Markov Land Cover Change Modeling Using Pairs of Time-Series Satellite Images

Abstract

Models of change processes created with the Markov chain model (MCM) can be used in the interpolation of temporal data and in short-term change projections. However, there are two major issues associated with the use of Markov models for land-cover change projections: the stationarity of change and the impact of neighboring cells on the change areas. This study addressed these two issues using an investigation of five time-series land-cover datasets generated between 1972 and 2009 for the Liverpool region of NSW, Australia. Four short term transition matrices were computed, and the results were used to predict land-cover distributions for the near future. The issue of neighborhood effects was addressed by incorporating spatial components in a Cellular Automata (CA)-based MCM, and the results were compared with those derived from a normal MCM. Given the marginal improvements in the simulation achieved with CA-MCM rather than MCM, and because of the ability of CA-MCM to incorporate spatial variants, CA-MCM was determined to be the more suitable method for predicting land-cover changes for the year 2019. The land-cover projection indicated that future land-cover changes will likely continue to affect the natural vegetation, which will in turn likely decrease through the continued conversion of natural to agricultural lands over the years.