Coastline shift analysis in data deficient regions: Exploiting the high spatio-temporal resolution Sentinel-2 products

Abstract

In most developing countries, coastline shift monitoring using in-situ (ground-based) data faces challenges due, e.g., to data unreliability, inconsistency, deficiency, inaccessibility or incompleteness. Even where practically applicable, the traditional “boots on the ground” methods are labour intensive and expensive, thus imposing burden on poor countries struggling to meet other urgent pressing daily needs, i.e., food and medicine. Remote sensing (RS) techniques provide a more efficient and effective way of collecting data for coastline shift analysis. However, moderate spatio-temporal resolution RS products such as the widely used Landsat products (30 m and 16 days) may be insufficient where high accuracy is desired. In 2015, Sentinel-2 Multi-Spectral Instrument (MSI) remotely sensed products with higher spatio-temporal resolution (10 m and 5 days) and high spectral resolution (13 bands), which promises to improve coastline movement monitoring to high accuracy, was launched. Using two war-impacted countries (Liberia and Somalia) as case studies of regions with data deficiency or of poor quality, for the period 2015–2018, this contribution aims at (i) assessing the suitability of the new freely available high spatio-temporal Sentinel-2 products to monitor coastline shift, (ii) assessing the possibility of filling the missing Sentinel-2 gaps with Landsat 8 panchromatic band (15 m) products to provide alternative data source for mapping of coastline movements where Sentinel-2 data is unusable, e.g., due to cloud cover, and (iii), undertake a comparative analysis between Sentinel-2 (10 m), Landsat panchromatic (15 m), and Landsat multi-spectral (30 m). The results of the evaluation indicate 23% (on average) improvement gained by using Sentinel-2 compared to the traditional Landsat 30 m resolution data (i.e., 32% for Liberia and 14% for Somalia). A comparison of 100 check points from Google Earth Pro (i.e., surrogate in-situ reference data) show 91% agreement for Liberia and 85% for Somalia, indicating the potential of using Sentinel-2 data for future coastal shift studies, particularly for the data deficient regions. The results of comparative studies for Sentinel-2, Landsat panchromatic (PAN), and Landsat multi-spectral (MS) show that the percentages of Sentinel-2 and Landsat PAN that falls within 10 m threshold is much higher than Landsat MS by 35% and 26%, respectively, and for the 2016–2017 period, they provide more detailed mapping of the Liberian coastline compared to Landsat MS (30 m). Finally, panchromatic Landsat data with 15 m resolution are found to be capable of filling the missing Sentinel-2 gaps, i.e., where cloud cover hampers its usability.