The advent of the new political dispensation in South Africa has seen an exponential growth in the rate of land transformation and encroachment by other land uses into valuable agro-ecological zones. Due to the socio-economic value of the often limited high-potential agricultural land in the country, a reliable determination of encroachment and transformation is necessary for effective monitoring and management of such agro-ecological resources. Using the robust support vector machine classification algorithm, this study adopted multi-temporal, remotely sensed datasets to assess the extent to which the physical development footprint in the uMngeni Local Municipality affected the existing agro-ecological zones from 1993 to 2003 and from 2003 to 2013. The results show a steady increase in built-up areas during the period under investigation. The study demonstrates the value of multi-temporal, remotely sensed datasets and techniques in mapping the vulnerability of existing agricultural land to urbanisation in the study area.