Using Google Earth Engine to detect land cover change: Singapore as a use case

Abstract

This paper investigates the web-based remote sensing platform, Google Earth Engine (GEE) and evaluates the platform's utility for performing raster and vector manipulations on Landsat, Moderate Resolution Imaging Spectroradiometer and GlobCover (2009) imagery. We assess its capacity to conduct space–time analysis over two subregions of Singapore, namely, Tuas and the Central Catchment Reserve (CCR), for Urban and Wetlands land classes. In its current state, GEE has proven to be a powerful tool by providing access to a wide variety of imagery in one consolidated system. Furthermore, it possesses the ability to perform spatial aggregations over global-scale data at a high computational speed though; supporting both spatial and temporal analysis is not an obvious task for the platform. We examine the challenges that GEE faces, also common to most parallel-processing, big-data architectures. The ongoing refinement of this system makes it promising for big-data analysts from diverse user groups. As a use case for exploring GEE, we analyze Singapore’s land use and cover. We observe the change in Singapore’s landmass through land reclamation. Also, within the region of the CCR, a large protected area, we find forest cover is not affected by anthropogenic factors, but instead is driven by the monsoon cycles affecting Southeast Asia.

Keywords

• Google Earth Engine
• big-data architecture
• land cover
• urban areas
• time series analysis