Recent advances in machine learning and computer vision have enabled increased automation in benthic habitat mapping through airborne and satellite remote sensing. Here, we applied deep learning and neural network architectures in NASA NeMO-Net, a novel neural multimodal observation and training network for global habitat mapping of shallow benthic tropical marine systems. These ecosystems, particularly coral reefs, are undergoing rapid changes as a result of increasing ocean temperatures, acidification, pollution, amongst other stressors. Remote sensing from air and space has been the primary method in which changes are assessed within these important, often remote, ecosystems at a global scale. However, such global datasets often suffer from large spectral variances due to the time of observation, atmospheric effects, water column properties, and heterogeneous instruments and calibrations. To address these challenges, we developed an object-based fully convolutional network (FCN) to improve upon the spatial-spectral classification problem inherent in multimodal datasets. We showed that with training upon augmented data in conjunction with classical methods, such as K-Nearest neighbors (KNN), we were able to achieve better overall classification and segmentation results. This suggests FCNs are able to effectively identify the relative applicable spectral and spatial spaces within an image, whereas pixel-based classical methods excel at classification within those identified spaces. Our spectrally-invariant results, based on minimally pre-processed WorldView-2 and Planet satellite imagery, show a total accuracy of approximately 85% and 80%, respectively, over 9 classes when trained and tested upon a chain of Fijian islands imaged under highly variable day-to-day spectral inputs.

The coloration of tropical reef corals is mainly due to their association with photosynthetic dinoflagellates commonly known as zooxanthellae. Combining High Performance Liquid Chromatography (HPLC), spectroscopy and derivative analysis we provide a novel approach to discriminate between the Caribbean shallow-water corals Acropora cervicornis and Porites porites based on their associated pigments. To the best of our knowledge, this is the first time that the total array of pigments found within the coral holobiont is reported. A total of 20 different pigments were identified including chlorophylls, carotenes and xanthophylls. Of these, eleven pigments were common to both species, eight were present only in A. cervicornis, and three were present only in P. porites. Given that these corals are living in similar physical conditions, we hypothesize that this pigment composition difference is likely a consequence of harboring different zooxanthellae clades with a possible influence of endolithic green or brown algae. We tested the effect of this difference in pigments on the reflectance spectra of both species. An important outcome was the correlation of total pigment concentration with coral reflectance spectra up to a 97% confidence level. Derivative analysis of the reflectance curves showed particular differences between species at wavelengths where several chlorophylls, carotenes and xanthophylls absorb. Within species variability of spectral features was not significant while interspecies variability was highly significant. We recognize that the detection of such differences with actual airborne or satellite remote sensors is extremely difficult. Nonetheless, based on our results, the combination of these techniques (HPLC, spectroscopy and derivative analysis) can be used as a robust approach for the development of a site specific spectral library for the identification of shallow-water coral species. Studies (Torres-Pérez, NASA Postdoctoral Program) are currently underway to further apply this approach to other Caribbean benthic coral reef features. The data will be used with planned and future airborne and satellite studies of the site and for algorithm development to advance the use of future airborne and satellite instrument capabilities (NASA PRISM and HyspIRI) for discrimination of coral reef benthic composition.

Reef corals typically contain a number of pigments, mostly due to their symbiotic relationship with photosynthetic dinoflagellates. These pigments usually vary in presence and concentration and influence the spectral characteristics of corals. We studied the variations in pigment composition among seven Caribbean shallow-water Scleractinian corals by means of High Performance Liquid Chromatography (HPLC) analysis to further resolve the discrimination of corals. We found a total of 27 different pigments among the coral species, including some alteration products of the main pigments. Additionally, pigments typically found in endolithic algae were also identified. A Principal Components Analysis and a Hierarchical Cluster Analysis showed the separation of coral species based on pigment composition. All the corals were collected under the same physical environmental conditions. This suggests that pigment in the coral’s symbionts might be more genetically-determined than influenced by prevailing physical conditions of the reef. We further investigated the use of remote sensing reflectance (Rrs) as a tool for estimating the total pigment concentration of reef corals. Depending on the coral species, the Rrs and the total symbiont pigment concentration per coral tissue area correlation showed 79.5–98.5% confidence levels demonstrating its use as a non-invasive robust technique to estimate pigment concentration in studies of coral reef biodiversity and health.

Land-based sediment stress represents a threat to many coral reefs in Puerto Rico primarily as a result of unrestricted land cover/land use changes and poor best management practices. The effects of such stresses have been documented along most coasts around the island. However, little attention has been paid to reefs located on the north coast, and very little is known about their composition and current state. Here, we present a study characterizing riverine inputs, water quality conditions, and benthic composition of two previously undescribed coral reefs (Tómbolo and Machuca reefs) located just eastward of the Río Grande de Manatí outlet in north-central Puerto Rico. This study utilizes a time series of remotely sensed ocean color products [diffuse vertical attenuation coefficient at 490 nm (Kd₄₉₀) and chlorophyll-a concentration (Chl-a) estimated with data from the Visible Infrared Imaging Radiometer Suite (VIIRS)] to characterize water quality in this coastal region. In general, the months with relatively high mean daily river streamflow also coincide with months having the highest proportion of eastward wave direction, which can promote the eastward influence of river waters toward the two coral reefs sites. Kd₄₉₀ and Chl-a showed a higher riverine influence closer to the watershed outlet. Kd₄₉₀ and Chl-a monthly peaks also coincide with river streamflow highs, particularly at those pixels closer to shore. Tómbolo Reef, located farther eastward of the river outlet, shows a well-developed primary reef framework mainly composed of threatened reef-building species (Acropora palmata, Pseudodiploria) and high coral cover (19–51%). The benthos of Machuca Reef, located closer to the river outlet, is dominated by macroalgae with a significantly lower coral cover (0.2–2.7%) mainly composed of “weedy” coral species (Porites astreoides and Siderastrea radians). Cover of major benthic components correlates with distance from the river outlet, and with gradients in Kd₄₉₀ and Chl-a, with higher coral cover and lower macroalgal cover farther from the river outlet. Coral cover at Tómbolo Reef is higher than what has been reported for similar sites around Puerto Rico and other Caribbean islands showing its ecological importance, and as up until now, an unrecognized potential refuge of reef-building threatened coral species.

Guánica Bay and the La Parguera Nature Reserve (LPNR) are ecologically and economically vital to Puerto Rico. Over time, however, these sites have faced a reduction in live coral cover presumably in part due to the combined effects of climate and impoverished water quality related to local land-based sources of pollution. The main objective of this multidisciplinary study was to analyze benthic composition data collected between 2014 and 2018 of six back-reef zones within the context of a ten-year (2012-2021) analysis of remotely sensed coastal water quality (chlorophyll-a [Chl-a]) and diffuse vertical attenuation coefficient at 490 nm [K_d490]) influenced by runoff of nearby rivers. Back-reef zones were mostly dominated by macroalgae (turf and filamentous) and octocorals (sea whips and sea fans). Most sites showed less than 10% calcifying Cnidarians (Scleractinian/Milleporids) cover. Coral cover at most sites was limited to weedy sediment-resistant species. Chl-a concentration and K_d490 were above proposed water quality thresholds for Puerto Rican coral reefs at most virtual stations and have apparently remained stable over the study period of 2012-2021 suggesting that the studied reefs have been exposed to these water quality conditions for more than a decade. This is a concerning factor as benthic composition is indicative of a phase-shift occurrence (or in process) at most back-reef areas from a Scleractinian/Milleporid dominance to macroalgae and/or octocorals. The incorporation of a larger temporal scale using additional remote sensing datasets may provide further insights into the seasonality and trends of coastal water quality linked to both natural and anthropogenic causes in southwest Puerto Rico.