Modelling pastoral livestock production: Problems and prospects

Introduction

One of the principal tasks of the International Livestock Centre for Africa (ILCA) is to identify opportunities for improving livestock productivity in sub-Saharan Africa. Therefore, the underlying causes of the variable and usually low output levels of livestock production systems need to be understood and quantified. From the beginning, ILCA staff believed that modelling should be complementary to other research activities as it would help to analyse the multiple interactions that exist between herd productivity, forage resources and management regimes. ILCA chose to develop further the Texas A & M model because of its high level of generalisation and flexibility, and its proven ability to simulate production systems in tropical environments. As described by Konandreas and Anderson (1982) several modifications have been built into the ILCA model to make it more suitable for modelling pastoral systems.

In this paper the focus is on the key determinant of pastoral herd productivity, i.e. the quantity and quality of available forage, which in turn generates the daily intake of metabolisable energy as the driving input of the model. It is argued that the model seeks to simulate the input and output of well-defined and location-specific production systems for which field data are available to validate and fine-tune the model. This leads to the contention that there is little merit in using a model in which the input component inadequately resembles the real world. Several shortcomings in the quantification of the required inputs were identified when the output of the model was validated for a number of West African pastoral production systems (de Leeuw and Konandreas, 1982), while further modifications to the output side have been proposed by Wagenaar and Kontrohr (1986).

The main premise of this paper is that a more realistic approach to the prediction of daily intake of forage is desirable. To identify the major problems and suggest solutions a stepwise scenario is proposed. Firstly, the general characteristics of the system that is to be modelled should be identified. Since by definition, each system is location-specific, the second step is to define its boundaries and assess its resource base. This requires delimiting the expected levels of primary productivity taking into account the environment, inter-annual variation and rates of exploitation. The third step involves the development of realistic profiles of the quantity and quality of forage on offer and the interactions of these variables with daily feed intake. To better identify and quantify the limitations to intake, the model should take into account the two primary determinants of intake: the amount of forage on offer daily, and the harvesting activities of the grazing animal.

Input requirements for predicting intake

The general structure of the model and the data needed to run it have been described in detail by Konandreas and Anderson (1982). For this paper, it suffices to discuss briefly the data needed for simulating intake: i.e. the driving force of the model.

Although it is recognised that voluntary intake of feed by grazing cattle is influenced by the quantity and quality of forage and by climatic and management factors, it is mainly forage quality that determines the prediction of intake. Forage on offer is simulated based on monthly data for an annual cycle of 12 months. Several year types are required to provide an adequate representation of the long-term resource variability. For each year type, monthly digestibility and crude protein content of forage on offer are identified. Year types are drawn randomly based on a given probability of the forage supply situation for each production system.

Intake is reduced when CP content is less than 6%, but given the close relationship between CP% and digestibility, the reduction of intake is defined in terms of digestibility alone as (d/0.4)^0.6. Thus, intake begins to decrease when forage digestibility is less than 40%. Similarly, when digestibility of the forage rises above 65%, intake is constrained by the physiological limit of the animal and is reduced to maintain metabolisable energy intake equal to that at 65% digestibility. Forage intake is reduced when standing biomass is less than 0.8 t/ha and when the distance grazing animals walk exceeds 14 km/day (or c. 4 hours of non-grazing activity).

In summary, the model requirements indicate that, in addition to climate and management, other factors such as forage palatability, species preference and regularity and quantity of water intake are not considered in the prediction of intake.

Forage resources in pastoral systems

Before discussing the monthly input of forage data required for the model, attention is focused first on the characteristics of three location-specific pastoralist production systems as examples of their complexity. These are: (1) transhumant Fulani in Mali: (2) semi-sedentary Fulani in the subhumid zone in Nigeria; and (3) semi-sedentary Maasai in Kenya. Their general features are summarised in Table 1 with regard to climate, degree of mobility and the importance of cropping, while components of the seasonal fodder supply are illustrated in Table 2. From these tables it is clear that pastoralists need to employ a wide range of management and movement strategies to exploit fully the different forages available and to minimise the effects of fluctuating supply between seasons and years.

Table 1. Major characteristics of three livestock production systems in Mali, Nigeria and Kenya.

Derived from de Leeuw (1984).

Table 2. Seasonal fodder utilisation of three livestock production systems in Mali, Nigeria and Kenya.

xxx = highly important: xx = important; x = low importance.

For example, in the rainy season transhumant herds in Mali mainly graze annual grasslands several hundred kilometres away from their home area, while during the dry season they gradually move across the floodplain of the Niger following the recession of the floods. Due to different levels and durations of flooding, plant cover on the floodplain is extremely variable and consequently stock have access, simultaneously, to tall-standing Andropogon gayanus stands, inundated Echinochloa stagnina grassland and regrowth on rice fields after harvest, some of which are burnt (de Leeuw and Diallo, 1983; Breman et al, 1978 ) .

In Nigeria the resource base of pastoralists in the subhumid zone is less diverse, since they rely on upland savanna for over 80% of their total grazing. Nevertheless, crop residues, browse and regrowth after burning are important grazing resources during the dry season (van Raay and de Leeuw, 1974; Bayer, 1984). The fluctuation in quality and quantity of the potential components of the diet of grazing cattle is shown in Table 3 for grazable residues of sorghum and millet and in Table 4 for a tall stand of perennial grass at the end of the growing season.

Table 3. Components and quality characteristics of forage from sorghum and millet fields at the start of grazing in the subhumid zone of Nigeria.

Component	Sorghum			Millet
	D%	CP%	%	D%	CP%	%
			yield			yield
Immature panicles	60	7.8	1	65	12.6	2
Upper leaves	60	7.3	6	60	11.4	7
Lower leaves	54	3.3	8	59	7.6	10
Upper stalk	49	1.4	16	48	2.4	23
Lower stalk	45	1.3	35	46	2.5	38
Total cereal	48	2.8	66	50	4.1	8.0
Grasses and weeds	55	7.0	34	55	7.0	20
Total	50	4.2	3.3a	51	4.7	2.1a

Derived from Powell (1984).

d% = digestibility:

CP% = Crude Protein Content.

a= t/ha

Table 4.Crude protein content (CP%) and in vitro digestibility (IVDMD%) of young and old leaves of Andropogon gayanus at the end of the growing season, Nigeria.

	CP%		IVDMD%
Date	Young	Old	Young	Old
12 September	6.8	2.1	66.0	49.5
3 October	8.7	3.1	56.6	29.2
26 October	6.6	3.4	54.4	39.4

Derived from Haggar (1970) .

The least complex of the examples chosen, as regards forage resources and management options, is that of the Maasai in semi-arid Kenya. They rely on cattle and small-stock husbandry for most of their subsistence and cash income, exploit a rather uniform and limited orbit mainly of perennial grasslands and rarely engage in cropping. However, their resource base is subject to tremendous variability in time and space, giving rise to unpredictable cycles of boom and bust periods (de Leeuw et al, 1984).

Range resources and primary productivity

For modelling specific range livestock production systems a prediction of the geographical and temporal distribution of range resources is required. As a first step, grazing resources are derived from secondary sources such as natural resource and vegetation surveys, which provide maps and descriptions of the major land uses and vegetation types of the area within which the livestock production system operates. Often, those regional surveys include estimates of end-of-season biomass, from which assessments of potential carrying capacity can be made. Furthermore, analysis of demand for and supply of forage resources can lead to identification of regional imbalances in utilisation (e.g. de Leeuw and Milligan, 1983; de Leeuw, 1976).

Equations linking annual or seasonal rainfall to end-of-season biomass are used to assess the grazing capacity of the land. Linear regressions were developed by Le Houerou and Hoste (1977) for West Africa and by Deshmukh (1984) for East Africa. Although the limitations of this approach are well recognised (cf Breman et al, 1984), these relationships appear to hold even when applied to smaller areas like the Tsavo National Park in semi-arid Kenya (Figure 1). Long-term weather variables have been used to explain past boom and bust periods for Maasai pastoralists in Kenya (de Leeuw, unpublished), and famine conditions in maize-growing regions in Kenya and Ethiopia (Stewart and Faught, 1984; Henricksen and Durkin, 1984). Fluctuations in carrying capacities are shown by the data in Table 5, which were generated to identify probable sequences of year types for modelling the Maasai system (see Figure 3).

Figure 1. Linear regressions of annual or seasonal rainfall on primary productivity of herbaceous cover in sub-Saharan Africa

Table 5. Annual carrying capacity based on biomass yield for different combinations of rainfall seasons in semi-arid Kenya (ha/TLU/annum).

First rains	Good---------------------------------------------------->Bad
Second rains1	Biomass
	t/IM/ha	3.0	2.0	1.0	0.5
Good	3.0	1.2	1.5	2.0	4.2a
	2.0		1.8	2.2	4.5
	1.0			2.7	5.0
Bad	0.5				7.2

¹First rains: October–December; second rains: March–May.

^a Based. on the assumption that 50% of standing biomass is consumed, and daily DM disappearance rate is 10 kg/day for a TLU of 250 kg.

After defining the overall demand and supply situation of the system to be modelled, the monthly average data on quantity and quality of forage on offer are required. For modelling African systems, mostly secondary data were used: e.g. Sullivan et al (1981) in Tanzania, de Leeuw and Konandreas (1982) for West Africa, and Konandreas et al (1981) for Botswana. The problems with this approach when applied to pastoral production systems have been discussed above.

There have been many efforts to simulate the forage supply available to grazing animals. These range from a simple pasture growth model using the length of the growing season (derived from a soil water balance model, e.g. McCown, 1981) to complex models that aim at simulating the entire soil-plant-animal complex (e.g. Wight, 1983). Although the ultimate aim of these models is to simulate animal productivity, the majority treat plant productivity as a separate subroutine and use plant production as an input for simulating the animal production component.

In Australia, McKeon et al (1980) developed an index of daily pasture growth from the product of separate soil, moisture, temperature and radiation indices. Daily indices were transformed into a mean seasonal growth index which was then multiplied with the potential pasture growth rate (McKeon et al, 1982).

Thus plant growth was related to the amount of green material capable of transpiring at a rate predicted by the daily soil moisture balance. Cornet (1984) used a similar model for predicting annual forage growth in the Sahel in Senegal, while Sullivan et al (1981) simulated changes in composition of perennial swards in subhumid Tanzania by partitioning biomass in standing green and dry material on a daily basis. The daily amount of green herbage (available to livestock) was a function of green forage at the beginning of the day, accounting for additional new growth and for losses of green growth to the dry biomass pool. Daily growth rate was influenced by soil moisture balance, the starting date of the growing season, the leaf-area index and stocking rate. The principal concept of these models is to relate transpiration to dry-matter yield and appears well suited to generating the input for simulating animal productivity. The approach is flexible end to model both annual and perennial species and their inter-annual variation as well as defoliation and grazing effects.

An alternative approach to modelling plant growth uses the same abiotic variables to drive the CO₂ assimulation rate and to simulate the flora of plant biomass and nitrogen in daily time steps (Hansons et al. In: Wight, 1983). These models were originally developed for the Grassland Biome Study and were modified to simulate the Serengeti grasslands in East Africa (Coughenour et al, 1984), monsoonal grassland in India (Parton and Singh, 1984), and the annual Sahel grassland (Penning de Vries and Djiteye, 1982).

It is not within the scope of this paper to indicate which of these primary production models is most appropriate for simulating plant–animal systems, but it seems that for this purpose, models are needed that combine "increased generality, less unnecessary complexity, easier data demand and greater validity" (van Keulen et al, 1981).

Forage quality profiles

Monthly values of forage quality for the two West African systems are shown in Figure 2. As already shown in Table 2, the forage resources are extremely variable and it is unlikely that these averages approximate the real world, even if between year variations are accounted for by inserting year types into the model.

Figure 2. Average monthly quality values of biomass on offer and expected intake levels for two pastoral production systems in West Africa

Therefore, it is proposed to construct monthly forage profiles which provide ideally a combined assessment of the quantity of forage on offer by quality class based on digestibility and/or crude protein content. This approach was tried for the Maasai system in Kenya, where fodder supplies are less diverse because rangelands with perennial grasses are the major grazing resource. Also, herd mobility is relatively low, implying that forage quality classes can be linked to known grazing pressure and expressed in kg/ha. Hence the monthly forage available in each quality class can be calculated (Figure 3).

To analyse the grazing resource situation, three data sets were compiled. The first provided monthly averages of digestibility and crude protein content. To illustrate the variability between year-types, the parameters are given for an above- and below-average rainfall (Figure 3a). Although differences between years are pronounced the annual curves follow similar trends. In a good year, average crude protein content is at 8 % or more during 8 months, in contrast to 5–6 months in a poor year.

The second set provides estimates of monthly averages of standing biomass derived front relationships between seasonal rainfall and biomass (Figure 3b) and supplemented by field measurements. The yield data in Figure 3 represent monthly estimates of standing biomass under a moderate level of stocking of 3-4 ha/TLU. It is sham that in a good year standing biomass rarely drops belay 1.5 t EM/ha whereas in a bad year, standing crop is less than that level for most of the year.

The third data set divides the standing biomass into three quality classes using CP% as a proxy for quality. At the onset of the rains there is a rapid increase in high-quality biomass concomitant with a rapid disappearance of old standing crop left from the previous season (Figure 3c) . With continued herbage growth, CP content in current growth declines together with further reduction in old standing forage, so that at the end of the rains only medium- and poor-quality forage are still available for grazing. Comparing the two year-types, it is evident that differences in CP supply become very pronounced; while in a good year there is 1 t of good-quality forage per hectare for 6 months in a poor year this period lasts only for 2–3 months. Hence for 9–10 months only poor- and medium-quality feed is available.

Figure 3. Grazing resource profile by month in semi-arid Kenya

Grazing behaviour of pastoral cattle

The concepts governing grazing behaviour and the search strategies which herbivores use when preferred species or the quantity of acceptable forage are limiting have not been well defined experimentally (Rice et al, 1983).

In pastoral systems, where livestock are herded, insufficient time to graze has been recognised as a factor limiting daily intake of forage. In the ILCA model, this limitation is simulated through reducing intake when the daily distance travelled by the herd exceeds 14 km (Konandreas and Anderson, 1982). Although distance travelled reduces time for grazing (14 km equates to approximately 4 hours of walking at 3.5 km/h), the actual duration and intensity of foraging may be more directly related to forage intake than the distance walked.

Most herded cattle walk less than 14 km per day. Exceptionally, longer distances (20–30 km) were covered by pastoral herds in northern Nigeria during the latter part of the dry season (van Raay and de Leeuw, 1974). In Mali, Dicko et al (1981) reported an average daily distance walked of 12.7 km, which consisted of 5.1 + 0.5 km of walking and 7.6 + 0.1 km of searching for forage. In Nigeria, semi-settled pastoralists rarely cover more than 10 km daily (Bayer, 1984; van Raay and de Leeuw, 1974), while Maasai herds walk 10–14 km during the dry season and less during wet periods.

There is a general trend for herders to extend the grazing orbit and the length of grazing day when grazing resources became scarce and of poorer quality. The grazing day may increase from 6–7 hours in the wet season to 9–12 hours in the dry season. Bayer (1984) recorded very short grazing days in the subhumid zone of Nigeria during the rainy season and concluded that this may contribute to the low productivity of this production system.

Another variable that may affect forage intake is grazing intensity. Van Raay and de Leeuw (1974) found that the proportion of high-intensity grazing was inversely related to total grazing time, indicating a self-adjusting mechanism in grazing behaviour. Using number of bites per minute as a measure of grazing intensity, de Leeuw and Peacock (1982) found that grazing intensity was negatively correlated with walking speed. Almost all grazing (93–96%) was done at walking speeds, of less than 1.5 km/h. Consequently, daily 'speed profiles' were used to determine actual. hours of grazing per day. Therefore, it seems necessary to adapt the model so that, in addition to the intake restriction due to distance walked, a second restriction is inserted that takes account of situations where grazing time is limiting.

Forage on offer and intake

Although Konandreas and Anderson (1982) included a correction factor to adjust voluntary intake for limitations in the quantity of acceptable forage on offer, this was never validated. Sanders and Cartwright (1979) incorporated the monthly forage dry matter available per animal per day and stated that forage consumption per unit time may be affected by forage density and several other factors, such as distance from water. However, no minimum quantity was given for acceptable forage per unit area below which intake is restricted. In a further development, Sullivan et al (1981) interfaced livestock productivity with forage quantity and quality, but in the application for Tanzania, available forage was assumed to be non-limiting. However, they used two levels of stocking which resulted in lower simulated growth rates at the higher level due to lower CP content and digestibility.

The SPUR model is probably the most comprehensive, since annual preferences for grazing sites and locations as well as for up to nine forage species or species groups were included together with a factor for physical availability of forage (Rice et al, 1983). This factor was defined as the proportion of the total above-grid biomass of a plant species group which is readily consumed by a single grazing event. Availability is therefore primarily a function of the growth form of the plant, rather than its quantity per unit area. From these, a plant supply matrix for grazing was developed. Supply and demand matrices for all sites were calculated on a daily basis, yielding daily intake of digestible dry matter. This daily removal was discounted in subsequent supply matrices.

Hendricksen et al (1982) in Australia related intake to dietary CP content, which was generated from a herbage growth model. This model estimated green herbage by age class and assumed that grazing progressed preferentially from young to old herbage¹. Intake was reduced when standing biomass was less than 230 kg/ha and also when the amount of herbage removed was greater than 30% of the standing dry matter.

¹ Konandreas (1980) outlined a similar procedure in the early phase of the development of the ILCA model.

Several validations of these simulation procedures have been published. Kothman and Smith (1983) used the Texas A & M model to evaluate management alternatives in cow-calf operations and stated that the model was adequate when forage quality and availability were available for different alternatives, but emphasised that there is a critical need to develop plant/animal interface models that will accurately predict availability and quality for forage and nutrient intake of grazing animals for different range and livestock management schemes. Validations of interfaced plant/animal models in Australia were satisfactory for sown grass/legume pasture. However, growth predictions of livestock grazing native pastures was poor, partially because the modelling of the quality of dead material on offer during winter periods was considered inadequate (McKeon et al, 1980).

To what extent the inclusion of other intake-limiting factors will improve simulations of pastoral systems remains to be seen. If the opinion expressed, by Cordova et al (1978) that 'no method has been devised by which intake of grazing livestock can be quantified' remains valid, validation of modelled intake will continue to be difficult. Variable levels of intake by pastoral cattle were recorded by Dicko et al (1983) using total faecal collection. Daily intake ranged from 1.9 to 3.9% of liveweight and from 0.5 to 1.2 kg per hour of actual grazing. These values are comparable to those given by Cordova et al (1978 ) .

Few data are available on the effect of low biomass yield on intake. Allison et al (1982), in comparing a wide range of short-term grazing pressures, showed that intake did not decrease even when daily forage availability was close to daily intake (± 10 kg/day) and there was very little standing biomass. It may be advantageous to consider the forage availability per grazing event, as proposed by Rice et al (1983), to define relationships between available forage and intake. If such events are equivalent to bites, it can be postulated that for freely ranging cattle, daily available forage is related to the area covered. If average number of bites/day equals 16 000 and the bite area is approximately 0.01 m², a mature animal can cover 160 m²/day. If 10 kg DM/day is assumed to be the minimum quantity required, standing acceptable forage should be more than 625 kg/ha. This is close to the 800 kg/ha mentioned by Konandreas and Anderson (1982) as the quantity below which intake is reduced.

Model validation

To date, few African pastoralist production systems have been modelled, due to the complexity of this mode of production coupled with the difficulties and costs involved in monitoring the long-term production parameters of traditional livestock enterprises.

De Leeuw and Konandreas (1982) validated the ILCA model for four West African pastoralist systems and found good correspondence between model output and the real world when productivity parameters were aggregated in a production index, but stated that simulated animal growth rates were over-estimated, and that mortality and monthly conception rates were difficult to simulate. Improvements to the simulation of these parameters have been proposed by Wagenaar and Kontrohr (1985) using the data from a 4-year study of transhumant herds in Mali (Wagenaar et al, 1984) . Therefore, in this section comparison between simulated and actual productivity will be confined to animal growth rates. This parameter has received more attention, because weight changes are easier to monitor than reproductive rates, mortality and milk output (Wilson and Semenye, 1983).

As shown in Table 6, simulated and actual weights of one-year-old calves and animals close to maturity are well matched, whereas the model over-estimated the growth of immature animals. Possible causes of these discrepancies have been given by Wagenaar and Kontrohr (1985). Also monthly weight changes usually showed a good fit both in West Africa (Table 7) and in Botswana (Kahn and Spedding, 1984). Unfortunately, none of the field data were for pastoralist cattle but were derived from grazing trials on experimental stations. The same applies to the growth data used by Sullivan et al (1981) to validate their model in Tanzania.

Table 6. Actual and simulated liveweights at four ages for female cattle in production systems in Mali and Nigeria.

System	Year				Reference
	1	2	3	4
	Liveweight (kg)
Transhumant, Mali
Actual	80	125	170	198	Wagenaar et al (1984)
Simulated	71	153	226	237	de Leeuw and Konandreas
					(1982)
Agropastoral, Nigeria
Actual	80	145	190	245	Pullan (1980)
Simulated:          1.	80	164	262	280	Konandreas and
2.	98	192	277	280	Milligan (1981.)

1. Born in December (mid dry season).

2. Born in May (early wet season).

Table 7. Actual arid simulated liveweight changes of growing cattle in semi-arid West Africa.

Actual	Initial	Dry season	Wet season	Total annual
Changes	weight (kg)	loss (kg)	gain (kg)	gain (kg)
Malil	255	35	90	55
	344	65	95	35
Niger2	150	15	75	60
Simulated changes
Mali3	250	40	104	64
Mauritania4	133	+10	82	92
	225	43	58	13

1 Derived from grazing trials on the Niono Ranch with steers during 1978/79 (de Leeuw and Hiernaux, 1980).

2 Summarized from grazing trials on four ranches in the Sahel zone of Niger (Wylie et al, 1983).

3 Simulation of the agropastoral livestock system in the 'Office du Niger' area in Mali (de Leeuw and Konandreas, 1982).

4 From Greenwood (1985) based on livestock nutrition and grazing trials in southern Mauritania.

The sensitivity of the Texas A & M model to changes in the nutritive value of forage on animal performance was tested by Kahn and Spedding (1984) in Botswana and by Sullivan et al (1981) in Tanzania. In Botswana, by lowering monthly digestibility by 5%, annual weight gain dropped from 105 kg to 90 kg per head, while weight losses (3 kg vs 23 kg) and subsequent growth rates were pronounced. In Tanzania, nutritive value of forage was influenced by stocking rate and annual rainfall; simulated annual growth in a good year at low stocking rate was 110 kg, dropping to 50 kg in a poor year at high stocking rate.

Simulated weight changes in adult female cattle in the subhumid zone in Nigeria were given by Konandreas and Milligan (1981). Season and parturition time interacted strongly, resulting in seasonal weight ranging from 350 to 260 kg. These weight changes seem in excess of those reported by Wilson (1983) for Mali.

In an attempt to predict animal performance across a climatic gradient in West Africa, Ketelaars (1984) used the length of the growing season and dietary N content as driving forces. For the 600–800 mm rainfall zone (southern Sahel and Sudan zone) his predictions are close to those presented in Tables 6 and 7. In contrast, annual weight gain in the subhumid zone reached 147 kg, because dietary N% was assumed to be more than 1% for 9 months, whereas in reality this period lasts only for 5–6 months. As a consequence, actual liveweight gains in long-term grazing trials are in the range of 60–80 kg per annum or fairly similar to those in the drier areas further north (de Leeuw, 1971) .

From this review, it appears that growth rates and seasonal liveweight changes can be simulated satisfactorily whenever data on forage quality and availability adequately resemble the real world (Kothman and Smith, 1983; Kahn and Spedding, 1984) . It also implies that the simulated output is as good as the input data that generate the driving force of the model.

References

Allison, C D, Kothman M M and Rittenhhouse L R. 1982. Efficiency of forage harvest by grazing cattle. J. Range Manage. 35:351–354.

Bayer W. 1985. Agropastoral herding practices and grazing behaviour of cattle. In: Livestock systems research in Nigeria's subhumid zone. Proceedings of the 2nd ILCA/NAPRI symposium, Kaduna, Nigeria, 29 Oct–2 Nov 1984. ILCA, Addis Ababa.

Breman H, Diallo A, Traore G and Djiteye M A. 1978. The ecology of the annual migrations of cattle in the Sahel. In: Proceedings of the 1st International Rangeland Congress. Denver, Colorado, USA.

Breman H, van Keulen H and Ketelaars J J M H. 1984. Land evaluation for semi-arid rangeland: a critical review. In: W Siderius, ed. Proceedings of the workshop on land evaluation for extensive grazing. ILRI, Wageningen. pp. 229–244.

Cordova F J, Wallace J D and Pieper R D. 1978. Forage intake by grazing livestock: a review. J. Range Manage. 31:430–438.

Cornet A. 1984. Utilisation de modéles simples de bilan hydrique et de production de biomasse pour déterminer les potentialités de production de parcours en zone sahélienne sénégalaise. In: W Siderius, ed. Proceedings of the workshop on land evaluation for extensive grazing. ILRI, Wageningen. pp. 207–228.

Coughenaur M B, McNaughton S J and Wallace L L. 1984. Modelling primary production of perennial graminoids-uniting physiological processes and morphometric traits. Ecol. Model. 23:101–134.

Deshmukh I K. 1984. A common relationship between precipitation and grassland peak biomass for East and southern Africa. Afr. J. Ecol.22:181–186.

Dicko M S, Lambourne J L, de Leeuw P N and de Haan C. 1981. Voluntary intake and livestock productivity in the Sahel zone of Mali. In: Proceedings of the 32nd annual meeting of EAAP. August 1981, Zagreb, Yugoslavia.

Dicko M S, Lambourne L J, de Leeuw P N and de Haan C. 1983. Nutrition animale, les bovins. In: R T Wilson, P N de Leeuw and C de Haan, eds. Recherche sur les systèmes des zones arides du Mali: résultats préliminaires. ILCA Research Report No. 5. ILCA, Addis Ababa, Ethiopia.

Greenwood G. 1985. Evaluation of Sahelian livestock strategies using regression models of cattle performance. (submitted for publication.)

Haggar R J. 1970. Seasonal production of Andropogon gayanus. J. Agric. Sci. Camb. 74:487–494.

Henricksen B L and Durkin J W. 1984. Moisture availability and prospects for early warning of famine in Ethiopia. Internal report, ILCA, Addis Ababa, Ethiopia.

Hendricksen R, Rickert K G, Ash A J and McKeon G M. 1982. Beef production model. Proc. Aust. Soc. Anim. Prod. 14:204–208.

Kahn, Hava E and Spedding C R W. 1984. A dynamic model for the simulation of cattle herd production systems. Agric. Systems 13:63–83.

Ketelaars J J M H. 1984. Evaluating rangeland as feed resource for livestock production. In: W Siderius, ed. Proceedings of the workshop on land evaluation for extensive grazing. ILRI, Wageningen. pp. 165–181.

van Keulen H, Seligman N G and Benjamin R W. 1981. Simulation of water use and herbage growth in arid regions. Agric. Systems 6:159–193.

Konandreas P A. 1980. Modelling forage intake and animal performance when the quality of forage on offer is highly variable: The case of transhumance in Mali. Unpublished report, ILCA, Addis Ababa, Ethiopia.

Konandreas P A and Anderson F M. 1982. Cattle herd dynamics: An integer and stochastic model for evaluating production alternatives. ILCA Research No. 2. ILCA, Addis Ababa, Ethiopia.

Konandreas P A and Milligan K. 1981. Evaluation of alternative production strategies within sedentary Fulani systems in northern Nigeria. Draft report. ILCA, Addis Ababa, Ethiopia.

Konandreas P A, Anderson F M and Trail J C M. 1981. Economic trade-offs between milk and meat production under various supplementation levels in Botswana. In: Proceedings of the IASTED conference. Cairo, September 1981.

Kothman M M and Smith G M. 1983. Evaluating management alternatives with a beef production systems model. J. Range Manage. 36:733–740.

de Leeuw P N. 1971. The prospect of livestock production in the northern Guinea zone. Samaru Agric. Newsl. 13:124–133.

de Leeuw P N. 1976. Fodder resources and livestock development in northeastern Nigeria. Savanna 5:61–74.

de Leeuw P N. 1984. Pastoral production systems and land utilization types. In: W Siderius, ed. Proceedings of the workshop on land evaluation for extensive grazing. ILRI, Wageningen. pp. 119–132.

de Leeuw P N and Diallo A. 1983. Conditions de pâturage et nutrition. In: R T Wilson, P N de Leeuw and C de Haan, eds. Recherché sur les systémes des zones arides du Mali: resultats préliminaires. ILCA Research Report No. 5. ILCA, Addis Ababa, Ethiopia.

de Leeuw P N and Hiernaux P. 1980. The secondary productivity of Sahelian rangelands. Internal report. IICA/Mali, Bamako.

de Leeuw P N and Konandreas P A. 1982. The use of an integer and stochastic model to estimate the productivity of four pastoral production systems in West Africa. In: Proceedings of the National Beef Conference. Kaduna, Nigeria, July 1982. pp. 184–212.

de Leeuw P N and Milligan K. 1983. Resources animales. In: R T Wilson, P N de Leeuw and C de Haan, eds. Recherches sur les systémes des zones arides du Mali: résultats préliminaires. ILCA Research Report No. 5. ILCA Addis Ababa, Ethiopia.

de Leeuw P N and Peacock C P. 1982. The productivity of small ruminants in the Masai Pastoral System, Kajiado district, Kenya. In: Proceedings of the SR-CRSP workshop on small ruminants. pp. 118–138.

de Leeuw P N, Bekure S and Grandin B E. 1984. Some aspects of livestock productivity in Maasai group ranches in Kenya. In: Proceedings of the International Savanna Symposium. Brisbane, Australia. (in press.)

Le Houerou H N and Hoste C H. 1977. Rangeland production and annual rainfall relations in the Mediterranean basin and in the African Sahelo-sudanian zone. J. Range Manage. 30:181–189.

McCown R L. 1981. The climatic potential for beef cattle production in tropical Australia. Part III. Agric. Systems 7:163–178.

McKeon G M, Rickert K G, Robbins G B, Scatini W J and Ivory D A. 1980. Prediction of animal performance from simple environmental variables. In: Proceedings of the 4th biennial conference of the Simulation Society of Australia. Brisbane, Australia. pp. 9–16.

McKeon G M, Rickert K G, Ash A J, Cooksley D and Scatini W J. 1982. Pastoral production model. Proc. Aust. Soc. Anim. Prod. 14:201–204.

Parton W J and Singh J S. 1984. Adapting a biomass simulation model to a tropical grassland.Ecol. Model. 23:151–163.

Penning de Vries F W T and Djiteye M A (Eds). 1982. La productivité des pâturages sahéliens: Une étude des sols, des végétations et de 1'exploitation de cette ressource naturelle. Agric. Res. Reports 918. PUDOC, Wageningen. 525 pp.

Powell M. 1984. Cropping and crop–livestock interactions. In: Livestock systems research in Nigeria's subhumid zone. Proceedings of the 2nd ILCA/NAPRI symposium, Kaduna, Nigeria. 29 Oct–2 Nov 1984. ILCA, Addis Ababa.

Pullan N B. 1980. Productivity of White Fulani cattle on the Jos Plateau, Nigeria. II: Nutritional factors. Trop. Anim. Health Prod. 12:17–24.

van Raay H G T and de Leeuw P N. 1974. Fodder resources and grazing management in a savanna environment. Occasional paper No. 45, Institute of Social Studies, The Hague, The Netherlands.

Rice R W, MacNeil M D, Jenkins T G and Koong L J. 1983. SPUR livestock component. In: J R Wight, ed. SPUR simulation of production and utilization of rangeland: a rangeland model for management and research. Misc. Publ. 1431, USDA.

Sanders J O and Cartwright T V. 1979. A general cattle production systems model. II. Procedures used for simulating animal performance .Agric.Systems 4:289–309.

Stewart J I and Faught W A. 1984. Response farming of maize and beans at Katumani, Machakos district, Kenya. East Afr. Agric. For. J. 44:29–51.

Sullivan G M, Cartwright T C and Farris D E. 1981. Simulation of production systems in East Africa by use of interfaced forage and cattle models. Agric. Systems 7:245–265.

Wagenaar K T and Kontrohr E. 1986. Validation of the ILCA model on cattle herd dynamics with data from pastoral systems in Mali and Kenya. In: N de Ridder H van Keulen, N G Seligman and P J H Neate, eds. Modelling of extensive livestock production systems. Proceedings of the ILCA/ ARO/CABO workshop, 5–9 Feb 1985, Bet Dagan, Israel. ILCA, Addis Ababa.

Wagenaar K T, Diallo A and Sayers A R. 1984. The productivity of transhumant Sudanese Fulani cattle in the Niger Inner Delta. Internal report, ILCA, Addis Ababa, Ethiopia.

Wilson R T. 1983. Production animale. In: R T Wilson, P N de Leeuw and C de Haan, eds. Recherches sur les systémes des zones arides du Mali: résultats préliminaires. ILCA Research Report No. 5. ILCA, Addis Ababa, Ethiopia.

Wilson R T and Semenye P P. 1983. Livestock productivity and management. In: Proceedings of a workshop on 'Pastoral systems research in sub-Saharan Africa.' ILCA, Addis Ababa, Ethiopia. pp.151–180.

Wight J R. (Ed.). 1983. SPUR simulation of production and utilization of rangeland: a rangeland model for management and research. Misc. Publ. 1431, USDA.

Wylie B R, Sencock L, Snyder B, Roentgen B and Porter B. 1983. Range research and results. USAID/NRL preliminary report, Niamey, Niger.

Discussion

Question– The upper line in Figure 1 is based on data from enclosed areas. All the plotted data from van Wyngaarden seem to fall below this line. Are those data collected from grazed areas?

Answer – I am not sure.

Question – Is your poor year in Figure 3 a really poor year? You measured 3.5 t/DM/ha with 450 mm of rainfall. In a really poor year I would expect higher CP%.

Answer – It was below average. We did not have enough data to compare standing crop of the same age with low and high rainfall, but we did look at the correlation between quality and quantity. Quality should be higher in a bad year than in a good year, but my figures do not show that, because the sampling system was not designed to prove that particular point.

Question – What do you mean by boom and bust periods?

Answer– That is the long-term fluctuation of good and bad periods of about 5 years each.

Question – Bad and good years in rainfall or also in productivity?

Answer – Difficult to say, since we only collected data for two years. But calculations of productivity indices for systems in Mali, Nigeria, Kenya and Sudan did indicate much similarity.

Question – Would you say that there is a strong stabilising factor in the system which maintains a fairly constant reproduction rate, and that all of the variations that we are looking at in detail are in fact elements of one system?

Answer – It is a very interesting hypothesis to test.

Comment – I do not agree with your conclusion that there is little difference in productivity of livestock system in semi-arid Africa. Your hypothesis is very dangerous, because it does not show whether the system is in a steady state or not, whether herds are growing or dying out. You should first check if your parameters of reproduction and mortality can create a viable herd or not.

Comment – I agree that we have to take into account good and bad situations. I calculated a productivity index for four or five systems in Africa and they all seem to have the same ratio, but I want to make the point that this index will not discriminate between a higher productivity of milk in one system and higher calving rates and lower calf mortality in another system because of the relationship between calving rates, calf mortality and milk offtake.

Reply – The final productivity index obscures the difference in values among its various factors. It is only useful as a broad comparison of productivity. Despite the uniformity over systems, you will find big differences between individual flocks and herds in the area, associated with some aspect of management or health. This indicates that the system can be manipulated.

Question – Are you sure that the difference between herds is not a matter of inequitable distribution of resources: some are doing well because others are doing badly?

Answer – As soon as someone loses 50% of his herd, flock productivity index is low as compared with those who did not lose stock, but productivity per animal is not so depressed.

Comment – I would like to add something to that question of large differences between the individual flocks. It is irrelevant to all systems which are close to maintenance because under such circumstances small improvements in feed conditions can have a tremendous effect on productivity. A 12% increase in energy could double the liveweight gain or make the difference between weight gain and weight loss.