Die Landschaft interpretieren: Interdisziplinäre Ansätze Interpreting Landscape: Interdisciplinary approaches

no generally valid, recognised definition of what landscape actually is. On the basis of its relatedness to the environment, to aesthetics, territory, society, politics, economics, geography, planning, ethnology and philosophy, the concept of landscape can be viewed as a “composite” notion shaped by a thousand years of Central European ideas and of literary and art history. In other words, the very term “landscape” is ambiguous and it is used in different ways both in the scholarly world and everyday speech. The aim of this to present various possible approaches to this phenomenon. be exhaustive have tried to do overarching interdisciplinary approach the Zentrum Landschaftsinterpretation und Tourismus Landscape Interpretation and Tourism) singling this fascinating multifaceted phenomenon. research media language and press freedom, stylistics and language in sport. her 2016), she compiled a newspaper corpus covering different regions and news domains and analysed it within a systemic functional framework, examining register variation and correlations between language and press freedom. Other areas of interest: empirical educational science and applications of corpus linguistics in other disciplines, mainly literary studies and social sciences.


Introduction
Climate change has long been a dominant topic in news coverage around the world and has reached new peaks since the "Fridays for Future" initiative started off in 2018. Opinions, especially regarding the anthropogenic nature of climate change, differ greatly and all groups are making their voices heard. In this context, social media has been established as a very popular platform for this discourse and its subgroups, as it makes it easy for everyone to participate in multiple ways and to various degrees. The special forms discussions and conflicts can take on social media have been studied in several analyses (see e.g. Fløttum et al. 2014;Cody et al. 2015;Anderson and Huntington 2017;Bloomfield and Tillery 2019;Pearce et al. 2019); however, despite the undeniable influence of social media, Nielsen and Schrøder (2014, 472) found that traditional media, including newspapers, "are still more widely used and seen as more important sources of news than social media". The rise of social media and its usage should therefore not lead to an underestimation of professional news outlets, which have had, and continue to have, a huge impact on the public's opinion and perception with regard to climate change (Boykoff and Roberts 2007, 1;Anderson 2009, 166;Carvalho 2010, 172;Speck 2010, 125).
With this responsibility and the huge audiences of news outlets in mind, it is clear that the way topics and opinions are presented is crucial to the perspective society takes in general (see e.g. Weingart, Engels and Pansegrau 2000;Storch 2009;Jaspal, Nerlich and van Vuuren 2016), and as for other areas of news, journalists follow norms of journalistic writing (see e.g. Bennett 1996;Boykoff and Boykoff 2004, 126;Fest 2016, 73-75). Unlike in other news domains though, some norms have proved to be problematic when reporting on climate change. The most extreme example is balance, which is defined as a norm aiming at objectivity and requiring that journalists present the viewpoints of all concerned parties when discussing an issue (cf. Entman 1989, 30). In theory, this serves to ensure that the audience receives as much information on an ongoing debate as possible and is therefore enabled to form a qualified opinion. Especially when looking at political news, this mechanism is highly desirable and a basic democratic necessity and right. When the issue under debate is one of science, however, this norm is difficult to put into practice because "it seems to demand that journalists present competing points of view on a scientific question as though they had equal scientific weight, when they actually do not" (Gelbspan 1998, 58).
This leads to the question of whether journalists have an alternative approach at their disposal. The main one coming to mind would be to present what is scientifically more accurate and thus stick to the scientific truth. This alternative brings with it two major issues: on the one hand, it means that journalists would have to be expected to understand (pseudo-)scientific evidence well enough to evaluate its accuracy and decide on the truth value. A lack of knowledge on the part of journalists has been identified as having an impact on the reporting (Wilson 2000). However, having the temporal resources and scientific background needed for such an approach is unlikely even in the case of science journalists (see Dunwoody and Peters 1992, 210). On the other hand, it stands to reason that this is a journalist's responsibility -or is it not mainly their duty to report on events and their right to expect that their recipients take the initiative of forming their own opinion instead of taking over one that is presented to them, thus adhering to the norm of balance? The second question is of course a difficult one which includes sociological, political and journalistic considerations, among others. This paper does not aim to answer it; however, since the first option -of journalists having to evaluate scientific output -is not a realistic one, aiming for balanced reporting remains the dominant path of action.
The norm of balance brings with it another problem, namely that of proportion. When different viewpoints are involved, presenting them equally elaborately (i.e. allocating them equal amounts of space in news coverage) is balanced in absolute terms, but does not reflect how widely these viewpoints are spread in reality. A representation of the actual proportion in reality is not possible -on the one hand, it is impossible to measure exactly how much of a discourse is based on the different perspectives, and on the other hand, news coverage itself is part of this discourse and shapes and changes it. Absolute balance, in turn, brings it with it the problem that some voices appear more prominent than they really are, while others seem less loud. The result is that what Boykoff (2013) terms "outlier voices" (i.e. those appearing more dominant in news coverage than they are in reality) are always the beneficiaries of the journalistic norm of balance, leading to information bias (see also Adger et al. 2001;Speck 2010, 130). This dilemma is amplified by the fact that scientific news is among the hardest to process for all involved. Coverage of science news in mass media has to undergo two stages of transmission: from scientist to journalist and from journalist to public (Boykoff and Boykoff 2004, 126). Both steps necessarily include simplification and thus loss of information, while other perspectives that are not as directly based on scientific evidence can be transmitted much more directly. As Cody et al. (2015, 1) summarise: "Newspaper articles may lack accuracy, while the severity of findings in scientific papers may be too opaque for the public to understand." Bailey, Giangola and Boykoff (2014) argue that journalists are becoming aware that their status is more than just that of passers-on of information, which means that they have to be regarded as consciously active participants in this discourse. Despite the issues mentioned above, journalists are not left with nothing in the way of tools and means to fulfil this role and actively frame different viewpoints. After all, having to present them, even in equal measure, does not mean having to present them using the same language. Different linguistic features have been shown to fulfil different functions, and via a deliberate use of them journalists are well able to shape the message of a text in more or less subtle ways.
On this basis, the present paper aims at taking a closer look at specific linguistic features that have been found to strengthen the impression of commitment on the one hand and uncertainty on the other. The analysis is based on a corpus of 120 newspaper articles from 2020, 60 each from the USA and Australia, covering both editorial and non-editorial news items (see Section 3.1 for a detailed description of the dataset). These countries were selected because both have faced, and are still facing, dramatic natural disasters in 2020, while also having been found to have a tradition of news reporting critical of anthropogenic climate change (Antilla 2005;Freudenburg and Muselli 2010;Bacon 2013;Jaspal, Nerlich and van Vuuren 2016;Schmid-Petri et al. 2017).
In the first part of the analysis, commitment (and avoidance of it) will be examined quantitatively. For this purpose, a set of three linguistic features -pronominal use, modality and passive voice -has been operationalised and extracted from the corpus. In the second part, these results will be combined with a set of keyword analyses, which will consider contrasts not only between editorial and non-editorial news pieces, but also between the two countries. In a last step, the results and their implications for the climate change discourse will be related back to previous findings and future research perspectives.

Climate change reporting in US and Australian print news
The changes observable in climate are happening on a global scale, but Australia and the USA have been very heavily affected by natural disasters connected to global warming throughout the past years. Droughts, storms and heat waves have been major issues in both countries, and most of all, wildfires destroyed huge areas of land and caused massive damage to nature and residential areas. The effects of such phenomena could be felt by most residents, which made them an important and constantly present topic in the news on both national and regional levels. In the political landscape, the pressure increased to act and find solutions to these problems before they become even more severe in the future, leading to debates about the causes and thus looping back to the bigger issue of climate change. In both countries, conservative leaders with critical stances towards anthropogenic climate change have been in charge for several years now (see Jaspal, Nerlich and van Vuuren 2016, 808-9; Pryck and Gemenne 2017), creating a constellation resulting in conflicting policies and public representation. Nonetheless, crucial differences between the two countries have been found. The USA is often considered the birthplace of scepticism towards climate change in the media (Antilla 2005;Boykoff 2007;Zehr 2000). Painter and Ashe (2012) contrasted the USA to five other countries and found it to be the country with the highest degree of scepticism in general, but also the only one in which all three types of sceptic as defined by Rahmstorf (2004) 1 , were represented to an almost equal degree. Boykoff (2007, 483) claims that these sceptical voices are not sufficiently put into the context of being contradictory to scientific insight and fact. This causes uncertainty regarding both the portrayal of climate change and reasonable ways to cope with it, which is amplified even further by a tendency of scientists to hedge a lot and rarely speak of their results as absolute and irrevocable (ibid., 482). Journalists interviewed in Boykoff's (2007) study criticised the scientific community, saying that those having the expertise need to take more responsibility and be more explicit: "[Scientists] often blame the media for not being prepared to go out and say things that they -who have the knowledge base -will not say in public" (ibid., 484). At the same time, journalists are trained to report events and stories and find it difficult to cover aspects and facts that do not (yet) have an outcome (Fedler et al. 1997, 94), meaning that hedging language by scientists and a need on the part of journalists for definite happenings collide in an inevitable conflict. On this basis, Rice, Gustafson and Hoffman (2018) analysed US news between 2005 and 2015 and came to more nuanced results. Despite the differences in language use between scientists and journalists, they found that "[t]he vast majority of the uncertainty, controversy, disagreement, and scepticism frames in climate journalism are not from scientists and are not in reference to topics on which there is scientific consensus" (ibid., 317; emphasis in original).
Possible explanations for this can be found when looking at results from studies analysing the role of mass media more closely. Bailey, Giangola and Boykoff (2014) examined how uncertainty is constructed in US news dealing with climate change, and found that overall, direct markers for uncertainty decreased slightly between 2001 and 2007. In contrast to this, indirect markers, e.g. elements of surprise -especially at the gap between predictions and observations -increased (ibid., 213), resulting in a subtler way of creating doubt and insecurity. Similar results are reported by Schmid-Petri et al. (2017), who found that while explicit scepticism had decreased over time, other means to express doubt are taking its place, for instance reminders that political actions should be cautious (ibid., 508).
Feldmann, Hart and Milosevic (2017), analysing the (non-)representation of climate change as a potential threat in US print news, came to similar conclusions; they found that only 41% of all articles dealing with climate change mentioned at least one possible threat arising from it, and although the vast majority of news items (85%) were found to contain some call for action, the responsibility for this was mainly placed with the government. Furthermore, actions and impacts were not often discussed together, meaning that "audiences are provided with information about the impacts of climate change without information about how to reduce those impacts or -more often -are informed of actions to address climate change without the context to appreciate why those actions are needed" (ibid., 492). All in all, the authors conclude that "leading US newspapers discuss climate change in ways that have the potential to leave readers feeling unconcerned about climate change, disempowered, or both" (ibid.). So, while scepticism and uncertainty directed at the issue itself (and the scientific consensus around it) have decreased, aspects with a less factual and scientifically founded character, such as recommended paths of action, have more often been framed by vague reporting.
This seems to show that complete denial of global warming is ebbing away in the USA, but readiness to take measures is also decreasing -a picture that is different from that observable in Australia. Morrison et al. (2013) ran a survey in Australia in 2011 to determine attitudes of citizens towards the issues of climate change and global warming, based on a scale created by Maibach et al. (2011) and used in 2008 for a similar study in the USA. Ranging from alarmed (as the highest degree of certainty that these issues are a threat) to dismissive (meaning that they are considered irrelevant), with the stages concerned, cautious, disengaged and doubtful in between, interviewees were asked to evaluate several aspects of their own behaviour and opinion. While for the USA, Maibach et al. (2011) reported 51% to be at least concerned, Morrison et al. (2013) found that three years later in Australia, only 33.3% were alarmed or concerned by climate change in general. The crucial point was not a lack of certainty about the event itself but rather that only about 45% of Australians believed in a dominantly anthropogenic nature of climate change (ibid., 92; see also Speck 2010, 129). This is in line with Bacon's (2013, 77) results, which show that in 2011/2012, climate change was mainly framed as a political topic in Australian media and sceptical voices were presented rather favourably.
The numbers have not changed much since; in 2018, 46% of Australians were reported to believe that human activity was the main driving force behind climate change (Ipsos 2019), as opposed to 49% in the USA (Funk and Hefferon 2019). At the same time, Australians were reported to be more certain than US Americans that climate change will affect them personally and will do so in the foreseeable future (Morrison et al. 2013, 92). In a second round in Australia of the survey described above, Morrison, Parton and Hine (2018) observed that more people believed in climate change than in 2011. Acceptance of human influence was found to have increased even more; however, personal involvement had decreased, revealing an atmosphere of "issue fatigue" in the Australian public (ibid., 7). While mass media is undoubtedly the main source of information for most people, news coverage of climate change has increased constantly over the last years and especially in Australia, this upward trend is very explicit (Schmidt, Ivanova andSchäfer 2013, 1241), which might have led to satiation and, eventually, defiance.
To fully understand this relation, analyses of newspaper language in relation to climate change reporting are necessary. As outlined above, a lot has been done in this area for US media, but for Australia, most studies conducted so far concentrated on yet more specific aspects. Bednarek and Caple (2010) took a multimodal approach and examined evaluative language, presentation of images and captions in news coverage dealing with climate change in the Sydney Morning Herald. Other studies analysed newspaper language in articles treating industries that are connected to climate change, for instance the fossil fuel (Holmes and Star 2018) and the coal industries (Bacon and Nash 2012). Robertson (2017) ran a pilot study tracing dominant vocabulary and recurring semantic themes in Australian climate change news from 2013 and concluded that "discourses surrounding climate change in the Australian news media are, ultimately, problematic for the Australian environment" (ibid., 52).
This study does not focus on any particular aspect of climate change -within the thematic field, a distinction is only made between opinion pieces and other news items. As scepticism towards climate change and human influence on it on the one hand and issue fatigue on the other are evidently problems the discourse is currently facing, but journalists have been reported to have fairly strong opinions on these matters (Brüggemann and Engesser 2014, 411-12), the analysis focuses on the commitment displayed by news writers. In so doing, the aim is to contribute to the general understanding of how climate change is presented in the news and whether certain linguistic markers create an influence beyond their original functions.

Methodology
In order to trace commitment in news writing, the study relies on different approaches. Three linguistic features -modality, passive voice and pronominal useare analysed quantitatively. Data points are created as relative frequencies of these features per article. Additionally, keyword analyses are done between the two countries as well as between the news domains of opinion pieces and other news to get a closer understanding of the lexis. In the following sections, the dataset and the operationalisations of the quantified features will be outlined in more detail before results are discussed in Section 4.

Dataset
The dataset the following analyses are based on consists of 120 articles from the websites of newspapers from the USA and Australia. The countries are represented by 60 articles each, the US American ones of which were collected from the New York Times, USA Today and the New York Post. Australia is represented by articles from the Sydney Morning Herald, the Brisbane Times, the Herald Sun, the Daily Telegraph and The Age. Of the 60 per country, 30 articles come from the "Opinion"-sections and 30 are other news. It was ensured that no newspaper contributes more than ten articles to either category.
All articles were selected within a predefined sampling frame. Since the focus of this research is on news coverage dealing with climate change from the year 2020, the search was restricted by these parameters. Furthermore, only articles that could be accessed freely were collected. Different newspapers were included in the corpus to ensure diversity. The difference in number of papers -three for the USA and five for Australia -is due to the fact that media ownership in Australia is distributed among only few corporations, a circumstance which is also mirrored in the country's press freedom index (Reporters Without Borders 2020). The five newspapers contained in the corpus are held by two owners, Nine Entertainment Co. (The Age, Brisbane Times, Sydney Morning Herald) and News Corp Australia (Daily Telegraph, Herald Sun) (cf. Harding-Smith 2011; Fisher 2019, 132). Articles appearing in one newspaper are often also published on the websites of others from the same corporation, which makes it difficult and redundant to assign individual news items to specific papers.
Within these boundaries, articles were sampled randomly. The US component of the corpus contains 50,357 words and 57,644 tokens, which means an average of 839.28 words and 960.73 tokens per article. The articles vary in length between 302 and 1,688 words. The Australian component consists of 46,910 words and 53,423 tokens, which results in an average of 781.83 words and 890.38 tokens per article. The articles consist of 273 to 2,407 words. In total, the corpus contains 97,267 words and 111.067 tokens. The corpus was annotated for parts of speech and lemmata using the TreeTagger tag set with 58 tags (Schmid 1994; University of Washington).
As can be seen right away, both the two parts as well as the articles within them vary regarding their length. In the analyses, this will be resolved by working solely with relative frequencies. In both parts, the opinion articles make up a larger amount of words than the news articles: in the USA, these items account for 28,324 words (56.25% of the whole), in Australia they make up 25,056 words (53.41%).

Modality
The use of modal verbs has often been connected to hedging and a vague style of writing (see e.g. Poynton 1989, 71, 79;Halliday and Matthiessen 2014, 176;Fest 2016, 191), as they can be used to weaken statements or implicitly question their truth value. However, not all modals function alike, as the following examples from the corpus demonstrate (emphases added): 1) If every country acted like Australia, temperatures might rise by close to four degrees. (Kelly 2020, Brisbane Times) 2) Climate models indicate further cooling is likely, with the majority of models suggesting sea surface temperatures will approach or exceed La Niña thresholds during the southern hemisphere spring. (Crockford 2020, Sydney Morning Herald) The statement in Kelly (2020) is clearly made less certain by the use of might, -it seems that the consequence of rising temperatures is possible, but by no means definite. The use of will in Crockford (2020) on the other hand suggests clarity. The author could have used a weaker modal in this position, like might or may, but chooses not to, thereby expressing that there is no doubt. The English language contains numerous modals in the form of central and semi-modals and marginal auxiliaries (cf. Biber et al. 2012, 483-484). For the present analysis, only the central modals were considered, all nine of which as defined by Biber et al. (ibid.) are present in the corpus at hand: can, could, may, might, must, shall, should, will and would. In all corpus queries, variations of the words such as contracted forms ('ll, 'd etc.) and negations (cannot, can't, couldn't, won't etc.) were included in the counts.

Pronominal use
The use of pronouns is a potentially direct way of creating commitment in writing. Like the use of modal verbs though, not all pronouns have the same effect. The 1 st person singular pronouns (I, my, me, myself, mine) are the most explicit references journalists can make to themselves and usually indicate that a personal opinion or attitude is being expressed. 2 nd person pronouns (you, your, yourself, yourselves, yours) imply addresses to the readers, which do not necessarily signal commitment of the writer to the topic, but a desire to create a connection with the addressee (Neumann 2014, 70;Fest 2016, 214). A similar effect can be attributed to 1 st person plural pronouns (we, our, ourselves, ours, us) when used in an inclusive sense. When expressing stances or opinions, we includes I, but stating that something is our opinion is less personal and therefore less committing for an individual than explicitly referring to oneself only, which is why it can be regarded as a downgrading of commitment.
Pronouns can also be used to draw a boundary between one's own group and others (see e.g. Fina 1995;Wodak 1996), an effect which can be achieved by using 3 rd person plural pronouns (they, them, their, themselves, theirs). They imply that what is being reported does not necessarily match the writer's own opinion, or is at least not supposed to appear as if it does. However, this should be treated with caution when looking at news writing. It is general writing advice for journalists not to repeat names too often, and instead use points of reference to a person in the text to include more information (Keeble 2006, 113;Papper 2013, 49-56;Fest 2016, 175). Australia's Prime Minister Scott Morrison can also be referred to as "the 52year-old", "the head of government" or "the leader of the Liberal Party", or, when all relevant information has been given or the text becomes too dense, simply as "he". This does not necessarily imply anything beyond the purely pronominal use, but might be a textual necessity, an aspect on which the inclusion of 1 st person pronouns is less dependent.

Passive voice
Passive voice, as the counterpart to active voice, is another option to shift the focus of an utterance and leave out certain bits of information. This is particularly true for short passives, whereas long passives (also called by-passives) usually contain all information and just reverse the syntactic order (emphases added): 3) For Australia, the choice between danger and opportunity is clear, and that choice must be made now. (Figueres 2020 Example 3 shows how short passives can be used to identify an action -in this case to call for decision-making -without assigning responsibility for it to anyone. Long passives, like example 4, contain an actor, but place it at the end of the clause. Example 5 is a clear case of attributing responsibility to specific people by using active voice; Scott Morrison is accused of "diverting our opinion", and his predecessor Tony Abbott is named as the person abolishing carbon tax. In news, passives have been found to make up between 13.8% and 15% of all finite verbs and were most frequent in hard news and political news (Biber et al. 2012, 476;Fest 2016, 225). Generally, passive voice can be considered the marked option, as it occurs much less frequently than active constructions. This implies that a writer consciously opts for making use of it and its effects.
For the present study, short and long as well as be-and get-passives are taken into account. Since passive constructions can vary in their length, relative frequencies per any amount of words would be misleading; also, a writer can decide between active and passive voice only once per verb phrase. All relative frequencies are therefore given in percentages of all finite verbs per article.

Modality
Looking at the overall frequencies and distributions of the modal verbs can, could, may, might, must, shall, should, will and would, the first obvious trend is that they are used more in opinion pieces than in other news items (see Figure 1). The two categories show similar median values across both countries, although the US-Opinion part displays a broader range and a slightly more even distribution than its Australian equivalent.

Commitment in Newspaper Coverage of Climate Change 286
Figure 1: Boxplot of relative frequencies of modal verbs across all corpus parts.
As was described in Section 3, not all modals can be connected to vagueness and a non-committing style of writing to the same degree. Of the nine central modals analysed here, will and must (and their respective variations) imply much more certainty than the others. Must is not used very often, but will is the most frequently used modal of all, both in opinion articles and general news. Also, it is consistently more frequent in Australian articles than in US-American ones. This observation is in line with previous findings; Biber et al. (2012, 486) and Fest (2016, 191) also report will to be the most common central modal both in general language use and in newspaper writing. What is remarkable is that in both countries, the difference in modality -both for will and for other modals -between opinion items and general news is statistically significant (for will: W=1223.5, p=0.002414; for other modals: W=1270, p=0.005448), whereas the countries as a whole do not differ significantly for any modal (for will: W=2169, p=0.05226; for other modals: W=1890, p=0.6385).
It is not surprising that in articles of an editorial nature and taken from newspaper sections that are explicitly labelled "Opinion", personal attitudes are included and expressed. That these are somewhat hedged by modality can also be expected, as writers still need to cater for a big audience and therefore phrase their articles as discussion starters rather than fixed news. Keeping in mind that scientists who are the core source for all reports on scientific news also tend to hedge a lot (see Boykoff 2007), it is even more understandable that journalists demonstrate hesitation to be more explicit when giving their opinion on a topic on which even the designated experts remain vague.
This makes the frequent use of will, especially in contrast to other, weaker modals, quite noticeable. Obviously, hedging through modality is combined with a trend towards stronger statements, which appears paradoxical at first sight. Possible explanations can only lie in the concrete topics that are treated differently. The results presented by Bailey, Giangola and Boykoff (2014) and Schmid-Petri et al. (2017), saying that explicit uncertainty regarding climate change is decreasing but vagueness regarding non-scientific facts such as reasonable laws or measures is increasing, suggest that this might be the cause of the discrepancy found here. A close semantic analysis of all individual articles would be needed to check this correlation reliably; however, a look at collocates proves a helpful first indicator.
The most frequent collocates to appear immediately before or after will are be, it and we. Be is easily explained by the modal nature of will, but the pronouns, especially the 1 st person plural we, signal that committing statements are apparently made for a group, not an individual. This in turn means that authors using this combination are not making their own opinion more definite, but are creating pressure for themselves and their audience. We is also a frequent collocate for other modals, so the implications should not be over-interpreted; for the other modal verbs though, other pronouns, most notably I and they, also appear among the most frequent collocates, making the we less conspicuous and showing that when referring to themselves, more careful statements are preferred by the writers. The full verbs most frequently collocating with will, apart from the aforementioned be, are have, require, make, do and change, ergo mainly verbs that signal processes and actions. This somewhat contradicts the assumptions that actions to be taken are being framed as cautious (Schmid-Petri et al. 2017), but makes sense for Australia when considering the fear uttered by many that climate change will affect them personally and therefore, actions are urgent.

Pronominal use
As was discussed in Section 3.2, pronouns can serve very different functions in a text. The distributions that are found across the four corpus parts show some clear patterns, but also display some interesting differences. Across the corpus, 3 rd person singular pronouns are the most common type of pronominal use, which is not surprising regarding their textual necessity as explained above and the neutral nature of it. The use of 1 st person singular pronouns, which can be seen as direct involvement of the author, is generally rather low in news writing -however, the US-Opinion section stands out with a remarkably high figure. This is partly caused by one text which contains 48 such pronouns (5.32% of its entire length) and must therefore be considered an outlier. Nonetheless, the category has a much higher median than the others (US-Opinion 0.21; US-News 0; Aus-Opinion 0.1; Aus-News 0) and also shows a broader distribution across the texts. In contrast to this, it contains the fewest 2 nd person pronouns. Although the differences regarding you and its variations are small, there is a slight tendency to use it more in Australian opinion articles. It is the only category of the four which has a median value of >0 (0.09).
The most interesting results can certainly be observed for the use of 1 st person plural pronouns. As for modality, there is not much difference between the two opinion-parts, and the two news parts are quite alike too. Between opinion and other news however, there is a statistically significant difference (W=1098, p=0.0002298): the news part displays an average of 0.78% 1 st p. pl. pronouns per text, while for opinion pieces, the average is 1.43%. There are outliers only for the news category, and only towards the upward end, which means that the average would be even less if these were left out.
It can thus be said that opinion articles create a much more explicit connection between the author and the readers. Although these news items are presented as expressing the authors' opinions, they do not as a rule make significantly more references to their writers than other news; rather, it seems their strategy is to frame any given attitude as one that is shared by others. This is more committing than not expressing any personal reference, but it demonstrates a degree of hesitation to take full credit for what is being said by placing oneself in an (assumed) ingroup. We share this opinion, thereby making it stronger and spreading it over more shoulders. Combined with the frequent co-occurrence with will, which was discussed above, it can be seen that the consequences of having a particular opinion, especially for taking action, are also distributed and passed on to the readers.

Passive voice
Similar patterns can be seen if we look at the use of passive voice. As was described above, what is mainly of interest for this study is the short passive as the form which allows the writer to omit information. Table 2 includes relative frequencies for both passive forms. As can be seen, the short passive occurs much more frequently than the long passive. There is no major difference between the categories, and although the data shows a slightly bigger discrepancy on the level of countries, this difference is significant for overall passive use (W=2181.5, p=0.04552), but not for short passives only (W=2135, p=0.07912).
The long passives, although much less frequent, thus make a crucial difference. Long passives include an actor and can therefore not be said to be an avoidance strategy per se, yet it is interesting to see who or what is being mentioned follow-ing the by. In the current dataset, 72.28% of all long passives do not refer to concrete persons, but to organisations, phenomena or abstract concepts (emphases added): 6) Carbon capture and storage is backed by the Australian coal industry […]. (O'Malley 2020, The Age) 7) NSW beaches, including those near Sydney, were battered by abnormally high tides […]. (Hannam and Bungard 2020, The Age) 8) Yet those long-term concerns have typically been outweighed by the demand for new homes […]. (Flavelle 2020, New York Times) In only 27.72% of all long passives in the corpus, the reference was explicitly made to a human entity, and only 6.93% contained one or more concrete names.
The majority of cases refer to groups, such as "Western Australians", "members of the community" or simply "most" or "many people". Although the passive constructions give an actor, it can therefore not be said that they necessarily imply more certainty of or commitment to the issue discussed in the article.

Keywords
The quantitative analyses outlined above have already pointed out several interesting aspects and particularities regarding reporting on climate change in different countries and news categories. Lastly, we will take a more detailed look at significantly frequent lexical items. Table 3 lists the top ten keywords for comparisons between the countries and the news domains, and it can be seen right away that several observations from the quantitative analyses are mirrored on the lexical level. Looking at opinion and news, we again come across several 1 st person plural pronouns. Our, us and we are key for opinion articles, which is not surprising since the significance of the difference between the categories was already discussed above. 1 st person singular pronouns are not key, which strengthens the hypothesis that opinions are expressed as shared by a community rather than being the attitudes of individual authors. Will also shows up as a keyword for opinion articles, but other modal verbs do not appear among the top 10. The other items that are key for opinion articles are technical terms, which is interesting since there is no apparent reason why they should be more dominant in opinion pieces than in general news; what is striking though is that three of the terms, hydrogen, wind and solar, refer to renewable energy sources. It thus seems that these alternatives to the more traditional fossil fuel and coal get more attention in news categories which deal with personal ideas rather than facts, which might strengthen their "alternative" character.
In general news on the other hand, terms that refer to events and environmental facts rather than technologies are dominant; we find ozone, ice and fire among the top ten keywords. The keyness of Dr and researchers indicates that in these news items, the inclusion of sources and their validity in terms of expertise are stressed, which seems to play a minor role in opinion pieces.
In the comparison between the two countries, it can be seen that many keywords can simply be accounted for by geographical differences and can be attributed to the news value of proximity (Bednarek and Caple 2014), which is to say that what happens close by tends to be more relevant for recipients and is therefore covered more extensively in news. This accounts for all locations and also includes the U which is key for the US part; the letter appears as part of USA, but since SA is short for South Australia and consequently occurs in Australian news fairly frequently, only the U remained as being key. The difference between per cent and percent is simply due to local spelling conventions. This leaves only very few items of interest, namely koala, gas, emissions and government as key for Australian news and hurricane, Trump, ozone, ice, administration and city for the USA.
Koala and Trump are again self-explanatory as being particularities of the country concerned. Government and administration have related meanings, but it is striking that in Australian news, the government is explicitly named while in the USA, the reference is either to some entity of administration or to the president directly. Also, US news appears to discuss specific natural phenomena more directly, as we can find ice, ozone and hurricane among the top ten keywords. In Australia on the other hand, the keyness of gas and emissions indicates a focus on human involvement rather than on nature.
The table only distinguishes between the categories on the one hand and the countries on the other, but of course, more dimensions for keyword analyses are possible in order to shed light on the differences. Comparisons were also made of the two opinion parts only, to see whether the articles from one country were mainly responsible for the trends visible for opinion in general in Table 2. This was not the case though; the only item from the combined list appearing as key for one opinion part in contrast to the other was gas, which the cross-country comparison already shows to be an Australian key term. The same check was done for the news category; here, ozone and ice indeed proved to originate mainly from the US part of the corpus, which again is reflected in the overall comparison on country level.

Discussion
The results presented above give answers to some questions and indicate even more interesting aspects for deeper investigations. Regarding the commitment of the authors, which was the starting point of this study, it can be said that there is not much difference between Australia and the USA. Rather, variation appears to depend on the news category; opinion articles show higher frequencies of modal verbs in general, but also of will in particular, which implies a tendency to hedge on the one hand and to make clear predictions on the other. Will turned out to cooccur frequently with we, which means that commitment can be found, but authors place responsibility on the audience as well as on themselves. Since 1 st person singular pronouns were not found to occur more frequently in any category, it is safe to assume that opinion articles do not so much aim at expressing an individual's opinion as at negotiating a shared opinion with the target audience. This can be interpreted as a decrease of commitment on the part of the authors, especially in contrast to non-editorial news; however, it should be said that in general news, neither audience nor authors are often referred to. The results for long passives including mainly non-human entities as actors further strengthen the impression that people, as well as their actions and influences, are not a focus in general news, meaning that the necessity to make decisions on personal commitment is simply reduced from the start. In contrast to this, news articles could be seen to emphasise scientific facts and their validity. We can therefore confirm Feldmann, Hart and Milosevic's (2017) conclusions that descriptions of climate change and calls for action or involvement do not usually go together -they appear, in fact, in different news categories.
This observation certainly holds a lot of potential to help increase our understanding of climate change news discourse in the future and leaves room for more elaborate studies based on larger or more diverse datasets. Nonetheless, a close analysis of the lexis has proven very valuable too. While some results gained from the keyword analyses simply confirmed what the quantitative approaches had shown, one very particular finding is that renewable energy sources receive significantly more attention in editorial than in non-editorial news. This can be interpreted in different ways; on the one hand, the opinion-based character of editorials (in contrast to the science-based nature of other news) suggests that topics discussed here are less scientifically founded or verified. This might have negative effects on the way renewable energy is perceived by the public, as it could decrease the seriousness and factualness connected to them. On the other hand, since the opinion pieces have been found to include more direct addresses and calls for personal involvement, readers might be able to make a more direct link to concepts that are mentioned in the same article.
To understand this properly, a more detailed analysis would be necessary connecting news categories to the perception of the topics they deal with. A more specific dataset, looking only at articles dealing with renewable energy, could also be beneficial to a comparison of how and to what degree they are portrayed. This is related to another question raised by the results presented here, namely that regarding the construction of expertise. News was seen to refer to researchers, and to include academic titles such as Dr to validate the sources, but opinion pieces displayed no such trends. Nonetheless, they, too, have to create the impression of "knowing what they are talking about" if they are to keep their readers and be taken seriously. One way to achieve that might be via external sources and references to information elsewhere, e.g. via hyperlinks. Also, the number of direct quotes and the framing of those cited could be an indicator, for which a transitiveity analysis might be useful.
To conclude, it can be said that the discourse of climate change is far from being clearly understood, but results from various perspectives are contributing not only to gaining a more complete picture, but also to identifying the most crucial aspects for future work.