Modelling and Predicting Future Urban Expansion of Lagos, Nigeria from Remote Sensing Data Using Logistic Regression and GIS
Aniekan Eyoh, Dupe Nihinlola Olayinka, Peter Nwilo, Onuwa Okwuashi, Mfon Isong, Daniel Udoudo

Abstract
This research explores the implementation of a loosely coupled logistic regression model and geographic information systems in modelling and predicting future urban expansion of Lagos from historical remote sensing data (Landsat TM images of Lagos acquired on 1984, 2000 and 2005). ArcGIS and MATLAB software are used for the modelling. The three Landsat images are classified using the k-means unsupervised algorithm in MATLAB. Ten salient explanatory land use variables are extracted for the calibration of the model. The model is calibrated by running a simulation for period 1984 to 2000. The computed logistic coefficients of the 10 explanatory variables show that all the 10 explanatory variables are significant at 95% confidence level based on a two-tailed test, since all the 10 variables yields p-values <0.05. The simulated map in 2000 is compared with the reference data in 2000; and evaluated using the Kappa statistic. The computed Kappa statistic is 0.7640; which implies a substantial agreement between the predicted and the reference data. The calibration model for 1984-2000 is used to predict 2005 map. A comparison of the predicted and reference data in 2005 yields Kappa statistics estimate of 0.6998; which indicates a substantial agreement between the predicted and the reference data. A prediction of 2030 is derived upon satisfactory result obtained for the 2005 prediction based on the 1984-2000 calibrated model. An urban expansion of 129.49% is predicted between 1984 and the forecasted 2030.

Full Text: PDF