Modelling extensive livestock operations which utilise common pastures and face weather risk in Ferlo, Senegal

Transcription

1 Modelling extensive livestock operations which utilise common pastures and face weather risk in Ferlo, Senegal Jarkko K. Niemi, Natural Resources Institute Finland (Luke) Kari Hyytiäinen, University of Helsinki Astou Diao Camara & Cheick Sadibou Fall, Institut Sénégalais de Recherches Agricoles Siwa Msangi, International Food Policy Research Institute 27 October 2017 This paper is part of the FoodAfrica Programme, financed as a research collaboration between the MFA of Finland, Natural Resources Institute Finland, IFPRI, ILRI, ICRAF, Bioversity International, University of Helsinki and HAMK University of Applied Sciences.

2 Abstract Semi-arid regions in the Sahel have faced increasing environmental pressure due to population growth and decreasing rainfall. Building on earlier research we develop a stochastic dynamic programming model that describes extensive, common-pasture-based seasonally moving (i.e. transhumant) livestock husbandry under stochastic and spatially varying weather. The model allows animals movement between two regions and takes into account that decisions can be adjusted when new information on the weather arrives. Our numerical analysis demonstrates that in the absence of efficient feed markets and under unpredictable weather, transhumance can be a rational livestock management strategy. Transhumance and stocking rate in more humid regions are buffers which increase resilience upon shortage of water in Ferlo. Decisions regarding herd size interact between the regions, and the option of transhumance can increase stock size in Ferlo and decrease its sensitivity to drought. Keywords: livestock, common pasture, grazing, climate change, climate variability, drought, resilience, feed, feed price JEL codes: D80, O13, Q12 2

3 Introduction Livestock serves as an important source of food and cash for many households in the Sub-Saharan Africa. Semi-arid regions in the Sahel, such as the Ferlo in Senegal, have faced increasing environmental pressure due to population growth and decreasing rainfall (Ickowicz et al. 2012). Households in the Ferlo region often practice transhumance, which means that the household or part of it moves seasonally with animals from a common pasture to another. These changes pose substantial challenges to extensive pastoral livestock management, which is a prevalent economic activity in the area. As a consequence of increased demand for food and reduced productivity of rangeland, the competition on feed and rangeland has increased and longer distances are travelled nowadays than in the past to feed the animals (Cesaro et al. 2010). We develop a dynamic programming model to describe rational livestock management under stochastic annual rainfall and spatial distribution of rainfall, and for various projections of future demand for meat. Management is organized so that it maximizes the total value of cattle as an asset of a representative household. Our model is built on an earlier model by Weikard and Hein (2011), which we extend in two important aspects. Firstly, we allow the movement of animals between two distinct regions (low vs. high rainfall region). While Ferlo is considered as the low rainfall area, Kaffrine is considered to receive more rainfall. Kaffrine is located in the Peanut basin area where there is more vegetation, higher market prices for livestock and more supplementary feeds are available for purchase than in the Northern Senegal. Secondly, we optimize the movement of animals between regions and the number of animals to be removed from the stock in both regions each year under uncertainty. These extensions allow a representative household to account for all information that is available each time period and adjust the decisions as necessary when new information about the weather arrives. Because transhumance is largely determined by the amount of rainfall, the decision to sell animals at the market and the rate of transhumance are made at the end of rainy season instead of deciding them before rainfall is observed as it was assumed by Weikard and Hein (2011). Also the longterm impacts on the production capacity of the soil are taken into account. Optimal decision rules to sell animals and move them between regions are investigated under exogenously given scenarios, which can be regarded to address different price, market and climate change conditions. The goal of our analysis is to examine how different factors contribute to the animal stock and transhumance in a setting where the stakeholder maximizes the value of his/her asset. The results will illustrate how transhumance, or lack of it, can impact households in the two regions and how different state of nature where decisions regarding the herd are made, can affect the decisions. The model Objective function Dynamic programming was used to analyze the pastoral livestock manager s decision problem. Partly similar approach was previously used by Weikard and Hein (2010). The 3

4 objective function of the pastoral livestock manager is to maximize the value of his/her herd by adjusting the stocking rate and the rate of transhumance (see Figure 1 for the description when each event takes place): subject to: transition and biophysical Equations 2-6 are given. where is the maximized value of livestock herd at time period t when the current stocking rate, carbon content of the soil and the current year s rainfall in the two regions indicated by the superscripts are given; is the control vector which contains decisions to sell animals from the stock in Ferlo ( ) and in outside Ferlo ( ) and the decision to move seasonally with animals (, transhumance is indicated as the percentage of Ferlo s livestock that is participating in transhumance) as indicated by the superscripts; is the pastoral livestock manager s annual net cash flow; is the annual discount rate; E is the expectations operator; and T is the number of years examined. Dry area More humid area Year 1 Rainy season begins Rainfall observed Decide transhumance Dry season begins Decide sales Herd returns to Ferlo New animals are born Stock in Ferlo Stock in Ferlo Sell Transhumance Sell Stock outside Ferlo Stock outside Ferlo Year 2 Rainy season begins Rainfall observed Stock in Ferlo New born Stock outside Ferlo New born Figure 1. Description of the timing of stock-related actions (transhumance, sale of animals, addition of new animals in the stock) and rainfall over one-year time period in the dry (low-rainfall) and more humid area. Stocking rate in this paper refers to the stocking rate (TLU/ha) in the beginning of each year. Change in the stocking rate in other figures refers to the change in the initial stocking rate from the beginning of year 1 to the beginning of year 2. 4

5 In our model the decision to sell animals at the market and the rate of transhumance are made at the end of rainy season (Figure 1) instead of deciding them before rainfall is observed as it was assumed by Weikard and Hein (2011). In addition, animals which move to the rainy region and are sold there are procured according to the rainy region s marker conditions. Transhumance directly affects soil carbon content in both regions and the amount of meat that can enter in the markets in the two regions, and it may affect the number of animals in stock outside Ferlo and in Ferlo. The stocking rate (TLU per ha, Tropical Livestock Units per hectare of land), rainfall and soil carbon content impact the production of vegetative biomass and therefore contribute to the transhumance and animal sales decisions. During the dry season, transhumance increases competition on vegetative biomass outside Ferlo whereas in Ferlo the competition is reduced. We have assumed that maximum 90 % of livestock population in Ferlo can move seasonally because people not moving with the animals may also need some livestock. The pastoral livestock managers annual profit,, is described by where superscript denotes region where the parameter is relevant, meat prices in period t, c is the variable costs of having the livestock, transhumance per TLU per ha and c 0 is fixed annual costs. s denote regional is the costs of Simulation of livestock and biomass in the rangeland This section describes the generic model for a single region when livestock population in transhumance is not taken into account. We model annual grass production (as dry matter),, in each time period t as a function annual rainfall,, rain-use efficiency of a semi-arid rangeland,, and carbon content of the soil, : where the initial soil organic matter content is denoted by, represents the relative impact of reduced carbon content on plant productivity, rainfall efficiency R is a parabolic function, is a scaling parameter, describes minimum rainfall required for plant growth and stands for (hypothetical) level of rainfall where productivity drops to zero. Empirical observations (Hein and De Ridder 2006) suggest that in years with high rainfall, R may be lower than in years with average rainfall, although R never goes close to zero. The model by Weikard and Hein (2011) was modified accordingly by adding a parameter that makes sure that R does not go below 90% of its maximum level even during the years with much rainfall. 5

6 Weikard and Hein (2011) neglected the impact of soil carbon in increasing the retention of nutrient and water, and thus improving the plant productivity. However, historical records show that intensity of land use alters soil productivity (e.g. Bauer and Black 1994), affects plant diversity (e.g. Müller et al 2012) and it is important to include such a relationship in a model. was set at for Ferlo rangelands. The dynamics of soil organic matter are described as: where and v are parameters and represents the stocking rate (TLU per ha). Rangeland productivity was translated into its annual capacity to support livestock grazing,, by diving grassland productivity by a parameter :. The dynamics of livestock are described as where denotes the state variable for livestock, is the chosen stock rate, denotes sold animals, represents the reproduction capacity of the livestock. When transhumance occurs, a proportion of animals are moved within a year from Ferlo to outside Ferlo and back, unless sold for food outside Ferlo. Solution method Numerical methods were used to solve the model because of their flexibility in future uses of the model. The state and control variables were discretized and interpolation was applied between the evaluation nodes. The numerical model was developed and programmed in Matlab R2013a ( , Mathworks inc.), and solved by using the value function iteration method (i.e. backwards, see Ljunqvist and Sargent 2000). Exogeneously given scenarios for prices and climate change We examine optimal stocking rates, level of transhumance and prospects for sustainable pastoral livestock management under stochastic annual weather and exogenously given scenarios for prices and climate change. The results are reported for three year types (average rainfall, dry and humid year) and for the two study areas. The baseline scenario is a two-region model where transhumance can occur. Several scenarios were examined and compared with the baseline scenario. The possibility of practicing transhumance was also examined. Our analysis examined scenarios where one the following characteristics were adjusted from the baseline model 6

7 while keeping other factors at the same level as in the baseline: the mean of rainfall, the standard deviation of rainfall, meat price, and discount rate (Table 2). Increases in the mean and the standard deviation of rainfall reflected climate change scenarios and they were based on UNDP climate change country profile (McSweeney et al. 2010). Climate change can decrease the average annual rainfall and also increase the variability of weather. For instance, McSweeney at al. (2010) projected that the rainfall in Senegal can decrease by approximately 3% per decade by Moreover, the variability of rainfall has been estimated to increase by 30%. To address possible effects of rising demand and prices (see e.g. OECD-FAO 2014), we considered a scenario in which the price level of meat increases over time. The increase was based on the projected and observed population growth rates in Ferlo. Data Monthly statistics on rainfall ( ) available for Dahra observation station located in the Southern Ferlo and for a Kaffrine (outside Ferlo) observation station were used to describe the weather. Mean, variance and covariance of weather data were estimated based on the statistics. Next, probability distributions for annual rainfall were simulated for each iteration. Future projections of rainfall were based on mutually independent draws from the random distributions but taking into account the correlation between the two regions. Apart from the parameters specified in the previous section, other parameters in equations 2 to 6 were similar to Weikard and Hein (2011). The parameter values are summarized in Table 1. The price of meat varies by season and region. The price is typically higher in the more humid region than in the dry region (CSA 2014). Different prices for animals sold in Ferlo and for animals sold outside Ferlo were used. The price scenario was based analysis of statistics and news about seasonal fluctuation of meat price. This analysis, in addition to interviews with the herdsmen in Ferlo, suggested that the price of meat is higher outside Ferlo than in Ferlo during the transhumance. As opposed to Turner and Williams (2002), an analysis conducted with the weather and FAO long-term price data do not suggest that locally observed drought would significantly affect meat prices in the region. Other price parameters were based on interviews with herdsmen in Ferlo during a field study. T was set at 30 years. 7

8 Table 1. Parameter values used in the model. Parameter Value Remarks m α 1.251*10-5 r t Ferlo 282 SD 83, Dahra observation station r t outside Ferlo 611 SD 38, Dahra observation station r 29 r 252 φ 2511 μ v 27.6 β 0.6 H 44 Mean household size, Weikard and Hein (2011) m - Currently not used Ferlo p t CFA/TLU, calculated from the market price statistics p t outside Ferlo c c move c 0 δ T CFA/TLU, calculated from the market price statistics 2700 CFA/TLU, cost of labor 3259 CFA/TLU, extra labor, water and vaccination costs 0 Not relevant in this study 5.5 % real discount rate 30 Number of years Results Colors in Figure 2 illustrate the percentage change of the animal stock (i.e. newly born animals minus off-take divided by initial population) during one year period. The change is represented for different combinations of initial stocking rates in Ferlo and outside Ferlo, and for three different year types. The sustainability of current animal population in both the Ferlo and outside Ferlo depends upon the rainfall and stocking density in both regions because there is a movement of animals between the regions. In the event of an average rainfall year, Figure 2 shows that Ferlo (Figure 2c) can accommodate less animals (i.e. a lower stocking rate) than the region outside Ferlo (Figure 2d). Stocking rate also influences the offtake of animals. The larger stocking rate in the beginning of a year the more animals are removed from the stock during the year. The result is linked to the number of animals that one hectare of common pasture can feed. The rainfall impacts so that in a rainy year (e.g. Figure 2a) the stocking rate can increase during the year to a higher level than during a dry year (e.g. Figure 2e). Figure 3 represents the economically sustainable stocking rates. The sustainable stocking rate in one region depends on the stocking rate in the other region. The higher is the 8

9 initial stocking rate in one region, the lower it will be in the other region. However, rainfall also has a major impact on the optimal stocking rate. Figure 4 describes the rate of transhumance (% livestock population moving from Ferlo to outside Ferlo) by the stocking rate within the Ferlo and outside Ferlo. The rate of transhumance increases rapidly when the stock of animals in the Ferlo increases. However, when stocking rate outside of the Ferlo is high enough, there is more competition on biomass and the rate of transhumance increases less rapidly when the stocking rate in Ferlo increases. A higher proportion of the population is participating in transhumance in a dry year than in a more humid year. Transhumance can act as a balancing factor and increase the aggregate stocking rate. As herds which move between regions are competing for the same resources as pastoral livestock outside the Ferlo, the option to practice transhumance increases the stocking density within the Ferlo by a maximum level of 20%, but decreases it outside Ferlo by a maximum level of 5%. The option to practice transhumance was simulated to increase the value of livestock activity by approximately 5%. However, the decrease in the stocking rate outside of the Ferlo depends upon the situation. If there is initially a large stock of animals in the Ferlo, the reduction in the stock outside Ferlo can be more dramatic than if there is a small level of stock in the Ferlo. The results indicate that in the absence of efficient feed markets and under unpredictable weather conditions, transhumance can be a rational livestock management strategy. Weather has an important role in herder s decision-making especially in years when it turns out to be a dry year. If supplementary feeds are available in Ferlo, they can stabilize animal stock and herders income. However, the results indicate that the price of feed plays an important role: using supplementary feed is profitable to the herder in Ferlo only when the price of supplementary feed falls below about 30 CFA per kg. This is because common pasture is available free of charge to the herders. By contrast, economic variables play an important role when rainfall is not limiting herd sales decisions. Increased frequency of extreme weather conditions, such as heavy drought or rainfall, may have more severe impacts on livestock husbandry than gradual changes in the mean annual rainfall or temperature levels suggest. Hence, policies should aim at mitigating the negative consequences of extreme weather. 9

10 Table 2. Change (%) in value function when the baseline scenario and alternative scenarios are compared Scenarios Percentage change in the value function B Mean rainfall -3% per decade -1 % Standard deviation of rainfall C +30% -3 % E Meat price +20% 21 % G Meat price -20% -21 % H Meat price +2% per year 21 % L Discount rate doubled -37 % a) Dry, % sold, in Ferlo b) Dry, % sold, outside Ferlo c) Average, % sold, in Ferlo 5 5 d) Average, % sold, outside Ferlo e) Humid, % sold, in Ferlo 5 f) Humid, % sold, outside Ferlo Figure 2. Change (%) in the stocking rate (TLU/ha) in Ferlo (Figures a, c, e) and outside Ferlo (b, d, f) for dry (a, b), average rainfall (c, d) and humid (e, f) year at different initial stocking rates in the respective regions in the beginning of the year

11 Sustainable stocking rate in Ferlo TLU/ha Sustainable stocking rate outside Ferlo TLU/ha Sustainable stocking rate in Ferlo Sustainable stocking rate outside Ferlo Dry 1Average 0.09 Humid 0.07 Dry Average Humid Initial Initial Figure 3. Economically sustainable stocking rate in Ferlo (on the left) as a function of initial stocking rate in the region outside Ferlo and year type, and economically sustainable stocking rate outside Ferlo (on the right) as a function of initial stocking rate in the Ferlo and year type. a) Dry b) Average c) Humid Figure 4. The rate of transhumance (% livestock population in Ferlo moving) by the stocking rate (TLU/ha) in Ferlo and in Outside Ferlo in the beginning of the year (dark colors imply stronger reduction in the animal stock over a year) and for three year types rainfall (average, low, high rainfall) Conclusions The adjustments of animal stock size and transhumance are determined by natural constraints (the availability of feed) and by economic decisions to maintain a stocking rate that his consistent with the market prices and resources available. Rainfall has a major impact on the optimal stocking rate and on the amount of livestock sold to the market. In the event of drought, more animals will be sold than in a normal year. In particular it affects the stocking rate in the Ferlo where water is the key limitation to the production of vegetative biomass. By contrast, when there is a plenty of rainfall, the availability of vegetative biomass may not be a limiting factor, and the animal stock is managed so as to maximize the value of the herd. Transhumance and stocking rates are buffers which are adjusted according to the levels of rainfall simulated within our model. In the absence of efficient feed markets, 11

12 transhumance can be a rational way to adjust the livestock numbers to conform to varying weather conditions and biomass levels. It increases the possibilities of herdsmen in Ferlo to maintain their livestock at an acceptable level of productivity and health. However, it can increase stocking rate to a level which is not sustainable in some years. Results suggest that market conditions and time value of money can have a small impact on the optimal stocking rates and transhumance patterns. For instance, a larger discount rate can reduce animal stock levels. Livestock managers in Ferlo could benefit from the option to purchase feeds from the markets instead of exclusively practicing transhumance. However, the uptake of purchased feeds is dependent upon the price. Purchased feeds and transhumance are alternative options but are not mutually exclusive. Transhumance is likely to continue although it could be carried out in a smaller scale with only some animals remaining in place with increased supplementation (i.e. for milk production). References BAUER, A., BLACK, A.L. (1994). Quantification of the Effect of Soil Organic Matter Content on Soil Productivity. Soil Science Society of America Journal 58, CESARO, J.-D., MAGRIN, G., NINOT, O. (2010). Petit atlas de l élevage au Sénégal. Commerce et territories. Publication du projet de recherché ATP ICARE. 36 p. ICKOWICZ, A., ANCEY, V., CORNIAUX, C., DUTEURTRE, G., POCCARD- CHAPPUIS, R., TOURÉ, I., VALL, E., WANE, A. (2012). Crop livestock production systems in the Sahel increasing resilience for adaptation to climate change and preserving food security. p in: Meybeck, A., Lankoski, J., Redfern, S., Azzu, N., Gitz, V. (eds.) Building resilience for adaptation to climate change in the agriculture sector. Proceedings of a joint FAO/OECD workshop, April CSA Bulletins mensuels sur le suivi des marchés agricoles au Sénégal. Various issues. Commissariat à la Sécurité Alimentaire, Dakar. LJUNQVIST, L., SARGENT, T.J. (2000). Recursive Macroeconomic theory. MIT Press, Cambridge. 701 p. MCSWEENEY, C., NEW, M., LIZCANO, G., (2010). UNDP Climate Change Country Profiles: Senegal. oxford.report.pdf [Accessed 16 December 2013]. MULLER, J., KLAUS, V.H., KLEINEBECKER, T., PRATI, D., HÖLZEL, N., FISCHER, M. (2012). Impact of Land-Use Intensity and Productivity on Bryophyte Diversity in Agricultural Grasslands. PLOSone 1371/journal.pone TURNER, M.D., WILLIAMS, T.O. (2002). Livestock market dynamics and local vulnerabilities in the Sahel. World Development 30: WEIKARD, H.-P., HEIN, L. (2011). Efficient versus sustainable livestock grazing in the Sahel. Journal of Agricultural Economics 62: