It is now halfway through the second decade of the 21 ^st century, and climate change impact has emerged as a “hot topic” in biodiversity research. In the early decades of the discipline of conservation biology (1970s and 1980s), effort was focused on studying and mitigating the four principal drivers of extinction risk since the turn of the 16 ^th century, colourfully framed by Diamond ¹ as the “evil quartet”: habitat destruction, overhunting (or overexploitation of resources), introduced species, and chains of extinctions (including trophic cascades and co-extinctions). Recent work has also emphasised the importance of synergies among drivers of endangerment ² . But the momentum to understand how other aspects of global change (such as a disrupted climate system and pollution) add to, and reinforce, these threats has built since the Intergovernmental Panel on Climate Change reports ³ of 2001 and 2007 and the Millennium Ecosystem Assessment ⁴ in 2005.

Scientific studies on the effects of climate change on biodiversity have proliferated in recent decades. A Web of Science ( webofscience.com) query on the term “biodiversity AND (climate change)”, covering the 14 complete years of the 21 ^st century, shows the peer-reviewed literature matching this search term has grown from just 87 papers in 2001 to 1,377 in 2014. Figure 1 illustrates that recent scientific interest in climate change-related aspects of biodiversity research has outpaced—in relative terms—the baseline trend of interest in other areas of biodiversity research (i.e., matching the query “biodiversity NOT (climate change)”), with climate-related research rising from 5.5% of biodiversity papers in 2001 to 16.8% in 2014.

Interest in this field of research seems to have been driven by a number of concerns. First, there is an increasing societal and scientific consensus on the need to measure, predict (and, ultimately, mitigate) the impact of anthropogenic climate change ⁵ , linked to the rise of industrial fossil-fuel combustion and land-use change ⁶ . Biodiversity loss and ecosystem transformations, in particular, have been highlighted as possibly being amongst the most sensitive of Earth’s systems to global change ^{7, 8} . Second, there is increasing attention given to quantifying the reinforcing (or occasionally stabilising) feedbacks between climate change and other impacts of human development, such as agricultural activities and land clearing, invasive species, exploitation of natural resources, and biotic interactions ^{2, 9} . Third, there has been a trend towards increased accessibility of climate change data and predictions at finer spatio-temporal resolutions, making it more feasible to do biodiversity climate research ^{10, 11} .

What are the major directions being taken by the field of climate change and biodiversity research in recent years? Are there particular focal topics, or methods, that have drawn most attention? Here we summarise major trends in the recent highly cited literature of this field.

To select papers, we used the Web of Science indexing service maintained by Thomson Reuters, using the term “biodiversity AND (climate change)” to search within article titles, abstracts, and keywords. This revealed 3,691 matching papers spanning the 3-year period 2012 to 2014. Of these, 116 were categorised by Essential Science Indicators ( esi.incites.thomsonreuters.com) as being “Highly Cited Papers” (definition: “As of November/December 2014, this highly cited paper received enough citations to place it in the top 1% of [its] academic field based on a highly cited threshold for the field and publication year”), with five also being classed as “Hot Papers” (definition: “Published in the past two years and received enough citations in November/December 2014 to place it in the top 0.1% of papers in [its] academic field”). The two academic fields most commonly associated with these selected papers were “Plant & Animal Science” and “Environment/Ecology”.

Next we ranked each highly cited paper by year, according to its total accumulated citations through to April 1 2015, and then selected the top ten papers from each year (2012, 2013 and 2014) for detailed assessment. We wished to focus on data-oriented research papers, so only those labelled “Article” (Document Type) were considered, with “Review”, “Editorial”, or other non-research papers being excluded from our final list. Systematic reviews that included a formal meta-analysis were, however, included. We then further vetted each potential paper based on a detailed examination of its content, and rejected those articles for which the topics of biodiversity or climate change constituted only a minor component, or where these were only mentioned in passing (despite appearing in the abstract or key words).

The final list of 30 qualifying highly cited papers is shown in Table 1, ordered by year and first author. The full bibliographic details are given, along with a short description of the key message of the research (a subjective summary, based on our interpretation of the paper). Each paper was categorised by methodological type, the aspect of climate change that was the principal focus, the spatial and biodiversity scale of the study units, the realm, biome and taxa under study, the main ecological focus, and the research type and application (the first row of Table 1 lists possible choices that might be allocated within a given categorisation). Note that our choice of categories for the selected papers was unavoidably idiosyncratic, in this case being dictated largely by the most common topics that appeared in the reviewed papers. Other emphases, such as non-temperature-related drivers of global change, evolutionary responses, and so on, might have been more suitable for other bodies of literature. We also did not attempt to undertake any rigorous quantification of effect sizes in reported responses of biodiversity to climate change; such an approach would have required a systematic review and meta-analysis, which was beyond the scope of this overview of highly cited papers.

Based on the categorisation frequencies in Table 1 (counts are given in the n columns adjacent to each category), the “archetypal” highly cited paper in biodiversity and climate change research relies on a database of previously collated information, makes an assessment based on future forecasts of shifts in geographical distributions, is regional in scope, emphasises applied-management outcomes, and uses terrestrial plant species in temperate zones as the study unit.

Many papers also introduced new methodological developments, studied montane communities, took a theoretical-fundamental perspective, and considered physiological, population dynamics, and migration-dispersal aspects of ecological change. Plants were by far the dominant taxonomic group under investigation. By contrast, relatively few of the highly cited paper studies used experimental manipulations or network analysis; lake, river, island and marine systems were rarely treated; nor did they focus on behavioural or biotic interactions. Crucially, none of the highly cited papers relied on paleoclimate reconstructions or genetic information, despite the potential value of such data for model validation and contextualisation ¹² . Such data are crucial in providing evidence for species responses to past environmental changes, specifying possible limits of adaptation (rate and extent) and fundamental niches, and testing theories of biogeography and macroecology.

At the time of writing, 5 of the 30 highly cited papers listed in Table 1 (16%) also received article recommendations from Faculty of 1000 experts ( f1000.com/prime/recommendations) ^{9, 13– 16} with none of the most recent (2014) highly cited papers having yet received an F1000 Prime endorsement.

A broad conclusion of the highly cited papers for 2012–2014 (drawn from the “main message” summaries described in Table 1) is that the pace of climate change-forced habitat change, coupled with the increased frequency of extreme events ^{15, 17} and synergisms that arise with other threat drivers ^{9, 18} and physical barriers ¹⁹ , is typically outpacing or constraining the capacity of species, communities, and ecosystems to respond and adapt ^{20, 21} . The combination of these factors leads to accumulated physiological stresses ^{13, 15, 22} , might have already induced an “extinction debt” in many apparently viable resident populations ^{14, 23– 25} , and is leading to changing community compositions as thermophilic species displace their more climate-sensitive competitors ^{13, 26} . In addition to atmospheric problems caused by anthropogenic greenhouse-gas emissions, there is mounting interest in the resilience of marine organisms to ocean acidification ^{27, 28} and altered nutrient flows ¹⁶ .

Although models used to underpin the forecasts of climate-driven changes to biotic populations and communities have seen major advances in recent years, as a whole the field still draws from a limited suite of methods, such as ecological niche models, matrix population projections and simple measures of change in metrics of ecological diversity ^{7, 12, 29} . However, new work is pushing the field in innovative directions, including a focus on advancements in dynamic habitat-vegetation models ^{30– 32} , improved frameworks for projecting shifts in species distributions ^{29, 33, 34} and how this might be influenced by competition or predation ^{35, 36} , and analyses that seek to identify ecological traits that can better predict the relative vulnerability of different taxa to climate change ^{37, 38} .

In terms of application of the research to conservation and policy, some offer local or region-specific advice on ecosystem management and its integration with other human activities (e.g., agriculture, fisheries) under a changing climate ^{18, 24, 35, 39} . However, the majority of the highly cited papers used some form of forecasting to predict the consequences of different climate-mitigation scenarios (or business-as-usual) on biodiversity responses and extinctions ^{20– 22, 33, 40} , so as to illustrate the potentially dire consequences of inaction.

The current emphasis on leveraging large databases for evidence of species responses to observed (recent) climate change is likely to wane as existing datasets are scrutinised repeatedly. This suggests to us that future research will be forced to move increasingly towards the logistically more challenging experimental manipulations (laboratory, mesocosm, and field-based). The likelihood of this shift in emphasis is reinforced by the recent trend towards mechanistic models in preference to correlative approaches ⁴¹ . Such approaches arguably offer the greatest potential to yield highly novel insights, especially for predicting and managing the outcomes of future climate-ecosystem interactions that have no contemporary or historical analogue. Along with this work would come an increasing need for systematic reviews and associated meta-analysis, to summarise these individual studies quantitatively and use the body of experiments to test hypotheses.

Technological advances will also drive this field forward. This includes the development of open-source software and function libraries that facilitate and standardise routine tasks like validation and sensitivity analysis of projection or statistical models ^{42, 43} , as well as improved access to data layers from large spatio-temporal datasets like ensemble climate forecasts ¹⁰ and palaeoclimatic hindcasts ⁴⁴ . An increasing emphasis on cloud-based storage and use of off-site high-performance parallel computing infrastructure will make it realistic for researchers to undertake computationally intensive tasks ³¹ from their desktop.

These approaches are beginning to emerge, and a few papers on these topics already appear in the highly cited paper list ( Table 1). This includes the innovative exposure of coral populations to varying carbon dioxide concentrations, and the meta-analyses of tundra plant response to experimental warming ⁴⁵ and marine organisms to ocean chemistry ²⁷ . Such work must also be underpinned by improved models of the underlying mechanisms and dynamic processes, ideally using multi-species frameworks that make use of ensemble forecasting methods for improved incorporation of scenario and climate model uncertainty ¹⁰ . Such an approach can account better for biotic interactions ⁴¹ via individual-based and physiologically explicit “bottom-up” models of adaptive responses ³¹ . Lastly, there must be a greater emphasis on using genetic information to integrate eco-evolutionary processes into biodiversity models ⁴⁶ , and on improving methods for making the best use of retrospective knowledge from palaeoecological data ¹² .