An ecological economic simulation model of a non-selective grazing system in the Nama Karoo, South Africa

Ecological Economics 42 (2002) 221– 242 This article is also available online at: www.elsevier.com/locate/ecolecon

ANALYSIS

An ecological economic simulation model of a non-selective grazing system in the Nama Karoo, South Africa P.C. Beukes a,*, R.M. Cowling b, S.I. Higgins c a

Department of Nature Conser6ation and Oceanography, Cape Technikon, Box 652, Cape Town 8000, South Africa b Terrestrial Ecological Research Unit, Department of Botany, Uni6ersity of Port Elizabeth, P.O. Box 1600, Port Elizabeth 6001, South Africa c UFZ Umweltfoschungszentrum, Leipzig-Halle, Sektion O8 kosystemanalyse, Permoserstrasse 15, 04318 Leipzig, Germany Received 5 July 2001; received in revised form 25 March 2002; accepted 25 March 2002

Abstract The Nama Karoo region of South Africa is characterized by low ( 200 mm) and variable annual rainfall, which results in grass and shrub biomass production, which is low and highly variable in space and time. These characteristics of Nama Karoo rangelands challenge the ability of the region’s livestock farmers to make a sustainable living. In this paper we model a farming system, which attempts to create an environmental buffer of forage reserves by restricting access of livestock within numerous small camps. This is achieved by using a multi-camp infrastructure, which forces the livestock to remove non-selectively most of the available forage within a camp. Non-selective grazing in small camps allows for long rest periods of each camp, and these rest periods build up forage reserves for the dry years. A computer model of a 7000 ha farm was used to simulate rainfall and above-ground plant biomass accumulation, and to test the economic merits of investing large sums of money in multi-camp infrastructure. The model shows that 60 camps or more allows time for forage reserves to build up, but that more than 150 camps becomes too costly. Our simulations suggest that given 250 mm yr − 1 rainfall and the agriculturally recommended stocking rate, camp numbers of 60–80 provide higher profits than other camp numbers investigated. However, with higher rainfall and more animals, increasing camp numbers up to 150 is economically viable and more ecologically desirable. At low rainfall ( B200 mm yr − 1) production is too low to warrant investment in multi-camp infrastructure. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Computer model; Farming systems; Forage reserves; Infrastructure; Semi-arid;

STELLA

model

1. Introduction

* Corresponding author. Tel.: +27-21-460-3203; fax: + 2721-460-3193 E-mail address: [email protected] (P.C. Beukes).

Annual precipitation is known to be the major determinant of primary production in arid and semi-arid regions (Seely, 1978; Rutherford, 1980). The Nama Karoo biome of South Africa is characterised by a low (100 –300 mm) and variable

0921-8009/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 8 0 0 9 ( 0 2 ) 0 0 0 5 5 - 1

222

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

annual rainfall (Venter et al., 1986). The uncertainty of rainfall, together with the large variety of terrain and soil characteristics of the Nama Karoo (Visser, 1986), results in grass and shrub biomass production that is highly variable in space and time. Variability in total rainfall and seasonality of rainfall results in a rangeland that shows episodic, event-driven dynamics, which may include shifts in the relative biomass of grasses and shrubs (Milton and Hoffman, 1994). The frequency of disaster-magnitude droughts (Vogel, 1994) with the consequent losses of perennial grasses and shrubs, invasion by alien weeds, and drastic decline in carrying capacity (Milton et al., 1995) adds further uncertainty (Ellis, 1994) to the survival of land owners who have to make a living in these extensive pastoral areas. In former times, the stochastic availability of food and water to domestic livestock was partially overcome by nomadic farming practices (Vorster, 1994; Dean et al., 1995). With the establishment in the late 19th century of the mining industry in South Africa, and accompanying industrial development, a large market for food and other agricultural commodities was created. This initiated the change from subsistence to commercial stock farming on privately owned rangelands (Vorster, 1994). Today many commercial farmers are faced with some form of dryland degradation (Dean et al., 1995), decreasing carrying capacities, lower product income, and higher input costs and interest rates (Vorster, 1994). Solutions to these problems include the establishment of simple, low-input farming systems using adaptive and opportunistic management (O’Reagain and Turner 1992; Scoones, 1994; Vorster, 1994; Wiegand et al., 1995); better adapted animal genotypes; invader plant control; drought insurance policies; and drought fodder crops (Vorster, 1994). Another approach is to devise farming systems that allow for both the accumulation of forage reserves (Pickup and Stafford Smith, 1993) and for the optimum utilization of available forage (Nolan, 1996). The prerequisite for any sustainable system is that the long-term capacity of the biological system to produce forage from rainfall must be maintained, and the system must produce an acceptable finan-

cial return for the owner and dependents, thereby providing an acceptable standard of living (Pickup and Stafford Smith, 1993). The owner of the farm ‘Elandsfontein’ near Beaufort West designed a system based on the grazing management principles of Acocks (1966) and Savory and Parsons (1980). The system incorporates a multi-camp infrastructure (147 camps on a 7000 ha unit) with three animal types (sheep, cattle and goats). This system, termed non-selective grazing (NSG), is based on the premise that by concentrating a large number of animals with different feeding habits into a small paddock for a short time, animals will be forced to defoliate most of the plants, irrespective of palatability. This should reduce the competitive advantage of the less-palatable species compared to a system where animals can graze more selectively (Acocks, 1966). The concomitant hoof action (trampling/ treading), dunging, and urination are assumed to increase biomass turnover rate, improve the topsoil environment, stimulate production, and allow more rapid shifts in species (Savory, 1991). The relatively long rests of this grazing system allows more time for plant recovery. NSG simulates forage-animal processes that occurred prior to the establishment of pastoralism, i.e. migratory game-induced grazing pulses followed by long absences (Acocks, 1966). Such a system is more likely to maintain indigenous biodiversity and its associated processes than other grazing systems that impose novel defoliation regimes (McCabe, 1987; McNaughton et al., 1988; Savory, 1991). NSG requires a high investment in capital and operating costs. Since costs can rapidly exceed the returns, careful assessment of viability is needed (Pickup and Stafford Smith, 1993). This paper reports on a model that was developed to simulate the ecological processes associated with NSG, and to test the economic merits of investing resources in infrastructure in order to achieve this system at the farm scale. Given that investing in fences helps to dampen the impacts of environmental stochasticity, how can the number of camps and the animal complement be manipulated to maximize profits at different levels of veld productivity?

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

2. The study system

3. Model description and methods

Most data for the model were collected on the farm ‘Elandsfontein’ situated 32 km north-east of Beaufort West in the Nama Karoo (32°15%S 22°45%E), South Africa. The landscape consists mainly of level pediments with occasional rocky outcrops. The soils are derived from dolorite that is intruded into the sedimentary rocks more typical of Nama Karoo landscapes. The vegetation is homogeneous on the pediments and comprises a dwarf, open grassy shrubland dominated by Stipagrostis and Eragrostis spp., while Pentzia and Rosenia are the dominant shrub genera. The farm has been occupied by the Lund family since 1923 when there were no internal fences. When the current owner, A. Lund, inherited the farm in 1964, it was already subdivided into 40 camps. At this stage, the legacy of continuous selective grazing in the past was still very much evident in the form of shrub encroachment (Lycium), signs of soil loss, and extensive bare areas. The landowner reasoned that, in order to combat selective defoliation and utilize the available forage resources more efficiently, he needed to develop the infrastructure and diversify the stock complement. Over a period of 30 years, the farm was subdivided into 147 camps, arranged around 38 watering points, using a wagon wheel layout. Homogenous vegetation units were kept separate as far as possible to reduce area selectivity by livestock. An indigenous cattle breed (Nguni) was stocked together with Merino sheep in order to utilize coarse grass material more efficiently. Boer goats were introduced to utilize more of the available shrub forage. The number of herds was reduced as far as possible in order to concentrate the impact of feeding and hoof action on soil and vegetation processes. Animals are supplied with cut shoots of saltbush (Atriplex nummularia), which grows wild on the farm and acts as a cheap, but satisfactory rumen stimulant. The saltbush improves rumen fermentation and therefore enables animals to ingest low quality, fibrous material. As a result of using rumen stimulants animals can be retained in a camp for a few days longer while utilizing more of the available forage and still remain in an acceptable condition.

3.1. Conceptual description of model

223

A dynamic model was built with the aim of integrating knowledge of Nama Karoo ecological processes with management options and consequent financial implications. The model was developed in STELLA (High Performance Systems Inc, 1993), a high-level programming language, which facilitated the interactive and collaborative development of the model (Costanza et al., 1993). To keep the model simple and consistent with current knowledge, we assumed a homogenous Nama Karoo landscape and expressed all results on a per hectare basis. An annual time step was selected in order to simulate the impact of annual rainfall on forage and animal production, and finances. These decisions obviously sacrifice spatial and temporal resolution, but since our model was designed to compare different management tactics and not to accurately predict long-term income, it was regarded as sufficient for the purpose. The length of the time horizon (50 years) in the scenario analysis is based on the need to view management decisions as having long-term implications on the one hand, and on the other to reap the benefits (if any) within a farming career. The model comprises three interactive submodels, namely, vegetation, production, and financial (Fig. 1). Grass and shrub biomass production is determined by rainfall. Decomposition and grazing and browsing animals remove biomass. Empirical data from biomass harvesting trials (Section 3.2) were used to estimate production and decomposition factors. The vegetation sub-model is based on the concept that NSG can be manipulated by forcing large numbers of animals into small grazing camps. The amount of time animals can spend in a grazing camp (‘camp time’) is based on grass biomass available when animals enter the camp, the density and intake of animals, and the proportion of the available grass biomass that can be removed. Because shrub biomass is more constant than grass biomass (Beukes, 1999) we do not consider shrub biomass to be an important factor influencing camp time. The proportion of the

224

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

Fig. 1. Conceptual model of the grazing system.

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

available grass biomass that can be removed depends, in turn, on camp size and whether rumen stimulants are provided or not. The relationship between the proportion of the available grass biomass that can be removed and camp size is a key function in the model. Camp time determines the recovery period for the camp, which, in turn, determines when it is a ‘graze year’ for that camp. Since the model was based on an annual time step the effects of grazing season on plant recovery are not considered. The production sub-model calculates the meat and wool production per year from animal density, intake per animal, camp time, and the area grazed per year (ha), which in turn depends on farm size and recovery period. The financial sub-model converts the meat and wool production into monetary terms using 1999 meat and wool prices. Expenses comprise establishment and maintenance costs of the infrastructure (which is determined by camp numbers), rumen stimulant costs, and the costs of stocking the farm. Here we assumed that the full investment into infrastructure and stock occurred at the onset of the modeled time frame. Money had to be borrowed and debt serviced through annual interest payment. The model omits the opportunity for staging of investment costs as well as debt reduction. Prices were also assumed to stay fixed over the 50-year time frame. These are important shortcomings of the model. However, the fundamental purpose of this model, which was to evaluate cost effectiveness of various management tactics for extensive pastoral enterprises in the Nama Karoo, warranted the sacrifice of system behavior in a qualitatively realistic and in a quantitatively precise way. Despite these weaknesses, the model’s value lies in articulating the economic risks of management tactics that strive for optimizing an ecological process i.e. building a forage buffer for dry periods. Stafford Smith and Foran (1990) argue that in the rangeland context, it is more relevant to integrate ecological data into farm scale finances than to provide detailed data that cannot be thus integrated.

3.2. Methods Two harvesting trials were conducted at Elands-

225

fontein to derive rough estimates of above-ground standing crop of grass and shrub, net accumulation of biomass over time, and proportions of biomass removed under different grazing intensities. In the first trial 3 camps (A, E and F) were identified which had 50× 50 m2 fenced exclosure plots erected inside in 1995. These camps were subjected to a NSG treatment in May 1997. The exclosure plots were ungrazed while 50× 50 m2 plots outside the exclosures were demarcated as the grazed plots. In May 1997, after the grazing treatment, 15 quadrats (1×0.5 m2) were positioned inside each grazed and ungrazed plot in a systematic random way. The above-ground grass and shrub biomass inside each quadrat was harvested to ground level, dried at 80 °C for 24 h, and weighed to the nearest gram. This procedure was repeated in October 1997 after a 5-month resting period. After another grazing treatment in June 1998 the same procedure was repeated. The data were converted to kg ha − 1 and the mean per camp was calculated (Table 1). In a second harvesting trial, a 91 ha camp was selected and 30 quadrats (1× 0.5 m2) were placed in a systematic random way. Above-ground grass and shrub biomass was harvested, dried and weighed. The camp was then divided into two (: 45 and 46 ha) and grazed at two different intensities. In another camp (45 ha), a large camp size was simulated by reducing animal numbers and lengthening the period of stay. After the grazing treatments the harvesting procedure was repeated. Data were converted to kg ha − 1 and averaged per camp (Table 2). Each of the sub-models is discussed in more detail below. Parameter names, units, symbols and estimates are listed in Appendix A.

3.3. Vegetation sub-model The mean, standard deviation and long-term periodicity at the site are used to parameterize the rainfall generator. Annual rainfall is simulated by assuming that annual rainfall is determined by the mean and standard deviation of annual rainfall and by the long-term periodicity of rainfall (Desmet and Cowling, 1999) such that the annual rainfall (Ra, mm) is,

A E F Mean

A E F Mean

A E F Mean

May 1997

October 1997

June 1998

33 108 32

33 108 32

33 108 32

Size (ha)

LSGD, large stock grazing days.

Camp

Cutting date

61 51 54

None None None

35 40 35

Grazing intensity (LSGD ha−1) Grazed

558 9288 1144 9 648 1048 9 534 916 9314 670 9304 1556 9 772 1078 9 526 1102 9 442 80 9120 314 9272 148 9186 180 9120

Ungrazed

786 9290 1830 9686 1364 9602 1326 9522 736 9392 1240 9612 904 9310 960 9256 414 9246 770 9458 1058 9386 748 9322

Grass

Shrub

81 59 86 76

– – – –

29 37 23 31

1350 9 464 908 9688 1356 9940 1206 9256

1304 91288 1218 91382 1202 9 928 1240 954

1194 9984 962 9 1034 1248 9 1010 1134 9152

Proportion removed Ungrazed (%)

Table 1 Grass and shrub standing biomass values (kgDM ha−1 9SD) for the farm Elandsfontein, Beaufort West

570 9460 1338 9 1020 1042 9746 982 9 388

1472 9 790 1536 91166 1526 9 1092 15129 34

1428 9 1050 1168 9860 1654 9 1410 1416 9 242

Grazed

58 – 23 19

– – – –

– – – –

Proportion removed (%)

226 P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

227

Table 2 Grass standing biomass values (kgDM ha−1 9 SD) before and after grazing at different intensities Grazing intensity

Before grazing

After grazing

Proportion removed (%)

46 ha camp grazed for 4 days by 222 LSU ƒ 19 LSGD ha−1 45 ha camp grazed 8 days by 175 LSU ƒ 31 LSGD ha−1 a 45 ha camp grazed 22 days by 61 LSU ƒ 30 LSGD ha−1

1602 9 694 1602 9694 1322 9578

1032 9700 858 9 614 1236 9668

36 46 7

LSU, large stock unit; LSGD, large stock grazing days. a Large camp size simulated.

Ra = NORMAL(Rm, Rm ×CV) +Rm ×0.1 ×sin(2 ×p× y/p)

(1)

where NORMAL(x, sd) is a STELLA function that generates normally distributed random numbers with mean (x =mean rainfall Rm, mm yr − 1) and standard deviation (sd= Rm ×coefficient of variation of rainfall, CV), y is the simulation year and p (= 20) is the long-term periodicity of the rainfall cycle. Annual grass production (Gp, kg ha − 1) and shrub production (Sp, kg ha − 1) are estimated using Ra such that: Gp =1.9× Ra Sp =0.9× Ra The production factors used in the above formulae were roughly estimated from the results of above-ground biomass harvesting trials conducted at Elandsfontein (Table 1). The only plots that showed a net accumulation of grass biomass between May 1997 and October 1997 were the grazed plots that accumulated grass at an average annual rate of 1.9×250 mm (being the long-term mean annual rainfall for Elandsfontein). The production factor of 1.9 was estimated using the average grass accumulation over the 5-month period (which incorporates a growth peak in September/October) scaled up to a 12-month period incorporating another growth peak in March/April. According to the data in Table 1, the shrub biomass accumulated in both the ungrazed and grazed plots between May and October 1997 at an average of approximately 100 kg ha − 1. This converts to an annual accumulation of approximately 240 kg ha − 1. An estimate of shrub production can therefore be calculated as 0.9×250 mm.

Using the data in Table 1, the starting biomasses for grass (Gb) and shrub (Sb) were estimated at 1000 and 1200 kg ha − 1, respectively. Annual loss of grass and shrub biomass through decomposition (Gd and Sd, kg ha − 1) is calculated as a ratio of the standing crop biomass: Gd = 0.4× Gb Sd = 0.08× Sb These decomposition factors were estimated from the decline in grass and shrub biomass values in the ungrazed plots over the period October 1997–June 1998 (Table 1). The decline in biomass over a full year was calculated using the ratio of months of observation to 12 months. In testing the model it showed an unrealistic build up of shrub biomass and the decomposition factor had to be adjusted from 0.04 to 0.08. Animals are more concentrated in smaller camps where a high grazing intensity (LSGD ha − 1 = large stock grazing days ha − 1 = large stock units, LSU× grazing days ha − 1) can be maintained with a short period of stay. Concentrating animals forces them to feed less selectively, thereby a wider spectrum of species is taken (Acocks, 1966) as well as more biomass per plant (Danckwerts, 1987; Du Toit, 1996). The advantage of a short period of stay (5–10 days) is that it minimizes the number of times that regrowth is removed (Acocks, 1966; Danckwerts, 1987). We consequently assume that the proportion of available above-ground grass biomass which can be removed by grazing animals (Gbprop) is determined by camp size (Cs, ha), which in turn is a function of farm size (Fs, 7000 ha) and camp numbers (Cn, adjustable management input):

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

228

Cs =

Fs Cn

and Gbprop = u1exp(−u2Cs) +u3

(2)

The data presented in Table 1 and Table 2 were used to roughly estimate the maximum proportion standing grass biomass that can be removed from a large camp (\200 ha; 20%), a medium camp (100 ha; 60%) and a small camp (30 ha; 80%). This was fitted in a non-linear regression. The parameter estimates for Eq. (2) were u1 =0.7, u2 = 0.00065, u3 =0.2 (n = 3, R 2 =0.99). Rumen stimulants improve fermentation rate, and therefore intake of low quality forage (Leng, 1984). We assumed that rumen stimulants (Rs) would increase the proportion of grass removed by 50%. Therefore: If

Rs = 1

then

Gbprop

else

Camp time (Ct, days) is determined by the available grass biomass (Gb) when animals enter the camp, the proportion of grass biomass which can be removed (Gbprop), the density of grazers (cattle [Cdens]; sheep [Sdens] in LSU ha − 1), and intake per LSU per day (ILSU, kg LSU − 1 day − 1). Densities depend on the numbers stocked by the grazier and Cs, while intake was taken as 10 kg LSU − 1 day − 1 (ARC, 1980). It was assumed that a sheep’s diet consists, on average over a year-long period, of 50% shrub and 50% grass (Du Toit et al., 1995), while cattle eat only grass, and goats concentrate mainly on shrubs and non-grassy herbaceous plants (Botha et al., 1983). Ct was determined by grass removal so goats were excluded in its definition: Ct = (Gb × Gbprop) (3)

The recovery period (Rp, years) for a simulated hectare depends on Ct and Cn and was calculated as the proportion of a year the animals spend in all the other camps before they return to the modeled hectare. Since the time step is annual, Rp had to be rounded to the nearest year: Rp =ROUND[(Cn −1)Ct/365]

Gy = If

Rp 0 1

or

Rp = COUNTER(1, 1 +Rp) 0

then

1

else (4)

Here COUNTER(x, y) is a built-in STELLA function that enables one to define time-dependent cycles. If it is a graze year for the simulated hectare, the amount of grass eaten (Ge, kg ha − 1) and shrub eaten (Se, kg ha − 1) is calculated as follows: Ge = If

Gbrop ×0.5

/[(Cdens × ILSU) +(Sdens ×ILSU ×0.5)]

Here ROUND(x) is a built-in STELLA function that rounds the argument, x, to the nearest integer value. Rp determines when it is a graze year (Gy, 0 or 1) for the hectare. If Rp is 1 year or less then the modeled hectare gets grazed every year. If Rp is e.g. 2 years then the hectare gets grazed every alternate year. Thus:

Gy = 1

then

(Ct × Cdens × ILSU + Ct × Sdens × ILSU × 0.5)/dt else Se = If

0 Gy = 1

(5) then

(Ct × Gdens × ILSU + Ct × Sdens × ILSU × 0.5)/dt else

0

(6)

In the vegetation sub-model we assume that veld remains in a relatively good condition, and we do not consider possible influences of annual rainfall fluctuations nor the impacts of the grazing system on vegetation composition, and therefore primary production and acceptability to livestock.

3.4. Production sub-model In this sub-model the grass and shrub forage ingested per hectare over the time that the animals stay on the hectare (Ct) is converted into marketable meat (Mp, kg) and wool production (Wp, kg) for the whole year by incorporating the area grazed per year (Ag, ha). Ag can be calculated from Fs and Rp as follows: Ag = If

Rp E 1

then

Fs/Rp

(7)

We further assume that livestock require 15 kg dry matter intake to gain 1 kg in lean body mass

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

229

(Meissner and Roux, 1984) and sheep require 90 kg dry matter intake to produce 1 kg wool (Boyazoglu, 1997). All the wool and 15% of the meat produced are marketable (A. Lund, personal communication, 1998). It then follows that:

the set-up costs (SUcost, Rands) for additional camps (Cincr = Cn − 10) can be calculated using the marginal cost per camp (Ccost, Rands).

Mp = If

where the average costs of fencing material (J. Stru¨ mpher, Progressive Wire, personal communication, 1998) and watering points (A. Lund, personal communication, 1998) were used to construct an exponential relationship between Cincr and Ccost (Fig. 2). With the marginal cost per camp declining exponentially from R26 000 to an average of approximately R4000. As camps are subdivided, the required fencing material per camp becomes less, and watering points can be shared between camps. According to the landowner, maintenance costs (Mcost, Rands) amount to approximately 2% of SUcost per annum;

Gy = 1

then

((Ct × Cdens ×ILSU +Ct ×Sdens ×ILSU)/15) ×Ag × 0.15

else

0

(8)

and Wp = If

Gy = 1

then

(Ct × Sdens × ILSU/90) ×Ag

else

0

(9)

3.5. Financial sub-model Annual gross income (Igross, Rands1) is calculated by converting Mp and Wp into monetary terms using meat (Pm, Rands and cents) and wool prices (Pw, Rands and cents). Igross =Mp ×Pm + Wp ×Pw

SUcost = Cincr × Ccost

Mcost = SUcost × 0.02

(10)

Four items comprise the expenses (Ex, Rands) on an annual basis, namely, stocking the farm, setting up the infrastructure, maintaining the infrastructure, and the costs of rumen stimulants (if provided). The recommended stocking rate for Elandsfontein is 300 LSU (24 ha LSU − 1 for a 7000 ha unit) (A. Lund, personal communication, 1998). The animal increment (Aincr, LSU) is calculated if the manager wants to increase total animals stocked (Atot, LSU) above the recommended:

If rumen stimulants are provided then rumen stimulant costs (Rcost, Rands) amount to R32 per LSU (A. Lund, personal communication, 1998) which gives; Rcost = If

Rs = 1

then

Atot × 32

else

0

Aincr =Atot − 300 For the purpose of the model we assumed that money would have to be borrowed to acquire these animals at the average value per LSU (Aval) of R2000 (P. Van Zyl, Dept. Agric., personal communication, 1998). If 10 camps are assumed to be the minimum for an acceptable rotational grazing system (Hobson, 1987), then

1

11.03 Rands = US$1 at the time of publication.

Fig. 2. The relationship between additional camps (Cincr) and the marginal cost per camp (Ccost; Rands).

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

230

Table 3 Comparison between observed annual wool and meat production figures (kg) from Elandsfontein and figures predicted by the model Year

1987 1988 1989 1994 1995 1996 a

Rainfall (mm)

200 290 310 230 230 370

Annual wool production (kg)

Marketable annual meat production (kg)a

Predicted

Observed

Predicted

Observed

6170 6745 6612 7077 7231 7680

5365 7354 9586 6121 7437 10 146

– – – – 10 087 10 906

– – – – 9547 12 028

Regarding 15% of total meat production as marketable.

With the assumption that the funds for extra camps and stocking the farm above the recommended number of animals have to be borrowed at an interest rate of 10% per year, then Ex can be calculated as follows: Ex =((Aincr ×Aval +SUcost) ×0.1) +Rcost +Mcost (11) This leaves annual profit (Iprofit, Rands) to be calculated as follows: Iprofit =Igross − Ex

(12)

4. Model behavior Ideally we would like to compare the production figures the model generates with production figures obtained from an independent data source, however, such data are not available. In Table 3 we run the model for the rainfall recorded at Elandsfontein and compare the production figures generated by the model with the production figures actually recorded at Elandsfontein. This analysis suggests that the model is able to simulate production figures of the correct order and that it is able to simulate how production is influenced by rainfall. Although this can be considered as a form of model validation, it is not ideal in that the data used to parameterize the model are not independent from the observed production data. Moreover, we only have data for comparing one aspect (animal production) of the model’s behavior.

5. Scenario analysis

5.1. Scenario definition The scenario analysis was used to answer four questions; (a) Does an increase in camp numbers create a forage buffer? (b) What number of camps generates the highest annual profit for a given set of conditions? (c) What is the best profit scenario with the number of camps selected in (b) in place? (d) Under which conditions are fewer, or more camps than selected in (b) economically viable? Table 4 summarizes the underlying rationale, choice of variables and range of values for the key questions listed above. Each scenario was simulated ten times with randomly generated rainfall figures. All results presented below are the mean of ten simulation runs. Rainfall scenarios covered the range from 300 to 150 mm yr − 1 (a range that encompasses the Nama Karoo) with coefficient of variation of rainfall (0.6, 0.5, 0.4) applied accordingly (Desmet and Cowling, 1999). Camp numbers were selected to vary from average (:20 camps) to exceptionally high (150 camps) for the region.

5.2. Scenario results 5.2.1. Forage buffer Changes in above-ground grass and shrub biomass over time for different camp numbers

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

231

Table 4 Key questions addressed in the scenario analysis showing the underlying rationale, choice of variables, and range of values Question

Rationale

Choice of variables

Range of values

(a) Does an increase in camp numbers create a forage buffer?

Farm size (7000 ha) stocking rate (350 LSU) animal type (sheep only) and rumen stimulants (none) were kept conservative and constant. With rainfall being the main driver, we analyzed for a forage buffer with camp numbers varying from average to exceptionally high

Mean rainfall (Rm) Coefficient of variation of rainfall (CV) Camp numbers (Cn)

150, 200, 250 mm 0.6, 0.5, 0.4

(1) Camp numbers (Cn)

20, 60, 150, 300

(2) Camp numbers (Cn) Mean rainfall (Rm) Coefficient of variation of rainfall (CV)

20, 40, 60, 80 150, 200, 250 mm 0.6, 0.5, 0.4

Number of sheep (LSU) Number of cattle (LSU) Number of goats (LSU) Rumen stimulants (Rs)

350, 390 0, 40, 80 0, 20 0, 1

Camp numbers (Cn) Number of sheep (LSU) Number of cattle (LSU) Rumen stimulants (Rs)

60, 150 350, 400 0, 80 0, 1

(b) What number of camps generates the highest annual profit for a given set of conditions?

Here we performed two analyses (1) and (2) (1) We analyzed for annual profit by varying camp numbers and therefore costs. Rainfall, CV of rainfall, stocking rate, animal type and rumen stimulants were kept conservative and constant (2) Analysis (1) discarded 150 and 300 camps as too costly. We attempted to find the highest profit with fewer camps under different rainfall regimes. For this analysis relative (to R150 000) mean annual profit was calculated in order to present results in a more comparable format

(c) What is the ‘best’ profit scenario The results of the above showed with the number of camps selected in approximately 60 camps to be a (b) in place? good option under a 250 mm rainfall regime. Keeping rainfall constant, we analyzed for annual profit by varying animal numbers, animal mix, and rumen stimulants. Animal numbers and mix were kept within the norm for the region (d) Under which conditions are fewer, or more camps than selected in (b) economically viable?

20, 60, 150

Analysis for question (a) showed that fewer camps, approximately 20–40, are better under lower (0200 mm) rainfall. We asked the question: under higher rainfall (e.g. 300 mm), is an exceptionally high number of camps (e.g. 150) economically viable? Here we had to stock more and supply rumen stimulants.

232

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

Under the rainfall regime where Rm = 200, and CV = 0.5, the 60, and 150 camp systems do maintain the grass and shrub biomass slightly longer than the 20 camp system, but again it appears as if 350 LSU sheep might be too many in the long run. However, it is already apparent at this rainfall level that increasing camp numbers does allow for forage buffers to be maintained for longer compared to a system with fewer camps (e.g. 20).

Fig. 3. Changes in above-ground grass biomass with time for three rainfall regimes (Rm = 150, 200, 250 mm; CV = 0.6, 0.5, 0.4) for a 1 ha patch of Nama Karoo veld under different management systems (Cn = 20, 60, 150 camps).

and for three rainfall regimes are shown in Fig. 3 and Fig. 4. With a low mean annual rainfall (Rm =150) and high coefficient of variation (CV = 0.6), there is virtually no biomass accumulation irrespective of camp numbers. Under these conditions a conservative stocking rate of 350 LSU sheep might be too high and grass and shrub resources will be depleted within 10– 20 years.

Fig. 4. Changes in above-ground shrub biomass with time for three rainfall regimes (Rm =150, 200, 250 mm; CV =0.6, 0.5, 0.4) for a 1 ha patch of Nama Karoo veld under different management systems (Cn =20, 60, 150 camps).

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

Fig. 5. Mean annual profit (with standard error margins) under a constant rainfall regime (200 mm) for four management systems (Cn = 20, 60, 150, 300 camps). 350 LSU sheep stocked and no rumen stimulants fed.

233

perior because of the lower set-up costs and the starting figures for grass and shrub biomass (Gb = 1000 kg ha − 1; Sb = 1200 kg ha − 1), but the 60 and 150 camp systems exceed the profit from the 20 camp system within 10–15 years, and thereafter prove superior most of the time. The marginally better buffering capacity of the 150 camp system (Figs. 3 and 4) does not convert into higher profits compared to the 60 camp system (Fig. 5) and the 150 camp system is therefore rejected. Under a low rainfall regime no management system succeeds in building a forage buffer

With the 250 mm rainfall regime, both grass and shrub biomass is maintained under the 60 and 150 camp systems compared to the depletion of biomass under the 20 camp system. The higher rainfall accentuates the ‘carry over’ effect generated by the longer recovery periods of the 60 and 150 camp systems. With the slower decomposition rate of shrubs, the build up of shrub biomass under long recovery period systems (e.g. 150 camps) might even become problematic in the sense that shrubs might outcompete the herbaceous layer in the long run (Fig. 4). Fig. 3 and Fig. 4, therefore, show that the benefits of longer recovery periods (Rp) created by increasing camp numbers applies to higher rainfall regions. In lower rainfall regions (0200 mm) stock numbers will have to be manipulated in order to maintain a forage buffer for dry years.

5.2.2. Most profitable number of camps Setting up and maintaining an infrastructure of 300 camps on a 7000 ha unit is too costly and does not compare favorably in terms of profit with a 20 camp system (Fig. 5). Although the 20 camp system has a very limited buffering capacity and experiences a drastic decline in profit in the early years, the 300 camp system is less profitable to set-up and maintain. With their buffering capacities, the 60 and 150 camp systems are less susceptible to an unpredictable series of dry years, and, therefore, better maintain biomass, which converts into higher profit, than the 20 camp system. The last mentioned system is initially su-

Fig. 6. Relative (to R150 000) mean annual profit under three rainfall regimes (Rm =150, 200, 250 mm; CV =0.6, 0.5, 0.4) for four management systems (Cn =20, 40, 60, 80 camps). 350 LSU sheep stocked and no rumen stimulants fed.

234

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

Table 5 Mean annual profit (Rands) over 50 years for three management systems (Cn =40, 60, 80 camps) Management system

Average after 50 years Standard deviation Coefficient of variation (%) (CV) Average for the last 25 years of the 50 year simulation runs Standard deviation Coefficient of variation (%) (CV)

40 camps

60 camps

80 camps

255 656 51 003 20 279 134 32 474 12

248 043 54 569 22 282 667 17 142 6

230 194 52 337 23 266 948 13 054 5

Rainfall regime and stocking rate remain constant. Rm = 250; CV =0.5; 350 LSU sheep; Rs =0.

and biomass and, therefore, profit fluctuates with rainfall (Fig. 6). However, as rainfall increases the buffering capacity of the 40– 80 camp systems generally perform better than the 20 camp system. Although the 80 camp system shows excellent buffering capacity under the 250 mm regime (Fig. 6), it is very difficult to choose between 40, 60 and 80 camp systems. In this regard it is useful to look at the results presented in Table 5. The 40 camp system has the highest average and lowest CV over a 50-year period because the lower set-up costs results in higher profits initially. However, in the long run, forage resources start to fluctuate more in the 40 camp system, and the 60 camp system with its higher average and lower CV over the last 25 years of a 50-year simulation, seems to be a better option. Although the 80 camp system has a lower CV in the long run, the average profit is substantially lower compared to the 60 camp system.

5.2.3. The ‘best’ profit scenario In an attempt to find the ‘best’ profit scenario under a specific rainfall regime (Rm =250, CV= 0.5) and with the 60 camp system in place, various management systems were simulated. Feeding of rumen stimulants greatly enhanced the buffering capacity of the 60 camp system, but at such a reduced profit that it did not appear to be an economically viable strategy (Fig. 7). The value of feeding rumen stimulants may be realized in the long-term when the manager is forced to maintain high animal numbers through unfavorable periods as a result of, for instance, low market prices.

Whereas with high animal numbers recovery periods would have been too short for a forage buffer to build up, animals can stay slightly longer in a camp when supplied with rumen stimulants, which then increases recovery period and allows for a forage buffer to accumulate. This argument is supported by the relatively high average annual profit and low CV of profit for the last 25 years of management system 5 (Table 6). An increase in stocking rate (from 350 LSU to 390 LSU, MS2) increased the variability and lowered the average annual profit (Table 6). Higher stocking rates neutralized the dampening effect of the camp system on the variability of forage production. However, bringing in browsers (goats

Fig. 7. Relative (to R150 000) mean annual profit (with standard error margins) from a 7000 ha unit with 60 camps stocked with 350 LSU sheep and subjected to different management strategies. Rs =rumen stimulants.

243 019 63 611 26 256 799 56 022 22

248 043 5332 2 282 667 17 142 6

Rs =0 MS2

Rs =0 MS1

Rs = 0 no rumen stimulants; Rs = 1 with rumen stimulants; MS = management system.

Average after 50 years Standard deviation CV (%) Average for the last 25 years of the 50 year simulation runs Standard deviation CV (%)

390 LSU sheep

350 LSU sheep

Management systems

52 012 19

257 180 64 735 25 274 934

Rs =0 MS3

390 LSU sheep 20 LSU goat

51 082 23

213 542 56 328 26 218 587

350 LSU sheep 20 LSU goat 40 LSU cattle Rs =0 MS4

20 406 8

208 750 59 404 28 252 250

350 LSU sheep 20 LSU goat 40 LSU cattle Rs = 1 MS5

53 330 26

181 401 58 284 32 201 802

350 LSU sheep 20 LSU goat 80 LSU cattle Rs = 1 MS6

Table 6 Mean annual profit (Rands) for a 7000 ha unit with 60 camps under a rainfall regime where Rm =250 and with CV = 0.5 and with different management systems

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242 235

236

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

(MS3)) can increase annual profit, but according to the model there was no evidence that more than 20 LSU goats (: 100 goats) is a viable option (unpublished analysis). Because cattle utilize only grass in our model, replacing 40 LSU sheep with cattle resulted in a drastic decline in forage availability and therefore profit (MS4). Adding rumen stimulants (MS5) reduced annual profit further because of the cost of these stimulants, but in the longer term it enabled the manager to maintain more animals with a lower variability in production resulting in greater average profits. There is a point where stocking rate is too high and not even supplying rumen stimulants buffered the variability in forage production (MS6). Overall, the ‘best’ management option for a 7000 ha unit is to invest in a 60 camp infrastructure and stock the farm conservatively with sheep and possibly a few goats. Rumen stimulants could be useful, but then only during dry periods to maintain animal condition and keep the animals for a few days longer in a camp. This scenario is likely to enable a pastoralist to endure an exceptionally severe 1-in-50-year drought without having to de-stock drastically and sacrifice income.

5.2.4. De6iations from the 60 camp scenario Fewer camps, e.g. 20– 40, appear to be more economically viable in drier regions (0 200 mm) (see Fig. 6). The low rainfall and high variability thereof overrides any buffering effects that the camp system might have on forage availability. In very dry areas (0 100 mm) the approach would be to have a highly flexible stocking rate, and to keep overheads, such as interest on infrastructure investments and maintenance costs, as low as possible. Higher rainfall enhances the buffering capacity of many camps (e.g. 150), and with the feeding of rumen stimulants to maintain animal condition under high stocking densities, more animals can be maintained successfully with an increase in gross annual income (Table 7, MS8). However, this system only makes economic sense if overheads are not as high as simulated in the model, i.e. the infrastructure of a 150 camp system has been set-up over time without borrowed money, stock are allowed to breed up over time, mainte-

Table 7 Average annual gross income (Rands) from a 7000 ha unit under two different management systems with a rainfall regime where Rm =300 mm and with CV =0.4 Management systems 60 camps 350 LSU sheep 20 LSU goat

Average after 50 years Standard deviation CV (%) Average for the last 25 years of the 50 year simulation runs Standard deviation CV (%)

Rs =0 MS7

150 camps 400 LSU sheep 20 LSU goat 80 LSU cattle Rs =1 MS8

310 344

312 911

57 405 18 351 454

83 384 27 364 997

17 231 5

51 233 14

nance costs are kept low by using quality materials and workmanship initially, and rumen stimulant costs are kept low by feeding, for instance, chopped fodder growing naturally on the farm, e.g. Atriplex nummularia (A. Lund, personal communication, 1998).

6. Discussion The interactions between ecological and economic systems are many and strong. Analysis and modeling of linked ecological economic systems can provide valuable insights, which will hopefully change the behavior of resource managers towards a sustainable pattern of resource use that works in synergy with the life-support ecosystems on which the economy depends (Costanza et al., 1993). The fundamental purpose of this model was to evaluate cost effectiveness of various management tactics for extensive pastoral enterprises in the Nama Karoo of South Africa. We found that ecological buffering mechanisms created by forcing NSG increased the ecological resilience of the system and that this increased resilience had economic benefits too. Below we discuss the eco-

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

logical buffering mechanisms created by forcing NSG and we present this in the context of the financial circumstances at the scale of the farming unit.

6.1. Coping with 6ariability —the forage buffer In uncertain environments, forage availability fluctuates widely over time and space (Scoones, 1994) and managers need to apply farming systems and techniques that can reduce the impact of seasonal and periodic droughts (Vorster, 1994). In this regard a manager has basically two options: to apply an opportunistic, highly flexible management approach in order to track the variable forage resource (Scoones, 1994; Milton et al., 1995; Campbell et al., 2000), or to use forage and financial buffers to minimize the impact of this variability (Pickup and Stafford Smith, 1993; Vorster, 1994). One way of achieving tracking, which is matching available food with animal numbers, is to de-stock in the face of drought and restock when forage is available after the drought (Riechers et al., 1989; Stafford Smith and Foran, 1992; Scoones, 1994; Hatch et al., 1996). However, the decision of when and to what extent to de-stock is not an easy one (Stafford Smith and Foran, 1992; Pickup and Stafford Smith, 1993), and stock prices may be low because many managers are taking the same action. Furthermore, high animal prices after drought-breaking rain may result in a strong incentive to retain stock during a drought, which may be damaging to the rangeland (Pickup and Stafford Smith, 1993; Vorster, 1994). Finally, de-stocking and restocking after drought is associated with considerable transaction costs (Campbell et al., 2000). Integrating all of these factors, Campbell et al. (2000) has demonstrated convincingly in communal grazing areas in Zimbabwe that such a tracking approach would come with considerable economic losses and is likely to result in considerable environmental degradation, which in itself, would incur additional costs. The second option of building forage and/or financial buffers for droughts involves making long-term changes to management that enhance the capability of the property to cope with

237

drought (Foran and Stafford Smith, 1991). One such strategy is to invest in waterpoints and fencing in order to make more, smaller camps and to implement a non-selective rotational grazing system. This permits a high degree of control over the impact of the animals (Pickup and Stafford Smith, 1993). It also enables the manager to separate vegetation units of differing characteristics to promote uniform veld utilization (Edwards, 1988). This strategy obviously entails increased capital and operating costs and the decision whether to invest and thereby increase management intensity, or not, is fundamental in terms of the sustainability of pastoralism as a land use in these variable environments (Pickup and Stafford Smith, 1993). The objectives of intensive, multi-camp systems are (1) to limit the period of stay (Ct, camp time) so as to avoid grazing of regrowth, (2) to increase grazing intensity and thereby reduce selectivity (Acocks, 1966), and (3) to allow for long-term rests for seeding, vigor and forage accumulation (Booysen et al., 1974). Many issues, such as whether animals actually graze regrowth, whether selectivity can be reduced by increasing grazing intensity, and whether extended periods of absence would result in an accumulation of forage, remain unresolved (Hoffman, 1988; O’Reagain and Turner, 1992). There is, however, general agreement that leaving as large a proportion as possible of the property unoccupied at any one time allows for a large forage buffer left from the last good rains, and whatever rain does fall during the drought can be effectively converted into forage over a large proportion of the farm. This forage buffer enables farmers to reduce uncertainty and, therefore, endure droughts without de-stocking at times of unfavorable stock prices. This strategy also enables the manager, who can comfortably maintain stock numbers through droughts, to sell animals after droughtbreaking rains at relatively high prices. Our model shows that there is no point in opting for the forage buffering strategy in lower rainfall (0 200 mm) areas. Under these environmental conditions, tracking (Scoones, 1994) or maintaining animals at a conservative, constant stocking rate (Campbell et al., 2000) are probably better options. However, under higher rainfall (\200 mm)

238

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

conditions, investing in infrastructure, and therefore better ecological control, does make financial sense. The model shows that effective buffering is achieved with multi-camp systems of 40– 80 camps on a 7000 ha farming unit, but the highest, most reliable economic return is realized in the order of 60 camps. This optimum is obviously affected by the current prices of meat, wool, fencing material, and maintenance.

6.2. Acceptability of forage buffering strategies Weighing the costs of intensification against ecological and economic benefits is one aspect. The social, political, and cultural forces against intensification is another. Scoones (1994) argues that the blueprint ranch model aiming at higher control and boosting single outputs (e.g. meat and wool) is unlikely to work in most of Africa’s pastoral areas, and that an emphasis on flexibility and mobility is the key to success in these areas. O’Reagain and Turner (1992) present an argument for simple grazing systems using adaptive and opportunistic management in Southern African rangelands. Technological requirements, e.g. for water distribution across the property, management expertise for implementing complex rotational grazing systems (Hoffman, 1988), and the long-term vision required for farm development, might be beyond the capacity of most African pastoralists, especially on communal lands. Unless a low-cost mechanism can be found to control animal density and movements in communal areas, our model probably only applies to commercial farming areas where some camp system already exists.

6.3. Stocking rate, animal mix, and rumen stimulants Stocking rate is generally regarded as the most important management tool under control of a pastoralist (Danckwerts, 1987), and several models have shown the relationships between farm outputs and stocking rates (Noy-Meir, 1978; Riechers et al., 1989; Stafford Smith and Foran, 1992; Hatch and Tainton, 1995; Hatch et al., 1996). Stocking rate generally has to be some-

where between maximum production per animal and maximum production per hectare (Danckwerts, 1987), and should not exceed the long-term carrying capacity for a particular vegetation type (O’Reagain and Turner, 1992). However, under drought conditions severe de-stocking might be necessary (Section 6.2) while under intensively managed systems very high stocking rates (2× recommended) might be sustainable and even ecologically desirable (Acocks, 1966; Savory, 1978). The model supports the findings of Danckwerts (1987) and Campbell et al. (2000) that conservative stocking rates increase the pastoralist’s resilience towards drought without adversely affecting profitability. Because camp time (Ct) and recovery period (Rp) in the model are based on grass biomass which is highly sensitive to stocking rates, even moderate increases in the stocking rate of grazers reduced the buffering capacity of the multi-camp system. Selecting the correct grazer:browser combination is necessary for efficient and ecologically balanced use of mixed herbaceous and woody vegetation (Nolan, 1996). Animal performance can be improved (O’Reagain and Turner, 1992; Nolan, 1996), carrying capacity increased (Botha et al., 1983), potential for veld degradation decreased (O’Reagain and Turner, 1992), and diversification of income improved (Nolan, 1996) when implementing a suitable mix of animal types. In order to select the correct mix, the manager needs information on dietary overlap, which in turn depends on vegetation type, animal breed, and season (Botha et al., 1983; Du Toit et al., 1995). Overlap on semi-arid rangeland can be generalized by assessing sheep diets as intermediate between cattle (graze-only) and goat (browse-only) diets (Nolan, 1996). The model incorporates this overlap, but only in calculating the total amount of grass and shrub biomass removed by each animal type (sheep, cattle and goats). This ignores the very important complementary effect of these animals by feeding on different plant species and plant parts (A. Lund, personal communication, 1998). Although the model shows that adding browsers (goats) to the animal mix can increase annual profit, the real value of goats lies in their

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

ability to utilize forage that is largely unacceptable or unavailable to other animals (e.g. Lycium spp). In a similar way the model ignores the value of cattle in removing the coarser grass material and thereby improving sheep performance (Danckwerts, 1987), and their possible value as ‘engineers’ for maintaining ecosystem health through dunging, urination, and hoof action (Savory, 1991). Product demand, and meat and wool price fluctuations are other important considerations when deciding on a suitable animal mix. The input costs of supplementary feeding can be excessively high, making it uneconomical at face value (Vorster, 1994). However, there are situations when the limited use of rumen stimulants, which improve rumen fermentation and roughage intake, can be advantageous. Through its effect on rumen microbial activities, these stimulants help ruminants to ingest more fibrous, less-palatable forage. Without rumen stimulants animal condition will drop dramatically when they are forced to remain in a camp in order to remove a larger proportion of the available forage reserve. It is towards the end of an occupation period (camp time) that rumen stimulants help animals to ingest lower quality forage while still maintaining an acceptable level of performance (A. Lund, personal communication, 1998). Especially during dry periods, rumen stimulants can be used to force animals to remove up to 80% of the above-ground grass biomass. A plentiful supply of clean, cool water is absolutely essential to achieve this (A. Lund, personal communication, 1998). With more forage per hectare that can be removed, more animals can be accommodated and, because camp time is lengthened, the overall buffering capacity of the camp system is enhanced. This enables the manager to maintain higher animal numbers for longer periods before forage deficits set in. The model shows that it is possible, within limits, to maintain high animal numbers through dry periods using rumen stimulants. However, this practice is not economically viable. The model assumes that rumen stimulants are supplied to all animals throughout every year, which is obviously unnecessarily costly. It also disregards the use of

cheaper, readily available Atriplex nummularia.

239

supplements

e.g.

6.4. Future research needs The model can be improved by expanding the plant and animal production sub-models. In the plant production sub-model different herbage pools, e.g. green, standing dead, and litter (McKeon et al., 1982), as well as flows between these pools, can be incorporated. The feedback effects of rainfall and patterns of herbivory on plant population dynamics, soils, and therefore productivity need to be considered (Owen-Smith, 1991). Herd structure (including age classes) could be more accurately reflected in the conversion of forage into animal products (see the HerdEcon model of Stafford Smith and Foran, 1988). A further challenge is to incorporate spatial heterogeneity on a farm scale, and seasonal distribution of rainfall. Reducing the time frame of the model might not be as difficult as coping with spatial heterogeneity. Existing models like Paddock (Stafford Smith and Foran, 1990) have taken up the spatial heterogeneity challenge. Another useful development would be to link this and similar models to an integrated decision-support system (Ludwig et al., 1992) for managers, land use planners, and extension officers. More importantly, it is necessary to apply our model to a wider array of environmental conditions by providing more accurate production and decomposition figures for the Nama Karoo. Refining the relationship between camp size and maximum proportion of standing biomass that can be removed by animals is important for improving confidence in the model. A study comparing NSG versus exclusion plots over a 4-year period showed positive impacts of NSG on physical soil properties, organic turnover, and forage production (Beukes, 1999). These ecological merits of NSG were not considered as monetary benefits in our model. Although no evidence could be found of either a gain or loss of biodiversity under NSG (Beukes, 1999), studies are underway to test the claim that NSG neutralizes the negative effects of selective defoliation and maintains native biodiversity.

240

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

cial returns becomes. Our results indicate that annual rainfall is a key determinant of the choice of strategy. In lower rainfall regions (0 200 mm) tracking forage resources (Scoones, 1994) or maintaining constant conservative stocking rates (Campbell et al., 2000) are appropriate strategies, whereas under higher rainfall (\ 200 mm), investing in a large camp system and stocking at the agriculturally recommended stocking rate appears to be the better option. The consistent income of this farming system can be further increased by manipulating the grazer:browser ratio with a suitable mix of cattle, sheep and goats, and by providing rumen stimulants during dry seasons in order to improve forage use efficiency and maintain animal performance levels.

7. Conclusions It is likely that every pastoralist in arid lands will have to cope with drought at some stage. There are basically two strategies to remain financially viable during drought events, (1) tracking the varying forage availability by, for example, applying a flexible stocking system; and (2) opting for a more capital intensive farming system which allows for a forage buffer to accumulate during good years. Most conservation-orientated commercial pastoralists want to maximize annual profit, minimize variability thereof, while maintaining the health of their biological systems. The more capital intensive a farming operation becomes, the more important predictability of finan-

Appendix A. Parameter names, units, symbols and estimates for the model Parameter name (units) Vegetation sub-model Annual rainfall (mm) Mean rainfall (mm yr−1) Periodicity of rainfall Coefficient of variation of rainfall Grass biomass (kg ha−1) Shrub biomass (kg ha−1) Annual grass production (kg ha−1) Annual shrub production (kg ha−1) Grass decomposition (kg ha−1) Shrub decomposition (kg ha−1) Camp numbers Farm size (ha) Camp size (ha) Proportion of grass biomass which can be removed Rumen stimulants Cattle density (LSU ha−1) Sheep density (LSU ha−1) Goat density (LSU ha−1) Intake per Large Stock Unit per day (kg LSU−1 day−1) Camp time (days) Recovery period (years) Graze year Amount of grass eaten (kg ha−1) Amount of shrub eaten (kg ha−1)

Symbol

Estimates

Ra Rm p CV Gb Sb Gp Sp Gd Sd Cn Fs Cs Gbprop

Eq. (1) Adjustable 20 Adjustable 1000 1200 1.9×Ra 0.9×Ra 0.4×Gb 0.08×Sb Adjustable 7000 or adjustable Fs/Cn Eq. (2)

Rs Cdens Sdens Gdens

1= yes; 0= no Adjustable Adjustable Adjustable

ILSU Ct Rp Gy Ge Se

10 Eq. (3) ROUND[(Cn−1)Ct/365] Eq. (4) Eq. (5) Eq. (6)

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

Parameter name (units)

Symbol

Estimates

Production sub-model Area grazed per year (ha) Meat production (kg yr−1) Wool production (kg yr−1)

Ag Mp Wp

Eq. (7) Eq. (8) Eq. (9)

Financial sub-model Meat price (Rands and cents) Wool price (Rands and cents) Annual gross income (Rands) Average value per Large Stock Unit (Rands) Total animals stocked (LSU) Animal increment (LSU) Additional camps Marginal cost per camp (Rands) Set-up costs (Rands) Maintenance costs (Rands) Rumen stimulant costs (Rands) Expenses (Rands) Annual profit (Rands)

Pm Pw Igross Aval ATot Aincr Cincr Ccost SUcost Mcost Rcost Ex Iprofit

11.94 12.76 Eq. (10) 2000 Adjustable ATot−300 Cn−10 Fig. 2 Cincr×Ccost SUcost×0.02 If Rs = 1 then ATot×32 Eq. (11) Eq. (12)

References Acocks, J.P.H., 1966. Non-selective grazing as a means of veld reclamation. Proc. Grassland Soc. S. Afr. 1, 33 –39. ARC (Agricultural Research Council), 1980. The Nutrient Requirements of Ruminant Livestock. Commonwealth Agricultural Bureaux, England. Beukes, P.C., 1999. Livestock impact assessment and restoration strategies in the semi-arid Karoo. PhD thesis, University of Cape Town. Botha, P., Blom, C.D., Sykes, E., Barnhoorn, A.S.J., 1983. A comparison between the diets of small and large stock on mixed Karoo veld. Proc. Grassland Soc. S. Afr. 18, 101 – 105. Booysen, P, De, V., Klug, J.R., York, B.S., 1974. Number of camps for rotational grazing of veld. Proc. Grassland Soc. S. Afr. 9, 145 – 148. Boyazoglu, P.A., 1997. Animal nutrition. Concepts and application. Revised edition. van Schaik, J.L., (Ed.). Campbell, B.M., Dore´ , D., Luckert, M., Mukamuri, B., Gambizo, J., 2000. Economic comparisons of livestock production in communal grazing lands in Zimbabwe. Ecol. Econ. 33, 413 – 438. Costanza, R., Wainger, L., Folke, C., Ma¨ ler, K.-G., 1993. Modelling complex ecological economic systems. Towards an evolutionary, dynamic understanding of people and nature. Bioscience 43, 545 –555.

241

else

0

Danckwerts, J.E., 1987. The animal/plant interaction. In: Danckwerts, J.E., Teaque, W.R. (Eds.), Veld Management in the Eastern Cape. Pasture Research Section (Eastern Cape Region), Stutterheim, pp. 92 – 108. Dean, W.R.J., Hoffman, M.T., Meadows, M.E., Milton, S.J., 1995. Desertification in the semi-arid Karoo, South Africa: review and reassessment. J. Arid Environ. 30, 247 – 264. Desmet, P.G., Cowling, R.M., 1999. The climate of the Karoo— a functional approach. In: Dean, W.R.J., Milton, S.J. (Eds.), The Karoo: Ecological Patterns and Processes. Cambridge University Press, pp. 3 – 16. Du Toit, P.C.V., 1996. Karoobush defoliation in the arid Karoo. J. Range Manage. 49, 105 – 111. Du Toit, P.C.V., Blom, C.D., Immelman, W.F., 1995. Diet selection by sheep and goats in the Arid Karoo. Afr. J. Range Forage Sci. 12, 16 – 26. Edwards, P.J., 1988. Principles of grassland management. In: Tainton, N.M. (Ed.), Veld and Pasture Management in South Africa, Second Impression. Shuter and Shooter, Pietermaritzburg, pp. 313 – 321. Ellis, J., 1994. Climate variability and complex ecosystem dynamics: implications for pastoral development. In: Scoones, I. (Ed.), Living with Uncertainty. New Directions in Pastoral Development in Africa. Intermediate Technology Publications, London, pp. 37 – 46. Foran, B.D., Stafford Smith, D.M., 1991. Risk, biology and drought management strategies for cattle stations in central Australia. J. Environ. Manage. 33, 7 – 33.

242

P.C. Beukes et al. / Ecological Economics 42 (2002) 221–242

Hatch, G.P., Tainton, N.M., 1995. The influence of stocking rate, range condition and rainfall on residual herbage mass in the semiarid savanna of KwaZulu-Natal. Afr. J. Range Forage Sci. 12, 76 – 80. Hatch, G.P., Tainton, N.M., Ortmann, G.F., 1996. Towards the development of a bioeconomic stocking model for the semi-arid savanna of KwaZulu-Natal. Afr. J. Range Forage Sci. 13, 67 – 71. High Performance Systems Inc, 1993. Getting started with STELLA II. A hands-on experience. 45 Lyme road, Hanover NH 03755. Hobson, F.O., 1987. False Karoo. In: Danckwerts, J.E., Teague, W.R. (Eds.), Veld Management in the Eastern Cape. Pasture Research Section (Eastern Cape Region), Stutterheim, pp. 286 – 310. Hoffman, M.T., 1988. The rationale for Karoo grazing systems: criticisms and research implications. S. Afr. J. Sci. 84, 556 – 559. Leng, R.A., 1984. Supplementation of tropical and subtropical pastures for ruminant production. In: Gilchrist, F.M.C., Mackie, R.I. (Eds.), Herbivore Nutrition in the Subtropics and Tropics. The Science Press, pp. 129 –144. Ludwig, J.A., Sinclair, R.I., Noble, I.R., 1992. Embedding a rangeland simulation model within a decision support system. Math. Comp. Simul. 33, 373 –378. McCabe, K., 1987. Veld management in the Karoo. The Naturalist 31, 8 – 15. McKeon, C.M., Rickert, K.G., Ash, A.J., Cocksley, D., Scattini, W.J., 1982. Pasture production model. Proc. Aust. Soc. Animal Production 14, 201 –204. McNaughton, S.J., Ruess, R.W., Seagle, S.W., 1988. Large mammals and process dynamics in African ecosystems. Bioscience 38, 794 – 800. Meissner, H.H., Roux, C.Z., 1984. Growth and feed intake patterns: 2. Application to feed efficiency. In: Gilchrist, F.M.C., Mackie, R.I. (Eds.), Herbivore Nutrition in the Subtropics and Tropics. The Science Press, pp. 691 –705. Milton, S.J., Hoffman, M.T., 1994. The application of stateand-transition models to rangeland research and management in arid succulent and semi-arid grassy Karoo, South Africa. Afr. J. Range Forage Sci. 11, 18 –26. Milton, S.J., Dean, W.R.J., Marincowitz, C.P., Kerley, G.I.H., 1995. Effects of the 1990/91 drought on rangeland in the Steytlerville Karoo. S. Afr. J. Sci. 91, 798 –804. Nolan, T., 1996. Perspective: animal –vegetation relations which optimise production, utilisation and protection of natural resources: an African experience. Afr. J. Range Forage Sci. 13, 42 –48. Noy-Meir, I., 1978. Grazing and production in seasonal pastures: analysis of a simple model. J. Appl. Ecol. 15, 809 – 835. O’Reagain, P.J., Turner, J.R., 1992. An evaluation of the empirical basis for grazing management recommendations for rangeland in Southern Africa. J. Grassland Soc. S. Afr. 9, 38 – 49. Owen-Smith, N., 1991. Veld condition and animal perfor-

mance: application of an optimal foraging model. J. Grassland Soc. S. Afr. 8, 77 – 81. Pickup, G., Stafford Smith, D.M., 1993. Problems, prospects and procedures for assessing the sustainability of pastoral land management in arid Australia. J. Biogeogr. 20, 471 – 487. Riechers, R.K., Conner, J.R., Heitschmidt, R.K., 1989. Economic consequences of alternative stocking rate adjustment tactics: a simulation approach. J. Range Manage. 42, 165 – 171. Rutherford, M.C., 1980. Annual plant production – precipitation relations in arid and semi-arid regions. S. Afr. J. Sci. 76, 53 – 56. Savory, A., 1978. A holistic approach to ranch management using short-duration grazing. Proceedings of the First International Rangeland Congress. Denver, Colorado, pp. 555 – 557. Savory, A., 1991. Holistic resource management: a conceptual framework for ecologically sound economic modelling. Ecol. Econ. 3, 181 – 191. Savory, A., Parsons, S.D., 1980. The Savory grazing method. Rangelands 2, 234 – 237. Scoones, I., 1994. New directions in pastoral development in Africa. In: Scoones, I. (Ed.), Living with Uncertainty. New Directions in Pastoral Development in Africa. Intermediate Technology Publications, London, pp. 1 – 36. Seely, M.K., 1978. Grassland productivity: the desert end of the curve. S. Afr. J. Sci. 74, 295 – 297. Stafford Smith, D.M., Foran, B.D., 1988. Strategic decisions in pastoral management. Aust. Rangeland J. 10, 82 – 95. Stafford Smith, D.M., Foran, B.D., 1990. RANGEPACK: the philosophy underlying the development of a microcomputer-based decision support system for pastoral land management. J. Biogeogr. 17, 541 – 546. Stafford Smith, D.M., Foran, B.D., 1992. An approach to assessing the economic risk of different drought management tactics on a South Australian pastoral sheep station. Agric. Syst. 39, 83 – 105. Venter, J.M., Mocke, C., De Jager, J.M., 1986. Climate. In: Cowling, R.M., Roux, P.W., Pieterse, A.J.H. (Eds.), The Karoo Biome: a Preliminary Synthesis. Part I —Physical Environment. South African National Scientific Programmes Report No. 124, CSIR, Pretoria, pp. 39 – 52. Visser, J.N.J., 1986. Geology. In: Cowling, R.M., Roux, P.W., Pieterse, A.J.H. (Eds.), The Karoo Biome: a Preliminary Synthesis. Part I —Physical Environment. South African National Scientific Programmes Report No. 124, CSIR, Pretoria, pp. 1 – 17. Vogel, C., 1994. (Mis)management of droughts in South Africa: past, present and future. S. Afr. J. Sci. 90, 4 – 6. Vorster, M., 1994. Sustainable commercial stock farming in the arid parts of the Karoo. Karoo Agric. 6, 20 – 24. Wiegand, T., Milton, S.J., Wissel, C., 1995. A simulation model for a shrub ecosystem in the semiarid Karoo, South Africa. Ecology 76, 2205 – 2221.