Markov chains and cellular automata to predict environments subject to desertification

Area of Science:

• Environmental Science
• Geographic Information Science
• Computational Geography

Background:

• Desertification poses a significant environmental threat, necessitating accurate prediction models.
• Land use/land cover (LULC) changes are key drivers and indicators of desertification.
• The Vieira River basin in Brazil is identified as an area susceptible to desertification.

Purpose of the Study:

• To evaluate the performance of a combined Markov chain and cellular automata (CA-Markov) model.
• To predict LULC changes in a desertification-prone region.
• To assess the model's accuracy in both quantitative and spatial predictions.

Main Methods:

• Utilized a CA-Markov model for LULC change prediction.
• Performed LULC prognosis for the year 2005.
• Employed Kappa indices to validate model predictions against ground truth data.
• Assessed cellular automata performance relative to the Markov chain model.

Main Results:

• The Markov chain component demonstrated efficiency in quantitatively predicting LULC changes.
• The cellular automata component showed average performance in the spatial distribution of LULC classes.
• The CA-Markov model effectively estimated the total area of 'Bare Soil,' a class highly susceptible to desertification.
• Spatial prediction of 'Bare Soil' by the cellular automata was found to be inefficient.

Conclusions:

• The CA-Markov model exhibits good overall predictive capacity for LULC changes, particularly in quantitative assessments.
• While effective for total area estimation of susceptible classes like 'Bare Soil,' the model's spatial prediction capabilities require further refinement.
• The study highlights the utility and limitations of CA-Markov models in understanding and managing desertification-prone environments.