The CAO Coral Reef Challenge
Gregory P. Asner, September 2017
Coral reefs are global hotspots of biological diversity and support the livelihoods of more than a billion people worldwide. Coral reefs cover roughly 500,000 km2 of the Earth’s surface, but are sparsely distributed over more than 200 million km2 of ocean (Figure 1). Field studies currently represent less than 0.01% of coral reefs worldwide, and although local monitoring is important, it provides little understanding of the trajectory of coral reefs undergoing regional and global environmental change.
Recent reports of rapidly declining coral cover and health are undoubtedly true, but spatial variability of coral resistance and resilience to climate and other changes is extremely high, even at the species level within a single reef ecosystem. Variation within and between reef habitats renders field-based observations prone to local bias caused by, for example, limited access for diving operations. This, in turn, greatly impedes current efforts to understand ongoing changes to coral reefs, and to seek, study, and assist corals that may be capable of surviving under novel environmental conditions including higher seawater temperature, ocean acidification, and local-to-regional pollution (wastewater effluent, sunscreens, etc.).
Satellite-based multispectral instruments have been effectively used to discern the presence and absence of shallow (< 10 meter depth) coral reef (e.g., Andréfouët et al. 2003). New satellite constellations, such as from Planet.com, provide near-daily multispectral observations of reef presence-absence at 3-m spatial resolution (Asner et al. 2017). However, no satellites provide assessments of reef condition at the resolution of individual corals deemed critical for understanding their resistance, resilience and response to multi-scale environmental changes ranging from coastal pollution to climatic change.
Two technological advances have opened new possibilities for mapping and monitoring coral reefs at high spatial and temporal resolution. First, high-fidelity imaging spectroscopy (HiFIS) has emerged and improved since 2011, providing the first highly detailed spectroscopic measurements of corals from manned aircraft. Although seawater limits the total wavelength range of spectroscopic signatures of benthic organisms to 400-700 nanometers, the few HiFIS in existence can provide these measurements at very fine spectral resolutions of 2-5 nanometers. This is critically important because only through such detailed and highly precise spectroscopic signatures can coral health and composition be determined (Figure 2).
A second revolution in analytical methods has facilitated automated processing of the massive amounts of spectral data collected by HiFIS sensors. The approaches now being used include deep learning neural network and support vector machine models for determining species from spectral signatures, and Bayesian linear mixture modeling to separate benthic groups such as live coral, dead coral, and algal turf (Figure 3; Thompson et al. 2017).
The unique combination of airborne HiFIS and big data computation provides a new and powerful toolkit with which to better understand the impact of regional and global environmental changes to coral reef ecosystems and their inhabitants. However, the airborne platform used to collect the data also plays a major role in determining the efficacy of the mapping and monitoring effort. Civil-sector drones (unmanned aerial systems) currently have neither the payload capacity for HiFIS (80+ kg) nor the mapping range (> 100,000 hectares per day desired) needed to transform our understanding of coral resistance, resilience and response to regional or global scale environmental change.
An ideal platform is one that can carry a HiFIS payload while maintaining a high spatial resolution of 0.5 meters or finer, and also long duration aloft for large-area mapping (> 100,000 hectares per day). Following more than 12 years of development, use, and improvement, the Carnegie Airborne Observatory-3 (CAO-3) has emerged as the best existing platform for high-resolution, spectroscopic mapping of coral reef ecosystems (Figure 4). CAO-3 carries the Airborne Taxonomic Mapping System (AToMS), which includes two high-fidelity imaging spectrometers. Both spectrometers measure benthic reflectance in the required 400-700 nanometer range at 5-nanometer bandwidth. At 1000 m (3300’) above sea level, the two spectrometers provide 0.5 and 1.0 m spatial resolution. In addition, CAO-3 carries a digital airborne camera capable of collecting 0.10 m resolution data from 1000 m altitude. CAO-3 also carries a high-resolution, dual-laser scanner (LiDAR) with an embedded high-precision Inertial Motion Unit (IMU) taken from defense sector technology. This system provides extremely accurate geospatial positioning information for the HiFIS and digital camera data, typically with uncertainties of less than 10 centimeters in three dimensions.
CAO-3 is a complete airborne laboratory based on a highly modified Dornier 228-202 aircraft (Figure 4), with global deployable range, and proven record of operating on three or more continents within a given year. For example, in 2016, CAO-3 deployed across the Pacific Ocean, mapping Bornean land and coral reef systems, as well as Ecuadorian rainforests, Hawaiian coral reefs, forests and volcanoes, and California’s Sierra Nevada mountains.
At 1000 m a.s.l., CAO-3 can collect up to 100,000 hectares (247,000 acres) of data at 0.5 m resolution per flight. In other words, each flight can yield up to 4 billion spectroscopic measurements of coral reef inhabitants. In comparison, a day of intensive benthic habitat mapping yields fewer than a hundred 0.5 m x 0.5 m plot-based measurements for a highly-experienced 4-person scientific dive team.
Advancing CAO Applications in Coral Reef Science and Conservation
Recent advances in ocean spectral data analytics using CAO-3 may break a long-term barrier in benthic habitat and coral species mapping. In particular, my collaborators and I have developed a comprehensive processing chain, from measured spectroscopic radiance signatures at the aircraft, to benthic reflectance and classification following atmospheric, ocean surface glint, and seawater column compensation (Figure 3; Thompson et al. 2017). The resulting maps provide spatially-detailed information on coral presence, cover, health and aspects of biological composition. They may also provide spatially-explicit information on coral bleaching, and resistance to bleaching, which may advance our understanding of how different coral species respond to regional and global environmental changes.
Despite our advances in spectral data-analytics, much more work is needed to associate and integrate the airborne benthic reflectance signatures with biological and ecological information drawn from field work. I have therefore established four coral reef mapping and analysis projects designed to link CAO HiFIS data to coral properties, and to more deeply integrate and understand the ways that spectroscopy can be used to infer key properties and processes in coral reef ecosystems. These range from phylogenetic controls on bleaching response, to the role of biogeochemical and hydrological dynamics determining coral health and composition.
Hawaii Extensive Reef Research Project
My team and I have developed a long-term ecological research project to explore the spatial ecology of coral reefs covering more than 200 km of coastline along the leeward side of Hawaii Island (aka Big Island). The Hawaii Island coastline is a valuable coral reef research asset for several reasons. First, the overall extent of Hawaii Island coral reefs is relatively well known, as are the environmental stressors that range from light to intense depending upon reef location. Stressors include wastewater sewage from hotel, golf, agriculture and suburban development, as well as intensive reef fishing and aquarium industry activities. These areas are readily juxtaposed against zones of relative protection from these stressors. The matrix of conditions allows for comparative analyses of coral reef ecological processes and health against a range of globally-relevant environmental stressors. Second, the pool of hard coral species is well known, and includes species in common Indo-Pacific genera such as Porites, Pocillopora, and Pavona. Of these, there are approximately six hyperdominant shallow-water species, which facilitates the development of new mapping algorithms and spatial ecological approaches due to the relative simplicity of the ecosystem. Third, project partners at NOAA in Honolulu and The Nature Conservancy of Hawaii are contributing regional-scale coral survey data and expertise to the project.
On a high-frequency basis since 2016, my team and I have mapped coral reefs along the leeward coast of Hawaii Island following the 2015 mass-bleaching event. The event was caused by sustained high water temperatures associated with El Nino as well as the east Pacific sea surface temperature anomaly (known as “The Blob”). This provided an opportunity to test, improve and apply the new CAO algorithms for coral reef monitoring at high spatial resolution. An example result is shown in Figure 5, highlighting the extremely high occurrence of algae-covered reef in red. Green indicates live coral cover, and cyan identifies areas of natural sand deposits mixed with live coral cover.
Figure 5. CAO-3 mapping of coral reefs along the leeward coast of Hawaii Island, 2016. Data indicate areas of high algal-covered (dead) coral in red, live coral cover in green, and natural sand patches in blue.
Hawaii Intensive Reef Research Project
We have developed two focal reefs study sites, each of more than 500 hectares, with ultra-high spatial resolution studies of the benthic habitat and the inhabitant coral and fish species. The goal of these sites is to improve our ability to map and monitor individual coral colonies, while developing methods to assess the health of coral polyps. Through this approach, we will learn much more about the resistance, resilience and recovery of corals undergoing environmental change (Figure 6). Planned long-term observations at the intensive sites include measurements of environmental conditions (temperature, salinity, light) as well as biological composition and coral condition (i.e. growth, fluorescence, algal cover).
One of our core research sites located at Papa Bay is ideal for benthic research in the 4-20 m depth range. We have extensive local knowledge of the site, and are developing a base for our long-term research studies. The second core research site located at Honaunau Bay (Figure 6) is near to the Papa Bay site. Honaunau bay is comprised of coral reef with similar composition to the Papa Bay site, but at shallower water depths of 1-15 m. This depth range is important in the context of hot water coral bleaching events and other coastal impacts that are more common in shallower waters.
Combined, our extensive and intensive research reefscapes along the Hawaii Island coast allow us to address critically important questions in coral reef ecology and remote sensing. Example questions include:
- What are the spatial-ecological patterns of biological diversity among corals, and how do natural environmental controls affect these patterns?
- How does the spatial configuration of corals, and the abiotic setting, mediate coral resistance and resilience to hot water/bleaching events?
- What are the temporal scales and spatial patterns of coral colony establishment, mortality, and bleaching, and how do these relate to long-term and episodic changes in the abiotic environment?
- Can CAO airborne imaging spectroscopy provide an efficient means for mapping and monitoring coral reef ecosystems at the organismic scale?
Coral Assisted-Evolution Project
A powerful use of the CAO capability rests in combining our mapping with genetic methods to assist coral evolution for increasingly warmer and more acidic ocean environments. An ideal reefscape for such work is the Kaneohe Bay on Oahu in the Hawaiian Islands. The Hawaii Institute for Marine Biology (HIMB), directed by Dr. Ruth Gates, is located on Coconut Island in the heart of Kaneohe Bay. With funding from the Paul G. Allen Foundation, Dr. Gates and I are focusing collaborative effort on high temporal and spatial resolution mapping and monitoring of Kaneohe reefs to develop a combined remote sensing and genetics-based approach to identify and propagate corals for warmer waters (van Oppen et al. 2015).
In phase-one of the project, we are determining whether there are features in the CAO maps that distinguish stress-resistant corals from stress-sensitive corals. A stress resistant coral is one that remains a healthy color or brown when exposed to conditions that cause stress sensitive corals to turn pale and bleach. The color of corals reflects the concentrations of symbiotic algae (called zooxanthellae) in coral tissues with brown indicating high concentrations of zooxanthellae, and blue to white colors indicating low concentrations of the symbiont. These differences in the densities of zooxanthellae in coral tissue, in turn, reflect in different concentrations of the pigments in the corals, such as chlorophyll, and these pigments can potentially be detected and captured in a spatially explicit way using the CAO. Our specific goals include:
- Test how accurately the CAO identifies stress resistance in corals by localizing stress resistant corals in the maps, using the map to guide the collection of these corals and verifying stress resistance using stress tests in the laboratory.
- Determine how stable features that distinguish stress resistance in corals are in the CAO maps are over time and through seasonal variations in seawater temperature. This analysis will pinpoint the optimal time window for CAO mapping to be done in other locations.
If successful, this project will enable stress resistant corals to be identified and mapped in coral reef landscape from the air. This will transform our ability to: 1) scale up climate optimization of interventions and restoration activities on reefs, 2) track and measure the success of interventions using stress resistant corals and 3) identify the distribution of climate resistance and climate refugia in shallow tropical marine landscapes.
Lighthouse Reef, Belize Project
Another addition to the CAO coral mapping and analysis effort centers on the largest reef atoll in the Caribbean. Known as Lighthouse Reef (LHR), the atoll was selected for its high biological diversity, field accessibility, robust collaborative opportunities, and its separation from mainland Belize thereby minimizing coastline-related environmental stressors (Figure 8).
We are working closely with colleagues from Wake Forest University and the Belize Audubon Society on the LHR environment to further test and apply our coral mapping capability, while also comparing it to what can be achieved using the latest high-spatial resolution satellite imagery. Our previous work using Planet satellite data in the South China Sea established our ability to map two basic benthic types common to all atolls – shallow coral reef and sand (Asner et al. 2017). Studies with Planet imagery, and indeed with many other types of spaceborne multispectral imagery, indicate that only 2-5 benthic types can typically be mapped, and none can be done at the spatial resolution approaching individual coral species or small communities of species.
Our latest work at LHR shows that CAO HiFIS imagery distinguishes no less than 15 benthic community types, compared to a maximum of four types that can be reliably derived from Planet satellite data (Figure 9). Additionally, we have successfully mapped the location and extent of live corals throughout the mosaic of patch reefs found across the atoll (Figure 10). Based on these and other results, future plans for global coral reef monitoring will likely fuse the Planet data for flight planning, followed by tactical CAO mapping of coral composition and health.
Mapping the composition of LHR in Belize has opened numerous pathways of biological and ecological research. For example, benthic mapping of patch coral reefs on the interior of LHR Atoll has facilitated questions about fish herbivory controls on seagrass communities surrounding patch reefs. These questions link directly to issues of over-fishing and the presence or absence of grazing fish species. Another example focuses on deeper reef ecosystems, and the geomorphology of the benthic habitat and its effects on biological diversity.
An additional set of findings common to both Hawaii and Belize centers the depth to which we are able to detect and map coral reefs. In Hawaii, our maximum depth of sensitivity is 18 meters. In Belize, it is 33 meters. Critically, the vast majority of coral bleaching occurs in near-surface waters that have undergone the large temperature anomalies. These typically occur at water depths of 1-10 meters.
Spectroscopy of Global Reef Ecosystems
The efficacy of airborne mapping and analysis of coral reefs is highly dependent upon our knowledge of the spectral-optical properties of coral species worldwide. Global databases are non-existent, so I am undertaking a global survey of coral spectral properties at 50 of the most phylogenetically diverse (and different) reefs on the planet. Analogous to our effort to develop the Spectranomics library of tree species worldwide (Asner and Martin 2009, Asner and Martin 2016), the coral reef spectral survey will result in a library-type database for use in study and mapping of corals using the CAO and future airborne and satellite imaging spectrometers. The 50 reefs have been preliminarily selected, and will be visited throughout 2017-2018. This project is being undertaken in collaboration with Stuart Phinn and Chris Roelfsema from the University of Queensland.
Andréfouët, S. et al. 2003. Multi-site evaluation of IKONOS data for classification of tropical coral reef environments. Remote Sensing of Environment 88, 128-143.
Asner, G. P. et al. 2012. Carnegie Airborne Observatory-2: Increasing science data dimensionality via high-fidelity multi-sensor fusion. Remote Sensing of Environment 124, 454-465.
Asner, G.P. et al. 2017. Coral reef atoll assessment in the South China Sea using Planet Dove satellites. Remote Sensing in Ecology and Conservation, 1-9, doi:10.1002/res2.42.
Asner, G.P. and R.E. Martin. 2009. Airborne spectranomics: Mapping canopy chemical and taxonomic diversity in tropical forests. Frontiers in Ecology and the Environment 7:269-276.
Asner, G.P. and R.E. Martin. 2016. Spectranomics: emerging science and conservation opportunities at the interface of biodiversity and remote sensing. Global Ecology and Conservation 8:212-219.
Thompson, D.R., E.J. Hochberg, G.P. Asner, R.O. Green, D. Knapp, B.C. Gao, R. Garcia, M. Gierach, Z. Lee, S. Maritorena, and R. Fick. 2017. Airborne mapping of benthic reflectance spectra with Bayesian linear mixtures. Remote Sensing of Environment 200:18-30.
Van Oppen, M.J.H., J.K. Oliver, H.M. Putnam, and R. Gates. 2015. Building coral reef resilience through assisted evolution. Proceedings of the National Academy of Sciences 112:2307-2313.