|Title||Modelling the dynamics of agricultural development : a process approach : the case of Koutiala (Mali)|
|Author(s)||Struif Bontkes, T.|
|Source||Agricultural University. Promotor(en): H. van Keulen, co-promotor(en): A. Kuyvenhoven. - S.l. : Struif Bontkes - ISBN 9789058080240 - 233|
Theoretical Production Ecology
|Publication type||Dissertation, internally prepared|
|Keyword(s)||landbouwontwikkeling - agrarische bedrijfsvoering - landgebruik - dynamica - simulatiemodellen - besluitvorming - agricultural development - farm management - land use - dynamics - simulation models - decision making|
|Categories||Agriculture (General) / Agricultural Economics (General) / Economics (General)|
Sustainability of agricultural production and food supply is threatened in many developing countries by human population growth. The increasing food requirement forces the population to extend the cultivated areas to less fertile areas, often without taking sufficient measures to maintain soil fertility, causing soil degradation and declining yields. Moreover, the ensuing competition for land and other resources increases differentiation between rich and poor. To change this course of development, appropriate measures have to be taken at the farm and at the policy level, requiring insight in the relevant agro-ecological and socio-economic processes and their interactions.
However, there is no generally applicable definition of sustainability, hence criteria need to be defined in relation to the area studied. In this study, the criteria pertain to agro-ecological aspects such as soil fertility, crop and animal production, and to socio-economic aspects such as income distribution and cereal prices. As the concept of sustainability has a temporal dimension, a time frame needs to be defined. A period of 25 to 30 years is considered suitable, as the chance of uncertain events increases with time. An ecologically uniform region within a country is considered an appropriate level of analysis of sustainability issues, as this allows to take decision making at both regional and farm level into consideration.
To obtain insight in the processes related to agricultural sustainability, it is useful to consider the problem situation as a system and to represent it as a quantitative model. It is thereby important to develop such models interactively with the stakeholders. These models should include agro-ecological as well as behavioural processes at the farm level and allow aggregation of these processes to the regional level.
In the past decennia, different types of agro-ecological models have been developed. Most of these models address only a limited number of aspects, such as crop production or organic matter dynamics and are often limited to one growing season. Moreover, these models are usually very detailed and have high data requirements. For the purpose of this study, insight in the interactions between ecological processes, such as soil fertility and crop growth over a longer period is required. As the availability of data is often limited in developing countries and as it is not necessary to make precise predictions in a regional study with a long time frame, summary models are used to simulate ecological processes. These summary models allow integration of many processes over a longer period and require a limited amount of data.
Farmer's behaviour can be modelled in several ways using econometric techniques, mathematical programming or decision rules. Econometric techniques are used to predict future behaviour based on historical data. Limitations of these techniques for this study are their limited suitability to deal with new phenomena and their extensive data requirements. Mathematical modelling is suitable for optimisation, but less appropriate to describe actual behaviour. The use of decision rules to represent human behaviour offers more flexibility and is less dependent on the availability of data.
There have been several attempts to develop models that provide insight in the ways to achieve sustainable agriculture at the regional level. However, they are not very satisfactory for one or more of the following reasons:
The current study has been undertaken to develop a modelling approach that is suitable to:
The empirical setting of this study is the Koutiala region in SouthEast Mali. The major crops grown in this area are millet, sorghum, maize, cotton and groundnut. Cotton has appreciably increased the incomes of the farmers in the area and as a result, both the number of farmers and the cultivated area have increased. However, sustainability of this development is being threatened: the area under continuous cultivation is rapidly increasing, very often without taking sufficient measures to maintain soil fertility and to prevent erosion, leading to soil degradation. Due to lack of alternative investment possibilities, farmers spend their surplus income on the purchase of cattle, causing overgrazing of the common pastures.
Two dynamic simulation models have been developed in this study:
The farm model
The farm model consists of one core model and four data sets, each representing a particular farm type. Four farm types (A, B, C and D) are distinguished, mainly based on herd size, area cultivated and level of equipment. By changing a number of parameters in the data sets, the effect of different management strategies on soil fertility, crop and livestock production, farm income and food availability can be simulated. As the farm is subdivided in a number of fields of 1 ha, the model permits also to examine the effects of various crop rotations.
Soil fertility indicators, used in the model are organic matter, nitrogen, organic and inorganic phosphorus, pH and soil depth. The model simulates changes in these indicators caused by e.g. the application of fertiliser and manure, decomposition of organic matter, removal by crops, erosion, leaching etc., using time steps of one year. Soil moisture content is simulated on a monthly basis. Crop production is determined by uptake of nitrogen and phosphorus, water availability, effect of pests and diseases and effect of labour input.
Animal production (growth rates, calving rates, death rates and milk production) is determined by the amount and quality of the available feed on a monthly basis. The feed consists of grass and browse from the common pastures, crop residues and some concentrate. The results show decreasing soil organic matter contents on all farms types. Phosphorus contents, however, are increasing except on D farms, as these farms do not apply fertiliser. Soil pH decreases due to the use of ammoniacal fertiliser.
Millet yields decrease over time due to the decrease in soil organic matter, the most important source of nitrogen for this crop. Maize appears to be susceptible to drought, partly explaining the reluctance of the farmers to grow this crop. Nevertheless, cereal supply can be maintained above the minimum requirement for all farm types. Results of model experiments suggest that stable feeding of millet straw positively influences animal production, which is further enhanced by the introduction of dolichos as an intercrop in maize. Introduction of dolichos, however, reduces maize yield and hence incomes.
Soil conservation measures such as ridging and tied ridging, increase maize yields in dry years by reducing run-off losses of water and fertiliser, and by increasing water infiltration. The increased labour requirement for the construction of tied ridges, however, renders this practice less attractive for the farmer than simple ridges.
Determination of the number of animals per farm that maximizes income, results in very large herd sizes per farm but also in large areas of pasture land required to feed these herds. This explains the interest of farmers to increase their herds, but also shows the consequences of this practice for the environment. Model experiments suggest that taxation of the use of pasture land at a rate of FCFA 3000 per ha would reduce the interest of the farmers to continuously increase their herds.
The regional model
The regional model is based on the farm model. Soil processes, labour requirements, farm income and crop and animal production are modelled in the same way. The way crops are rotated over the different fields, however, has not been included in the regional model. In the regional model, the farmers are regarded as actors. This implies that their behaviour has become endogenous such as crop choice, purchase of fertiliser and purchase and sale of cattle.
The model simulates agricultural development over the period 1980 - 2025. The development of the number of farms per farm type plays a central role in the regional model. The number of farms per farm type may change for several reasons:
farmers may change type if the changes in their herd size are such that the new herd size does not match the criteria of the current type. At the start of the simulation, the number of farms per farm type is known as well as their household sizes, the areas of sandy and loamy soil occupied, the number of animals per farm and the soil fertility per farm and soil type.
Farmers determine the area per crop to be cultivated on the basis of their food requirement, expected yields, net revenues per crop, taste preferences, credit availability, input supply, etc. The application of animal manure depends on availability and on the crop. The use of fertiliser is determined by the crop and the fertiliser-crop price ratio. Cereal prices are endogenously determined on the basis of the surplus production and the demand of the non-farm population.
Depending on his income, the farmer may use part of it to invest in cattle. If the average herd size of a particular farm type increases, part of the farms belonging to that type, move to a 'higher' type. On the other hand, if average herd size decreases, part of the farms move to a 'lower' type. When a farmer moves to another farm type, he takes his household, land and herd with him. These changes, along with the effects of population growth and migration, result in changes in the number of farms per farm type, average household size, herd size, area and soil fertility per farm type. Land, required for the new farms and for the expanding farms is withdrawn from the common pasture area.
Model results show a continuous increase in cultivated area until all land is occupied, and decreasing levels of organic matter and, hence, of millet yields. Decreasing millet yields and increasing urban demand lead to higher cereal prices and, hence, to a larger share of cereals cultivated. Due to favourable incomes, farmers invest in cattle, resulting in an increase in the number of large (A) farms. However, as the common pasture area is shrinking, animal feed supply falls short of the requirement, increasing animal death rates and reducing the herd size and, hence, A farms become B farms.
In the basic model, the behaviour of the different farm types is described by different sets of decision rules. Running the model over a number of years, however, results in changes that create a new situation, such as a structural shortage of feed and decreasing millet yields. It is likely that farmers will adapt their behaviour to the new circumstances, e.g. they may change their selling and investment strategy and improve feed supply by growing a fodder crop or start to apply fertiliser on their millet crop when yields drop below a certain level. Therefore, the model has been adapted by including such behaviour.
Finally, a number of policy experiments has been carried out: changes in prices of fertiliser and cotton, introducing a tax on the use of pasture land and increased off-farm wages. Increasing cotton price and reducing the fertiliser price both by 20 %, increase the area of cotton, the use of fertiliser and income, stimulating farmers to increase herd size. The larger number of animals in the area results in lower animal growth rates, reducing herd sizes per farm and causing the number of A farms to decrease and the number of B farms to increase.
Increasing off-farm wages reduces the number of farms, as young people leave the farms to find a remunerative job outside agriculture. As higher wages positively affect off-farm incomes of the farm households, enabling farmers to increase their herds, the number of A farms increases. Hence, the decrease in the total number of farms is compensated by the increasing share of A farms, maintaining cereal production and prices at approximately the same level. Experiments to explore the effects of taxation of the use of common pasture land at the regional level suggest that a taxation of the use of pasture land of FCFA 5000 per ha reduces total herd size and total number of farms, especially A farms, and improves feed availability.
Evaluation of the approach
The regional model presented in this study integrates biophysical and socio-economic aspects: farm management decisions affect the resource base of the farmer and the changes in the resource base affect farm management. In addition to that, the model integrates the farm and the regional level: the behaviour of the farmer influences cereal prices and land availability, which influence the behaviour of the farmer in the subsequent period. The descriptive approach to the modelling of decision making provides flexibility in the simulation of farmer's behaviour and offers possibilities for sociological research, a discipline that ought to play a more important role in land use studies.
Questions may be raised however regarding the predictive power of such models. Is it possible to make reliable predictions on agricultural development of a large region, comprising many farms of different types? This is further complicated by a lack of tested theories, relevant to the situation, and by a paucity of reliable longitudinal data that are required to construct and validate the model. Moreover, unpredictable events, such as droughts, devaluation, changes in world market prices of cotton and political changes constitute sources of uncertainty. The model should therefore be considered as a hypothesis, that may be applied for decision support, rather than an instrument that enables the decision maker to predict the future with some certainty.
As a model cannot capture the real world and as the real world continuously changes, the model should be repeatedly subjected to a process of testing. In this process, model predictions are compared with real world data, followed by an adjustment of the model. This implies that the model should not be considered as a fixed but rather as an evolving representation of the real world. The validity of such models may be further enhanced if developed in continuous interaction with stakeholders, including farmers, researchers and policy makers. Hence, this approach emphasises the importance of processes in two ways: the processes that are part of the model and the processes related to the way the model is developed.
Such models may be helpful in improving understanding of the dynamics of the system, allowing decision makers at farm and regional level to improve the quality of their decisions. The model may also help to discover discontinuities in behaviour when conditions change as shown above. Similarly, the model may be useful in discovering undesirable trends and permits exploration of effects of various policies or identification of topics for agricultural research that may contribute to avoid or remedy problems in the future.
This approach might also be used in combination with approaches using mathematical programming, where the latter may be used to generate technically feasible options for sustainable land use, while the approach used in this study could serve to determine how to stimulate adoption of such land use systems, starting from the present situation.