The ISI Web of Science WOS Core Collection (https://apps.webofknowledge.com/) and Scopus (https://www.scopus.com/home.uri) database were chosen for the collection of academic articles. On January 18, 2023, we retrieved all publications related to climate change in Chile using the following Boolean search strategy: [(climat^* chang^* OR global chang^* OR “climat^* emergenc^* OR “climat^* crisis OR “global warming) AND Chile^*]. A comprehensive search strategy was employed to identify relevant publications from 2012 to 2022, without any language restrictions Following the search based on these criteria, a total of 1,758 articles from Web of Science (WOS) and 1,730 articles from Scopus were retrieved. The search results were downloaded in.xlsx format for further analysis. To ensure data accuracy, a manual comparison was conducted between the SCOPUS and WOS records, which involved examining the title, primary author, source title, and year of publication. All the articles obtained, including their titles and abstracts, were exclusively in English. Duplicate articles were discarded. We next used the title and abstract- when available- of each article to ensure we only included studies aimed at understanding climate change in Chile either by Chilean or international scientists. We include original articles and reviews, but not conference proceedings or books/book chapters, in our analysis. Articles without an abstract were also excluded. This resulted in 1,261 articles used to build the academic corpus, which comprises the following metadata for each document: database, title, abstract, and publication year.

Climate Change coverage from Chilean electronic news platforms was also studied over the 10-year period from 2012 to 2022. This time period was determined by the availability of items on the selected platforms. The sample included eight electronic platforms: La Tercera, Meganoticias, CNN Chile, El Mostrador, T13, CHV Noticias, El Desconcierto and Diario Financiero. The platforms were chosen based on their national coverage, their high circulation and accessibility without a subscription fee. The approach to retrieve the articles was as follows. First, tags directly related to climate change were identified: “climate change,” “global warming,” “climatic crisis,” and “climatic emergency.” This strategy allows for a systematization of sampling. For each article, the name of the media, tag, headline, date, and URL of the source page were retrieved using the Rvest (Wickham, 2016) and RSelenium (Harrison and Harrison, 2022) R-packages. The URLs were then used to extract the articles' full text (body). Those articles that were not retrievable using this method due to forbidden access or any other restrictions in the source page were discarded from the collection. A total of 6,056 news articles were retrieved between January 06 and 15, 2023. Because a news item may include different tags, we removed duplicate articles for each of the platforms. Articles in which the date could not be retrieved were also discarded. After this filtering process, we obtained 5024 articles, which were used to build the news media corpus (Table 1).

The corpora were preprocessed as follows: performing tokenization into unigrams (one word) using the “tidytext” R-package (Silge and Robinson, 2016), normalizing text into lowercase and removing punctuation, symbols, numbers, and HTML tags. English and Spanish lists of stop words were applied to the academic (Puurula, 2013) and news media (a proposed list of Spanish stop-words was used; Díaz, 2016) corpus, respectively. Additional terms (e.g., academic corpus: “mission”, “b.v”, “rights”, “reserved”; news media corpus: “tags”, “u-uppercase”, “video”, “cnn”, “iphone”) were added to the list of stop words as frequent words present across many documents that are expected not to be related to any topic and whose presence might hinder the interpretation of the results. Also, plural words were converted to singular (e.g., academic corpus: “glaciers” to “glacier”, “southern” to “south”; news media corpus: “gases” to “gas”, “emissions” to “emission”). To preprocess the corpora, we used the “quanteda” R-package (Benoit et al., 2018).

The Mann-Kendall trend test was used to detect an increase, decrease or no difference in the number of articles published for both academic and news media corpora. Mann-Kendall test is a distribution-free test that can be used to identify monotonic trends for as few as four samples (Mann, 1945; Kendall, 1975). This is relevant for our purposes, given the results of our study were limited by a small sample size (n = 10). In brief, we tested the null hypothesis if the data are identically distributed (i.e., non-trend). The alternative hypothesis was that the data follow a monotonic trend. This monotonic trend could be positive or negative. We fitted the Mann-Kendall model using the “Kendall” R-package (McLeod and McLeod, 2015).

Latent Dirichlet Allocation (LDA), a probabilistic topic-modeling technique, was used to identify the most common topics and themes in both corpora. Briefly, topic modeling is an unsupervised machine learning technique which can identify co-occurring terms and patterns from collections of text documents (Kherwa and Bansal, 2019). Latent LDA is a well-suited unsupervised algorithm for general topic modeling tasks, particularly when dealing with long documents, which is the case with analyzing academic or news media articles (Anupriya and Karpagavalli, 2015; Goyal and Kashyap, 2022). LDA is a three-level hierarchical Bayesian model that employs three basic elements, namely the corpus which is constituted from a set of documents that is composed from a group of words (Blei et al., 2003; Blei, 2012). LDA can infer probabilistic word clusters, called topics, based on patterns of (co) occurrence of words in the documents that are analyzed. LDA models each document as a mixture of topics and the model generates automatic summaries of topics in terms of a discrete probability distribution over words for each topic, and further infers per-document discrete distributions over topic. LDA output can be used logically to classify the documents according to the topic it belongs to.

Before performing the LDA, the number of topics needs to be estimated. In this study, we used two metrics from the R-package “ldatuning” (Nikita, 2016): CaoJuan2009 and Deveaud2014. Whereas measure CaoJuan2009 has to be minimized (Cao et al., 2009), Deveaud2014 has to be maximized (Deveaud et al., 2014). Both metrics showed a plateau in the curves at 9 and 13 topics (k) for both academic and news media corpora, respectively (Figure 1). For each corpus, we fitted the LDA model using the “topicmodels” R-package (Grün and Hornik, 2011). The collapsed Gibbs sampling method was used to estimate the LDA parameters with 1,000 iterations for k = 13 and k = 9 topics for academic and news media corpora, respectively). Once generated, we assigned a label that adds an interpretable meaning to each of the inferred topics. It is important to note that the news media corpus was analyzed in its original language (i.e., Spanish), but the results (i.e., topics and themes) are presented in English.

Lastly, we used a variation of Vu et al. (2019) and Keller et al. (2020) procedures to sort the topics into five overarching themes: climate change indicators (e.g., warming, temperature, glaciers, sea-level, oceans, coastal, weather, wildfires, drought, etc.); climate change impacts (e.g., water, food, agriculture, livestock, biodiversity, ecosystems, financial etc.); climate change and society (e.g., health, wellbeing, pollution, education, humanity, population, etc.); climate politics (e.g., government, law, policy, regulation, U.N., COP, agreement, etc.); and addressing climate change (e.g., adaptation, mitigation, action, renewable, GHG, emissions, fuel, management, etc.). Figure 2 summarizes the steps of data retrieval, corpus creation and content analysis.

Data visualizations were performed using R (R Core Team, 2022) in conjunction with the software package ggplot2 (Wickham et al., 2016) and dplyr (Wickham et al., 2022).

National and international authors published 1,261 research academic articles related to climate change in Chile during the 2012–2022 period. More than half of these articles, approximately 66.0%, were published from 2019 onwards. In terms of news media, we retrieved 5,024 articles over the period 2012–2022. Of these articles, 76.6% were published in the past 4 years. Figure 3 shows trends in the number of articles for both the academic and news media corpus. Note that the scales of the y-axis are different between corpora. Mann-Kendall trend analysis showed a significant and upward trend for the number of academic articles (τ = 1, p < 0.01, Figure 3A) and news media articles (τ = 0.85, p = < 0.05, Figure 3B) articles. The number of articles published per year follows a similar trend in both corpora, however, news media articles showed a sharp increase in 2019. After these peaks, the number of published media articles decreased before an additional increase was observed.

The output of the LDA for the academic and news media corpora are displayed in Table 2. Topics were labeled based on the top 15 keywords with the largest probabilities in topics vectors (Figures 4, 5) and content in most relevant articles. In the academic corpus, the nine topics extracted were categorized into three overarching themes: “climate change indicators” (Topic A 2, A3 and A 4), “climate change impacts” (Topics A 7, A 8, and A 9), and “addressing climate change” (Topics A 1, A 5, and A 6). No topics in the academic corpus were classified as “climate change and society” or “climate politics”. The 13 topics extracted from news media corpus were classified in five themes: “climate change indicators” (Topic NM 1, NM 4, NM 7, and NM 9), “climate change impacts” (Topic NM 8 and NM 12), “addressing climate change” (Topics NM 5 and NM 13), “climate change and society” (Topics NM 2 and NM 11), and “climate politics” (Topics NM 6 and NM 10).

Our study explores the coverage of climate change in Chile by news media and research academia during the 2012–2022 period. We found a significant increase in the number of academic and news media articles published on climate change in Chile over the past decade, indicating growing interest and urgency surrounding the issue (Figure 3). The rise in Chilean literature suggests an increased interest by the scientific community in understanding climate change in Chile, which is crucial for understanding global environmental changes and their impacts on natural, social, political, and economic systems. Our findings are consistent with previous studies that have mapped the evolution of climate change science worldwide (Klingelhöfer et al., 2020; Nalau and Verrall, 2021; Reisch et al., 2021; Rocque et al., 2021). The media coverage of climate change in Chile also increased significantly since 2012, reaching a peak during 2019 before decreasing sharply in 2020 and increasing again thereafter. In 2019, the peak coincided with the climate summit (COP 25) held by Chile, generating great interest among civil society, scientists, and the private sector to share their plans for mitigating and adapting to climate change (Hjerpe and Linnér, 2010). This event occurred at the same time as the #FridaysForFuture campaign, which mobilized an unprecedented number of youths worldwide to join the climate movement, including Chile (Fisher, 2019). The campaign was instrumental not only for its potential impact on policy but also for raising public awareness about climate change and promoting action to address it. However, the media landscape experienced a notable shift in priorities due to the global COVID-19 pandemic. The pandemic brought about unprecedented challenges and uncertainties, leading to changes in media coverage patterns and public attention. News media had to allocate significant resources to reporting on the pandemic, including public health information, policy responses, and updates on the spread of the virus (Krawczyk et al., 2021; Mach et al., 2021). This shift in media priorities affected the extent and prominence of climate change coverage. Consequently, the media coverage of climate change in Chile experienced a temporary decline in 2020. However, as the world gradually adapted to the ongoing pandemic, news media resumed their coverage of climate change, and the topic regained attention. Additionally, the upcoming international conferences, such as COP 26 in England (2021) and COP 27 in Egypt (2022), may have contributed to the increased media coverage observed since 2021, as these events serve as key moments to discuss global climate action.

Using LDA topic analysis, we found that both academic and news media articles covered three of the five evaluated themes—“climate change indicators”, “climate change impacts”, and “addressing climate change”—as shown in Table 2 and Figures 4, 5. The themes “climate change and society” and “climate politics” were covered by news media but has been relatively underexplored in academic research.

Our study focused on assessing the alignment between climate change coverage in news media and academic research in Chile, revealing significant gaps in the framing of climate change between these two domains. Academic research and media coverage of climate change often focus on different aspects and utilize distinct methodologies. Academic sources offer rigorous scientific investigations, providing in-depth analysis and evidence-based insights into the complexities of climate change (Cook, 2019; Farrell et al., 2019; Masson-Delmotte et al., 2021). In contrast, media sources serve as a bridge between scientific findings and public understanding, shaping public opinion and influencing societal actions (Boykoff, 2009; Drews and Van den Bergh, 2016; Boykoff and Luedecke, 2017; Stecula and Merkley, 2019; Merkley, 2020; McAllister et al., 2021; Okoliko and de Wit, 2023). The complementary nature of academic and media sources allows for a more comprehensive communication and understanding of climate change (Goldstein et al., 2020; Lewandowsky, 2021). Through analyzing both academic and media sources, discrepancies and gaps in climate change coverage can be identified, uncovering biases and insufficient attention to certain aspects. This analysis significantly enhances public understanding by facilitating the development of targeted communication strategies that bridge these gaps, ultimately promoting informed public debates and driving effective actions. However, it is crucial to recognize that the level of media influence on public opinion depends on the level of audience engagement with climate change discourse (Wonneberger et al., 2020). Consequently, aligning academic and media coverage becomes even more essential as it enables a more accurate and balanced portrayal of climate change, thereby facilitating the implementation of necessary policies and practices to address this pressing global concern. Our findings have important implications for future research and climate communication in Chile, suggesting the need for increased attention to the challenging dimensions of climate change, such as the social dynamics and political factors associated with this global issue.

This study has several limitations that should be taken into account when interpreting the findings. Firstly, the academic corpus only included articles published in English, while the news media corpus only included articles published in Spanish. As a result, topics' keywords had to be translated into English for comparison between corpora, which could have an effect on the results. Secondly, we selected eight Chilean electronic news media sources with high readership and free accessibility without subscription fees; however, future studies should consider including other paid subscription news media as well. Thirdly, our research does not take into account other mass media platforms that can provide information about climate change (Tandoc and Eng, 2017; Becken et al., 2022). Future research could explore this topic further. Lastly, this study analyzed two corpora inherently different in terms of their coverage; news media tends to cover climate change from an international perspective, while academia focuses on a more local or regional level. These limitations do not diminish the significance of our findings. Our study highlights the need for better communication and dissemination of scientific findings to the general public. The findings of this study are not only relevant to Chile but also have global implications in addressing the pressing issue of climate change. It is crucial to bridge the gap between academic research and news media coverage to promote effective solutions for tackling this issue.