Land use and land cover change detection and assess future change using remote sensing and machine learning algorithms: a study of uMgeni river catchment.

Abstract

Water catchment areas are the key regional strategic water source with a range of ecological benefits such as providing habitat for biodiversity, climate change mitigation, sport, and recreation. Detailed temporal information on the spatial attributes of the catchment areas is paramount for water resource managers to support water resource conservation strategies. Data from remote sensing provide a good playground for the advanced statistical tools to study land cover patterns in the water catchments. Thus, our study leveraged remote sensing data and explored options of Machine Learning Algorithms (MLAs) for the classification of thematic Land Use and Land Cover (LULC) classes within the uMngeni River Catchment area. This study used Landsat 8 Operational Land Imager (L8-OLI) data along with Naïve Bayes (NB), Random Forests (RF), Support Vector Machine SVM), and Artificial Neural Networks (ANN). The RF model proved its robustness with high accurate overall accuracy and kappa of 97.02% and 0.96, respectively. Hence, the RF was subsequently utilised to assess LULC change detection for 20 years (from 2003-2023). The interplay among the land use classes was observed, with common transitions to built-up class and a noticeable constant decline of the forest land use class. Change detection maps enabled the modelling of 2032 land cover maps using the hybrid model from Multi-Layer Perceptron Neural Network (MLPNN) and Markov Chain (MLNN-Markov). The combination of RF and L8-OLI is suitable for mapping the LULC in the case of a South African complex landscape. The uMngeni River catchment possesses fair land cover distribution with an alarming built-up land cover class at the cost of indigenous forests, which is expected to increase by 3.7% in 2032. The findings of this work will also demonstrate the practical application of MLAs in mapping LULC and support the water resource managers in formulating management strategies aimed at safeguarding the integrity of the water resources.