Geospatial Land cover change analysis using the CA-Markov Chain Model in Chikwawa District, Malawi

Changes in Forest cover impact the local climate by altering energy and water exchanges, assessing and quantifying these changes is crucial for sustainable resource management and the protection of ecosystems. The study evaluated IsoData, Support Vector Machine, Random Forest, and Maximum Likelihood classifiers using Kappa Coefficient and accuracy metrics for Landsat images (1979-2023). CA-Markov projected future land cover for 2035, 2045, and 2065, while Spearman rank correlation assessed the statistical significance of changes. The Random Forest classification yielded the highest accuracies, with Kappa coefficients of 85%, 86%, 86%, and 90%, and overall accuracy of 88%, 90%, 90%, and 93% for 1979, 1995, 2009, and 2023. Forest, vegetation, and water decreased by 21%, 3%, and 0.3%, while built areas and bare land increased by 22% and 16% over 44 years. The predicted results of Land Cover changes indicate a sharp decrease in Forest cover (- 28.6%, -33.8% and -51.0%), Vegetation (-12.9%, -19.8% and -24.7%) and Water (-28.2%, -12.3% and -14.0%) with an increase in the Built up area (+23.2%, +22.2% and +22.8%) and Bare land (+13.0%, +15.3% and +37.1%) for the period 2035-2023, 2045-2035 and 2065-2045 respectively. Validation of predictive outcomes and the assessment of model accuracy were conducted using the CA-Markov and Kappa index. The CA-Markov model demonstrated a level of agreement ranging from good to perfect for the three testing years: 2009 (CA-Markov = 83.4, Kno = 0.79, klocation = 0.81 and Kstandard=0.81), 2011 (CA-Markov = 82.7, Kno = 0.81, klocation = 0.79 and Kstandard=0.82) and 2023 (CA-Markov = 81.6, Kno = 0.80, klocation = 0.84 and Kstandard=0.83). Over the past 44 years, the study area has seen a significant decrease in forest land, a trend that is expected to continue without intervention. This ongoing deforestation will exacerbate the absorption of runoff and increase flood severity in the district.