carried out at 21wells that are reasonably well distributed over the studyarea. The groundwater level measurements at the sevenwells used for the water-table fluctuation method are partof a larger network of 41 groundwater level monitoringwells which showed that the levels range from a high ofabout 1,220 m above sea level in the southeast to a low onthe order of 1,100 m above sea level in the northwest atSawmills.

Figure 7 shows the groundwater level contour lineswhere flownet computations have been carried out usingEq. (4) for each individual stream tube. By summation ofthe flows obtained for each tube, the total flow through theForest Sandstone aquifer at the selected contour line wasobtained. Transmissivity values representative for each setof selected contour lines have been considered. Thecomputations have been done for the dry and rainyseasons of the periods from 1997 to 1998 and from 2000to 2001 which can be considered to represent an averageand a wet year respectively. The average flow in theaquifer using the contour lines of 1,190 and 1,200 mabove sea level has then been computed in the range of

1,600–2,100 m3/day and for the lines 1,120 and 1,130 mabove sea level, from 11,500–14,300 m3/day.

Recharge values for the area of 960 km2, in betweenthe mentioned sets of contours, have been computedconsidering that this parameter is equal to the outflow atthe downstream set of contour lines minus the inflow atthe upstream set of lines plus the net groundwaterabstractions and any transpiration losses by deep-rootedplants reaching the water table. The domestic abstractionsin the area for the city of Bulawayo have been measured at20,000 m3/day and agricultural abstractions for irrigationschemes have been estimated at 25,000 and 12,000 m3/day (Beasley 1983; Martinelli and Hubert 1996, respec-tively). For the whole study area, net agriculturalabstractions of about 17,000 m3/day have been consideredtaking into account a 10% return flow based on theposition of flow monitoring points and the use of efficientsprinklers. Since some wells are located ‘downstream’ ofthe groundwater level contour lines used for the flownetcomputations, the actual net agricultural abstraction takenfor recharge calculations amounted to 12,000 m3/day. Theabstraction values have further been confirmed in themodeling process.

Transpiration losses by deep-rooted plants is an issuethat is also elaborated on in the modeling section, but forrecharge computations using the flownet approach, caseswhereby roots do not reach and do reach the water tablehave been considered. In the first case, the losses are notconsidered in the recharge computation. However, in thesecond case, deep-rooted vegetation may take water at arate of 31,000 m3/day (or 12 mm/year, Obakeng 2007)from the water table. This amount of water has beenincluded in the recharge computations. Filling in the dataand expressing the parameter as a yearly rate the averagerecharge over the periods from 1997 to 1998 and from2000 to 2001 have been computed at 16.1 and 16.5 mm/year, respectively, for the case whereby the roots do notreach groundwater and from 27.9 to 28.2 mm/year for the

20 0

N

EW

S

Recharge rate (mm/year )if Sy = 0.002 and if Sy = 0.008

Monitoring well

Village 810-42

Umguza 6-25

6-25Cawston 2-9

Khami 6-22

Dilkosch 12 -50

Shirville2-10

Mpandeni 6-24

Umguza River

Khami River

20 kilometers



Basement Granite


Forest Sandstone

Railway

Legend

Sawmills

Fig. 6 Recharge determined with the water-table fluctuation method

1435


case where the deep roots tap the water table. Theserecharge estimates amount to 2.9 and 5% of the averageannual precipitation, respectively.

The recharge values calculated are a combination oftwo components, recharge from precipitation and influentflow from river-bank infiltration. River-bank infiltrationonly occurs after heavy rains when the streams are flowingand, as explained in the hydrogeological conceptualiza-tion, contributions to the total recharge can be assumed tobe small.

Groundwater dating using 14C gives an averagerecharge rate of 25 mm/year. This corresponds to 4.5%of the long-term average annual rainfall of 555 mm/year.The estimation is based on samples taken in boreholes inthe recharge areas. All 14C ages were determined usingequations for mean residence time computations describedby Vogel (1970) and Mook (1989). The results show thatrecent groundwater is located in the central and easternpart of the study area, while older groundwater dominatesin the west (Fig. 8). The mean residence time of the wateris a function of the flow distance from the recharge areaand the flow velocity in the aquifer. Groundwater agesincrease with distance following the flow gradient indi-cating that the assumption of Darcian flow in the mainForest Sandstone aquifer is defendable.

Groundwater flow modeling

A numerical groundwater flow model of the Nyamandh-lovu aquifer system was prepared to complement andconfirm the findings of the estimations of the recharge ratesusing the four methods discussed in the previous sections.The modeling activities comprised the formulation of theconceptual model, the specification of the model code, gridand input data, and the model calibration.

Conceptual modelIn the first place, the model conceptualization comprisesthe schematization of the model layers. Three distinctlayers were considered for the groundwater model of theNyamandhlovu area. These include the Kalahari Sand asthe upper layer (layer 1), the Karoo Basalt as the middlelayer (layer 2) and the Forest Sandstone as the bottommodel layer (layer 3). The Basement Complex mainlyconsisting of impermeable granitic-gneissic and locallydoleritic material acts as the impermeable bottom in themodel. Layer 1 was modeled as unconfined while layers 2and 3 were modeled as convertible between confined andunconfined.

Second, the conceptual model entails the formulationof a groundwater balance which can be formulated asfollows for the Nyamandhlovu aquifer system:

Qprec þ Qsurfin

� �� Qwell þ Qlso þ ET½ � ¼ 0 ð6Þ

Qprec (mm/year) is groundwater recharge into theForest Sandstone, Karoo Basalt and Kalahari Sandcomplex originating from precipitation and Qsurfin (mm/year) is the recharge supplied from surface water. Qwell

(mm/year) is the net abstraction at domestic and agricul-tural wells, mainly in the eastern and central part of themodel area. Qlso (mm/year) is the subsurface outflowwhich mainly takes place at the western margin of themodel area, and ET (mm/year) is the transpiration of thedeep-rooted vegetation possibly tapping water from deepwater tables. Obakeng (2007) proves that deep-rootedvegetation in the Serowe area, about 300 km southeast ofthe study area, is taking up transpiration water fromdeeper layers in the unsaturated zone and possibly fromthe water table. The similarity between both areas from ameteorological, hydrogeological and land use (vegetation)point of view justifies the consideration of an ETcomponent in the groundwater balance.

(

(

Umguza River

Khami River

20 kilometers20 0

N

EW

S

Groundwater level contourline in meters above sea level

1190

11301130

1120

1170

1150

1190

1120

1170

1150

11301130

11801180

1150115011401140 1140

11601160

1140

1140

11801160



Basement Granite


Forest Sandstone

Railway

Legend

Sawmills

Fig. 7 Groundwater level contour map and flownet computation method

1436


Third, the model boundaries were defined. In the eastthe model boundary is roughly the contact between theForest Sandstone and the impermeable Basement Com-plex. In the north and south, the model boundaries follow,in most places, groundwater flow lines north of theUmguza River and just south of the Khami River. Themaps in Fig. 9, which cover the active model domain,demonstrate that flow lines roughly being perpendicular tothe contour lines follow the model boundaries. Finally, inthe west the boundary mainly follows groundwater levelcontour lines west of Sawmills.

Fourth, the groundwater flow condition was deter-mined. Groundwater levels considered for the long-termmodeling period from 1989 to 2004 showed a downwardtrend from 1989 to 1999 and levels were restored from1999 onwards. For the overall period there is not asignificant up- or downward trend and steady-statemodeling was therefore performed.

Model code, grid and input of dataA numerical groundwater flow model was prepared usingthe MODFLOW-PMWIN code. A regular mesh and ablock centered finite difference grid with 1-km2 cell areaswith a total of 50 rows and 90 columns was used. With atotal grid area of 4,500 km2, the active model area is1,900 km2. This area is considerably larger, mainlystretching out towards the west, than the middle andeastern part of the study area (960 km2) for which therecharge estimation methods were mostly applied. Usingthe copying and field interpolator facilities of PMWIN, thedata input could be conveniently accomplished.

Based on ample well information, the top and bottomelevations of the three model layers were inserted in themodel. Permeabilities for the Forest Sandstone, rangingfrom 0.2 to 1.2 m/day and determined from pumping testscarried out at 21 wells, were distributed over the modelarea. Permeabilities for the Karoo Basalt and the Kalahari

Sand were taken from literature and input into the modelas single values for the whole unit.

Recharge rates, computed by the recharge estimationmethods, have been considered as input for the model.Since none of the recharge estimation methods were ableto show clear differences in recharge rate between theForest Sandstone and the Karroo Basalt areas, the samerecharge rates, mainly based on the flownet computations,have been applied for these two formations. The KalahariSands have been assigned slightly lower recharge ratesthan the sandstone and basalts (Fig. 10). Although thesands have a larger permeability than the rocks, the lowerrecharge rate is partly due to the absence of preferentialflow and the abundant vegetation growing on this unit,giving rise to high transpiration rates (see also Martinelliand Hubert 1996).

The recharge rates based on the recharge estimationmethods have been assigned to the middle and eastern partof the Nyamandhlovu model area. The other slightlysmaller part, in the west, is expected to receive a lowrecharge and very low rates have been assigned to this partof the model area. The allocation of higher recharge ratesin the middle and eastern part of the model area followsthe hydrogeological conceptualization where the reductionin recharge towards the west was already suggested.

Net abstractions from wells, which are on the order of37,000 m3/day, have been assigned to the Forest Sand-stone Formation represented by the third model layer. Thevalue is based on abstractions of 20,000 m3/day fordomestic purposes and net withdrawals of 17,000 m3/dayfor irrigation.

Model calibrationSimilar to the approach adopted for the flownet compu-tation, a case whereby the deep-rooted vegetation does notreach the water table and a case where the roots do reachthis water table have been considered for model calibra-

(

20 0 20 kilometers

N

EW

S

1100

Recent

Sampled well for carbon 14 methodResidence time in years

Recently recharged groundwater

Umguza River

Khami River

Nyamandhlovucentre

6706700

RecenRecentRecenRecent

RecenRecent

RecenRecent

125001300

1101100

> 30000

236023600

> 30000

> 30000> 30000

23600

6700

RecentRecent

Recent

Recent

125001300

1100



Basement Granite


Forest Sandstone

Railway

Legend

Sawmills

Nyamandhlovucentre

Nyamandhlovucentre

Fig. 8 Groundwater ages determined with the 14C method

1437


tion. In the latter case, it has been assumed that all thetranspiration water at a rate of 12 mm/year (Obakeng2007) is taken from the water table and not from otherdeeper layers in the unsaturated zone.

Model calibration was accomplished by varying a setof parameter values that produce simulated groundwaterlevels that match field observations. In many cases,permeabilities and recharges are considered as parametersthat should be selected for calibration and this selectionoften leads to non-unique modeling results (Scanlon et al.2002). Due to the large number and reasonable distribu-tion of pumping tests carried out in the Nyamandhlovuaquifer, it has been assumed that its permeabilities arerather accurate and recharge has been the main parameterconsidered for adjustment. In addition, some minor andlocal adjustment of abstraction rates was implementedwithout changing the total amount.

For the Nyamandhlovu model area, with a totaldifference in groundwater level on the order of 120 m, amaximum of 5 m difference between observed andcalculated groundwater levels was thought to be anacceptable calibration target. After calibration, the ground-water level patterns of observed and computed ground-water levels were quite similar (see Fig. 9) and 95% of the

level observations taken from a total of 24 wells werewithin 5 m of the computed values. The other 5% of thelevel observations were within 10 m of the modelcomputed values.

Calibrated recharge rates for the middle and easternpart of the modeled Nyamandhlovu aquifer for the casewhereby the deep-rooted vegetation does not tap the watertable were found to be 14 mm/year for the ForestSandstone and Karoo Basalt area and 10.5 mm/year forthe higher strips with Kalahari Sand. Recharge rates forthe western model area are low and on the order of 1.9–2.5 mm/year. For the case where the deep-rootedvegetation takes its transpiration water from the watertable in the recharge area (approximately 12 mm/year),this component is simulated as an outflow component ofthe model’s water balance leading to higher calibratedrecharge values. For the middle and eastern part of themodel area these recharge rates are 26 and 22.5 mm/year,respectively for the Forest Sandstone and Karoo Basalt,and Kalahari Sand areas.

Discussion

Comparison of resultsFour basic methods and groundwater modeling toestimate groundwater recharge from precipitation in theNyamandhlovu aquifer showed results that do not differthat much from each other. Table 2 summarizes therecharge rates obtained through the different methodsapplied in the study area. Despite the similarity there aredifferences in the results that can be explained as follows.The chloride mass balance method which is oftenrecommended in semi-arid regions may over-estimate

1110

1110

1120

1120

1130

1130

1140

1140

1150

1150

1160

1160

1170

1170

1180

1180

1190

1190

1100

1100 Umguza River

Khami River

Computed values

Umguza River

Khami River

Observed values

Groundwater level contourline in meters above sea level

Flow line

1140

N

EW

S

Legend

20 0 20 kilometers

Sawmills

Sawmills

Fig. 9 Groundwater levels for the Forest Sandstone aquifer

Umguza River

Khami River

Calibrated recharge rate

14 mm/year

10.5 mm/year

2.5 mm/year

1.9 mm/year

N

EW

S

Legend

20 0 20 kilometers

Sawmills

Fig. 10 Calibrated recharge rates; case with no transpiration fromthe groundwater system

1438


recharge rates in fractured rock areas. The methoddetermines point recharge which is supposed to be diffuserecharge originating from weathered and fractured parts ofthe rock at or ‘upstream’ of an observation well wherechloride samples have been taken. In these areas, therecharge estimation is correct, but the estimation can notbe extended to areas where the rock is not weathered andaquifers are overlain by impermeable layers. Such areasare present in parts of the Nyamandhlovu study area.Therefore, the computed recharge range from 19 to62 mm/year and the average rate of 37 mm/year are toohigh to be considered as averages for the typical rechargearea covering the middle and eastern part of theNyamandhlovu study area.

The water-table fluctuation method also determinespoint recharge. The highly variable specific yield in thestudy area, determined at only two locations, also hasresulted in a very broad range of recharge values rangingfrom 2 to 50 mm/year. The application of the water tablerise method in Nyamandhlovu does not allow theestimation of useful aerial recharge rates for the studyarea, but is useful in the sense that it provides wide limitsto the recharge values which can be compared with theresults of other methods.

The flownet computational method is not a methodwhereby point recharge is determined, but it integratesrecharge over a larger area. Large variations and lack ofinformation regarding transmissivity values and ground-water level contour lines may limit the application of thismethod, but for the Nyamandhlovu area, these parametersfall within fairly well-defined ranges allowing the estima-tion of recharge rates on the order of 16–28 mm/year.Being an integrating method, the computed rates providean idea on the average recharge for the middle and easternpart of the model area.

The groundwater dating method to estimate rechargealso determines point recharge. The rough estimate on theporosity (0.1) of the Forest Sandstone aquifer required forthe recharge estimation with this method could lead toerrors. However, the recharge rate estimations on the orderof 23–28 mm/year are close to the results of the othermethods. The only two recharge estimates determined

with the groundwater dating method are just locally validin terrain characterized by flow through weathered andfractured rock.

Groundwater modeling is usually engaged to provide acheck on the other recharge estimation methods, but canalso be considered as an additional method by itself. Agroundwater model provides recharge estimates for thevarious hydrogeological units. Model calibration for thestudy area resulted in recharge estimates on the order of10.5–26 mm/year for the typical recharge area in themiddle and eastern parts of the modeled study area. Areasonable set of data was available for the model and inparticular the good set of permeability values givesconfidence in the estimation of the model estimatedrecharge values.

Evaluation of methodsThe confidence in recharge estimation improves whenapplying several methods and this point of view was alsoreported by other authors (e.g. Beekman et al. 1996; DeVries and Simmers 2002). Nevertheless, the differentmethods to estimate point recharge can be rated in termsof accuracy, ease and costs following a method introducedby Bredenkamp et al. (1995). The ratings are based on thepresented results and studies carried out in SouthernAfrica.

Table 2 also summarizes the ratings following thesuggested method. The chloride mass balance methodrates high, as it is considered as one of the best pointrecharge estimation methods in terms of ease of applica-tion and costs (Bredenkamp et al. 1995). The water-tablefluctuation method also rates high and also for this methodthe simple application and low costs are strong points. Theapplied groundwater dating method is the least accurate,most difficult to apply and the most expensive method toestimate recharge. Isotope data required for the ground-water dating method are often not readily available andthe cost involved in the acquisition of the data is high. Forexample, analyzing one carbon-14 sample costs aboutseveral hundreds of Euros, and the time involved insampling and analyzing the samples is enormous.

Table 2 Recharge estimations and rating of different point recharge methods

Method Rechargerate(mm/year)

Limitations RatingAccuracya Easeb Costc

Chloride mass balance 19–62 Long-term atmospheric chloride deposition unknown. 2 1 1Water-table fluctuation 2–50 Specific yield (Sy) difficult to determine 2 1 1Darcian flownetcomputations

16–28 Poorly known transmissivities and contour lines – – –

Groundwater dating 23–28 Poorly known porosity and correction of dead carboncontribution

3 2–3 3

Groundwater modeling 11–26 Lack of data – – –

a Accuracy rating: 2 difference from true value within a factor of 5; 3 within a factor of 10 or moreb Ease of application: 1 easy to use; 2 not so easy to use; 3 difficult to usec Cost: 1 inexpensive; 3 expensive

1439


Conclusions

The hydrogeological conceptualization with regard torecharge is confirmed by the results of the rechargeestimation methods. Recharge in the Nyamandhlovuaquifer is dominated by direct and localized percolationfrom precipitation during the rainy season whereas theinfluent flow from rivers and irrigated fields is small.

Where Forest Sandstone is exposed at the surfacerevealing unconfined conditions, recharge from precipita-tion takes place directly into this unit, which can beconsidered the main aquifer of the area. In the placeswhere weathered and fractured Karoo Basalt crops out,overlying the Forest Sandstone, recharge also takes placeinto the basalt and subsequently the water is transferred tothe sandstone. Finally, recharge may be accommodatedthrough the Kalahari Sands and is also transmitted to theForest Sandstone where the Karoo Basalt is permeable.

The study has revealed that the main contribution ofrecharge to the Nyamandhlovu aquifer occurs in themiddle and eastern parts of the study area. The increasingthickness of the basalt towards the west indicates areduction in recharge in this direction. However, theresults of groundwater dating showing the virtual absenceof recently infiltrated water into the Forest Sandstoneaquifer in the western part of the study area and theoutcome of the model calibration confirms the very lowrecharge rates in these areas.

This study has also shown that each of the differentrecharge estimation methods has its own individual merit.The chloride mass balance method, the water-tablefluctuation method and the groundwater dating methodare well suited to identify the existence of recharge andenable one to determine good estimates of point rechargevalues. Especially in fractured rock terrain, however, theextension of the point recharge values determined withthese methods to a whole study area tends to over-estimatethe aerial recharge. Because the methods integrate therecharge over selected areas, the flownet computationmethod and groundwater modeling provide better esti-mates of aerial recharge.

For the middle and eastern parts of the studiedNyamandhlovu aquifer, the estimates for recharge fromprecipitation using the flownet computation method andthe groundwater modeling approach range from respec-

tively 16–28 and from 10.5 to 26 mm/year. Since thetransmissivity values and groundwater level contour linesused for the flownet method are less accurate thangenerated in groundwater modeling, the results of themodeling approach should be considered more reliable.The lower ends of the recharge ranges apply to the casewhereby there is no uptake of water from the water tableby deep-rooted vegetation and the upper ends relate to thecase whereby all transpiration water obtained from deeperlayers is obtained from the water table. Both cases arebelieved to apply for the Nyamandhlovu aquifer system.In the areas with shallow water tables on the order of 5–10 m below surface near the streams, and where the rockis soft, the direct uptake of water from the water table mayprevail and in the higher areas where the water table is30–40 m deep, and the rock becomes hard, the roots maytake the majority of the deep transpiration water from theunsaturated zone. It is reasonable to assume for therecharge area covering the middle and eastern part ofNyamandhlovu, aerial recharge rates in the range of 15–20mm/year representing 2.7–3.6% of the long-term averageannual rainfall of 555 mm/year.

The range of recharge rates obtained in this study mostlyagrees with the findings of similar studies in the region.Larsen et al. (2002) predicted for adjoining areas in the MidZambezi basin recharge rates on the order of 20–25 mm/year,Obakeng (2007) found for its similar region in northeastBotswana a recharge rate of 15 mm/year, and Beekmanet al. (1996) adopted recharge values for Botswana in therange of 10–25 mm/year. Comparing with previousrecharge studies carried out in the Nyamandhlovu area, itcan be concluded that the computed recharge rates underthis study are not far off from the estimates by Martinelliand Hubert (1996) ranging from 14 to 28 mm/year. Theestimate of 125 mm/year adopted by Beasley (1983) isclearly too high and also does not fit into the predictions ofthe other researchers.

With the groundwater flow model, total recharges,including some inflow from streams, of 16.7×106 m3/yearand 29.0×106 m3/year, respectively for the cases withoutand with the uptake of deep transpiration water from thewater table in the recharge area, have been computed forthe modeled Nyamandhlovu aquifer (Table 3). For thesecond case, the uptake of deep transpiration wateramounted to 12.2×106 m3/year. Assuming (again) that

Table 3 Model groundwater balance for the Nyamandhlovu aquifer

Groundwater inflows (m3/year) Groundwater outflows (m3/year)

Case 1: no deep root transpiration from water tableRecharge from precipitation 15.9×106 Well abstraction 13.5×106

Recharge from streams 0.8×106 Net lateral sub surface outflow 3.2×106

Transpiration 0Total 16.7×106 16.7×106

Case 2: with deep root transpiration from water table in recharge areaRecharge from precipitation 28.2×106 Well abstraction 13.5×106

Recharge from streams 0.8×106 Net lateral subsurface outflow 3.3×106

Transpiration 12.2×106

Total 29.0×106 29.0×106

1440


both cases of water uptake by deep-rooted system areapplicable in the area, then a total recharge in the range of20×106–25×106 m3/year could be realistic for the studyarea. If at all possible, it seems unwise to use theseestimated quantities of recharge all for domestic andagricultural abstractions in view of the ecological damagethat will be done to the deep-rooted vegetation. Theoptimum groundwater availability for abstractions in thearea is the quoted 16.7×106 m3/year. This means that anysubstantial extension of the abstractions would not bepossible in view of the existing abstractions. Thisavailability assessment will be a crucial input for the“Catchment Outline Plan” for the Gwayi catchment.

Acknowledgements The authors would like to thank the Dutchgovernment for providing sponsorship for the studies of the firstauthor at UNESCO-IHE, Delft, The Netherlands. We would alsolike to thank members of staff of the Zimbabwe National WaterAuthority for their support and assistance during data collection andfieldwork. Many thanks go to the University of Zimbabwe, GeologyDepartment, in particular to Mr. Pride Mangeya for providingadditional data and advice during the research.

References

Allison GB, Barnes CJ, Hughes MW, Leaney FWJ (1984) Effects ofclimate and vegetation on oxygen-18 and deuterium profiles insoil. In: Isotope hydrology 1983, Proceedings of a symposium,September 1983, IAEA, Vienna, pp 105–123

Banda WM, Hindson L, Wurzel P (1977) Nyamandhlovu explor-atory drilling report. Ministry of Water Development, Harare,Zimbabwe

Beasley AJ (1983) The hydrogeology of the area around Nya-mandhlovu, Zimbabwe. PhD Thesis, University of London,London

Beekman HE, Sunguro S (2002) Groundwater recharge estimation -Suitability and reliability of three types of rain gauges formonitoring chloride deposition. Proc. Conf. Tales of a HiddenTreasure, Somerset West, South Africa, 16 September 2002, pp225–233

Beekman HE, Gieske A, Selaolo ET (1996) GRES: groundwaterrecharge studies in Botswana 1987–1996. Botsw J Earth SciIII:1–17

Bredenkamp DB, Vogel JC (1970) Study of a dolomite aquifer withcarbon-14 and tritium. In: Isotope hydrology 1970, Proceedingsof an IAEA symposium, Vienna, 9–13 March 1970

Bredenkamp DB, Botha LJ, van Tonder GJ, van Rensburg HJ(1995) Manual on quantitative estimation of groundwaterrecharge and aquifer storativity. WRC Report TT 73/95,WRC, Pretoria, South Africa, 407 pp

De Vries JJ, Simmers I (2002) Groundwater recharge: an overviewof processes and challenges. Hydrogeol J 10(1):5–17

De Vries JJ, Selaolo ET, Beekman HE (2000) Groundwaterrecharge in the Kalahari, with reference to paleo-hydrologicconditions. J Hydrol 238:110–123

Fetter CW (2001) Applied hydrogeology, 4th edn. Prentice Hall,Englewood Cliffs, NJ

Gieske ASM (1992) Dynamics of groundwater recharge: a casestudy in semi-arid eastern Botswana. PhD Thesis, FreeUniversity, Amsterdam

Holman JP (2006) Climate change impacts on groundwaterrecharge: uncertainty, shortcomings, and the way forward?Hydrogeol J 14(5):637–645

Kinzelbach W, Aeschbach W, Alberich C, Goni IB, Beyerle U,Brunner P, Chiang W-H, Rueedi J, Zoellmann K (2002) Asurvey of methods for groundwater recharge in arid and semi-arid regions. Early Warning and Assessment Report Series,UNEP/DEWA/RS.02–2. United Nations Environment Program,Nairobi, Kenya

Larsen F, Owen R, Dahlin T, Mangeya P (2002) A preliminaryanalysis of the groundwater recharge to the Karoo formations,mid-Zambezi basin, Zimbabwe. Phys Chem Earth 27:765–772

Lerner DN, Issar AS, Simmers I (1990) Groundwater Recharge: aguide to understanding and estimating Natural Recharge.International Contributions to Hydrogeologists, vol 10, IAH,Goring, UK

Macgregor MA (1937) The Geology of the area around Queen’smine, Bulawayo District, Geological Survey Bulletin 30,Reprinted 1972, Government printers, Harare, Zimbabwe

Mangeya PM (2003) Groundwater Resource evaluation of theUmguza Aquifer, Zimbabwe. University of Zimbabwe, Harare,Zimbabwe

Martinelli GL, Hubert GL (1996) Consultancy services forgeophysical and hydrogeological investigations, aquifermodeling and monitoring of the Nyamandhlovu aquifer,Bulawayo, Zimbabwe. Royal Danish Embassy, Harare,Zimbabwe

Mook WG (1989) Principles of isotope hydrology. Centre forIsotope Research, Groningen, The Netherlands, 153 pp

Nonner JC (2006) Introduction to hydrogeology. IHE Delft Lecturenotes series. Taylor and Francis, London

Nyagwambo L (2006) Groundwater recharge estimation and waterresources management in a tropical crystalline basementaquifer. PhD Thesis, Delft University of Technology, TheNetherlands

Obakeng OT (2007) Soil moisture dynamics and evapotranspira-tion at the fringe of the Botswana Kalahari, with emphasis ondeep rooting vegetation. PhD Thesis, Free University, theNetherlands

Scanlon R, Healy RW, Cook PG (2002) Choosing appropriatetechniques for quantifying groundwater recharge. Hydrogeol J10(1):18–39

Simmers I (1987) Estimation of natural groundwater recharge.Reidel, Amsterdam, The Netherlands

Siwadi J (2001) Redesign of groundwater level monitoring networkin the Nyamandhlovu aquifer, Zimbabwe. MSc Thesis, IHE,Delft

Vogel JC (1970) Carbon-14 dating of groundwater. In: Isotopehydrology 1970, Proc. Symp. IAEA, Vienna, pp 225–237

1441


Documents

Comparison of groundwater recharge estimation methods for ... · rect recharge refers to water that percolates to the groundwater through the beds of surface watercourses. Localized