UTTERWORTHIE I IM E M A N N

'<<•< fnni'n. Vd) 19. No I. pp. 7) -XX IW.S

l:.Ui-VliT Stii'TKV l.HlC'opvri{.hi © I'W5 Ini i id N.H«>n<

Primed in Gri.il Brilain. All rithls rcM-ncd

The impact of small farmreservoirs on urban watersupplies in Botswana

Jeremy Meigh

In eastern Botswana there are many small farm reservoirs within the catchments of the majorwater supply reservoirs, and there is increasing demand for more small reservoirs. Theincreasing development of farm reservoirs has an impact on the availability of water from themajor reservoirs, which supply urban and industrial users, and this creates a conflict betweenthe needs of the rural water users and the urban and industrial users. This paper describes amodel which has been developed to allow the effects of the existing small reservoirs and thepossible impacts of future proposed ones on the water resources of the major reservoirs to bequantified. The model provides a planning tool, enabling guidelines for future small reservoirdevelopment to be determined. The model is a general one which could also be calibrated andapplied in other areas with a broadly similar climate. The results of a series of model runsindicate the rate of decline of runoff and yield from the major reservoirs as the total capacityof small reservoirs within the catchment increases. It also shows how this decline is affectedby secondary factors such as the relative location of the small reservoirs within the catchment,the typical size of small reservoirs and the type of use to which they are put. The results clearlyindicate the adverse effect which uncontrolled development of farm reservoirs would have onthe water supplies from the major reservoirs. By quantifying these effects, planners have someof the necessary information to determine the optimum balance between development of small-scale rural water supplies and large-scale urban supplies-

Urban water supply in Botswana is largely dependent onsurface water sources and three major reservoirs havebeen constructed to supply the urban centres. However,within the catchments of these reservoirs there is alsoconsiderable demand for agricultural water supply, andover rhe years large numbers of small dams have beenconstructed.1 Clearly the small dams have some impacton the flows into and the yields from the major reser-voirs, but the size of this impact is unknown. There isincreasing demand for more small dams in the catch-ments, leading to a conflict of interest between the rural

Institute of Hydrology. Wallingford. Oxfordshire. OX 108BB.UK

'in Botswana the word dam refers to the reservoir itself ratherthan just the water-retaining wall, and it is used with this mean-ing in ihis paper.

communities and the urban users. In order to quantifyboth the effects of the existing small dams and the possi-ble impacts of future proposed dams, a model of thesystem has been developed which allows these effects tobe simulated. The model provides a planning tool,enabling guidelines for future small dam development tobe determined which provide the maximum benefit torural users of small dams while safeguarding the urbansupplies from the major reservoirs.

This paper briefly outlines the methodology behindthe niodel structure and its calibration, but its main pur-posdr is to present some of the results obtained fromapplying the model and to discuss the implications forrural water resources development.

The niodel was developed for the BotswanaDepartment of Water Affairs (DWA) by the Institute ofHydrology in cooperation with Sir Alexander Gibb andPartners (Gibb/lnstitute of Hydrology. 1992). It was

71

- \3353

Small farm reservoirs in Botswana: J Meigh

Shashe catchment

Kumakwane

Gaborone dam

KEY

Major water supply dam

Flow record

Small dam

Sub-catchment division

Sand river

Wetland

Figure 1 Location plan

constructed and calibrated for the catchments of thethree existing major water supply reservoirs inBotswana (Figure I), but provision has been made for itto be easily reconfigured to run for additional catch-ments when more major dams are developed. It has beendesigned as a user-friendly piece of software, running

ueder Windows 3.1. and has been installed in DWA'soffices in Gaborone.

There have been a number of previous studies of theeffects of farm dams in semi-arid climates, for instanceby Moolman and Maaren (1986) and Chunnet Fount-(1990) in South Africa. Mostert (1990) in Namibia and

72 Natural Resources Forum 1995 Volume 19 Number I

I I

Small farm reservoirs in Botswana:./ Mei^h

Neil and Srikanthan (1986) in Australia. These studieswere conducted on broadly similar principles andshowed broadly similar results lo the present one. All.however, were tied to a specific network of dams, andnone went so far as to produce a generalized modelwhich can be used to assess the impact of a range ofproposed dams.

The study catchmentsThe location and some of the main features of thethree catchments are shown in Figure 1; they lieclose to the eastern borders of Botswana where themajority of the population is concentrated. The cli-mate of the area is semi-arid, with annual rainfall of450-550 mm, falling almost exclusively fromOctober to April. Rain generally falls as short intensestorms which are often highly localized. The soils ofthe catchments are mostly sandy, and it is usuallyonly after several days of storms that any significantrunoff occurs, the rivers being dry much of the timeeven in the rainy season. Runoff events typicallyoccur as spates lasting two or three days up to aweek, with 75% or more of annual runoff concen-trated into the period December to March, but withvery high variability in flows from year to year.Annual open water evaporation rates are about2000 mm, greatly in excess of rainfall. This patternof runoff and evaporation has significance when con-sidering the effect of small dams. At the beginning ofthe rainy season almost all dams are dry. The damshave their greatest impact at this time as they fill upwith the first few runoff events. If more runoff fol-lows fairly soon afterwards it tends to pass throughthe dams little changed, but if as often happens thereis a longer dry period, subsequent flows will again beconsiderably affected as the drawndown dams fill uponce more.

The small dams are typically between 1000 and100 000 m3 in size, with a few considerably largerthan this. This compares to the three major damswhich range from 19 to 144 million cubic metres(Mm3). A total of 320 small dams were identified inthe catchments; the majority are used for stock water-ing, with a lesser number for small-scale irrigationschemes and other purposes.

Model developmentThere were three main stages in the development ofthe model: estimation of the dam characteristics (forwhich very few data are available): construction of themodel itself; and model calibration. Each of these areoutlined briefly here (for more detail, refer toGibb/Institute of Hydrology, 1992).

J Catcfirnvnt: • Gaborone

I o Srusha

£ IOO -

Surface area (ha)

Figure 2 Dam capacity against surface area

Estimation of dam characteristics

For most of the small dams, no data were available ontheir physical characteristics, but to model them it wasnecessary to have estimates of at least the followingfor each: the surface area and capacity, the area/capac-ity relationship as the dam is drawn down, and theabstractions. The locations of the dams were identifiedfrom 1:50 000 scale mapping and from records heldby government departments. For a small sample of thedams, physical surveys were carried out and additionalinformation was sought on the usage of water by thelocal people. This information was used to derivemethods to estimate the dam characteristics.

Surface areas and capacities

The surface areas of all the dams were estimated fromthe size of the blue area shown on the 1:50 000 maps.It might be expected that the area estimated from themaps would be a poor measure of the actual areabecause the aerial photography on which the maps arebased is unlikely to correspond to times when thedams were full, and because many of the dams arevery small and are indicated by symbols which maynot be representative of the actual areas of the dams.However, when compared to the measured surfaceareas for the sample of dams surveyed, there appearedto be negligible bias, and it was concluded that it wasadequate to use the map areas. This result was alsoconfirmed by measurements made from satelliteimages of the region.

Heaving established the surface area of a dam, it isnext.necessary to estimate its capacity. Figure 2 showsthe capacities plotted against area for the surveyeddams. The best (it relationship was

capacity = 7.381 area1-251 {r 2 = 93.1%)

where capacity is in thousand m3 and area in hectares.This fairly strong relationship results because geology

Natural Resources Forum 1995 Volume 19 Number I 73

Small farm reservoirs in Botswana: .1

i

Eo

Bokaa catchmentGaborone catchmentShashe catchment

04 06

Capacity as proportion of maximum, PV

Figure 3 Typical dam area/capacity curves

and topography are generally uniform over the threecatchments, with the great majority of the dams lyingin similar relatively broad, flat valleys.

Area/capacity curves

The area/capacity curves for a selection of the surveyeddams were made non-dimensional by expressing bothcapacity and area as proportions of their values when thedams are full, and were then compared as shown inFigure 3. Although there is a fairly wide variety of dif-ferent curves, many are quite similar and these areconcentrated at the centre of the others. A mean curvewas estimated by eye. and this was taken as the typicalcurve, used to represent all dams for which full surveydata are not available. It is clear that in many cases thecurve will represent actual sites poorly, but thearea/capacity curve is of secondary importance com-pared to the estimates of dam capacity and surface area.

Abstractions

The majority of the dams are used for stock watering,and although it is likely that the amount abstracted isvery much less than losses to evaporation, it wasthought worthwhile to include some estimate of it. In thesurvey it was found that no systematic data are kept ofthe numbers of cattle using particular dams, and in anycase numbers vary widely as stock are moved aroundaccording to the availability of water and grazing.However to make some allowance for abstractions asomewhat arbitrary assumption was made: that 10 cattledrink at a dam for each 1000 nv1 of capacity. Goats and

other animals were ignored. Each head of cattle wasassumed to consume 31.5 I/day (SMEC et al, 1991).This approach gave cattle numbers which were gener-ally in accordance with the typical numbers thought tobe in the area.

For dams used for irrigation, the monthly water usefor a typical range of crops and typical irrigation effi-ciency was estimated from work carried out for otherstudies (SMEC et al, 1991; Arup Botswana, 1991:Chunnet Fourie, 1990). For testing the model with irri-gation dams, yield estimates for the dams were alsoneeded. Some curves relating yield to dam capacity.both expressed as proportion of mean annual runon.were derived from detailed yield analyses which havebeen carried out in the past for a number of sites (egGibb. 1985).

Model constructionThe model was required to simulate the daily flows intothe major water supply reservoirs over a period of sev-eral years. This needed to be done for each of the threeconditions of no small dams, the existing small dani^

jpnly, and existing plus proposed dams in the upstreamcatchments. It was necessary to allow any combinationof proposed small dams to be examined, and to providefacilities for comparisons with observed data to permitmodel calibration. The model was designed so that itcould'be easily reconfigured to run for other catchments.

Because of the large number of small dams and theneed to allow new dams to be added, a fully distributedapproach where the runoff into each dam would be esti-mated separately, was not practicable. But to model the

74 Natural Resources Forum 1995 Volume 19 \nmhcr I

1

Small farm reservoirs in Botswana: .1 Meigh

STARTrCreate composite dams

in separate categories.

_LRainfall/ runoff module:

generate flows for the local sub-catchment

Dam module (dams on side branches only):route appropriate proportion of local flows through

the dams and add back the unrouted part

:- Loop for EX.EX+P dams i

Is there an upstream sub-catchment? •©If necessary sum flows

from upstream sub-catchment(s)

Channel routing and transmissionloss modules

Loop for No, EX, EX+P dams iOam module: route flows from upstream sub-catchment(s)

through dams on main channel

Add flows from local sub-catchment

Is there a special dam ?

Loop for EX.EX+P dams Dam module: special dams

Store flows for this sub-catchment

tIf in calibration mode, compare to observed flows

KEY

Summary and plots of results

FINISH

Figure 4 Simplified overall siructure of the model

Loop for eachsub-catchment

EX = Existing dams

EX+P = Existing + proposed dams

dams with some degree of realism and because of thetemporal and spatial variability of rainfall in Botswana,treating the whole catchment as a lumped unit was alsoinadequate. Therefore a semidistributed approach wastaken. Initially each catchment was divided into a number

of subcatchments whose boundaries were then fixed: thischoice of subcatchment divisions is a key initial decisionin setting out to model a catchment. Each of the subcatch-ments is treated as a lumped system, within which it isassumed that hydrological processes are approximately

Natural Resources Forum IW5 Volume 19 Number I 75

Small farm reservoirs in Botswana: J

uniform. The proportion of runoff which is routedthrough each dam is taken as the proportion of the sub-catchment area controlled by that dam. To make someattempt to account for the spatial variability of rainfall, aseparate rainfall input is used for each subcatchment.

The model runs with a daily time step because it wasrequired that the effects of the dams on short-periodflows could be examined as well as on monihly totals.This choice was a compromise: to represent the tempo-ral variability of the processes as fully as possible,hourly or even shorter time intervals would be required,but considerations such as the computation time andlack of suitable data, mean that a daily time step was theshortest practicable.

The model consists of a number of modules for rain-fall/runoff transformation, for routeing through dams,and for channel routeing and transmission losses; theseare described in the following sections. The overall struc-ture (illustrated in Figure 4) can best be understood byfollowing the processes involved in a typical model run.Each subcatchment is processed in turn, working in thedownstream direction. For each, the steps are as follows:

(1) The dams in the subcatchment are divided into anumber of different categories, and within thesethey are amalgamated into composite dams.

(2) A rainfall/runoff model generates a flow series forthe local subcatchment.

(3) A proportion of these flows is routed through thecomposite dams which are on side branches of themain channel and the unrouted portion of theflows is added back to the routed part.

(4) The next part of the process is only carried outwhen there are other subcatchments upstream ofthe current one. The summed flows from theupstream subcatchments are passed through achannel routeing and transmission loss model toaccount for time delay and attenuation of the floodwave and for any significant transmission losses.

(5) The flows are routed through the dams on themain channel. Again, this is only when there areupstream subcatchments.

(6) Dams can also be defined as special dams whichcan only be at the downstream end of a subcatch-ment. The total flow from upstream plus the localsubcatchment runoff are routed through this ifthere is a special dam.

At this point the processing for a single subcatchmentis complete and the flows are saved to provide the inputsto other subcatchments further downstream. Each re-maining subcatchment is processed in the same manner.

The processing of the appropriate parts of the systemis carried out for the three cases of no dams, existingdams only and existing plus proposed dams. Thus, thefinal results are the flow series for these three cases:

summary statistics are then calculated and a range ofplotting options are available to allow the flow series tobe compared. An additional option allows the model tobe run in calibration mode. In this case the How seriesfor existing plus proposed dams is not computed:instead the flows for no dams and for existing dams onlyare compared to the observed flows at a river gauge.

Within the overall model, there are three separatemodules (to describe the processes of rainfall/run offconversion, dam routeing, and channel routeing/trans-mission losses). These modules are not described indetail here, but a very brief outline is given of each.

Rainfall!run off module

The rainfall/runoff module used was the Pitman dailymodel (Pitman, 1976). This is a physically based modelwhich, for each subcatchment, converts the rainfall inpu:from a single raingauge into the runoff from that sub-catchment. The Pitman model was chosen because thereis a monthly version of it (Pitman, 1973) which has beenused on most water resources studies in Botswana and isknown to give generally good performance for the con-ditions there. The daily model has essentially the samestructure as the monthly model, and would therefore beexpected to do likewise.

Dam module

The dam module carries out the function of determininghow much water is lost from the dams (by evaporationand abstraction) and how much continues downstream.To make the modelling practicable, and because thereare a large number of dams, they were combined intocomposite dams within each subcatchment. Because ofthe different behaviour of dams in different location*and with different uses, the dams were divided into ,inumber of categories before creating the compositedams. The categories used were, first, whether the damis on a main branch of the river (which receives flowfrom an upstream subcatchment), or on a side branch(which does not receive flow from an upstream sab-catchment). And, secondly whether the dam is used forstock watering or for irrigation. An additional categoryof special dams was also used to deal with the \c\\much larger dams which behave rather differently to thetypical small dams. The assumption, which is beliewi.!lo be a reasonable one, in treating the dams as compos-kes of many small dams is that all the dams in aparticular category behave in the same way: each isfilled and drawn down at the same rate in relation to ii>own capacity.

Within each category the lumped or composite dam iscreated by summing the capacities, surface areas andabstractions of each, and a composite area/capacity

76 Natural Resources Forum 1995 Volume 19 .\'«ml'<i I

Small farm reservoirs in Botswana: .1 Meigh

o

" E

|

50 -

40 -

30 -

20 -

10 -

0 -

I

A 1

11 1

fI'

Observed 1Simulated!

I A..A.,70 71 72 73 74 75

Year

76 77 78 79

60

50 A

30 A

20 A

10 H

ObservedSimulated

80 81 82 83 84 85 86 87 88 89

Year

Figure 5 Model calibration for Gaborone catchment: comparison of monthly flows

curve is created as a weighted average of the individualarea/capacity curves for each dam.

To determine the proportion of the runoff from thesubcatchment which is routed through the dam, the pro-portion of the subcatchment controlled, AC, is used.This assumes that the runoff generated for a subcatch-ment is uniformly distributed over it. A composite valueof AC is found by summing the values for each dam.This is satisfactory provided that no dam is upstream ofany other dam, but when dams lie within the area whichis controlled by a downstream dam, AC is set to zero forthe upstream dams. If this were not done, the flows to berouted through the dams and thus their impact, would beoverestimated. This approach will still cause a loss ofaccuracy as the upstream dams are effectively trans-

ferred to the location of the downstream dam in thelumping procedure, altering the ratio of the dam capac-ity to its mean annual runoff. However, this occurred fora relatively small number of dams and its effect should

f be relatively minor.

GJ\annel romeing and transmission loss moduleChannel routeing and transmission loss models arerequired for the subcatchments which have othersupstream to provide a time delay and some attenuationof the upstream flows, and to account for significanttransmission losses in some of the main channels. It isimportant to model the losses as in some cases they havea very considerable impact on total catchment runoff

Natural Resources Forum J 995 Volume 19 Number I 77

Small farm reservoirs in Botswana:

Table 1 Model calibration: comparison of annual statistics

H hale calilitncnisGaborone

Bokaa

Shashe

Parts of catchmentsNnywane

Thamaga

Kumakwane

Mooke weir

Catchmentarea (km2)

3983

3570

3650

238

1320

166

2460

Period(years)

20

15

20

20

4

9

11

ObsSimObsSimObsSim

ObsSimObsSimObsSimObsSim '

MAR(Mm')

30.230.313.113.1

106.0106.2

2.212.27

14.614.71.191.05

48.847.7

Mean(log)

1-0961.0760.8480.8381.8541.840

0.0050.0301.0160.947

-0.223-0.250

1.5871.590

SD(iog)

0.7900.7800.5570.5290.4100.413

0.6380.5920.3910.5370.5650.5140.3500.276

(approximately 50% of the total runoff in the Bokaacatchment).

Channel routeing was carried out by the simpleMuskingum procedure using hourly flows, estimatedfrom the daily values.

There are two main sorts of transmission lossesoccurring in Botswana. One is losses into the bed ofsand rivers, where the sand bed stores significant quanti-ties of water through the dry season and is rechargedwhen the river flows. The second, which is found in theBokaa catchment, is a wetland area beside theMetsemotlhaba river which again is recharged by riverflows and loses water by evapotranspiration through thedry season. The main areas where these two types oflosses occur are indicated in Figure 1. Both types of losswere modelled in a fairly simplistic manner by treatingthem as storages. These absorb a proportion of the riverflow until the store becomes full. After this point, allflow is passed downstream. In periods of no flow, thestore is gradually depleted, and this was treated as if it iscaused solely by evaporation. Although other processessuch as seepage losses and downstream flow within thesand bed also occur, these are probably much lessimportant, and for simplicity they are included with theevaporation. The determination and initial calibration ofthese models was assisted by survey data on the size andstorage capacity of the sand beds. and. for the wetland, asatellite image of the region helped to determine itsoverall extent (about 12 km-).

In the Bokaa catchment where both sand rivers andwetlands occur, there are some periods of simultaneousdata at flow gauges upstream and downstream of theseareas (see Figure 1), and these were used to test thetransmission loss model. The results indicated that the

simple storage approach performs reasonably well. Itwas found that the wetland was a major cause of waterloss from the Bokaa catchment, losing about12Mm3/year compared to a downstream annual runoffof about 13 Mm3. The sand rivers both in this case andin the Shashe catchment were relatively insignificant,losing about 1 Mm3 each.

Model calibrationThe model was calibrated for the three study catchmentsupstream of Gaborone, Bokaa and Shashe reservoirs,using the subcatchments as indicated in Figure 1. Aproblem arises with the calibration because the observedriver flows include the effects of the existing dams, butthe number of dams in use must have varied during thecalibration period. Ideally, the date of construction ofeach dam should be known so that its effect can beincluded in the simulation at the appropriate time.However, very little data on construction dates wereavailable, but it did appear that the majority werealready in existence before the start of the calibrationFor the purposes of calibration therefore, it was assumedthat all the dams existed for the whole period. Generally,the model tended to overestimate observed flows in theearlier part of the calibration and underestimate in thelater period (Figure 5), whereas if significant numbers of•small dams had come into use during this period theopposite would have been expected. Thus this assump-tion does not appear to cause any inconsistencies,indicating that the change in the number of dams duringthe calibration period must have been small.

The main emphasis in the calibration was on obtain-ing good agreement for statistics of annual runoff, and

78 Natural Resources Forum 1995 Volume 19 \wither I

I I

Small farm reservoirs in Botswana: J Mei);li

Table 2 Summary of effects of existing small dams

GaboroneBokaa (a)

(b)Shashe

Existing damsNumber

20310310314

CapacityiMm-'l

26.093.583.580.13

MAR (Mm') forNodams

40.4315.0117.96

106.45

Existingdams

30.2813.0715.57

106.24

Percentagechangein MAR

-25.1-12.9-13.3-0.2

(a) using calibration period; (b) using test period with reduced wetland area.

these results for each of the main catchments and for anumber of locations within the catchments are given inTable 1. It can be seen that excellent agreement has beenobtained, and this means that both the mean flow and itsvariability are reproduced by the model.

The seasonal distribution of flows was also examined,and this was not as good as hoped with a tendency forunderprediction in December and overprediction inApril, but this could not be improved without loss ofagreement for the annual statistics which were consid-ered to be the most important indicators of modelperformance. An example of the calibration is also indi-cated by the comparison of simulated and observedmonthly flows for the Gaborone catchment (Figure 5).This shows that the model is not able to cope with thedry period of the mid-1980s, tending to underpredict atthis time. One of the main factors contributing to thismay be the general inadequacy of the rainfall data.Rainfall in Botswana is generally very localized andthere is a great shortage of unbroken rainfall records,which means that rainfall records often do not give agood representation of the actual rain in any particularlocation. In these circumstances, the agreement betweenindividual monthly flow values is as good as can reason-ably be expected.

Model parameters were also needed for subcatchmentareas which have no observed flow data, and generally,it was assumed that, lacking any better information, thesame parameters should be applied to each subcatch-ment. In light of the broadly uniform topography,vegetation and soils over the catchments, this assump-tion is reasonable. The exception is the Bokaa catchmentwhere observed data within the catchment did allowsome differentiation between subcatchments.

Assessment of impact of small damsA set of runs of the hydrological model over the 20year period 1970-71 to 1989-90 was carried out to testthe impact of small dams. The principal factor investi-gated was the total capacity of the small dams. In orderto make the tests comparable between the differentcatchments, the total capacities expressed as a percent-

age of the mean annual runoff, MAR, of that catchmentwere approximately the same for each; small dams ofabout 2, 5. 10, 20 and 40% of MAR were used. Somesecondary factors in determining the effects of smalldams were also investigated. These were the size of thetypical small dam tested; the proportion of the catch-ment controlled by the dams; and the type ofabstraction from the dams. A typical small dam wastaken as having a capacity of 50 000 m3 as this is aboutthe size that is commonly built, and smaller and largerdams of 5000 and 500 000 m3 were compared to this.For the proportion of the catchment controlled by thesmall dams, AC, a value of 0.25 was taken as typical,but smaller and larger values of 0.1 and 0.5 were alsoexamined. In effect, this is looking at the impact of theposition of the dams in the catchment; as AC isincreased the same total capacity of dams are placedfurther downstream. To look at the effect of abstrac-tions from the dams, the tests were first carried outassuming that all proposed dams are used for stockwatering, and then repeated assuming they are used forirrigation.

The results of the model runs are flow series with dif-ferent combinations of proposed dams, which indicatethe impact on MAR, runoff patterns and seasonal flowdistribution.

Impact of existing dams

The effects of the existing dams on the runoff from thecatchments were determined in the course of the modelcalibration work since simulated flows taking intoaccount the existing dams were matched to the observedflows. The results are listed in Table 2; the effects varyfrem a negligible change in the Shashe catchment to a25% decline in MAR for Gaborone which has 26 Mm'of existing dams.

In'the case of Gaborone. 18 Mm3 of this total capacityof small dams is contained in a single large dam whichcontrols only a small part of the catchment, and becauseof this the decline in MAR is less than would otherwisebe expected. The simplification inherent in the modelmeans that the effect of small dams mav be slichtlv

Natural Rcstiui'ces /99.5 Volunte 19 Number 1 79

I

Small farm reservoirs in Botswana: J Meigh

(a) Gaborone

"75.

3ss§>

I

0

•1

•2

•3

-4

-5

•6

-7

(b) Bokaa

IO

(c) Shashe

1

Total capacity of proposed dams (Mm3)

14— I -

Prop, catchmentcontrolled. AC =- 0.1- 0.25<• 0.5

—i r 1 r10 20 30 4C

Total proposed dams as percentage of MAR3

Total capacity of proposed dams (Mm )2 4 6

• 10 cQ>a>

I - - 2 0

o.-30

0

•0.5

•1

-1.5

-2

•2.5

-3

-3.5

-4

•4.5

Prop, catchmentcontrolled, AC°• 0.1- 0.25• 0.5

1u<D

•20 i

<DQ .

10 20 30Total proposed dams as percentage of MAR

Total capacity of proposed dams (Mm3)10 20 30

40

40

o -+-

-15 '

•20

•25 ~i

•30

-35

Prop, catchmentcontrolled, AC=- 0.1- 0.25<• 0 .5

1 i 110 20 30

Total proposed dams as percentage of MAR

-10

• 2 0

L -30

40

Figure 6 Change in catchment MAR with increasing small dams and varying pro-portion of catchment controlled: (a) Gaborone, (b) Bokaa, (c) Shashe catchment(stock-watering dams with a typical capacity of 50 000 m3 are assumed)

toverestimated when many dams are nested, so that theseestimates of MAR for no dams should be thought of asupper limits for the undeveloped catchment MARa.

Impact of proposed dams

The changes in MARs for increasing total capacity ofsmall dams in the three study catchments are plotted inFigure 6. This also shows Ihe effect of varying the pro-

* portion of the catchment controlled by the dams. AC. Inall cases it is assumed that each proposed dam has aCapacity of 50 000 m-' and is used for stock watering. Theeffects of small dams are basically similar in each of thethree catchments; the total capacity of small dams is !hcoverriding factor affecting the runoff. Broadly, dams hav-ing a total capacity of 10% of MAR cause roughly thesame decline of between about 8 and 10% in catchmentMAR. For greater total capacities of small dams the effect

Natural Resources Forum 1995 Volume 19 Number I

Small farm reservoirs in Botswana: .1 Meigh

(a)o

-i -

•2 -

-3 -

-4 -

•5 -

•6 -

•7 -

•8 -

-9

Total capacity of proposed dams (Mm3)4 6 B 10 12

Slock damsAC= 0 25

Typical dam size =- 500,000 m 3

* 50,000 m 3

" 5,000 m3

a.<

_ .10

- -20 •£uo3

Q.

• 30

10 20 30Total proposed dams as percentage of MAR

40

(b) Total capacity of proposed dams (Mm )2 4 6 8 10 12

• 1 -

• 2 -

-3 -

-4 -

-6 -

•7 -

-8 -

• 9

Irrigation dams ,Typical dam = 50,000 rr?

Prop, catchmentcontrolled, AC=• 0.1- 0.25• 0.5

-10 oD)Cre

•5<DO>(0

-20 S

fotal20 30

proposed dams as percentage of MAR

- -3040

Figure 7 Change in catchment MAR with increasing small dams for Gaborone catch-ment: (a) showing the effect of varying typical dam size, (b) using irrigation dams ratherthan stock-watering (compare to Figure 6(a))

increases approximately linearly, but becomes relativelysmaller with more small dams. An examination of thehydrographs showed that the effect was considerablymore marked in dry years when the small dams absorb agreater proportion of the available flow. The same totalcapacity of small dams has a greater effect if they areplaced further downstream (that is, if AC is increased).This is because the flows entering the dams are greater,and the dams are therefore likely to fill more frequently,and thus will tend to lose more water.

In addition, Figure 7a shows, for Gaborone catch-ment, the additional effects of varying the size of thetypical proposed small dams, indicating that a smallnumber of large dams has a relatively smaller effectthan a large number of small dams with the same totalcapacity. This is because larger dams tend to have amore efficient shape which has lower evaporationlosses.

Figure 7b, when compared to Figure 6a. shows theeffect of the dams being used for irrigation rather thanstock watering. In this example this difference is quitelarge, but this was found to be largely a result of the waythe model was constructed. The proposed stock damsare treated as controlling a part of the catchment which

has no existing dams. This is perhaps rather unrealisticconsidering the large numbers of existing dams in thiscase, and these results may have tended to overestimatethe effects of stock dams for the Gaborone catchment. Inother examples where there were few existing dams inthe catchment, the difference between stock and irriga-tion dams was not great and this may be more typical ofthe general situation. Since water loss from dams isdominated by evaporation, differences in the abstrac-tions would not generally be very significant.

An example of the effects of small dams on the sea-sonal distribution of flows is given in Figure 8. Theeffects were similar for each catchment but vary throughthe year. The reduction in flows is relatively greater earlyin the hydrological year, as at this time the dams will bedry so that a greater proportion of the runoff is trapped.

Implications for watershed development

There are considerable implications of these results forwatershed development. They show clearly that devel-opment of small farm dams upstream of majorreservoirs should never be allowed to proceedunplanned. Small dams can have significant effects on

Natural Resources Forum 1995 Volume 19 Number 1 81

I I I

Small farm reservoirs in Botswana: .1 Meii;li

5

01enra

10

9

8

7

6

5

4

3

2

1

0

Key

'

• No damsf~"l Existing dams only7 n Existing + proposed dams

Oct Nov Dec Jan' Feb Mar Apr May Jun Jul Aug Sep

Figure 8 Impact of small dams on seasonal distribution of flows for Gaboronecatchment

the run off into and the yield from the large reservoirs,and therefore it is necessary to assess these effects andweigh the benefits of rural developments against theadverse effects to urban water users.

Where development of small dams is to be carriedout, the results of the study also help to show how thedams can be designed to minimize their impact. As faras possible, small dams should be located in the upperparts of the catchment, where they will have a smallereffect on downstream flows compared to placing themfurther downstream. Also the development of fewersomewhat larger dams, which minimizes evaporationlosses, should be preferred to numerous smaller dams ofthe same total capacity, as it is found that this reducesthe downstream impact. While this is the case whenexamined from a purely water resources point of view,there may well be environmental and socioeconomicfactors which would favour the option of a greater num-ber of smaller dams, and these factors will need to bebalanced against the greater impact on downstreamresources of smaller dams. These results relate to theparticular catchments in Botswana, and they cannot beautomatically extrapolated to other regions. However,the main conclusions are likely to remain valid in othersimilar semi-arid regions.

ConclusionsA model has been developed which allows the impactsof both existing small dams and of a range of proposedsmall dams within the catchments of major water supplyreservoirs to be assessed. The model provides a planningtool, enabling guidelines for future small dam develop-ment to be determined. It is a generalized model whichcan be reconfigured to other catchments with a similarsemi-arid climate. By using a semidistributed approach.

the model tries to take account of the spatial variabilityof the hydrological processes, and, to a reasonableextent, allows the high temporal and spatial variabilityof rainfall in Botswana to be represented.

Methods have also been developed to estimate thephysical characteristics of the small dams using only thesurface area shown on maps. This means that the major-ity of dams, for which detailed information is lacking.can still be included in the model without the need tocarry out extensive field surveys.

The results for the three study catchments show thatthe total capacity of small dams is the overriding factorcausing decline in catchment runoff. Dams having atotal capacity of 10% of mean annual runoff. MAR.cause roughly the same decline of between about 8 and10% in catchment MAR.

This decline in MAR h affected by other factorswhich are of secondary importance but still significant.These are first the location of the dams; for instance, thesame total capacity of small dams has a greater effect it'they are placed further downstream. This is because theflows entering the dams are greater, and the dams aretherefore likely to fill more frequently, and thus willtend to lose more water. The second factor is the size otthe typical proposed small dams; a small number oflarge dams has a relatively smaller effect than a largenumber of small dams with the same total capacity. This

.is because larger dams tend to have a smaller surface(area in relation to their capacity (from Figure 2. it can beseen that a dam which has a capacity twice that otanoiher, has, on average, only about 70% larger surfacearea). Thus, larger dams have a more efficient shapewhich produces lower evaporation losses. The third fac-tor, the type of use to which the dams are put, whetherfor stock watering or for irrigation, also has a minoreffect. In addition, it was found that small dams have a

82 Natural Resources Forum 1995 Volume 19 Number I

I

greater impact in dry years and al the start of the hydro-logical year.

AcknowledgementsThe author would like to thank the staff of SirAlexander Gibb and Partners (Botswana), particularlyfor their work in preparing the inventory of small dams.He is also grateful to staff of the Department of WaterAffairs and numerous other people both in Botswanaand in neighbouring countries for supplying data and foruseful discussions which helped in formulating themodel.

ReferencesArup Botswana (1991) Loisane dam feasibility study Report to

Department of Water Affairs, GaboroneChunnet, Fourie and Partners (1990) Lephalala river catch-

ment water resources development study Report toRSA/Lebowa Permanent Water Commission

Gibb, Sir Alexander and Partners (1985) Feasibility study ofKolobeng and Metsemoilhaba dams Report to WaterUtilities Corporation, Gaborone

Gibb, Sir Alexander and Partners (Botswana)/Institute of

Small farm reservoirs in Botswana: J Meigh

Hydrology (1992) A study of the impact of small dam con-struction on downstream water resources Report toDepartment of Water Affairs. Gaborone

MacDonald, Sir M and Partners (1988) MetsemoilhabaTransfer Scheme - Feasibility/ Preliminary Design StudyReport to Department of Water Affairs, Gaborone

Moolman, J and Maaren, H (1986) The Effect of Farm Damson the Hydrology of a Catchment Hydrological ResearchInstitute, Department of Water Affairs Scientific Services,Pretoria

Mostert, A C (1990) The effect of farm dams in the Omatjcnnedam catchment area on the yield of the Omatjcnne damHydrology Division, Department of Water Affairs,Windhoek, Namibia

Neil, D T and Srikanthan, R (1986) Effect of farm dams onrunoff from a rural catchment Technical memorandum86/18, CSIRO, Canberra, Australia

Pitman, W V (1973) A mathematical model for generatingmonthly river flows from meteorological data in SouthAfrica Report No 2/73, Hydrological Research Unit,University of the Witwatersrand, Johannesburg

Pitman, W V (1976) A mathematical model for generatingdaily river flows from meteorological data in South AfricaReport No 2/76, Hydrological Research Unit, University ofthe Witwatersrand, Johannesburg

SMEC, WLPU and SGAB (1991) Botswana National WaterMaster Plan (NWMP) Report to Department of WaterAffairs, Gaborone

Natural Resources h'nrum 199$ Volume 19 Number I