Matthew G. Betts

Determining variation in behavioral traits is critical for predicting how animals cope with anthropogenic climate change, including those that buffer reproductive output against changing environmental conditions. We investigated how contrasting ocean conditions influenced parental care in the Marbled Murrelet (Brachyramphus marmoratus), a secretive seabird that nests in coastal coniferous forests along the northeastern Pacific Ocean. Using >12,000 hr of video recordings, we quantified chick provisioning at 18 nests that represented 1,332 feeding events across 2 yr of poor (2018, 2019) and 2 yr of more favorable ocean conditions (2021, 2022). Adults provisioned chicks with a single forage fish during nest visits, which occurred primarily during early morning and late evening. Chick diet was consistent across ocean conditions: Pacific sand lance (Ammodytes personatus) and species in the smelt family (Osmeridae) comprised >90% of identified prey, with Pacific herring (Clupea pallasii) and Northern anchovy (Engraulis mordax) also present. Chicks received significantly fewer fish per day during poor ocean conditions (2.0 fish day−1; 95% CI: 1.6, 2.4) relative to more favorable ocean conditions (2.7 fish day−1; 95% CI: 2.4, 3.2). Fewer fish per day were also delivered as nest distance from the coast increased, independent of ocean conditions. However, estimated daily energy intake of chicks was slightly greater during more favorable ocean conditions, likely because Pacific sand lance delivered to chicks during poor ocean conditions were larger and estimated to have a greater energy density. Despite this partial compensation during poor ocean conditions, chicks took 25% longer (8 days) to fledge, increasing their risk of being depredated. Our study indicates that successful murrelet reproduction is dependent on a small number of prey species and that adults did not seek alternative prey to buffer the impacts of poor ocean conditions when rearing offspring. Therefore, climate-driven changes in prey availability are an important limitation on murrelet reproductive output.

Emerging evidence shows that responses to deforestation can differ both across species and among populations of the same species. The reasons underlying these complex patterns are unclear, and this lack of clarity is a barrier to accurate monitoring and prediction of biodiversity change. Using data from 2,262 bird species across 7,326 sites from all forested continents, we show that between-species and within-species variations in responses to forest cover are mediated by temperature. Populations in warmer macroclimates tend to decline in incidence in deforested landscapes because hotter microclimates push them closer to their species’ realized upper thermal tolerance limits. By contrast, populations in cooler macroclimates tend to be less affected by or even benefit from deforestation, as microclimatic temperatures are pushed closer to more optimal temperatures. Our findings offer an empirically grounded framework for biodiversity models that move beyond fixed species responses and incorporate interactions between temperature and land-use change.

Theory predicts that the strength and direction of species interactions can shift from being competitive in benign environments toward being facilitative in stressful environments. However, the environmental context dependency of species interactions has rarely been tested in animal communities. We capitalized on a 15‐year, landscape‐scale dataset, collected annually in a relatively stable old‐growth forest environment, to test the long‐held hypothesis that the strength and direction of species interactions might be mediated by climatic conditions. It is generally accepted that competitive and facilitative interactions drive the distributions of many species, but are rarely included in climate envelope models, and may partly explain poor predictive performance of species distribution models in the face of climate change. Using multi‐species dynamic occupancy models applied to long‐term data, we tested whether annual settlement by bird species could affect either the persistence or settlement by other phylogenetically related species, and whether these interactions are mediated by microclimate. We found that species interactions were influenced by microclimate for a high proportion of species pairs (80%). Related species pairs more often showed settlement dynamics that were indicative of attraction rather than repulsion (55 versus 30%, respectively). In some cases, competitive interactions in colder microclimates flipped to become facilitative in warmer ones (15%) and vice versa (5%). Furthermore, species pairs that were closely related were more likely to exhibit competitive relationships along at least part of the microclimatic gradient. Our results highlight the importance of using long‐term data to incorporate competitive and facilitative interactions into species distribution models, and support the notion that the strength and direction of species interactions can be dependent on microclimatic conditions.

Generative artificial intelligence (GenAI) tools are increasingly incorporated into ecological workflows at every stage of the scientific process, from hypothesis generation and data collection to synthesis, analysis, manuscript preparation, and peer review. This rapid integration presents new opportunities for discovery but also poses fundamental challenges to the epistemological foundations of the field. We discuss how the uncritical adoption of GenAI risks eroding core principles of ecological inquiry, including idea generation, reproducibility, transparency, and engagement with natural systems. By prioritizing scale, speed, and statistical pattern detection over rigorous theory development and testing, GenAI may shift research away from theory‐driven questions and toward large‐scale data‐mining approaches. GenAI tools offer clear benefits when used thoughtfully; these include expanded access to information and the ability to synthesize across otherwise disparate disciplines and datasets. But overemphasis on these approaches could have adverse impacts on the field, including reducing reproducibility and ecological inference and reinforcing existing inequities in knowledge production. Here, we identify key challenges associated with GenAI integration in the field of ecology and provide seven actionable guidelines to support its responsible and effective use. We call for a concerted effort to engage in reflective and intentional use of GenAI in ecological research, resisting pressures to prioritize efficiency over the rigor, creativity, inclusivity, and societal relevance of our field.

Generative artificial intelligence (GenAI) tools are increasingly incorporated into ecological workflows at every stage of the scientific process, from hypothesis generation and data collection to synthesis, analysis, manuscript preparation, and peer review. This rapid integration presents new opportunities for discovery, but also fundamental challenges to the epistemological foundations of the field. We discuss how uncritical adoption of GenAI risks erode core principles of ecological inquiry, including idea generation, reproducibility, transparency, and engagement with natural systems. By prioritizing scale, speed, and statistical pattern detection over rigorous theory development and testing, GenAI may shift research away from theory-driven questions and toward large-scale data-mining approaches. GenAI tools offer clear benefits when used thoughtfully; these include expanded access to information and the ability to synthesize across otherwise disparate disciplines and datasets. But over-emphasis on these approaches could have adverse impacts, including reducing reproducibility and ecological inference and reinforcing existing inequities in knowledge production. Here, we identify key challenges associated with GenAI integration in the field of ecology and provide seven actionable guidelines to support its responsible and effective use. More broadly, we call for a concerted effort to engage in reflective and intentional use of GenAI in ecological research, resisting pressures to prioritize efficiency over the rigor, creativity, inclusivity, and societal relevance of our field.

The models that produce most existing forest height maps minimize mean individual prediction errors quantified by metrics such as Root Mean Square Error (RMSE). However, to whatever degree the predictor data do not explain all of the height variability in the training sample, this objective function necessarily leads to prediction toward the mean at the expense of distorting population-level prediction of height range and variability. Several existing forest height maps have used training data from NASA’s GEDI (Global Ecosystem Dynamics Investigation) lidar mission, and while the mission’s retrievals are subject to measurement error, we used GEDI to both: 1) evaluate population-level errors of three prominent global height maps; and 2) produce a new set of maps designed to more accurately represent the full distribution of heights. Specifically, we created the GEDI L4D Imputed Waveform product, which assigns a high-quality waveform and all constituent measurements to each 30 m pixel in the tropical and temperate latitudes that GEDI measures. While the previous maps systematically homogenized tree heights over large regions, L4D did not, and tests verified that L4D preserved appropriate covariance among GEDI-derived forest vertical structure metrics. L4D top height RMSE values were on the order of only 1 m greater than other maps. Applications as diverse as predicting timber yield, modeling species distributions, and simulating fire spread rates may benefit from more demographically accurate tree height maps. As modeling methods evolve, attention to population-level accuracy metrics may increase the practical value of remotely sensed maps.

The species–area relationship (SAR) has long been used to predict extirpation rates from habitat loss, but these rates depend not only on habitat area but also on the surrounding landscape and species’ habitat specialization. We collated global data from forest islands created by river damming and forest fragments resulting from clear-cut deforestation to examine the effects of matrix type (aquatic or terrestrial) and tree cover on avian SARs. Unlike oceanic islands, which are often millions of years old, anthropogenic forest islands provide a contemporary analog to forest fragments to understand matrix effects on SARs and serve as a baseline for worst-case scenarios of forest fragmentation. Our database comprises 50 datasets from 45 studies conducted in tropical and subtropical regions, totaling 1,954 bird species detected through 39,197 incidence records from 336 forest islands and 669 forest fragments. We found that bird extirpation rates were lower in fragments than on islands, especially for forest-dependent species compared to all species. Species losses were further reduced by increasing tree cover around forest remnants at local landscape scales of 300 m, highlighting the importance of small-scale conservation strategies. Moreover, even small forest fragments with greater nearby tree cover held high conservation value, emphasizing the crucial role of the surrounding landscape in mitigating avian extirpations from forest remnants. Beyond protecting forest remnants themselves, area-based conservation efforts would therefore be greatly enhanced by improving matrix quality and expanding tree cover in otherwise hostile landscapes.

Rising timber demand is transforming forest structure globally, profoundly affecting biodiversity and climate resilience. Logging‐driven fragmentation is potentially a major driver of biodiversity loss in production landscapes, yet its interactions with escalating climate stressors remain poorly understood. We combine two decades of Landsat‐derived habitat metrics with 29,000 surveys of the marbled murrelet ( Brachyramphus marmoratus )—an iconic Pacific Northwest old‐forest specialist seabird affecting management of >10 million hectares. Controlling for habitat amount and detection probability, increasing landscape‐scale forest edge amount sharply reduces murrelet occupancy, with impacts worsening under unfavourable climate‐driven ocean conditions. Comparing alternative landscape‐scale timber harvest strategies, spatially concentrated logging consistently supports higher murrelet populations than fragmented approaches producing equivalent wood volumes, with benefits amplified under adverse ocean conditions. However, historical harvesting policies in the Pacific Northwest have instead driven severe habitat fragmentation, which we show is eroding the value of core set‐aside forests on federal and conservation lands and ultimately rendering murrelets more vulnerable to climate change. Synthesis and applications: We map key opportunities to boost populations by reducing edginess around remaining nesting habitat and investigate these opportunities' spatial distribution across land ownership and timber productivity gradients. Concentrating logging could be critical for mitigating fragmentation and climate threats for murrelets and potentially other forest‐dependent species amid rising timber demand.

Context: Pollen-mediated gene flow is crucial for ecological and evolutionary processes and understanding its disruption by anthropogenic disturbances is essential for conservation. Objectives: In this study, we developed landscape connectivity metrics that incorporated hummingbird movement behaviour to assess how structural (amount and configuration) and functional (species-specific behavioural response) landscape connectivity influence pollen-mediated gene flow in a tropical plant species. Methods: We adapted the incidence function model (IFM) to develop a set of functional landscape connectivity metrics that integrated field estimates of pollinator movement behaviour. We evaluated whether these metrics outperform structural landscape connectivity metrics for explaining contemporary pollen-mediated gene flow. Results: The performance of landscape connectivity metrics as predictors of contemporary pollen-mediated gene flow is scale dependent. At the focal patch scale, pollen-mediated gene flow was better predicted by structural connectivity metrics, specifically the area of contiguous mature forest. At the local landscape scale, pollen-mediated gene flow was better predicted by functional connectivity metrics that accounted for hummingbird movement, including gap-crossing probabilities. We found that including regenerating forests and narrow forest elements better explained pollen-mediated gene flow than a focus solely on mature forest. Conclusions: By integrating hummingbird movement behaviour, we offer a more realistic and nuanced understanding of how landscapes influence pollen-mediated gene flow. We underscore the importance of accounting for pollinator movement behaviour in conservation strategies that aim to preserve pollen-mediated gene flow, and demonstrate that forest regeneration is critical for maintaining functional landscape connectivity in tropical fragmented landscapes.