The resource of a corpus may be used in several fields, especially in linguistics, as a broader area, and natural processing language, for research in electronic lexicography, for general-purpose speech¹, and in terminology², for specialized speech. For each field, or research objective, the use of corpora differ, according to the parameters set for the specific research.

For example, in lexicographic or terminological studies, it is important to have a somewhat freer approach, to enable better systematization and more comprehensive understanding of meaning, while in a grammatical approach it is relevant to have an annotated corpus, to enable a more specific syntax analysis. On the other hand, data provided are more objective for the lexicographer or terminologist, while a grammarian has to formulate more hypotheses regarding formal and functional relations between words.

Within this context, the support of authentic texts has been largely used mainly in historical approaches to language, but only as a means to find evidence for an already set theoretical foundation, based on intuition. This way of doing research was later formalized as a ‘corpus-based perspective’ or, as Tognini-Bonelli (2001) points out, “a pre-corpus perspective”. Later, along with technological advances, it was possible to compile large amounts of discourse, enabling better systematization from the corpus evidence itself, a perspective named ‘corpus-driven’.

This idea of systematization takes into account what a given pattern of language use means, and how it can be evaluated in terms of typicity in a given system. In other words, language features have to be understood as the probability to happen in a general context or in a certain domain, not as a “fatalistic” evaluation of occurrences, of happening vs. not happening at all. As Tognini-Bonelli (2011: 57) points out, “presence or absence is one thing, but frequency is another”, which prompts the understanding of making efforts on how to analyze language patterns in a more contextualized way, also taking into consideration how a given corpus is compiled.

When it comes to specialized fields, establishing a normalized terminology is one of the most relevant elements to ensure information is being disseminated as most accurately as possible. In this sense, in the field of aviation, a safety-critical environment, the broad use of corpora may pose very positive influence to best relate normative and descriptive use of language, especially in the Aeronautical Meteorology subfield, which even comprises possible other classifications of a term within the same field (Cf. PEIXOTO 2020b, in press).

By taking those assumptions into consideration, based on terminology (PAVEL; NOLET 2001; CABRÉ 1999 and 2003) and corpora (SANTOS 2008 and 2014; TOGNINI-BONELLI 2011) theoretical foundations, this paper aimed at discussing definitions and translation to Portuguese of expressions and terms contained in the Table 4678, concerning the main aeronautical meteorology codes, in order to highlight nuances of meaning which could interfere in proper communication of content.

This mentioned table is relevant because it is used as a reference in the Aeronautical Meteorology field, as prescribed by the World Meteorological Organization (2011). As detailed later in this paper, Table 4678 is organized in two parts ― qualifiers and the phenomenon itself ―, so as to combine different aspects of a given weather phenomenon, represented by letter codes, in addition to a positive or negative sign sometimes.

The proposed analysis is based on a systematization for terminological procedures, by using corpora with set validation standards, as the discussion on corpora has reached more sophisticated levels throughout the years, especially in terms of how computer tools may help the terminologist to process large amounts of specialized discourse.

However, this paper follows the approach that corpora must be used as a process of systematization, to help the researcher with weighing options regarding validation. It is not a strict process of quantification, whether established previously or later in the elaboration of a terminological database, but a process of evaluating how relevant the corpus is and, especially, how to set specific parameters for each intended use.

For that, as the compiled Aeronautical Meteorology corpora is still under development, the parameters for the validation process were applied “manually”, and are intended to contribute to avoid misunderstandings regarding the criticality level of meteorological situations being communicated during air traffic operations.

In 1972, Terminology was established as a field of study, following the ideas disseminated by Eugen Wüster, who idealized the General Theory of Terminology (GTT), according to which terms were considered to express concepts³ (and not meanings).

After this pioneering moment of Terminology studies, a functionalist perspective of language started to be more disseminated, comprising a more descriptive approach to the specialized lexicon. This tendency was also followed by the Canadian School (Cf. PAVEL; NOLET 2001) and by the Catalan María Teresa Cabré, who proposed the Communicative Theory of Terminology (CTT).

For Cabré (1999 and 2003), the meaning of the term is relative, as it depends on its communicative scenario⁴ and, within the scope of specialized communication, it may acquire terminological status. In other words, a same given word considered to belong to the general lexicon may also be considered a term, as long as, in a given specialized segment, it has a well-defined sense.

It is also interesting to note that, even in a same contextualized specialized segment, as aviation, where standardization is critically paramount, there are cases of terms which might have a different meaning based on localization criteria, i.e. different locations with language variation, as Brazil and Portugal⁵.

For example, in his research, Rabello (2011: 45) stated the difference in meaning for the term ‘arremeter’, in Portuguese:

As an analogy, in English, ‘arremeter’ may be equivalent to ‘go-around’ and ‘missed approach’, with some nuances: while ‘go around’ is generally an instruction by an air traffic controller and may also lead to another attempt to land, ‘missed approach’ is an aborted landing maneuver carried out due to technical issues.

These differences corroborate the fact that the process of translation is not an end in itself, but presupposes a situated activity, in which there will be correspondence of data, whether corpora or terms, by a specific culture or a given technical segment (Cf. KRIEGER; FINATTO 2004). The translation process is, therefore, a form of rewriting, in which ideological and/or sociological factors are also determined by the dissemination of the same idea in the target language, even in the case of specialized translation.

Nonetheless, as mentioned by Krieger and Finatto (2004), the ‘specialized language’⁷, as it started to be called in the 18th Century, used to be seen as in vitro, static, and started to be seen as in vivo, adaptable to the communicative moment. As a result, the traditional approach of designing policies and standards for specialized terminology led to the institution of the study objects, in Terminology, term, phraseology and definition.

For the analysis of such study objects, the most important ability for the terminologist is high-skilled research, and not full understanding of the field, as a specialist (THELEN 2015; TAGNIN 2013). In that, the expertise of the terminologist is key for the quality of this work, since (s)he is a trained professional for managing language resources in a more comprehensive way, to search, investigate and evaluate relevant sources, in any field; and the specialist is also a very resourceful partner to clarify specific issues. As stated by Thelen (2015: 373), "for translators it suffices that they know the language of a subject field, and that they know to find their way in the field; they need not be specialists". Taking this into account, the following topic addresses an expanded discussion regarding terminology objects.

Along with the CTT perspective (CABRÉ, 1999 and 2003), the term is considered to have a polyhedric structure, comprising the linguistic, cognitive and sociocultural nature of it. Phraseology refer to nominal and verbal syntagmas, as idiomatic expressions, sayings and proverbs; and the definition is a description based on presence or absence of semic traces, as opposing genus (G) and differentia (D)⁸. For example, an ‘airplane’ would be a means of transportation (G) designed to travel through air (D).

In this way, a series of semic traces may be listed, and the absence or presence of those semic traces will define the specialized sense. For example, in a sort of binary polarization, the same word ‘airplane’ may be classified regarding categories of means of transportation(+), ground(-), motorized(+), piloted(+ ou -)⁹, etc. This classification would be more objective when trying to automate the definition or term equivalence work, for use in computational applications (Cf. SAGER; L’HOMME 1994).

Within this scope, defining a term is crucial for the development of terminological and terminographical work. According to Krieger and Finatto (2004), a term is a complex unit “whose recognition is based on one of the hardest tasks of a terminological work.” ¹⁰ (2004: 76). For Pavel and Nolet (2001), a term may be

In a broader sense, Krieger and Finatto (2004: 77) consider terms to be lexical entities which may express concepts, objects or processes. In this way, as also defended by Cabré (1999 and 2003), such lexical entities could also be common words which may have undergone a terminologization process

In this way, it is a fact that the classification as a term is not necessarily derived from the identification of morphosyntactic components, once it is not a case of structural equivalence (Cf. TOGNINI-BONELLI, 2001), but definition of specialized senses, as in the case of ‘holding point’, directly translated to Portuguese as ‘ponto de espera’, and ‘grip tester’, adapted to Portuguese as ‘equipamento medidor de atrito’, both considered terms.

The compilation of term definitions for the elaboration of terminological products such as glossaries, dictionaries, databases and automated systems for terminology recognition is an essential task mainly for Terminography, an applied application of Terminology.

On this issue, as a support for the extraction of terms and building of terminological products, Electronic Lexicography has been standing out, as an applied area which deals with management of terms, whether or not specialized, in a computational environment. Such an area is currently one of the most fruitful environments to seek for methodological renovations, since it makes broader use of corpora and other technological resources which are very productive for the management of terminology¹², as discussed in the following topic.

The use of corpora and lexicographical tools has several applications both for linguistic and computational purposes. In the field of aviation, specifically, corpora have been more broadly used in this past decade, as the works by Tosqui-Lucks (2018), Prado (2018), Prado (2015), Rabello (2011), Gonçalves (2017), Lepri (2017), Prado and Tosqui-Lucks (2019), Prado and Tosqui-Lucks (2017) and Prado (2010) show.

There has been an exponential improvement of corpora resources since the first ones were developed. At the beginning, electronic corpora used to have about 1 million words, such as the one created by the Brown University in 1967, then computational advances allowed words to be added to online databases to a greater extent: the Birmingham University developed a corpus with 10 million words in 1987; the British National Corpus was created with 100 million words in 1994; and nowadays corpora commonly have a quantity of words of the order of billions (Cf. LEW 2009).

During that period, there have been changes in theoretical perspectives, as Tognini-Bonelli (2001) very well describes, from corpus-based approaches to corpus-driven approaches and, still today, new insights are triggered depending on more and more specialized target audiences, which started to comprise professional linguists, translation specialists and specialists from other fields.

Although these two approaches seem to be compared in a more dichotomic way, they are actually part of a continuum of linguistic possibilities of analysis, as mentioned in the introduction of this paper. In other words, the corpus-based approach was not limited in its conceptualization but had the purpose to make best use of corpora available at the time, which were much shorter. Therefore, it was not possible to define language patterns, since representativeness was a major issue: it was more common to use only text samples, not full texts. When computational tools became more available, language processing developed into so many possibilities, also based on statistical analysis, and it was possible to refine the so-called corpus driven methodology (Cf. TOGNINI-BONELLI, 2001; KILGARRIFF, 1997), and focus on systematization taking into consideration the compilation process (SANTOS 2008).

In this perspective, it was possible to relate corpora and tools in several scopes of activity, and potentialize the linguistic analysis in several study areas, among which translation. However, as mentioned by Anthony (2013), it is important to highlight that corpora are linguistic data that need suitable analytical tools. As different tools allow a differentiated analysis of the same data set¹³, proper utilization of corpus data is directly related to tools used and understanding how it works (Cf. HUNSTON 2002).

For that, it is of paramount importance to delimitate definitions for each element (word, etc), according to a given theoretical model, as divergence regarding some definitions could result in different statistical numbers (ANTHONY 2013), mostly due to frequency issues, as best explained in the following topic.

The validation process of terminological work is one of the most important aspects of such work, especially when considering sensitive areas, as in the case of aviation. In this way, the productivity of corpus research, when it comes to such a validation process of terminology data, is directly related to frequency issues.

At first, it is necessary to highlight that a major number of occurrences do not necessarily mean a major level of terminological expressiveness. Although the well-known statistical logic indicates that “big data is good data”, the occurrence (or not) in a corpus may not be considered an irrefutable evidence because the value or validity of a corpus is not restricted to the size of this corpus, but to the purpose and type of information which may be extracted from it.

As theorized by Zipf (SANTOS 2008; MANNING; SCHÜTZE 1999; LEW 2009),

sometimes a higher number of occurrences only represent connecting elements or other featureless words, intrinsically more common in a language, while a lower frequency may make evident a term with a relevant terminological semantic load. As a result, the absolute frequency is not necessarily the most important or the most productive; in fact, “Zipf's law is a power law stating that the frequency of any word is inversely proportional to its rank” (PUSTEJOVSKY; STUBBS 2013: 66). It is related to the occurrence of hapax legomena, id est the single occurrence of an ordinary word in a given corpus, which should be disregarded.

Santos (2008) agrees with this point of view and states that the representativeness of a corpus is the “information (...) on the relative weight of a given question regarding the represented universe” ¹⁴ (SANTOS 2008: 54). In a language, it is shown that several different cases (in theory, non-normative) always occur in a low frequency while few cases occur in a high frequency.

Under this scope, it is possible to understand that, in linguistics with corpus (SANTOS 2008 and 2014), systematization nuances are actually a criterium of analysis, and definitions of algorithms have to be clearly explained. An example of how this could pose obstacles to the productivity of a tool is the case of WebAnno, which, although pointing out the statistical method of kappa (CARLETTA 1996), does not distinguish some indicators in a clear way, making impossible the suitable interpretation of results obtained.

When corroborating this relation between linguistics and systematization enabled by suitable statistics, Santos (2014) mentions, according to the defended by Guiraud (1960), that linguistics is a statistical science par excellence (SANTOS 2014: 197), as it is not possible to dissociate quantitative and qualitative natures. As stated by Santos (2014):

Therefore, although linguistics with corpus is mostly qualitative, there are also empirical studies using corpora, which may be exploratory or experimental. In the first case, there is a collection of samples and some data are identified, such as “there are more verbs X than verbs Y”, while, in the second case, there is already a given hypothesis (“language favours the abstract expression”).

In the case of terminology work, as discussed by Thelen (2015) when comparing a theory-oriented terminology vs. a translation-oriented terminology, a theory-oriented approach will help the terminologist to define his/her strategy, instead of using only a translation-approach, mostly based on practice.

It is also important to have a high standard when compiling texts which may be relevant for a specific field, in order to organize it in a more specific or more balanced way, to avoid bias. This is not an easy task, not just because it requires strategical planning but also due to the fact more practical computational issues may arise when dealing with the compiled data, as discussed in the following item.

The methodology of this research was based on a systematization by using corpus, to set validation standards for the proposed analysis, considering both quantitative and qualitative parameters, i.e. corpus-based and corpus-driven premises for this intended objective (Cf. SANTOS 2008).

Generally speaking, the basic distinction between those two approaches relies on the theoretical perspective adopted: having a previous assumption to be confirmed by corpus evidence or having a freer analysis to find language patterns, i.e. create categories.

For that, annotation or tags serve different purposes, depending on the intended use. Since the corpus-driven approach has a more specific stance, it does not use annotation, as there could be bias for data analysis, contaminating them. In the case of a corpus-based approach, annotation or tags are considered necessary to test a given linguistic theory. On this, more modern tools provide layers of data in different formats, in a way annotation or tags may be deactivated at the researcher’s discretion, to show the original text again, for a freer analysis.

Within this context, corpus linguistics derived from the first approach of corpus-based research, considered a sort of tool, complementary to language studies, then a corpus-driven approach has established itself as a very robust field of study heavily relying on statistics.

Santos (2008) offers a more flexible standpoint, taking into consideration both perspectives of corpus-based and corpus-driven, as found necessary for a specific study object. This approach was named “Linguistics with Corpus” (Linguistics supported by corpora), and focused on the compilation process. As a result, Santos (2008) defended that “a corpus is a collection of classified linguistic objects to be used in the Natural Language Processing / Computational Linguistics / Linguistics fields” ¹⁶ (SANTOS 2008: 45), created for specific purposes, to assist in the linguistic study, and analysis of occurrence of words and syntactical structures, among other structures, which may be counted and categorized, but not in a very strict methodological pattern.

From that theoretical stance, this paper aimed at analyzing the weather phenomena specified in Table 4678 (WMO, 2011), a reference in the Aeronautical Meteorology field, by investigating meanings and nuances found in the compiled corpus, in order to discuss and validate the nomenclatures in Portuguese.

For that, publications from six categories were selected for the Aeronautical Meteorology (AER MET) corpus, taking into consideration AER MET publications are much scarcer than general aviation ones, especially in the case of Portuguese documents. As a result, it was possible to find many institutional publications in English, whereas in Portuguese it was necessary to search for more academic studies, in order to comprise a broader range for the AER MET field. Therefore, the balance of the corpus was a bit different for the English and Portuguese languages.

The corpus architecture of this research comprised the following categories: (a) terminological databases; (b) institutional publications; (c) commercial documents; (d) institutional didactic instructions; (e) academic publications; and (f) publicity documents, as shown in the following Table 1:

Table 1:

CAT	Description	Nr. PT	Nr. EN
terminological databases	glossaries and dictionaries (print, online or in .pdf)	5	7
institutional publications	regulations and technical reports	52	62
commercial documents	documents from companies selling meteorological instruments	0	75
institutional didactic instructions	training material published by institutional organizations and publications for proficiency purposes	1	4
academic publications	theses, dissertations and books	361	40
publicity documents	journals for general dissemination of information	0	1
TOTAL		419	189

Legend: CAT - category; Nr. PT - number of publications in Portuguese; Nr. EN - number of publications in English. Source: Author (2020)

After the compilation of .pdf files, they were converted into .txt, to enable better search, primarily using the AntConc suite. However, file conversion itself posed some issues as some documents, especially the academic ones, showed some OCR errors or could not be converted at all. In those cases, some different online converting tools were tested for each file and it was possible to eventually convert most publications¹⁷.

However, as the compilation of this Aeronautical Meteorology corpus is still ongoing, more data will be compiled and duly organized according to set validation standards proposed in this paper. For that, each category of the corpus architecture will have a different weight regarding its “level of authoritativeness”, depending on the purpose of the research. In this way, this Aeronautical Meteorology corpus intends to acknowledge a more comprehensive scope of specialized sources while emphasizing the higher relevance of a given category for a given purpose.

In the case of this research, although the aim is analyzing terminology used by official institutions, language used by the professional community in other institutions and companies were also taken into account, as they rely on official guidelines. For this paper, as the corpus is still under development, this validation standard was applied “manually”.

In order to provide better familiarization with the AER MET area and specificities of publications in this field, the following sections will provide some explanation regarding the main AER MET institutions, then the code nomenclatures will be discussed.

In terms of worldwide institutional guidelines, the Aeronautical Meteorology field follows standards and recommendations prescribed by the World Meteorological Organization (WMO) and the International Civil Aviation Organization (ICAO). In the case of the WMO, founded in 1919, it has a specialized group for Aeronautical Meteorology, named Commission for Aeronautical Meteorology (CAeM), which was formerly named Commission for the Application of Meteorology to Air Navigation (CAMAN). As explained in the Aeronautical Meteorology Programme Homepage, in the WMO website:

The formal interrelation between the ICAO, founded in 1947, and the WMO¹⁸ was first introduced by Doc 7475, edited in 1963, where it is defined as a “formal interagency agreement” (ICAO 1963, Foreword), with working arrangements approved in 1953 by both organizations, to be effective as of 1 January 1954.

The purpose of this arrangement was establishing a delimited scope of work for each organization, in a way to employ resources more effectively, as stated in Doc 7475 (ICAO 1963):

Along with this purpose of interaction, it was established the ICAO would be in charge of “defining and encouraging the provision of services” (ICAO 1963: 3) whereas the WMO would be responsible for “specifying the technical methods and practices recommended for use in providing required meteorological services” (ibidem), as well as promoting scientific research in the Aeronautical

Meteorology field. However, this classification of activities would not necessarily be sharp, as both organizations could actually work jointly to solve an issue of shared interest, and also resort to other institutions which might help.

As a result, States currently follow WMO standards for methodological actions, including use of terminology, whereas the ICAO is the reference institution for recommended services, in alignment with other countries. It is important to emphasize, though, that the rules prescribed by the ICAO are classified as standard and recommended practices and procedures, where 'standard' refers to actions which shall be taken by States and 'recommended' refers to actions which should be taken by States.

In Brazil, the Department of Airspace Control (DECEA) is the official regulating agency for the Aeronautical Meteorology field and it issues publications according to actions prescribed by the ICAO, along with the WMO standardized methodology. More recently, precisely on March 26, 2020, a new center named Integrated Center of Aeronautical Meteorology (CIMAER) was inaugurated at Galeao Airport (SBGL), with the purpose of centralizing all Air Force units working in this field. With this new configuration, it is important to make efforts to try to ease Aeronautical Meteorology communication, shedding light on specific differences in key terms regarding weather phenomena, especially the ones used in the Aeronautical Meteorology Codes Table, as discussed in the next topic.

As stated before, the field of Aeronautical Meteorology is very specific and does not have as many publications available as general aviation documents. Within this context, it is important to highlight that Aeronautical Meteorology terms may be used in a different way when it comes to the general use of terms in the field and the codes themselves, as prescribed by the WMO, in English; and in the DECEA publications, in Portuguese.

The detailed way those Codes are to be used in Aeronautical Meteorology is explained in Doc 782, edited in 2019, a user's guide on how codes should be used, in which situations and how the format should be. Other publications, both from the ICAO and the WMO, also show complementary information on those procedural guidelines.

To allow better visualization, information from tables 4677 and 4678 from the WMO (2011) and Brazil (2017a) were organized in Figure 1 and Figure 2 in this paper, in order to focus on the codes and the nomenclatures, respectively.

Figure 1: Source: Adapted from WMO (2011)

The nomenclature of each code shown in Figure 1 is specified in Figure 2, in English and in Portuguese, as follows:

Figure 2: Source: Adapted from WMO (2011) and BRAZIL (2017a).

As the meaning of terms regarding weather phenomena is very specific, it is important to analyze its use in order to best understand nuances and avoid miscommunication when reporting a weather issue.

Some of the major aspects of Aeronautical Meteorology nomenclatures are the nuances regarding fog, mist and haze, and the precipitation scale to define intensity of weather phenomena. As those aspects pose a real impact on visibility, it is deemed very hazardous to aircraft operations (Cf. GULTEPE 2009).

Regarding ‘fog’ (FG), ‘haze’ (HZ) and ‘mist’ (BR), the classification depends on humidity and visibility issues. ‘Fog’ is reported when the air is at about 100 per cent humidity and the visibility is less than 1000 m (Cf. ICAO 2005), while ‘mist’ presents visibility ranging from 1000 m and 5000 m, and relative humidity above 90 per cent. (Cf. ICAO 2005). On the other side, ‘haze’ are “extremely small particles invisible to the naked eye and sufficiently numerous to give the air an opalescent appearance [...], usually only a few thousand feet thick, but may extend upwards to 15,000 feet (4,600 meters) [...]” and visibility may vary “greatly, depending on whether the pilot is facing into or away from the sun” (FAA 2016: 16-5). Concerning those terms, although “mist may be considered an intermediate between fog and haze” (ibidem), identifying those phenomena may be critical, as “there is no distinct line between any of these categories” (ibidem). In Portuguese, fog (FG), haze (HZ) and mist (BR) are translated as ‘nevoeiro’, ‘névoa seca’ and ‘névoa úmida’, respectively.

The occurrence of fog, haze and mist may be full or localized (in patches or bank), especially regarding ‘fog’. While the nomenclature specifies patch(es) (BC) as the descriptor, to be used in a structure like ‘patch(es) of ’, this word is used in the AER MET literature in a more general sense, with other sorts of combinations, such as ‘patches of blue sky’, ‘patches of cloud’, ‘patches of convection’, ‘patches of light’, ‘patches of anomalous structure’, ‘patches of dry snow’, ‘patches of foam’, ‘patches of greater or lesser development [such as clusters of apartments, shops, factories or parks, open areas or water]’, patch(es) of ground, and ‘patches of heavy rain’.

As qualifiers used before the noun ‘patch(es)’, words such as ‘snow’, ‘ice’, ‘wet’, ‘water, ‘grease’ are largely used, as well as ‘great’, ‘large’, ‘small’ and ‘circular’. In addition to that, ‘patch’ itself in a qualifier form (‘patchy’) was also found in the corpus, with a varied range of collocates, as ‘patchy band of convective cloud’, ‘patchy cloud’, ‘patchy fog’, ‘patchy grey’, ‘patchy ground’, ‘patchy ice’, and ‘patchy troposphere’, as illustrated in Figure 3 below:

Figure 3: Source: Author (2020)

Concerning ‘precipitation’, it may occur in a sort of uniform way (intermittent or continuous) or suddenly (as showers). While the nomenclature mostly uses ‘shower’ when mentioning the strong and short duration of a weather phenomenon, generally associated to convective clouds (Cf. ICAO 2005), there is a grading scale for the characteristics and types of precipitation. As described in United States (2003: 2-14),

Precipitation can take the form of drizzle; freezing drizzle; rain; freezing rain; hail or graupel; ice pellets or sleet; snow; or snow grains (Cf. UNITED STATES 2003: 2-14).

In the case of ‘shower’, specifically, the AER MET nomenclature may be composed of the terms and expressions ‘shower’, ‘showery precipitation’ or ‘showers in the vicinity’. On this, according to WMO (2008: II.4-20), 'showers'

Then,

In accordance with WMO (2008), 'shower' is strongly associated with hail, as "hail always occurs in the forms of showers and is generally observed during heavy thunderstorms" (WMO, 2008: II.4-21] and never associated to 'snow grains', as "they [snow grains] usually fall in small quantities, mostly from stratus or from fog and never in the form of a shower" (Cf. WMO 2008: II.4-21).

The term ‘unidentified precipitation’ is particularly interesting because there is no translation to Portuguese in ICA 105-16 guidelines (BRAZIL, 2017a). The ICAO classifies ‘unidentified precipitation’ as cases “where intensities are very low (< 0.1 mm/h), [and] precipitation type is not well identified” (ICAO 2011: 6-4). In such cases, “the code ‘unidentified precipitation (UP)’ is often used and is preferable to an identification error” (ibidem). As corroborating this, the Annex 3, by ICAO (2018), recommends that

In Portuguese, the equivalents for ‘shower’ or ‘showery precipitation’ would be ‘pancada de ’, while ‘precipitation’ is translated as ‘precipitação’. It is important to note that there could be a slight difference in terms of intensity between ‘shower’ and ‘showery precipitation’ but this is not transferred to the nomenclature in Portuguese as it could sound less natural.

Another interesting point to mention is the difference between ‘hail’ (GR) and ‘graupel’, not contained in Table 4678, which seems to make ‘graupel’ equivalent to ‘small hail and/or snow pellets’ (GS), in a more simplified way.

‘Hail’ and ‘graupel’, translated to Portuguese, within a general AER MET context, as ‘granizo’ and ‘graupel’ respectively, appear in the academic literature in Portuguese, where 'graupel' is considered to be a prior stage of hail formation (Cf. MEDINA 2015) or light hail (Cf. HALLAK 2007), and ‘hail’ is also considered a high intensity 'graupel' (Cf. RAMIREZ 2018); and in the corpus in English, where ‘graupel’ occurs very few times, mostly in academic and institutional training references, and a WMO publicity document, not intended to be used as a guideline.

The term ‘hail’ is strongly associated with ‘thunderstorm’, to the extent the ICAO (2016) defines that thunderstorm without hail should be classified as ‘thunderstorm’ only, and thunderstorm with hail should be referenced as ‘thunderstorm with hail’.

The term ‘thunderstorm’ is translated in the Portuguese nomenclature as ‘trovoada’, in ICA 105-17 (BRAZIL 2017b: 114), explained in the following way: “a thunderstorm is a succession of electric discharges and thunders, generally accompanied by precipitation and always associated with CB clouds [cumulonimbus clouds]”. ¹⁹ Thunderstorms are considered a less complex occurrence of ‘squall lines’, which are defined as “a line of violent thunderstorms” (UNITED STATES 2008: 3-14).

The definitions of Table 4678 nomenclatures highlighted here intend to shed light on some critical terms which may impact the reporting of weather phenomena in the Aeronautical Meteorological (AER MET) field, due to the fact specialized terms certainly have different uses in a daily situation and, in the case of AER MET, even in different contexts of this same field.

As discussed in this paper, having a more sophisticated use of corpora may help to make best use of resources available, so this paper aimed at presenting a research methodology with set validation standards and debating nomenclatures of Aeronautical Meteorology codes.

By assuming categorization is key for an in-depth analysis, a systematization was carried out, based on linguistics with corpus assumptions (Cf. SANTOS 2008), aimed to find patterns of use in the compiled corpora, but not using a very structured statistical analysis at the moment.

Then, relying on a terminological theoretical foundation aligned with Pavel and Nolet (2001) and Cabré (1999 and 2003), this research considered a more functional approach to terminology, based on context of use and categories of documents compiled in the corpus.

In this way, the proposed corpus architecture systematized the data in six categories and used set validation standards, applied “manually”, to account for the “level of authoritativeness” of each category of sources for the intended research.

This was particularly relevant due to the fact the Aeronautical Meteorology field is very specific and does not have many publications available, especially in Portuguese. So, a broader scope of sources had to be considered for this research, aimed at analyzing terminology used by official institutions, also considering language used by the professional community in other institutions and companies, which also rely on official guidelines.

By using this methodology, this paper intended to draw attention to specificities of some Aeronautical Meteorology terms, particularly on nuances regarding fog, mist and haze, and the precipitation scale to define intensity of weather phenomena. However, as the scope of this research was limited to terminology used in Table 4678, discussions were not intended to be exhaustive but only provide better understanding on the scope of use in the field, and contribute to avoid misunderstandings in communication operations.

In addition to that, this paper emphasizes the understanding that terminology work should be based on a theory-oriented approach to define the best strategy of compilation and analysis, and not just a translation-approach, mostly based on practice (THELEN 2015). Along with this perspective, having a high standard when compiling specialized texts and setting validation standards for a customized specialized corpus will enable proper systematization (SANTOS 2008 and 2014), by means of a scientific strategic perspective, not just a practical one.