Modeling and characterizing agricultural expansion: a case study in Senegal, West Africa.

Eric Channing Wood - 2002
[UMI Proquest Full Citation]

The overall goal of this study was to better understand the drivers of the expansion of agriculture in rural Senegal, and was attained through research focusing on two major objectives. (1) To determine an accurate approach for modeling the expansion of agriculture in southern Senegal. This objective required the comparison of multiple runs of the artificial neural network based Land Transformation Model, using a variety of parameterizations. Model modifications were made that maximized temporal generalization through the incorporation of land use/land cover change trajectories. (2) To explore the factors that are driving the expansion of agriculture in southern Senegal. Addressing this objective involved both modeling and detailed field investigations. A theoretical framework that emphasizes “extensification vs. intensification” of agriculture along the lines of the neo-Boserup-Malthus contradiction was used for assessing findings resulting from these investigations. Results of the study indicate that the department of Velingara is meeting its agricultural production needs through extensification, rather than intensification, of agriculture. All evidence, both from the field and through analysis of remotely sensed data, suggests that shifting agriculture continues to be the primary agricultural system in place in the department. Velingara fails to meet several of the criteria necessary to move from extensification, or labor-led intensification of agriculture, to a more sustainable capital-led move towards intensification. Another key result was the determination that an artificial neural network (ANN) based model was able to accurately model the expansion of agriculture in Velingara. KHAT statistics indicate that this model predicts transitions to agriculture at a level significantly greater than random chance. It was also shown that trajectory of change concepts can be incorporated into ANN-based models and provide significant increases in predictive accuracy.