Use of multi-temporal optical, radar, and fused optical-radar images for mapping land cover changes in the north-central Panama

Keywords: remote sensing, radar, optical, land cover change, Panama
Abstract Type: Paper Abstract

Abstract

The Colon Province on the Caribbean side of central Panama has experienced remarkable land cover changes in recent decades, in part due to the expansion of a large open-pit copper mine, Cobre Panamá. There is significant concern among infectious disease experts in the country that these land cover changes are modifying the risk of mosquito-borne diseases such as dengue and chikungunya. To better understand the relationships between land cover and disease risk, these experts need multi-temporal land cover maps that can be integrated in spatially explicit disease risk models. Our main objectives were to generate these maps and assess spatio-temporal changes in landscape composition and configuration in the Cobre Panamá mine region. To meet these objectives, we 1) applied three classifiers (convolutional neural networks, random forest, and support vector machine) to cloud-free Sentinel 2 image composites (optical data), Sentinel 1 image composites (radar data), and fused Sentinel 1 and 2 image composites (optical-radar) from the year 2021; 2) used the most accurate classifier / image combination to map land cover for two additional years (2015 and 2018); 3) created change maps and statistics for the 2015-2018, 2018-2021, and 2015-2021 time periods using post-classification comparison and intensity analysis; and 4) quantified changes in landscape configuration during the same time periods using FRAGSTATS and MSPA. In this paper presentation, we will describe in more detail the methodological approach as well as the results.