Temporal-Spatial Prediction of Land Use Changes Using LCM Model in Doroodzan Dam Watershed

Abstract

Predicting future landuse changes is an important step in the proper planning and management of watersheds. Therefore, in this study, the land use modeling of the Doroodzan Dam watershed was discussed. First, the land use maps of 1996, 2005, 2016 and 2021 were extracted using Envi software, and the maximum likelihood method. The transition potential maps were modeled for each of the sub-models by using multi-layer perceptron artificial neural network and 6 variables and Markov chain was used to calculate the allocation change to each land use. Then, the land use map for 2021 was predicted using Land Change Modeler (LCM) and 2005-2016 calibration period. To verify modeling accuracy, Hits 1.92%, Misses 8.8%, False Alarms 2.94% were calculated. The ratio of Hits to the total pixels has changed 14% indicates that model results are acceptable. Then, the land use map for the year 2050 was predicted. The results showed that given the predicted LC for year 2050 in comparison with the year 2021, bare land, agricultural land, residential areas and orchards will increase by 54290, 7621, 4494, and 2391 ha, respectively. Whereas grassland and forest cover will decrease 68441, and 689 ha, respectively.

Keywords

• artificial neural network
• landuse
• lcm
• markov chain