A general deep learning model for bird detection in high-resolution airborne imagery.

Advances in artificial intelligence for computer vision hold great promise for increasing the scales at which ecological systems can be studied. The distribution and behavior of individuals is central to ecology, and computer vision using deep neural networks can learn to detect individual objects in imagery. However, developing supervised models for ecological monitoring is challenging because it requires large amounts of human-labeled training data, requires advanced technical expertise and computational infrastructure, and is prone to overfitting. This limits application across space and time. One solution is developing generalized models that can be applied across species and ecosystems. Using over 250,000 annotations from 13 projects from around the world, we develop a general bird detection model that achieves over 65% recall and 50% precision on novel aerial data without any local training despite differences in species, habitat, and imaging methodology. Fine-tuning this model with only 1000 local annotations increases these values to an average of 84% recall and 69% precision by building on the general features learned from other data sources. Retraining from the general model improves local predictions even when moderately large annotation sets are available and makes model training faster and more stable. Our results demonstrate that general models for detecting broad classes of organisms using airborne imagery are achievable. These models can reduce the effort, expertise, and computational resources necessary for automating the detection of individual organisms across large scales, helping to transform the scale of data collection in ecology and the questions that can be addressed.

Rethinking atoll futures: local resilience to global challenges.

Atoll islands are often perceived as inevitably lost due to rising sea levels. However, unlike other islands, atoll islands are dynamic landforms that have evolved, at least historically, to vertically accrete at a pace commensurate with changing sea levels. Rather than atoll islands' low elevation per se, the impairment of natural accretion processes is jeopardising their persistence. While global marine impacts are deteriorating coral reefs, local impacts also significantly affect accretion, together potentially tipping the scales toward atoll island erosion. Maintaining atoll island accretion requires intact sediment generation on coral reefs, unobstructed sediment transport from reef to island, and available vegetated deposition sites on the island. Ensuring the persistence of atoll islands must include global greenhouse gas emission reduction and local restoration of accretion processes.

Transforming Palmyra Atoll to native-tree dominance will increase net carbon storage and reduce dissolved organic carbon reef runoff.

Native forests on tropical islands have been displaced by non-native species, leading to calls for their transformation. Simultaneously, there is increasing recognition that tropical forests can help sequester carbon that would otherwise enter the atmosphere. However, it is unclear if native forests sequester more or less carbon than human-altered landscapes. At Palmyra Atoll, efforts are underway to transform the rainforest composition from coconut palm (Cocos nucifera) dominated to native mixed-species. To better understand how this landscape-level change will alter the atoll's carbon dynamics, we used field sampling, remote sensing, and parameter estimates from the literature to model the total carbon accumulation potential of Palmyra's forest before and after transformation. The model predicted that replacing the C. nucifera plantation with native species would reduce aboveground biomass from 692.6 to 433.3 Mg C. However, expansion of the native Pisonia grandis and Heliotropium foertherianum forest community projected an increase in soil carbon to at least 13,590.8 Mg C, thereby increasing the atoll's overall terrestrial carbon storage potential by 11.6%. Nearshore sites adjacent to C. nucifera canopy had a higher dissolved organic carbon (DOC) concentration (110.0 µMC) than sites adjacent to native forest (81.5 µMC), suggesting that, in conjunction with an increase in terrestrial carbon storage, replacing C. nucifera with native forest will reduce the DOC exported from the forest into in nearshore marine habitats. Lower DOC levels have potential benefits for corals and coral dependent communities. For tropical islands like Palmyra, reverting from C. nucifera dominance to native tree dominance could buffer projected climate change impacts by increasing carbon storage and reducing coral disease.

Globally important islands where eradicating invasive mammals will benefit highly threatened vertebrates.

Invasive alien species are a major threat to native insular species. Eradicating invasive mammals from islands is a feasible and proven approach to prevent biodiversity loss. We developed a conceptual framework to identify globally important islands for invasive mammal eradications to prevent imminent extinctions of highly threatened species using biogeographic and technical factors, plus a novel approach to consider socio-political feasibility. We applied this framework using a comprehensive dataset describing the distribution of 1,184 highly threatened native vertebrate species (i.e. those listed as Critically Endangered or Endangered on the IUCN Red List) and 184 non-native mammals on 1,279 islands worldwide. Based on extinction risk, irreplaceability, severity of impact from invasive species, and technical feasibility of eradication, we identified and ranked 292 of the most important islands where eradicating invasive mammals would benefit highly threatened vertebrates. When socio-political feasibility was considered, we identified 169 of these islands where eradication planning or operation could be initiated by 2020 or 2030 and would improve the survival prospects of 9.4% of the Earth's most highly threatened terrestrial insular vertebrates (111 of 1,184 species). Of these, 107 islands were in 34 countries and territories and could have eradication projects initiated by 2020. Concentrating efforts to eradicate invasive mammals on these 107 islands would benefit 151 populations of 80 highly threatened vertebrates and make a major contribution towards achieving global conservation targets adopted by the world's nations.

Stomach nematodes (Mastophorus muris) in rats (Rattus rattus) are associated with coconut (Cocos nucifera) habitat at Palmyra Atoll.

Black rats ( Rattus rattus ) and their stomach nematodes (Mastophorus muris) were historically introduced to islets at Palmyra Atoll in the central Pacific Line Islands. To investigate patterns of parasitism, we trapped rats and quantified nematodes on 13 islets of various sizes and habitat types. Most rats were parasitized (59%) with an average of 12 worms per infected rat. Islet size did not greatly influence parasite population biology. Nematodes also did not appear to affect rat condition (weight to skull length). The only strong and consistent factor associated with the mean abundance of nematodes in rats was habitat (dominant cover and locally dominant plant species). Thus, nematodes were much more abundant in rats from sites dominated by coconut trees (Cocos nucifera). Coconut trees may also be an introduced species at Palmyra Atoll.