Deep learning Based Derivative for Shoreline Change Detection in Cape Town, South Africa

Abstract

Coastal environments are influenced by bio-chemical and physical processes that lead to the ecosystem services that support and influence the fishery, mining, real estate, and tourism industries. Coastal developments often adversely affect coastlines by disrupting water movement, degrading land quality, hindering biodiversity by reducing habitats and introducing pollutants, more so greenhouse gases that lead to increased ocean temperatures and thus sea level rise. Coastal areas are also affected by other factors such as underlying geology, near shore hydrodynamics and localized climate; sea level rise has still not been definitively modelled and as such is considered a hazard. The City of Cape Town has significant socio-economic dependence on its 240km coastline and is therefore invested in coastal processes and its effects on livelihood. This study has applied remote sensing to Sentinel-2 satellite imagery to determine and compare the shoreline change along four of Cape Town’s coastlines namely Nordhoek-Kommetjie, Milnerton, Strandfontein-Monwabisi and Strand using the Deep Learning derived CoastSat Python toolkit for Modified Normalized Difference Water Index (MNDWI) based shoreline detection at the sub-pixel level. The results indicate that Strandfontein-Monwabisi experienced the highest shoreline change of approximately 284m. The Nordhoek-Kommetjie beach had 110m of change whilst the Milnerton and Strand beaches had almost similar changes of 88m and 91m respectively. Research of this nature is useful in guiding the distribution of coastal management resources and consistent data collection and analysis can lead to the development of interactive coastal monitoring tools.