IMPROVED URBAN GROWTH PREDICTION WITH INTEGRATED MARKOV CHAIN AND KALMAN FILTER MODEL

Effective and inclusive urban planning is the key to sustainable development. An important aspect of urban planning is future land use land cover (LULC) change simulations. Broadly speaking there are two types of LULC change prediction models, namely, temporal and spatio-temporal. In this paper we focus on temporal models that quantify land change without predicting the co-ordinates of change. We start with Markov Chain, a probabilistic model, that is extensively used in the literature and highlight its disadvantages. We then systematically propose Kalman filter based richer temporal model. A key contribution of this paper is the ability to incorporate impact of the external driver variables (for example macro-economic factors, demography) that impact LULC changes as control inputs. This allows us to simulate scenario based future growth with varying growth rates, a lacunae created by the Markov Chain model. The proposed model has been implemented and validated on a sample dataset. The prediction error rate for our model on the dataset is less than 1.5%, compared to about 21.5 % for the corresponding Markov Chain model. For the same dataset we simulate two futuristic growth scenarios with varying growth rates. 1