The initial data set was composed of plot observations stored in the database “sudamerica” (https://www.givd.info/ID/SA-CL-001, Alvarez et al. 2012). This database already includes assignments of plot observations to phytosociological associations. While the first version of this database focused on grassland vegetation in Chile, its geographic and syntaxonomic focus have now been expanded to all kind of vegetation across the South American continent. The initial collection of references started by searching the phrases “vegetation survey”, “phytosociological classification”, and “syntaxonomy” in combination with the term “Chile” in electronic libraries (e.g. ScienceDirect, Wiley, Springer and Scielo) as well as in Google Scholar. Those searches were done in Spanish, English, and German, which are, according to our experience, the main languages used to publish vegetation surveys done in Chile. After the assessment of each publication, we also screened the respective reference lists and repeated a search using the names of handled syntaxa, leading to additional references, especially from the gray literature. All references including relevés were considered.

In a second step, we reviewed the literature for additional vegetation surveys meeting the following two criteria: 1) abundance of plant species in plots was recorded, 2) plot observations were assigned to associations.

Only header information was considered, avoiding the time-demanding digitizing of abundance tables and further assessment of species nomenclature and synonymy. The header data of each plot observation, which is usually a table column in the original source, include page, table and column numbers, data source (bibliographic reference), assignment to a plant community, elevation and plot coordinates. If not provided by the source, coordinates of plot observations were inferred from information of study sites using geonames (http://www.geonames.org/) and indications of directions and roads by locating them in Google Maps (https://www.google.com/maps).

For further bioclimatic descriptions, we also collected information on annual temperature and total annual precipitation from CHELSA version 1.2 (Karger et al. 2017).

Our syntaxonomic assessment used, as an initial stage, the classification of Oberdorfer (1960). This classification was expanded and adjusted as we assessed more recent publications (Suppl. material 1). Each plot record was inserted in the database by its original syntaxonomic assignment and was linked to a syntaxonomic concept using the relational model proposed by Alvarez and Luebert (2018). For the delimitation of higher syntaxonomic ranks we accepted the newest proposals as syntaxonomic references (see Suppl. material 2).

We respected all syntaxonomic ranks mentioned in the consulted references (principal and secondary ranks according to Theurillat et al. 2021). Relevés where the principal rank order or alliance was not assigned, were marked with question marks (i.e., “order?” and “alliance?”, respectively).

Plot observations were assigned to plant formations according to the EcoVeg classification (Faber-Langendoen et al. 2014). For this we used the keys and definitions by Faber-Langendoen et al. (2016). We also considered the additions done by Luebert and Pliscoff (2022).

For the comparison of formations and phytosociological units, we produced Sankey diagrams, weighting the assessed units by their relative number of phytosociological associations and linking Braun-Blanquet classes and alliances with EcoVeg formations.

Data management, statistical assessment and plotting was done in R version 4.1.0 (https://cran.r-project.org/). To build and adjust syntaxonomic classifications, we used the R-package “taxlist” version 2.2.3 (https://docs.ropensci.org/taxlist/, Alvarez and Luebert 2018). For linking plot observations with the syntaxonomy, we used the R-package “vegtable” version 0.1.8 (https://github.com/kamapu/vegtable). Data sources and syntaxonomic views (references for syntaxonomic concepts) were formatted by using the R-package “biblio” version 0.0.7 (https://github.com/kamapu/biblio). To draw Sankey diagrams we used the R-package “networkD3” version 0.4 (https://github.com/christophergandrud/networkD3). The final data set was stored in a PostgreSQL database.

The compiled database contains 1,776 plot observations extracted from 37 publications (see also Suppl. material 1). Most of the publications (34) are journal articles, while the rest are a book (Oberdorfer 1960), a book chapter (Roig et al. 1985) and a thesis (Ackermann 2001). The most frequent journals in this list are Agro Sur (6 articles), edited at the Universidad Austral de Chile in Valdivia, and the International Journal of Geobotanical Research (5 articles), edited in Spain. The year of publication spans from 1954 (Schmithüsen 1954) until 2019 (Amigo et al. 2009; Amigo et al. 2019).

The collected plot observations account for a total of 29 classes, 43 orders, 66 alliances, and 175 associations (full list included as Suppl. material 2). The most diverse classes in terms of number of associations are Wintero-Nothofagetea (31 associations), Lithraeo causticae-Cryptocaryetea albae (15 associations), and Myrteolo nummulariae-Sphagnetea magellanici (14 associations). While the distribution of plot observations is very clustered along the latitudinal gradient, they cover a large proportion of the country from the northernmost to the southernmost extremes (Figure 1). Nevertheless, huge gaps are observed from the 30°S northwards and from the 45°S southwards. The classes with the broadest latitudinal distribution are Lemnetea minoris, Wintero-Nothofagetea, and Nothofagetea pumilionis-antarcticae.

Figure 1. 

Latitudinal distribution of recorded phytosociological classes. In all panels the positions of vegetation plots are overlaid as red dots. The bars of the two middle panels correspond to mean annual temperature and total annual precipitation values extracted from CHELSA version 1.2 (Karger et al. 2017) for all pixels included in the respective latitudinal segments. The right panel shows the latitudinal distribution of recorded plot observations for each class. All observations belonging to the same class are connected by a vertical line. Classes are numbered by their latitudinal distribution: 1 - Polylepidetea tarapacano-besseri, 2 - Parastrephio lepidophyllae-Fabianetea densae, 3 - Anthochloo lepidulae-Dielsiochloetea floribundae, 4 - Plantagini rigidae-Distichietea muscoidis, 5 - Opuntietea sphaericae, 6 - Lemnetea minoris, 7 - Ambrosietea chamissonis, 8 - Gutierrezio paniculatae-Trichoceretea chilensis, 9 - Ephedro chilensis-Chuquiragetea oppositifoliae, 10 - Mayteno boariae-Salicetea humboldtianae, 11 - Helianthemetea guttati, 12 - Tessario integrifoliae-Baccharidetea salicifoliae, 13 - Lithraeo causticae-Cryptocaryetea albae, 14 - Stellarietea mediae, 15 - Molinio caerulae-Arrhenatheretea elatioris, 16 - Plantaginetea majoris, 17 - Phragmito-Magnocaricetea, 18 - Potametea, 19 - Littorelletea australis, 20 - Bidentetea tripartiti, 21 - Senecionetea chilensis, 22 - Loasetea, 23 - Nanojuncetea australis, 24 - Empetro rubrum-Pernettyetea, 25 - Wintero-Nothofagetea, 26 - Nothofagetea pumilionis-antarcticae, 27 - Myrteolo nummulariae-Sphagnetea magellanici, 28 - Rostkovietea magellanicae, and 29 - Deschampsio-Asteretea vahlii.

Vegetation plots recorded in Chile were assigned to 7 formation classes including 10 subclasses and 20 formations. From 29 phytosociological classes, 11 are associated to more than one formation, with Wintero-Nothofagetea linked to 5 formations, Phragmito-Magnocaricetea to 4, Parastrephio lepidophyllae-Fabianetea densae and Potametea to 3, and other 7 classes to 2 formations. On the other hand, 12 out of 20 formations are linked to more than one phytosociological class (Figure 2).

Figure 2. 

Correspondence between phytosociological classes and EcoVeg formations. The height of columns represents the relative number of associations included in each unit. The first and second columns from the left are phytosociological classes and alliances. The third column corresponds to formations. Green bars show phytosociological classes linked to more than one formation. Orange bars show alliances linked to more than one formation. Blue bars show formations linked to more than one phytosociological class.

Out of 66 recorded alliances, 19 are associated to more than one formation. Wintero-Nothofagetea is the phytosociological class including the highest number of alliances, which is 10, whereas the rest of the classes include 4 or less alliances.

The formation classes, sorted by decreasing number of assigned associations, were Shrub & Herb Vegetation class (66 associations), Forest & Woodland class (50 associations), Polar & High Montane Scrub, Grassland & Barrens class (16 associations), Agricultural & Developed Vegetation class (15 associations), Aquatic Vegetation class (11 associations), Desert & Semi-Desert class (9 associations), and Open Rock Vegetation class (1 association). The most important formations in term of assigned associations were Cool Temperate Forest & Woodland formation (33 associations), Temperate to Polar Bog & Fen formation (17 associations), Fallow Field & Weed Vegetation formation (15 associations), Tropical High Montane Scrub & Grassland formation (15 associations), and Temperate Grassland & Shrubland formation (14 associations).

In spite of the lack of a plot-based summary of the vegetation syntaxonomy in Chile, we found a surprisingly high number of recorded syntaxa. While we may focus on acquiring resources to continue digitizing data (see Suppl. material 1 for some references still in the queue), discovering further data sources may further increase the vegetation diversity from the perspective of the Braun-Blanquet approach. This is also expected in consideration of the strong bioclimatic gradients of the country and the huge data gaps from 30°S northwards and 45°S southwards (Figure 1). The database presented here may be an important complement to classifications applying expert systems (Bruelheide et al. 2021) and vegetation maps (Luebert and Pliscoff 2022). We also show the usefulness of this database for comparing different classification approaches based on plot observations.

Our assessment focused on principal syntaxomic ranks, namely class, order, alliance, and association (Theurillat et al. 2021), but secondary ranks are also distinctly applied in this scheme. On the other hand, there are gaps in the classification, where orders or alliances are not yet defined, as in the case of Bidentetea tripartiti, Helianthemetea guttati, Littorelletea australis, Phragmito-Magnocaricetea, Senecionetea chilensis, Stellarietea mediae, and Tessario integrifoliae-Baccharidetea salicifoliae (Suppl. material 2).

Some classes are cosmopolitan in their native distribution, especially aquatic and semi-aquatic vegetation as in the case of Lemnetea minoris, Phragmito-Magnocaricetea, and Potametea. These classes are defined by their physiognomy and ecology and have few species in common (Landucci et al. 2015).

Other classes collate associations dominated by introduced plant species and therefore of anthropogenic origin, as in the case of the classes Molinio caerulae-Arrhenatheretea elatioris, Plantaginetea majoris and Stellarietea mediae, which represent grassland, ruderal and weed vegetation, respectively. Most of the associations assigned to these communities are characterized by a mixture of native and introduced plant species, whereby the allochthonous elements use to be dominant. For instance, Oberdorfer (1960) classified the Chilean temperate grasslands into an own alliance called Agrostion chilensis, which was later assigned to an own order called Agrostietalia capillaris (San Martín et al. 1993). Oberdorfer (1960) argued for the benefits of this classification because of the occurrence of characteristic South American species as well as the generally species poor communities in comparison with their European counterparts. The classification of such communities is not yet aligned to the current state of the European syntaxonomy (compare with Mucina et al. 2016), nor assessed by a jointly review to date.

Conversely, some classes are endemic to the continent or even smaller areas. For instance, Wintero-Nothofagetea and Nothofagetea pumilionis-antarcticae represent South American temperate forests and woodlands.

Oberdorfer (1960) proposed some names for native vegetation corresponding to their ecological counterparts from the Northern Hemisphere. For instance Littorelletea australis corresponds to the southern hemisphere counterpart of Littorelletea uniflorae, as well as Nanojuncetea australis with respect to Isoëto-Nanojuncetea (see also Amigo 2009).

Oberdorfer (1960) established a backbone syntaxonomy for the transition area between the Mediterranean and temperate zones in central Chile. Additional publications complemented the geographical coverage for the Braun-Blanquet approach. Here we highlight the project “Transecta Botánica de la Patagonia Austral” (Roig et al. 1985) for the extreme South and the works by Ruthsatz (1995) and Luebert and Gajardo (2000, 2005) for the extreme north of the country.

Our analysis also highlights the lack of phytosociological work in north-central Chile between 25° and 32°S, where only one azonal class (Lemnetea minoris) was recorded (see Figure 1). While this gap may be partially filled when including further references (see Suppl. material 1), this zone has already been acknowledged as one of the least known in Chile from the point of view of its vegetation (Luebert and Pliscoff 2017), suggesting it as a priority for further field studies. Another big gap is observed from ca. 45°S southwards, which may be caused by the low population density in these areas, resulting in few research institutions or vegetation ecologists residing and researching in these areas, as well as due to the low accessibility, rugged topography, and harsh weather conditions. As consequence, less than a 50% of the country’s surface is covered by the current database, advising for more efforts on new, complementary vegetation surveys.

Plot observations are associated with 20 formations belonging to 10 subclasses and 7 classes from the EcoVeg approach. Considerable disagreement is observed between formations and phytosociological classes (Figure 2). The main reasons for this mismatch is probably the physiognomic heterogeneity of plant communities belonging to the same phytosociological class. For instance, the class Wintero-Nothofagetea, which mainly corresponds to the Cool Temperate Forest & Woodland formation, currently also includes swamp forests (order Myrceugenietalia exsuccae), which is azonal vegetation belonging to the Temperate Flooded & Swamp Forest formation (Figure 2). Excluding swamp forests from Wintero-Nothofagetea into a new class can be a solution to this problem. In this context, the implementation of the EcoVeg approach can be very useful for enhancing the consistency in the classification of Chilean vegetation as complementary to the Braun-Blanquet approach. Such kind of comparison have been done by Willner and Faber-Langendoen (2021) for phytosociological units in Europe. They also observed discrepancies between EcoVeg formation and phytosociological classes and proposed some adjustments to solve some of those issues.

As a consequence of the bioclimatic diversity of the country, plant formations observed in Chile are very diverse, including formations from all 7 classes proposed by Faber-Langendoen et al. (2016). However, part of this diversity is due to the fact that syntaxa also capture intrazonal and azonal vegetation, thus contrasting with the lower diversity of zonal vegetation formations identified for Chile by Luebert and Pliscoff (2022).

The database presented here also covers other countries in South America and can be accessed at https://syntax.kamapu.net. This database can be potentially used to support the consolidation of a syntaxonomic classification at the national or continental level, providing access to floristic tables for meta-analyses, and for implementing the development of expert systems.

The database can also be used to compare different plot-based classifications, as shown here. Further comparisons may include the mid and lower levels of the EcoVeg approach (Faber-Langendoen et al. 2014) or the classification of plant formations by Ellenberg and Mueller-Dombois (1966). In Chile, this database may be used to complement intrazonal units to the proposal of zonal vegetation formations by Luebert and Pliscoff (2022) and to incorporate EcoVeg units at the lower levels of alliance and association, not covered by Luebert and Pliscoff (2022).