Logistic modelling to derive agricultural land use determinants: A case study from southeastern Nigeria

Abstract

Human activity has a strong impact on spatial land cover patterns. Land cover characteristics show significant differences between the five local agricultural land use (ALU) systems at Ikem, southeastern Nigeria. Aerial photograph interpretation (1:6000), participatory rural appraisal and logistic modelling were combined to elicit spatial determinants and to model ALU using a partially nested strategy. The binary logistic model for estimating probabilities of private ALU correctly predicted 95.7% of the 300 sample plots and 89.8% of the 88 validation plots, all located within nine strip transects. The odds for private ALU increased with 21.3% per 10 m distance closer to the settlement and were highest for a location on the accumulation glacis. An ordinal logistic model for predicting four communal ALU classes correctly classified 83.3% of the sample plots and 78.3% of the validation plots. The odds for longer fallow periods increased with 21.8% per 100 m distance away from the settlement and were higher on both the mixed and erosion glacis. Landform, distance to the settlement and customary ownership were the most important of the local determinants for ALU. The ALU models could be incorporated into a land use framework for planning purposes, scenario analysis or impact assessment at the local government level.