LCM Logistic regression modelling of land-use changes in Kouhmare Sorkhi, Fars province

Abstract

Model-based prediction of potential land use patterns can support environmental planners and natural resource managers for more effective planning decisions. This research aimed at modeling land use change across kouhmare Sorkhi plain of Fars province by means of logistic regression. To do this, land use maps were first retrieved from multi-date Landsat imagery of 1987, 2000 and 2012. A change detection analysis was performed following the validation of classified imagery. Results of change detection between 1987 and 2000 returned a Kappa coefficient of 0.83 showed the highest increment for rangeland class (4224.24 ha). In contrast, the highest rate of decrease was observed on forest area (3953.75). Having this in mind, 10 explanatory variables were selected to model potential land use changes in 2012 by LCM logistic regression. The land use map of 2012 was modelled using a Markov chain. This resulted in a 0.74 Kappa coefficient for an error matric between the modeled and the 2000 Landsat-based land use maps. In addition, the revealed changes for the 2000-2012 period with 0.88 Kappa coefficient show a highest increment rate for rangeland class (1807.02 ha), whereas the greatest decrease was observed for forest (2132.82 ha). In accordance to this, the land use map of 2024 was predicted, for which the irrigated agriculture would be associated with the highest rate of change.

Keywords

• Change Detection
• Logistic regression
• LCM model
• Kouhmare Sorkhi region