Landcover classification and change detection using remote sensing and machine learning: a case study of Western Fiji

As a developing country, Fiji is facing rapid urbanisation, as evidenced by massive development projects that include civil works such as housing and roads. In this study, we present a machine learning-based framework that utilises remote sensing data to analyse land use and land cover changes from 2013 to 2024 in Nadi, Fiji. We used Landsat 8 satellite imagery for the study region and created a training dataset with labels for supervised machine learning. We use Google Earth Engine and unsupervised machine learning via K-means clustering to generate the land cover map. We utilise a framework that uses convolutional neural networks (CNNs) and compares with conventional machine learning models to classify the land cover types of the selected regions. We present a visualisation of change detection, highlighting urban area changes over time to monitor map changes. Our results indicate that the CNN model performs similarly to other machine learning models (0.96 F1-score) in terms of classification performance, but better captures the development of urban areas as verified by qualitative analysis. Our study ascertains that Nadi has experienced a rapid urbanisation process, and the expansion extended outward, taking over the sugar farms.