Résumé
BackgroundCoral reefs are facing increasing environmental pressures along with anthropogenic stressors and disturbances, leading to a worldwide degradation of their status. There is an urgent needfor developing adequate tools to assess reef state and resilience at large scale. This can greatly help to propose adequate management strategies for coral reef degradation mitigationand promoting reef resilience. MethodsUsing a spatial approach based on high resolution multispectral and hyperspectral image analyses, the present study aims at drawing, for the first time, a baseline of current coral reefstructure and status in a highly heterogeneous shallow reef flat in Reunion Island (South-West Indian Ocean).We first developed an automatic image processing methodology which could be repeated over time to extract accurate spatial quantitative metrics on benthic coral reefs. Geometricand atmospheric corrections were performed prior to image processing. Benthic cover estimates were derived from airborne hyperspectral image processing using water columncorrection and unmixing models implemented with the four main coral reef benthic components: corals, algae, seagrasses and sand. Coral reef health status was evaluatedthrough both coral cover estimates and a new index called Hyperspectral Coral to Algae Index (HCAI), defined as the ratio of living coral cover to the sum of living coral and algalcovers. Ground truth data acquired simultaneously were used to validate processing accuracy. A significant positive correlation was obtained between coral cover recorded in situ and estimated from image analysis.We then illustrated, through a diachronic analysis between 2009 and 2015, the value of hyperspectral and multispectral imagery in quantifying spatial-temporal changes of benthiccomponents for metrics extracted using the developed method.ResultsOur results reveal an overall decrease of the HCAI index and percent cover of all the dominant coral classes along the reef (e.g., −28.24% for the coral class), while habitat classesdominated by algae strongly increased during the period. The comparison of HCAI with the “live coral cover” extracted from Global Coral Reef Monitoring Network (GCRMN) stations showed similar trends although these two approaches differ fundamentally in many aspects regarding spatial and ecological scales (resolution, parameters measured, precision of the information obtained). Local and high-frequency measurements, as conducted by GCRMN, can be supplemented with broader spatial measurements of habitat dynamics and fragmentation through low-frequency spatialmonitoring. This Spatial approach is also more suitable for describing natural patterns emerging from a range of gradients.ConclusionRemote sensing analyses based on hyperspectral spatial approach, as developed in this particular study, opens new perspectives for coral reef monitoring, which have profoundimplications for coral reef management and conservation in the light of rapid environmental changes.