https://orcid.org/0000-0001-8449-8392
Figures
Abstract
Climate change remains a primary threat to inland fishes and fisheries. Using topic modeling to examine trends and relationships across 36 years of scientific literature on documented and projected climate impacts to inland fish, we identify ten representative topics within this body of literature: assemblages, climate scenarios, distribution, climate drivers, population growth, invasive species, populations, phenology, physiology, and reproduction. These topics are largely similar to the output from artificial intelligence application (i.e., ChatGPT) search prompts, but with some key differences. The field of climate impacts on fish has seen dramatic growth since the mid-2000s with increasing popularity of topics related to drivers, assemblages, and phenology. The topics were generally well-dispersed with little overlap of common words, apart from phenology and reproduction which were closely clustered. Pairwise comparisons between topics revealed potential gaps in the literature including between reproduction and distribution and between physiology and phenology. A better understanding of these relationships can help capitalize on existing literature to inform conservation and sustainable management of inland fishes with a changing climate.
Citation: Lynch AJ, DiSanto A, Olden JD, Chu C, Paukert CP, Gundermann D, et al. (2023) Climate impacts to inland fishes: Shifting research topics over time. PLOS Clim 2(12): e0000326. https://doi.org/10.1371/journal.pclm.0000326
Editor: Jennifer Lee Wilkening, US Fish and Wildlife Service, UNITED STATES
Received: August 14, 2023; Accepted: November 16, 2023; Published: December 29, 2023
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: The data used in this study are publicly available at https://rconnect.usgs.gov/ficli. The data and code to create the user interface are also publicly available at https://doi.org/10.5066/P9973SMC and https://doi.org/10.5066/P9VKJ23R, respectively. The code to run the analyses is publicly available at https://doi.org/10.21429/n7z0-b813. Code for the analysis was implemented in R (version 4.1.1; https://r-project.org) using the packages ‘topicmodels’ (0.2.12), ‘tm’ (0.7.8), ‘slam’ (0.1.50), ‘ggplot2’ (3.3.6), ‘bib2df’ (1.1.1), ‘scales’ (1.1.1), ‘reshape2’ (1.4.4), ‘ggrepel’ (0.9.1), ‘vegan’ (2.6.2), and ‘ldatuning’ (1.0.2). There are no restrictions to the access or use.
Funding: Funding for workshops that led to the FiCli database was provided by the American Fisheries Society with in-kind support from U.S. Geological Survey, National Climate Adaptation Science Center (NCASC) and U.S. Geological Survey, Missouri Cooperative Fish and Wildlife Research Unit. Maintenance of the FiCli database is supported by NCASC and the University at Buffalo. A. DiSanto, D. Gundermann, and M. Lang were funded by a University of Virginia Environmental Resilience Institute internship.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
Freshwater fishes comprise over half (51%) of all fish species and almost a quarter of all vertebrate species [1,2]. Globally, inland fisheries provide an economic value of US $38 billion for capture fisheries and up to US $78 billion for recreational fisheries [3]. The substantial value in inland fishes stems from diverse services that include livelihoods, income, food, recreation, cultural services, ecosystem function, and biodiversity [4]. Despite this importance, one-third of freshwater fish species are threatened with extinction [2], with the rate of decline for freshwater vertebrates double that observed in terrestrial and marine realms [5].
Climate change is one of the most critical and pervasive threats to inland fishes and fisheries [6]. Changes in climate-driven environmental regimes will have direct effects on water temperature, stream flows, water levels, oxygen concentrations, and other drivers of fish physiology [7], populations and assemblages [8], as well as fisher behavior [9]. Documented and projected effects of climate change on inland fisheries around the world are well recognized [10–12]. Reducing or mitigating these impacts remain fundamental to most management and policy strategies.
Monitoring trends in research on climate change and inland fishes and fisheries can help track the state of knowledge on these important topics, identify persistent and emerging research gaps, and understand deficiencies where the science is failing to support management needs. In doing so, quantitative analysis of literature trends can help prioritize science needs to optimize use of limited resources. Fisheries science is a mission-driven discipline. However, the scientific focus of this discipline, as reflected by the global peer-reviewed literature, may not always reflect perceived fisheries management and policy priorities. Past investigations have demonstrated this disconnect in the field of conservation science [13–15].
Tracking popular topics and trends can also highlight reciprocity between science and policy trends and global initiatives, with climate change as a frequently high-profile topic. For example, global efforts such as the United Nations Convention on Biological Diversity [16], Kunming-Montreal Global Biodiversity Framework [17], and Nature-based Solutions [18] include wording which recognizes that biodiversity conservation efforts must consider the direct influence of climate change on habitats and ecosystems as well as the cumulative effects of climate change and other stresses on ecosystems. Further, the effectiveness of actions under these frameworks may be better served with knowledge of local, regional, and global research to understand potential changes in habitats, identify critical and sensitive habitats, and quantify species and ecosystem vulnerability to climate change.
Topic modeling is a flexible and widely used tool that enables text mining / recognition and unsupervised topic discovery from a corpus of documents, e.g. [19]. Our objective is to apply topic modeling approaches to examine a comprehensive global dataset of inland fish climate change studies to determine which topics are being researched, the prevalence of those topics in the literature, the trajectories of those topics over time, and the relationships among those topics. We also relate the trends in topics through time to global climate change and biodiversity initiatives and compare the topic modeling results to artificial intelligence (AI) generated results. This exercise is intended to identify key knowledge gaps and future opportunities for research on climate impacts to inland fishes with the end goal of helping improve conservation and management of these important resources.
2. Methods
The Fish and Climate Change (FiCli; pronounced “fick-lee”) database is a comprehensive and growing global database that currently contains information from over 1,200 documented and projected responses to impacts of climate change by inland fishes extracted through a systematic review process from 323 peer-reviewed papers published in English [11,12]. Papers included in this database documented or projected fish responses to climate change (even if there was no relationship or statistically significant response). The database can be filtered by a combination of taxonomic, biological, physical, ecological traits, and geographic parameters including climate change response type (https://rconnect.usgs.gov/ficli).
Topic models [20] have emerged as an effective method for retrieving information and discovering latent topical structure of a collection of published papers [21]. We used Latent Dirichlet Allocation (LDA) modeling, the most widely used variant of topic modeling, to identify common topics using text presented in the title, abstract, and keywords of the 323 papers in FiCli database published between 1985 and 2021 in R (version 4.1.1; https://r-project.org; see Data Availability Statement for full access to the data, code, and R packages used to run this analysis). Evidence for topics is identified by sets of co-occurring words that often have similar meaning and represent a shared theme using combined phrases (e.g., “climate_change”) with non-informative words removed (i.e., stop words, words that were represented greater than 400 or fewer than 15 times; see S1 Appendix). LDA identifies sets of co-occurring words that are more frequently presented within the same linguistic context than expected by chance alone [22]. The LDA model follows the assumption that papers exhibit multiple topics in mixing proportions, thus capturing the heterogeneity of the research topics within scientific publications, akin to a mixed-membership model [23]. For example, a paper might be 50% about the topic “A”, 30% about the topic “B” and 20% about the topic “C”. A paper is usually referred to be “about” the dominant topic (topic with the highest proportion), but this can be misleading when two or more topics occur in the same paper in similar proportions or the difference between the top topics in a paper is small.
Outcomes from the LDA modeling included a list of the most common words and their topic probabilities for each paper. We provided a descriptive name to each topic by inspecting the 20 most unique words (thus most representative) from each topic. When necessary, we also inspected the papers in which each topic was dominant. Due to the unsupervised nature of LDA, the number of topics is either known a priori, or chosen based on a performance metric. Using the R package “ldatuning” [24], we optimized the number of topics by creating 15 different LDA models varying the K-parameter (i.e., number of topics) by two from two to 30. The final LDA “best” model was fitted using the R package “topicmodels” [25].
The LDA model generated a matrix of the weight (occurrence probability) of each word within each topic. Following Luiz et al. [19], this matrix was summarized by an association metric and subsequent multivariate ordination analysis. We analyzed topic similarity by calculating the Bray-Curtis distance between each topic pair using the word weight matrix in each topic to generate a word distance matrix. This distance matrix was then visualized in ordination space using non-metric multidimensional scaling (NMDS).
The LDA model generated a matrix of the weight of each topic within each paper (paper weight matrix). General topics had papers with relatively constant topic weights while specific topics had papers with one topic being predominantly weighted. To investigate topic specificity, we selected and averaged the highest topic weight of each paper within a topic. Then, we calculated the mean weight of all but the highest weighted topic for each paper (i.e., unselected topics), averaged by topic, and plotted against the highest weighted topic (i.e., selected topic). General topics had low selected weight and high unselected weight, while specific topics had high selected weight and low unselected weight.
Following Luiz et al. [19], we defined topic gaps as two topics that had both dissimilar content and papers (i.e., they were rarely investigated together), so topics with larger distances between their words and papers would be identified as gaps. To determine areas with gaps between topics, we generated the distance matrix of the paper weight matrix using the Bray-Curtis distance. Then, we scaled the paper distance matrix and word distance matrix and multiplied to generate a topic gap distance matrix with values ranging from 0–1 for each topic pair, with higher values representing higher gaps.
We also evaluated how topic modeling results compared to output generated using the recently released artificial intelligence (AI) application, ChatGPT, via OpenAI.com (ChatGPT query run by TJK on a University at Buffalo computer). The two AI search prompts were: (1) What are ten literature themes in published research on climate change effects on inland fish? And (2) What are ten literature topic areas in published research on climate change effects on inland fish? (S2 Appendix). We conducted the search on 12 June 2023.
3. Results
Following a period of low publication activity through the 1990s (with no publications in the early 2000s), the total number of FiCli papers increased rapidly in the mid-2000s with 2021 producing more than twice as many papers as any other year (Fig 1). There was a stepwise increase in the proportion of studies documenting the impacts of climate change on fish beginning in the mid-1990s and again in the mid-2000s with a steady increase thereafter (Fig 1). Through 2010, 30% of the studies were documented; from 2010 to 2021, 45% were documented (Fig 1).
The top panel bar graph is the total number of papers in the Fish and Climate Change (FiCli) database by year (brown = documented studies; teal = projected studies). The top panel line graph is the cumulative proportion of documented papers over time. The bottom panel bar graph is the proportion of papers in each topic by year. The vertical lines indicate important milestones in global biodiversity and climate change policy (IPCC = Intergovernmental Panel on Climate Change founded; CBD = Convention on Biological Diversity ratified; MDGs = United Nations Millennium Development Goals ratified; MEA = Millennium Ecosystem Assessment published; IPBES = Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services founded; SDGs = United Nations Sustainable Development Goals ratified).
The topic modeling resulted in a set of ten topics that we defined based on the top words represented in each grouping (see Table 1). We present them in order based on their hierarchical relationships to one another (Fig 2A):
- populations (e.g., abundance, density),
- growth (e.g., bioenergetics, prey availability),
- reproduction (e.g., spawning, recruitment),
- phenology (e.g., emergence, developmental timing, migration timing and spawning),
- distribution (e.g., range shifts, changes in population distribution),
- physiology (e.g., thermal responses, thresholds),
- assemblages (e.g., species composition, species richness, evenness, species interactions),
- invasive species (e.g., non-native species, species interactions),
- drivers (e.g., stressors, environmental pressures), and,
- climate scenarios (e.g., future scenarios, projections).
A) Dendrogram of topics of papers in the Fish and Climate Change (FiCli) database. The height (y-axis) of the dendrogram represents the dissimilarity between topics. B) The topic gap distance between two pairs of topics of papers. Blue cells are topic pairs with high overlap of papers between the two topics; red cells have little to no overlap in papers between the two topics.
We acknowledge there is subjectivity in defining these topic names manually based on the top words included (e.g., €, Euros, in invasive species may highlight that costs are frequently discussed in this topic), but the relationships hold regardless of the topic names. Based on the FiCli database which includes studies from 1985 to 2021, studies on invasive species, reproduction, and distribution were more common in the earlier portion of the dataset while climate scenarios, drivers, assemblages, and phenology are more common recently (Fig 3).
Mean annual change in topic popularity between consecutive years of topic weights within each paper expressed as a percentage. Blue topics grew and red topics shrank in popularity within the Fish and Climate Change (FiCli) database between 1985 and 2021.
Papers focusing on the climate change impacts on species’ distributions, physiology, and growth were predominantly projected studies whereas timing papers (e.g., spawning, migration) were largely documented studies (Fig 4A). Although the phenology and reproduction topics grouped together, the remaining other topics did not form strong clusters (Fig 4A). The phenology topic was highly specific, meaning that papers grouped in this cluster typically did not include elements found in other topics and were also predominantly documented studies (Fig 4B). By contrast, the assemblages topic was general and often included papers addressing additional topics (Fig 4B).
Based on word distributions of title, abstract, and keywords for papers in the Fish and Climate Change (FiCli) database summarized using non-metric multidimensional scaling (NMDS): A) Similarity in word distributions between topics as indicated by their proximity in ordination space, and B) Topic specificity with more general papers found in the upper left; more specific papers are found in the lower right. Bubbles are scaled by the total number of papers in a topic and color relates to the proportion of documented studies in a topic (brown = more documented studies; teal = more projected studies).
‘Populations’ included 32 papers focusing on subjects such as arctic and high latitude fisheries, e.g. [26], long term studies, e.g. [27], warming, e.g. [28], and adaptation, e.g. [29]. The topic has experienced some decline in popularity (Fig 3). Papers within ‘populations’ were relatively unique but most similar to those within ‘reproduction’ and ‘growth’ (Fig 4A) likely because these topics are related to fish population dynamics. ‘Populations’ papers were categorized as more general literature (Fig 4B) and were mostly projected studies (Fig 4). These papers were also predominantly focused on lentic habitats (while lotic habitats are more common across the FiCli database generally; Table 2).
‘Growth’ included 31 papers on topics concerning body mass fluctuations on seasonal and annual scales. Papers generally focused on food availability, and how shifting thermal regimes may affect prey densities as well as inherent fish growth. Illustrating this, [30] modeled how the growth of Great Lakes fish may be affected by the water temperature changes due to climate change. They found (through incorporating fish thermal guilds), that fish growth would generally increase with climate change should prey availability increase but decrease if prey availability did not change. Growth was further tied to specific seasons, which may alter in length as climate change becomes more influential. Similarly, to the paper described above, a majority of papers including ‘growth’ tend to be predictive in nature (Fig 4A and 4B). ‘Growth’ appears to be more prevalent within topically specific papers (like the other ‘predictive’ topics, such as ‘physiology’) (Fig 4B), but generally retains a mostly unique word distribution (Fig 4A).
‘Phenology’ included 33 papers, predominantly focused on shifts in the timing of seasonal migrations and reproduction. Most of these papers were on salmonids, e.g. [31], but some focused on other taxa, such as walleye (Sander vitreus), e.g. [32], and golden mahseer (Tor putitora), e.g. [33]. The relative number of papers in this topic area has increased more than any other topic from 1985 to 2021 (Fig 3), suggesting increased importance or data availability related to phenology. The ‘phenology’ topic area is unsurprisingly most like ‘reproduction’ (Fig 4A) and is the most specific topic area in the list (Fig 4B).
‘Reproduction’ included 40 papers. Papers within this category examined how climate change, primarily through thermal regime and discharge alterations, impacts population reproduction rates. Exemplifying papers include [34], which examines how discharge and temperature impacts recruitment of European grayling (Thymallus thymallus), and [35], which examined how thermal sensitivity and flow changes have impacted migration and recruitment for coastal sockeye salmon (Oncorhynchus nerka). Reproduction showed the second-strongest decrease in topic popularity over the study period (Fig 3). The topic of ‘reproduction’ was evenly split between projected and documented studies and was neither strongly specific nor general; this suggests that the topic of ‘reproduction’ is used broadly by varying research efforts (Fig 4B). ‘Reproduction’ possessed the greatest word distribution similarity with ‘phenology,’ likely because spawning frequently occurs at highly specific times (Fig 4A).
‘Distribution’ included 36 papers on range shifts of fish species that have been either observed [36] or projected based on future climate scenarios [37]. ‘Distribution’ papers decreased between 1985 and 2021 (Fig 3). A greater proportion of the distribution papers were projected rather than documented and as such are most like ‘climate scenarios’ papers (Figs 2 and 4A). The ‘distribution’ papers fell between the specific and general papers, which suggests that while some of the papers focus on distributions, there was some overlap with papers focused on ‘physiology’ and ‘invasive species’ among other specific topics (Fig 4B).
‘Physiology’ included 35 papers. The topic featured studies on temperature and how it affects the function of fish species, e.g. thermal carrying capacity [38], thermal sensitivity [39]. ‘Physiology’ decreased slightly over the study period (Fig 3). Most papers in this topic were projected studies, with the topic more specific than general (Fig 4B) and having similar word distributions to ‘growth’ and ‘distribution’ and large gaps to ‘assemblages,’ ‘phenology,’ and ‘reproduction’ (Fig 4A). These studies were predominantly in lotic habitats (Table 2).
‘Assemblages’ included 32 papers focusing on subjects including fish community structure, biodiversity, and assemblages. The topic is the second most increasingly popular topic in the FiCli database between 1985 and 2021 (Fig 3). ‘Assemblages’ papers were most like the ‘drivers’ and ‘phenology’ papers and least like the ‘physiology’ papers (Fig 4A), largely projected studies and a general topic overall (Fig 4B).
‘Invasive species’ included 28 papers that generally focused on the ecology and management of nonnative and invasive species in a changing climate. Brown trout (Salmo trutta) and smallmouth bass (Micropterus dolomieu) are among the most common invasives studied [40,41]. We found evidence for a striking reduction in the popularity of invasive species studies over time, demonstrating the greatest downward trend compared to all other topics (Fig 3). Studies focusing on invasive species tended to be more specific in their research focus, concentrating less on other recognized topic areas (Fig 4B). These studies were predominantly in lotic habitats (Table 2).
‘Drivers’ included 27 papers primarily on how land use change and landscape patterns affect fish. Several papers focused on more direct drivers of how climate change may result in changes to water temperature and streamflow that, in turn, would alter the distribution, abundance, or presence in fish and macroinvertebrates. For example, Kovach et al. [42] analyzed 92 populations of bull trout (Salvelinus confluentus) in North America and found that increased water temperature drove declines of bull trout, but the presence of the additional driver of invasive species are more immediate threats that can be proactively managed. Although the proportion of papers about ‘drivers’ has increased in the FiCli database in recent years (Fig 3), they were relatively ubiquitous and do not directly align with other topics and fell between specific or more general topics (Fig 4). Of the four topics that had their topic weight increase in recent years, ‘drivers’ was the only one where documented studies were a lower proportion than projected studies (Fig 4A).
‘Climate scenarios’ included 25 papers predominantly related to future or projected scenarios, typically corresponding to fish responses to temperature or hydrology changes. The studies tended to have a reference time range that happened in the past or concurrently and a future scenario typically decades into the future, e.g. [43]. ‘Climate scenarios’ increased slightly over the FiCli database time series (Fig 3). The topic did not show any strong similarities with other topics (Fig 4A), perhaps unsurprisingly, predominantly projected studies, and a general topic overall (Fig 4B). These studies were almost exclusively in lotic habitats (Table 2).
By examining the similarity and specificity of the topics, we assessed how closely related the topics group within NMDS space (Fig 4A) and overlap of papers between the topics (Fig 4B). Using the topic gap distance between topic pairs, we can discuss synergies between topics and knowledge gaps (Fig 2B). Pairwise examination of topics reveals potential gaps in the literature where few studies are grouped in both topic clusters. The ‘Drivers’ topic had strong pairings with ‘Climate scenarios,’ ‘Distribution,’ and ‘Assemblages’ but the gap distances were not large across any other topics, suggesting that there are thematic commonalities between this topic and most of the others. Similarly, ‘Growth’ was most topically like ‘Populations,’ ‘Physiology,’ and ‘Reproduction’ (likely because all are related to fish age demographics) but it did not have substantial topic gaps among the other topics, indicating that it is a well-integrated topic.
Six of the topics identified using the topic modeling approach were also identified with AI approaches (Table 1). The four topics that were not common were ‘Climate Scenarios’, ‘Populations’, ‘Physiology’, and ‘Invasive species’. Interestingly, broader topics of biodiversity loss, ecosystem functioning, ecosystem services, and conservation and management strategies were identified as important literature topics by AI.
4. Discussion
Trends in fisheries research have implications for understanding the threats of climate change and, consequently, the allocation of management effort and policy reform. The major thematic topics we identified represent the major areas of study within the English-language peer-reviewed literature on climate change effects on inland fishes. Though our analysis was narrower in scope than Syed et al.’s [44] topic analysis of fisheries science, a number of topics are common between the two studies: reproduction, growth, and assemblages to the topic modeling analysis and conservation and management to the AI search (Table 1). By examining how these topics have changed in popularity over time, we can explore potential misalignments between research interests and the needs of existing policy initiatives (Fig 1). For example, prior to the establishment of the International Panel on Climate Change, very few studies were published on climate change and inland fishes but this corpus grew exponentially by the mid-2000s concurrent with the publication of the Millennium Ecosystem Assessment. Though it is more difficult to directly attribute, these scientific studies may drive new policy initiatives and vice versa. FiCli, for example, was featured in the “Statement of World Aquatic Scientific Societies on the Need to Take Urgent Action against Human-Caused Climate Change, Based on Scientific Evidence” [45], and the rise in the ‘Drivers’ topic could reflect rising interest in cumulative effects highlighted in global conservation frameworks.
Other topics did show large gaps. For example, studies that focus on ‘Reproduction’ rarely also focus on ‘Distribution.’ This result may be attributable to the tendency of distributional studies to focus on occupancy associated with habitat change rather than the changes in reproductive success that may result in range expansions or contractions. Likewise, studies that examine ‘Physiology’ rarely also examine ‘Phenology’. This could be because physiological changes from temperature are less commonly associated with a specific timing of an event and are generally just warmer or colder.
Comparisons with AI approaches
We also evaluated our topic modeling results compared to AI approaches using the recently released ChatGPT tool via OpenAI.com. S2 Appendix includes AI search prompts and results (search attempted on 12 June 2023). The AI results are largely similar to those revealed in the topic modeling. For the first prompt “What are ten literature themes in published research on climate change effects on inland fish?”, approximately six of the ten topics in each approach had 1:1 correspondence (Table 1). Similarly, when prompted to provide a list of ten literature topics dominant in climate change and inland fish studies, AI provided eight which matched our results (S1 Table). One area where the approaches deviated is the presence of management/conservation themes and social/human impacts in the AI output. We expect this is because this theme is actually integrated across many of the topic modeling topics (e.g., resourc,” “manag,” “restor,” and “conserv” appeared as top 20 words across multiple topics; Table 1). While papers included in the FiCli database are explicitly related to organismal responses to climate change, discussion of conservation and management implications are often present in these papers and FiCli even includes a specific field on management recommendations [12].
Research needs and future opportunities
The ten thematic topics identified in this study reflect the foci of much of the inland fish primary literature published since 1985. Although some of the papers within the topics are not mutually exclusive, as indicated by the specificity and gap distance results, classification into ten topics allows areas to be highlighted for future research needs and opportunities. These gaps can be classified as minor, moderate or major. ‘Reproduction’ and ‘Distribution’ were not closely related but were the two most common topics with 79 of the 323 (24%) papers yet, they showed a decline in popularity through time. The shift from these types of studies suggests that they have been well researched and perhaps that the methods to determine climate change effects on distribution and reproduction are established. However, minor taxonomic and geographic gaps remain, e.g. bias of research towards salmonid species and temperate ecosystems [11], that could be addressed through application of the methods in existing studies to species and geographic regions where understanding may be lacking.
Although the remaining topics that have been researched provide some understanding of the effects of climate change on the biology and ecology of fishes, moderate gap areas for future research are evident. For example, many of the studies in ‘Phenology’ (and the closely related ‘Reproduction’) have been conducted on fishes important to commercial, recreational, and subsistence fisheries because the timing of spawning coincides with population assessments, harvest practices, and informs harvest regulations. While important from a cultural and economic standpoint, native ‘rough’ fishes and species at risk also provide important ecosystem services, and their conservation compels assessments of the climate change on their biology and ecology [46,47].
The large topic gaps between ‘Physiology’ and ‘Reproduction’, and ‘Physiology’ and ‘Phenology,’ are curious and present a research opportunity. For example, stock productivity models (related to ‘Reproduction’) could be improved by the inclusion of organismal physiological responses to environmental change [48]. Similarly, studies linking the ‘Distribution’ and ‘Reproduction’ topics could identify if areas newly occupied by a species had the conditions present to reproduce and not be a population sink. As new breeding sites are established and old breeding sites are abandoned, a potential study could compare reproductive timing and success changes concurrent with the range expansions/contractions.
Major gaps remain regarding the studies linking trophic dynamics to multiple stressors. ‘Assemblages’ papers focus on fish assemblage-level dynamics related to climate change. However, studies of climate change effects on trophic dynamics that include lower trophic levels and nutrient dynamics and those focusing on functional group responses are important avenues for research as warming can rewire freshwater food webs [49–51]. The increasing popularity of ‘Drivers’ papers reflects the increasing availability and application of landscape-level datasets, e.g. [52] which presents an opportunity for watershed and multi-ecosystem assessments of climate effects. It also reflects the need to understand the relative influence of climate change versus other stressors on inland fishes [53,54] to improve fish and fisheries conservation and management [55].
The distribution of habitat types (i.e., lotic and lentic) across the topics highlights that some system–topic pairs are understudied and provide an opportunity for future work. For example, just one of 25 ‘Climate Scenarios’ papers is focused on a lentic habitat whereas 72% of the ‘Populations’ papers are focused on lentic habitats, counter to the general habitat proportions across the whole FiCli database (73% lotic, 27% lentic; Table 2). While there may be constraints on some study approaches based on habitat type, these gaps highlight opportunities to expand understanding of system-specific climate impacts.
Study limitations
Topic modeling studies are useful tools to critically evaluate the output of a bounded scope of study, however, there are some recognized limitations, generally and with the particular dataset we used. First, labeling topics in a topic modeling exercise is somewhat subjective because it is presently a manual process [44]. Consequently, some topics could have been labeled differently and that may have influenced the interpretation of the relationships among the topics. For example, we considered ‘Biodiversity’ and ‘Stressors’ as alternative names for the ‘Drivers’ topic and ‘Vulnerability’ and ‘Disturbance’ as alternatives for ‘Invasive Species.’ Second, using topic modeling as an exercise to identify gaps is somewhat myopic because only identified topics can be compared. Understudied and underrepresented topics are not present. Using a complementary analysis like simple AI queries (e.g., Table 1) can help identify additional themes that may be well-represented in the current literature. These may be true gaps in study or they may be not captured by the review process (e.g., grey literature).
The FiCli database has some additional limitations that may impact the results of this analysis. First, as it only includes English-language literature, this is not a truly global review. While English is often considered “the global language of science” and meta-analyses for conventional medicine found no evidence of systematic bias from use of language [56], approximately 35% of biodiversity conservation literature is published in languages other than English [57]. Consequently, FiCli may be representing only around two-thirds of the available literature on documented and projected climate impacts to inland fishes. This language gap is likely to perpetuate geographic biases as well (i.e., higher representation from locations where authors publish in English). Second, the systematic review process used to generate the FiCli database is strictly bounded to studies that aim to document or project responses of inland fish to climate change [12]. This results in exclusion of other literature relevant to this discussion (e.g., experiments, laboratory studies, studies on climate variability; [58]). Consequently, it is important to caveat that any findings here are also bound by that same scope. Again, a supplementary analysis like AI queries can highlight potential gaps (Table 1).
5. Conclusions
Topic modeling is an increasingly popular tool to examine trends and relationships within bodies of scientific literature. Here, reflecting on almost 40 years of studies on documented and projected climate impacts to inland fish, we can observe the growth and evolution of this literature from just a handful of studies through the mid-2000s to a well-established field by the 2020s (Fig 1). The emergent topics of assemblages, climate scenarios, distribution, drivers, growth, invasive species, populations, phenology, physiology, and reproduction (Table 1) have been important areas of inquiry for climate impacts to inland fish. The relationships between these topic areas highlight that some of these topics are more closely linked than others (Fig 2A) and that some are more singular (Fig 2B). But, perhaps more importantly, the topic gaps indicate important areas of existing research and potential opportunities for future research (Fig 4). Climate change is and will continue to be an important driver of change for inland fish. Learning from the existing body of literature on the key topics in this field can help conserve and sustainably manage these important fishes with a changing climate.
Supporting information
S1 Table. Comparison of topic modeling topics and artificial intelligence (AI) generated themes from a ChatGPT query via OpenAI.com from prompt 2.
At least eight of these theme areas are similar between topic modeling and AI output.
https://doi.org/10.1371/journal.pclm.0000326.s001
(DOCX)
S1 Appendix. Words and phrases found in the title, abstract, and keywords for papers in the Fish and Climate Change (FiCli) database that were removed from the analysis.
These include stop words (e.g., “the,” “and,” “but”), words and phrases used too frequently (>400 times) or rarely (< 15 times) to be informative to the exercise, punctuation, and numbers.
https://doi.org/10.1371/journal.pclm.0000326.s002
(DOCX)
S2 Appendix. Output from an artificial intelligence (AI) application, ChatGPT, query via OpenAI.com as a comparison to topic modeling.
https://doi.org/10.1371/journal.pclm.0000326.s003
(DOCX)
Acknowledgments
We thank Sarah Burton for producing the word cloud image. We also thank all authors from the previous FiCli associated manuscripts (Lynch et al. 2016b; Myers et al. 2017; Krabbenhoft et al. 2020; Lynch et al. 2022) and Lauren Craige, Candace Engel, and Ashley Robertson for their help in identifying and reviewing literature on the FiCli team. We thank Bonnie Myers for an internal USGS review which improved the manuscript. The Missouri Cooperative Fish and Wildlife Research Unit is jointly sponsored by the Missouri Department of Conservation, the University of Missouri, the U.S. Geological Survey, the U.S. Fish and Wildlife Service, and the Wildlife Management Institute. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
References
- 1. Fricke R, Eschmeyer WN, Van der Laan R. Eschmeyer’s Catalog of Fishes: Genera, Species, References. 2021. Available: https://www.calacademy.org/scientists/projects/eschmeyers-catalog-of-fishes.
- 2.
WWF. The World’s Forgotten Fishes. World Wide Fund for Nature; 2021 p. 47. Available: https://wwf.panda.org/discover/our_focus/freshwater_practice/the_world_s_forgotten_fishes/.
- 3.
Funge-Smith S. Review of the state of the world fishery resources: inland fisheries. Rome, Italy: Food and Agriculture Organization of the United Nations (FAO); 2018 p. 376.
- 4.
Lynch AJ, Arthur RI, Baigun C, Claussen JE, Kangur K, Koning AA, et al. Societal Values of Inland Fishes. In: Mehner T, Tockner K, editors. Encyclopedia of Inland Waters. Elsevier Inc.; 2022. pp. 475–490. https://doi.org/10.1016/b978-0-12-819166-8.00030-x
- 5. Thieme ML. Inland Waters Finally Get the Mention They Deserve. Sustainability Works. 2023. Available: https://www.worldwildlife.org/blogs/sustainability-works/posts/inland-waters-finally-get-the-mention-they-deserve.
- 6. Paukert CP, Olden JD, Lynch AJ, Breshears DD, Christopher Chambers R, Chu C, et al. Climate Change Effects on North American Fish and Fisheries to Inform Adaptation Strategies. Fisheries. 2021;46: 449–464.
- 7. Whitney JE, Al-Chokhachy R, Bunnell DB, Caldwell CA, Cooke SJ, Eliason EJ, et al. Physiological basis of climate change impacts on North American inland fishes. Fisheries. 2016;41: 332–345.
- 8. Lynch AJ, Myers BJE, Chu C, Eby LA, Falke JA, Kovach RP, et al. Climate change effects on North American inland fish populations and assemblages. Fisheries. 2016;41: 346–361.
- 9. Hunt LM, Fenichel EP, Fulton DC, Mendelsohn R, Smith JW, Tunney TD, et al. Identifying alternate pathways for climate change to impact inland recreational fishers. Fisheries. 2016;41: 362–372.
- 10. Comte L, Buisson L, Daufresne M, Grenouillet G. Climate-induced changes in the distribution of freshwater fish: Observed and predicted trends. Freshwater Biology. 2013;58: 625–639.
- 11. Myers BJE, Lynch AJ, Bunnell DB, Chu C, Falke JA, Kovach RP, et al. Global synthesis of the documented and projected effects of climate change on inland fishes. Reviews in Fish Biology and Fisheries. 2017;27: 339–361.
- 12. Krabbenhoft TJ, Myers BJE, Wong JP, Chu C, Tingley RW, Falke JA, et al. FiCli, the Fish and Climate Change Database, informs climate adaptation and management for freshwater fishes. Scientific Data 2020 7:1. 2020;7: 1–6. pmid:32317639
- 13. Lawler JJ, Aukema JE, Grant JB, Halpern BS, Kareiva P, Nelson CR, et al. Conservation science: a 20-year report card In a nutshell: Front Ecol Environ. 2006;4: 473–480.
- 14. Stroud JT, Rehm E, Ladd M, Olivas P, Feeley KJ. Is conservation research money being spent wisely? Changing trends in conservation research priorities. Journal for Nature Conservation. 2014;22: 471–473.
- 15. Di Marco M, Chapman S, Althor G, Kearney S, Besancon C, Butt N, et al. Changing trends and persisting biases in three decades of conservation science. Global Ecology and Conservation. 2017;10: 32–42.
- 16.
CBD. Global Biodiversity Outlook 3. Montreal, Canada: Secretariat of the Convention on Biologival Diversity (CBD); 2010 p. 94. Available: https://www.cbd.int/gbo3/.
- 17.
CBD. DECISION ADOPTED BY THE CONFERENCE OF THE PARTIES TO THE CONVENTION ON BIOLOGICAL DIVERSITY 15/4. Kunming-Montreal Global Biodiversity Framework. 2022; 1–23.
- 18.
Cohen-Shacham E, Walters G, Janzen C, Maginnis S. Nature-based solutions to address global societal challenges. Cohen-Shacham E, Walters G, Janzen C, Maginnis S, editors. Nature-based solutions to address global societal challenges. Gland, Switzerland: International Union for Conservation of Nature and Natural Resources; 2016. https://doi.org/10.2305/iucn.ch.2016.13.en
- 19. Luiz OJ, Olden JD, Kennard MJ, Crook DA, Douglas MM, Saunders TM, et al. Trait-based ecology of fishes: A quantitative assessment of literature trends and knowledge gaps using topic modelling. Fish and Fisheries. 2019;20: 1100–1110.
- 20. Blei DM. Probabilistic topic models. Communications of the ACM. 2012;55: 77–84. https://doi.org/10.1145/2133806.2133826.
- 21. Liu L, Tang L, Dong W, Yao S, Zhou W. An overview of topic modeling and its current applications in bioinformatics. SpringerPlus. 2016;5. pmid:27652181
- 22. Murakami A, Thompson P, Hunston S, Vajn D. “What is this corpus about?”: Using topic modelling to explore a specialised corpus. Corpora. 2017;12: 243–277.
- 23. Erosheva E, Fienberg S, Lafferty J. Mixed-membership models of scientific publications. Proceedings of the National Academy of Sciences of the United States of America. 2004;101: 5220–5227. pmid:15020766
- 24. Ldatuning Murzintcev N. [R]. 2014.
- 25. Hornik K, Grün B. 17 topicmodels: An R Package for Fitting Topic Models. Journal of Statistical Software. 2011;40: 1–30.
- 26. McDonald ME, Hershey AE, Miller MC. Global warming impacts on lake trout in arctic lakes. Limnology and Oceanography. 1996;41: 1102–1108.
- 27. Kangur K, Kangur P, Ginter K, Orru K, Haldna M, Möls T, et al. Long-term effects of extreme weather events and eutrophication on the fish community of shallow Lake Peipsi (Estonia/Russia). Journal of Limnology. 2013;72: e30.
- 28. Cohen AS, Gergurich EL, Kraemer BM, McGlue MM, McIntyre PB, Russell JM, et al. Climate warming reduces fish production and benthic habitat in Lake Tanganyika, one of the most biodiverse freshwater ecosystems. Proceedings of the National Academy of Sciences of the United States of America. 2016;113: 9563–9568. pmid:27503877
- 29. Jeppesen E, Mehner T, Winfield IJ, Kangur K, Sarvala J, Gerdeaux D, et al. Impacts of climate warming on the long-term dynamics of key fish species in 24 European lakes. Hydrobiologia. 2012;694: 1–39.
- 30. Hill DK, Magnuson JJ. Potential effects of global climate warming on the growth and prey consumption of Great Lakes fish. Transactions of the American Fisheries Society. 1990;119: 265–275.
- 31. Juanes F, Gephard S, Beland KF. Long-term changes in migration timing of adult Atlantic salmon (Salmo salar) at the southern edge of the species distribution. Canadian Journal of Fisheries and Aquatic Sciences. 2004;61: 2392–2400.
- 32. Schneider KN, Newman RM, Card V, Weisberg S, Pereira DL. Timing of Walleye Spawning as an Indicator of Climate Change. Transactions of the American Fisheries Society. 2010;139: 1198–1210.
- 33. Joshi KD, Das SCS, Pathak RK, Khan A, Sarkar UK, Roy K. Pattern of reproductive biology of the endangered golden mahseer tor putitora (Hamilton 1822) with special reference to regional climate change implications on breeding phenology from lesser himalayan region, india. Journal of Applied Animal Research. 2018;46: 1289–1295.
- 34. Bašić T, Britton JR, Cove RJ, Ibbotson AT, Gregory SD. Roles of discharge and temperature in recruitment of a cold-water fish, the European grayling Thymallus thymallus, near its southern range limit. Ecology of Freshwater Fish. 2018;27: 940–951.
- 35. Atlas WI, Seitz KM, Jorgenson JWN, Millard-Martin B, Housty WG, Ramos-Espinoza D, et al. Thermal sensitivity and flow-mediated migratory delays drive climate risk for coastal sockeye salmon. Facets. 2021;6: 71–89.
- 36. Alofs KM, Jackson DA, Lester NP. Ontario freshwater fishes demonstrate differing range-boundary shifts in a warming climate. Diversity and Distributions. 2014;20: 123–136.
- 37. Buisson L, Grenouillet G, Casajus N, Lek S. Predicting the Potential Impacts of Climate Change on Stream Fish Assemblages. American Fisheries Society Symposium. 2010;73: 327–346.
- 38. Ayllón D, Nicola GG, Elvira B, Parra I, Almodóvar A. Thermal carrying capacity for a thermally-sensitive species at the warmest edge of its range. PLoS ONE. 2013;8. pmid:24282584
- 39. Campos DF, Braz-Mota S, Val AL, Almeida-Val VMF. Predicting thermal sensitivity of three Amazon fishes exposed to climate change scenarios. Ecological Indicators. 2019;101: 533–540.
- 40. Al-Chokhachy R, Schmetterling D, Clancy C, Saffel P, Kovach R, Nyce L, et al. Are brown trout replacing or displacing bull trout populations in a changing climate? Canadian Journal of Fisheries and Aquatic Sciences. 2016;73: 1395–1404.
- 41. Sharma S, Jackson DA, Minns CK. Quantifying the potential effects of climate change and the invasion of smallmouth bass on native lake trout populations across Canadian lakes. Ecography. 2009;32: 517–525.
- 42. Kovach RP, Al-Chokhachy R, Whited DC, Schmetterling DA, Dux AM, Muhlfeld CC. Climate, invasive species and land use drive population dynamics of a cold-water specialist. Journal of Applied Ecology. 2017;54: 638–647.
- 43. Segurado P, Branco P, Jauch E, Neves R, Ferreira MT. Sensitivity of river fishes to climate change: The role of hydrological stressors on habitat range shifts. Science of The Total Environment. 2016;562: 435–445. pmid:27100019
- 44. Syed S, Borit M, Spruit M. Narrow lenses for capturing the complexity of fisheries: A topic analysis of fisheries science from 1990 to 2016. Fish and Fisheries. 2018;19: 643–661.
- 45. Bonar SA. Development of the “Statement of World Aquatic Scientific Societies on the Need to Take Urgent Action Against Human-Caused Climate Change, Based on Scientific Evidence.” Fisheries. 2021;46: 411–412.
- 46. Woo-Durand C, Matte JM, Cuddihy G, McGourdji CL, Venter O, Grant JWA. Increasing importance of climate change and other threats to at-risk species in Canada. Environmental Reviews. 2020;28: 449–456.
- 47. Rypel AL, Saffarinia P, Vaughn CC, Nesper L, O’Reilly K, Parisek CA, et al. Goodbye to “Rough Fish”: Paradigm Shift in the Conservation of Native Fishes. Fisheries. 2021;46: 605–616.
- 48. Cooke SJ, Sack L, Franklin CE, Farrell AP, Beardall J, Wikelski M, et al. What is conservation physiology? Perspectives on an increasingly integrated and essential science. Conservation Physiology. 2013;1: 1–23. pmid:27293585
- 49. Bartley TJ, McCann KS, Bieg C, Cazelles K, Granados M, Guzzo MM, et al. Food web rewiring in a changing world. Nature Ecology and Evolution. 2019;3: 345–354. pmid:30742106
- 50. Angeler DG, Allen CR, Birgé HE, Drakare S, McKie BG, Johnson RK. Assessing and managing freshwater ecosystems vulnerable to environmental change. Ambio. 2014;43: 113–125. pmid:25403974
- 51. Elmqvist T, Folke C, Nyström M, Peterson G, Bengtsson J, Walker B, et al. Response diversity, ecosystem change, and resilience. Frontiers in Ecology and the Environment. 2003;1: 488–494.
- 52. Isaak DJ, Wollrab S, Horan D, Chandler G. Climate change effects on stream and river temperatures across the northwest U.S. from 1980–2009 and implications for salmonid fishes. Climatic Change. 2012;113: 499–524.
- 53. Sharma S, Vander Zanden MJ, Magnuson JJ, Lyons J. Comparing climate change and species invasions as drivers of coldwater fish population extirpations. PLoS ONE. 2011;6. pmid:21860661
- 54. Gutowsky LFGG Giacomini HC, de Kerckhove DT, Mackereth R, McCormick D, Chu C. Quantifying multiple pressure interactions affecting populations of a recreationally and commercially important freshwater fish. Global Change Biology. 2019;25: 1049–1062. pmid:30580472
- 55. Jackson MC, Loewen CJG, Vinebrooke RD, Chimimba CT. Net effects of multiple stressors in freshwater ecosystems: a meta-analysis. Global Change Biology. 2016;22: 180–189. pmid:26149723
- 56. Morrison A, Polisena J, Husereau D, Moulton K, Clark M, Fiander M, et al. The effect of English-language restriction on systematic review-based meta-analyses: A systematic review of empirical studies. International Journal of Technology Assessment in Health Care. 2012;28: 138–144. pmid:22559755
- 57. Amano T, González-Varo JP, Sutherland WJ. Languages Are Still a Major Barrier to Global Science. PLoS biology. 2016;14: e2000933. pmid:28033326
- 58. Lynch AJ, Myers BJE, Wong JP, Chu C, Tingley RW, Falke JA, et al. Reducing uncertainty in climate change responses of inland fishes: A decision‐path approach. Conservat Sci and Prac. 2022;4: e12724.