Mapping the benthic habitat of coastal reef using hyperspectral imagery

• Grant number: 25/03699-2
• Support Opportunities: Scholarships abroad - Research Internship - Post-doctor
• Start date: October 01, 2025
• End date: March 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Geosciences
• Principal Investigator: Milton Kampel
• Grantee: Thais Andrade Galvao de Medeiros
• Supervisor: Vittorio Ernesto Brando
• Host Institution: Instituto Nacional de Pesquisas Espaciais (INPE). Ministério da Ciência, Tecnologia e Inovação (Brasil). São José dos Campos , SP, Brazil
• Institution abroad: Consiglio Nazionale delle Ricerche (CNR), Italy
• Associated to the scholarship: 23/13544-0 - Temporal dynamics of benthic habitats in the coastal reef of Abrolhos Bank, BA, BP.PD

Abstract

Coral reefs are among the most biodiverse and ecologically important ecosystems, providing coastal protection and supporting millions of people socially and economically. However, they are increasingly threatened by climate change, rising ocean temperatures, ocean acidification, pollution, and overfishing. This highlights the urgent need for improved monitoring tools to assess reef health and guide conservation efforts. Remote sensing has emerged as a powerful tool for mapping and monitoring coral reefs. While multispectral data offer regional-scale assessments, hyperspectral imagery enables the detection of spectral features that help differentiate various benthic types and complex habitats, minimizing confusion caused by optically active water column parameters. Several studies have demonstrated the effectiveness of hyperspectral imaging in reef monitoring, but its high cost has historically limited its large-scale application. However, recent advancements in satellite hyperspectral sensors, such as PRISMA and EnMAP, offer new opportunities for reef monitoring. This project aims to evaluate the capability of hyperspectral data to classify and identify benthic habitats in the Sebastião Gomes Reef (ABR, Brazil). The study will apply semi-analytical and machine learning (ML) approaches to analyze the imagery and assess its ability to detect variations in benthic composition. By leveraging the spectral richness of these sensors, this research seeks to improve habitat classification accuracy and contribute to more effective reef conservation strategies. The study area of this project is Sebastião Gomes Reef, located within the Abrolhos Bank (16°40'S-19°40'S, 37°20'W-39°10'W), an extensive reef system on the eastern Brazilian shelf. It is influenced by nearby estuaries and rivers, resulting in optically complex waters that pose challenges for remote sensing. Two in situ datasets are required: bio-optical measurements from the water column and benthic substrate data. The first dataset was collected between March and April 2024, covering 18 sites for bio-optical parameters and 46 sites for benthic cover. A second field campaign is planned for early 2025. PRISMA and EnMAP imagery will be downloaded from their respective catalogs. PRISMA, operated by the Italian Space Agency, provides 30 m spatial resolution across 234 spectral bands (400-2500 nm). EnMAP, developed by Germany, offers similar resolution and spectral coverage. A PRISMA image of the study area was acquired on May 3, 2023, and additional PRISMA and EnMAP images have been requested to coincide with the 2025 field campaign; however, their utilization will depend on data availability and quality. Atmospheric correction will be performed using the ACOLITE package, which employs the Dark Spectrum Fitting (DSF) algorithm to account for surface and atmospheric effects, ensuring accurate retrieval of water column and benthic reflectance data. A combination of semi-analytical and ML approaches will be used for habitat classification. The SAMBUCA model, which estimates bottom reflectance and water column properties, will be parameterized using in situ bio-optical data. The SAMBUCA results, along with benthic cover data, will serve as inputs for a Random Forest classifier, known for its effectiveness in benthic habitat mapping. The classification accuracy will be evaluated using standard metrics, including overall, producer, and user accuracy. The developed methodology will contribute to future reef conservation strategies and can be adapted for other marine environments facing similar challenges.