Research Paper |
Corresponding author: Elvira Casagranda ( elvira.casagranda@gmail.com ) Academic editor: Gwendolyn Peyre
© 2023 Elvira Casagranda, Andrea E. Izquierdo.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Casagranda E, Izquierdo AE (2023) Modeling the potential distribution of floristic assemblages of high Andean wetlands dominated by Juncaceae and Cyperaceae in the Argentine Puna. Vegetation Classification and Survey 4: 47-58. https://doi.org/10.3897/VCS.95779
|
Aims: The aim of this work was to model the distribution of suitable environmental conditions of vegas with specific floristic characteristics. Vegas are high Andean wetlands that represent the main sequestered carbon stocks, biodiversity hotspots, and water regulating systems in the region. In these wetlands, plant communities are the main biological factor that determines functional processes, and plant species assemblages are associated with different ecogeographic features. Study area: Argentine Central Andean Puna ecoregion. Methods: For two different floristic assemblages of vegas, we develop ecological niche models of n-dimensional minimum volume ellipsoids through NicheToolBox, then obtain potential distribution maps. One floristic assemblage was dominated by the cushion-structured plant Oxychloe andina (Juncaceae) and the other by plants of the Cyperaceae family. Results: Elevation and precipitation were the main environmental factors determining the distribution of the two floristic assemblages. Juncaceae dominated vegas tend to be located in high, humid, and cold places, while Cyperaceae vegas are found at a lower elevation, with less humidity, and higher temperatures. According to the dominant climatic gradient in the region, potential distribution maps show that vegas of Juncaceae are commonly found towards the Northeast of the Puna while Cyperaceae vegas are more frequent at lower elevations to the South of the region. Conclusions: This study represents the first approach to niche modeling based on plant communities in vegas of the Argentine Puna, providing knowledge on the environmental factors that limit their distribution. This information could serve as a planning tool in a region exposed to growing perturbations such as mining and climate change.
Taxonomic reference:
Abbreviations: AUC = Area Under the ROC Curve; NDVI = Normalized Difference Vegetation Index; ROC = Receiver Operating Characteristic.
Central Puna ecoregion, Cyperaceae, floristic assemblages, Juncaceae, niche models, plant communities, vegas, vegetation distribution
Vegas (also known as bofedales; see
In these wetlands, plant species dominance gradually changes north-south throughout their distribution, and is determined by ecological factors and the different histories of colonization of the species (
Several studies have shown that the distribution of the vega forming species seems to respond to changes in environmental conditions, indicating that they present a high environmental specificity, as is the case of the Andean Oxychloe andina cushion plant, which is commonly found in high, wet and cold vegas (
Ecological niche modeling is useful in estimating the environmental requirements of the species, which can be projected geographically to help identify areas suitable for their survival (
The present contribution provides an initial model that predicts the environmental variables that determine the ecological niche of different floristic assemblages of vegas at regional scale. For this, we: a) model the potential niches of two floristic assemblages of vegas, and b) analyze their potential spatial and ecological distribution into the study area. This approach provides information about environmental requirements of ecological engineer species and allows us to identify potential sites throughout the region with the environmental conditions that these vegas need to exist.
The study area includes the Argentine sector from the Central Andean Dry Puna, Central Andean Puna and part of Southern Andean Steppe of the terrestrial ecoregions of the world (
For this study we used the classification of vegas that was obtained by
For niche modeling we built two groups using data belonging to three of these five floristic assemblages, covering a total of 41 vegas (Fig.
On the left side the map of the studied area is shown with the location of the vegas classified by
Vega groups classified by
Juncaceae vegas | Cyperaceae vegas | ||||
---|---|---|---|---|---|
Group 1 species | Mean cover (%) | Group 2 species | Mean cover (%) | Group 3 species | Mean cover (%) |
Oxychloe andina (Juncaceae) | 27.52 | Oxychloe andina (Juncaceae) | 34.48 | Eleocharis pseudoalbibracteata (Cyperaceae) | 14.19 |
Festuca nardifolia (Poaceae) | 9.30 | Zameioscirpus atacamensis (Cyperaceae) | 8.16 | Zameioscirpus atacamensis (Cyperaceae) | 9.62 |
Deyeuxia hackelli (Poaceae) | 6.58 | Triglochin concinna (Juncaginaceae) | 3.99 | Juncus balticus (Juncaceae) | 7.43 |
Distichia muscoides (Juncaceae) | 5.21 | Deyeuxia eminens (Poaceae) | 3.73 | Eleocharis atacamensis (Cyperaceae) | 4.90 |
Trichophorum rigidum (Cyperaceae) | 5.07 | Eleocharis pseudoalbibracteata (Cyperaceae) | 3.40 | Triglochin concinna (Juncaginaceae) | 4.38 |
Zameioscirpus muticus (Cyperaceae) | 3.44 | Eleocharis atacamensis (Cyperaceae) | 3.10 | Oxychloe andina (Juncaceae) | 4.30 |
Deyeuxia curvula (Poaceae) | 2.67 | Deyeuxia curvula (Poaceae) | 2.13 | Festuca argentinensis (Poaceae) | 4.17 |
Phylloscirpus deserticola (Cyperaceae) | 2.36 | Festuca argentinensis (Poaceae) | 1.83 | Lobelia oligophylla (Campanulaceae) | 3.96 |
Zameioscirpus atacamensis (Cyperaceae) | 1.79 | Juncus stipulatus (Juncaceae) | 1.62 | Phylloscirpus acaulis (Cyperaceae) | 2.36 |
Rockausenia pygmaea (Asteraceae) | 1.51 | Calandrinia acaulis (Montiaceae) | 1.39 | Carex macrorrhiza (Cyperaceae) | 2.16 |
We modeled the ecological niche of the floristic assemblages using the biotic-abiotic mobility (BAM) theoretical approach (
Despite the importance of biotic interactions in determining the distribution of species and species assemblages at regional, continental and global scales (
Variables used for niche modeling. WorldClim 2.0 bioclimatic variables are derived from temperature and precipitation data for the period 1970–2000.
BIO1: mean annual temperature | BIO11: average temperature of the coldest quarter |
BIO2: daytime temperature range | BIO12: annual precipitation |
BIO3: isothermality (BIO 2/ BIO 7)*100 | BIO13: precipitation of the rainiest month |
BIO4: seasonality of temperature (σ*100) | BIO14: precipitation of the driest month |
BIO5: maximum temperature of the warmest month | BIO15: seasonality of precipitation |
BIO6: minimum temperature of the coldest month | BIO16: precipitation of the rainiest quarter |
BIO7: annual temperature range (BIO 5 - BIO 6) | BIO17: precipitation of the driest quarter |
BIO8: average temperature of the rainiest quarter | BIO18: precipitation of the warmest quarter |
BIO9: average temperature of the driest quarter | BIO19: precipitation of the coldest quarter |
BIO10: average temperature of the warmest quarter | Altitude |
To perform the niche models we used the NicheToolBox package (
The presence points we used correspond to the coordinates of the center of each vega in each of the groups, obtaining two sets of occurrence data, one for the Juncaceae (17 points) and another for the Cyperaceae group (24 points). Each occurrence dataset was randomly divided into training and test data in a 70:30 ratio, and environmental information was extracted for both training and test points. For each set the correlations between the environmental variables were estimated in order to identify the variables that were least correlated to each other to avoid redundant information. With these variables the models were fitted, specifying the number of them to be used (i.e. the number of dimensions in which the ellipsoid models will be constructed). Other important parameters that were defined before running the models are the proportion of training points that will be used to fit the model (
Once the best model was chosen for each vegas group, minimum volume ellipsoid and the map of potential habitat suitability in the geographic space (or potential distribution map) were obtained. Habitat suitability varies between 0 and 1, with 1 indicating that the cell contains environmental conditions more similar to the sites where the vegas are present. The suitability map was clipped with the vega layer of the studied area, and then a cut-off threshold was applied to conserve all those vegas that presented suitability ≥ 0.1, considering that beyond this threshold, there are already suitable conditions for a vega to be present. From the maps obtained after applying the threshold, for each group, the number of vegas with suitability ≥ 0.1, surface area, average, minimum and maximum altitude were obtained.
Data on the species that compose the plant communities of the studied vegas at this study are openly available in GBIF API at https://api.gbif.org/v1/, reference number https://doi.org/10.15468/gdkn99.
The environmental variables used to fit the models (i.e. the least correlated with each other) in Juncaceae vegas were: altitude, mean annual temperature (BIO1), diurnal temperature range (BIO2), isothermality (BIO3), minimum temperature of the coldest month (BIO6), annual precipitation (BIO12), precipitation of the driest month (BIO14) and seasonality of precipitation (BIO15). In Cyperaceae vegas, the variables used were: altitude, mean annual temperature (BIO1), isothermality (BIO3), seasonality of temperature (BIO4), annual temperature range (BIO7), annual precipitation (BIO12), precipitation of the rainiest month (BIO13) and seasonality of precipitation (BIO15). For both groups, ellipsoid models were constructed in three, five and seven dimensions using all possible combinations of the mentioned least correlated variables, generating 120 models in both cases. The best niche model for Juncaceae vegas obtained was constructed with three variables: mean annual temperature (BIO1), diurnal temperature range (BIO2) and annual precipitation (BIO12), whose values at the centroid were 5.3°C (BIO1), 18.5°C (BIO2) and 108 mm (BIO12). For Cyperaceae vegas, the best model was built with five variables: altitude, mean annual temperature (BIO1), seasonality of temperature (BIO4), annual temperature range (BIO7) and seasonality of precipitation (BIO15); the centroid of this model was located at 3,762 m a.s.l. (altitude), 7.2°C (BIO1), 3.2°C (BIO4/100), 23.4°C (BIO7) and 97.4 mm (BIO15). Of the 120 models obtained in each group, two of the models generated for Juncaceae vegas exceeded the omission rate criterion of less than 6% for the training data, while none of the models exceeded this criterion for the test data, so the selection criterion was to choose the one with the highest AUC among the two models that exceeded the omission rate for training data. In the case of Cyperaceae vegas, 51 of the 120 models passed the omission criterion for both training and test data. The parameters of the best selected models are shown in Table
Best minimum volume ellipsoid models for Juncaceae and Cyperaceae groups using the model selection and calibration protocol in NicheToolBox. For each group, the model that had omission rates ≤ 0.06, significant partial ROC value (p < 0.01) and the highest AUC value are shown.
Vegas group | Best model variables (BIO) | Omission rate (training) | Omission rate (test) | Partial ROC p-value | AUC | Total number of calibrated models |
---|---|---|---|---|---|---|
Juncaceae | 1, 2, 12 | 0.00 | 0.28* | 0.00 | 0.71 | 120 |
Cyperaceae | Altitude, 1, 4, 7, 15 | 0.05 | 0.00 | 0.00 | 0.85 | 120 |
Potential distribution maps for both vegas groups are shown in Fig.
Fig.
In this study we modeled the ecological niche and potential distribution of two floristic assemblages of the Argentine Central Andean Puna vegas according to the environmental variables that determine their distribution ranges. Niche modeling of functionally close species groups allows us to study the niches and potential distribution of these plant communities even with relatively few data (
On the other hand, NicheToolBox has performed well in predicting sites with ideal environmental conditions for the existence of plant communities in both studied floristic assemblages. The reliability of the models is indicated by the relatively high AUC values reported, 0.71 and 0.85 for Juncaceae and Cyperaceae vegas respectively, with which it is possible to consider that they identify the sites where these vegas have been reported quite efficiently (
The environmental suitability map for Juncaceae vegas places them in sites mainly located in the NE region of the study area, while that of Cyperaceae shows a potential distribution that covers areas of lower altitudes and reaches sites further south in the region (Fig.
In general, organisms respond to complex interactions between environmental variables (
The overlap of the niche models in predicting vegas with suitability values greater than 0.1 for both groups could be due to the fact that these groups share both dominant cushion plant species of the communities (e.g. Oxychloe andina and Zameioscirpus atacamensis) and accompanying flora (e.g. Eleocharis pseudoalbibracteata, Triglochin concinna). In addition, these vegas classified to both floristic assemblages contain at least 20 vegas with an altitudinal variation of more than 100 to 200 m, which could be partly the reason why both floristic assemblages can be found in these vegas. This overlap can also be understood taking into account the gradual replacement of dominant species (and therefore associated flora) along the distribution gradient of these wetlands in the tropical and subtropical Andes (
Despite the importance of climatic variables in determining the distribution ranges of the different types of plant communities in vegas, in the future it would be interesting to have other variables that contribute to the understanding of their habitat requirements. It has been shown, for example, that in similar wetlands in the northern hemisphere, the distribution and ecological niche of Cyperaceae species are not only determined by the climatic gradient and the chemical components of the water (especially salinity levels), but also by shade and the level of the water table (
This contribution represents an initial step towards ecological niche modeling based on plant communities for vegas in the Argentinean Central Andean Puna. The results obtained contribute to the knowledge of the environmental factors that limit the distribution of vegas dominated by different species assemblages, which give them particular characteristics in terms of productivity, organic matter accumulation and carbon storage capacity, among others. Knowledge of the current and potential distribution of these wetlands is a valuable tool, useful for planning in a region exposed to growing anthropogenic disturbances such as lithium mining and climate change.
E.C. performed the niche models and led the writing. A.E.I. conducted the field sampling and project administration. Both authors have contributed to the development of the ideas discussed in this article, the writing, and critically revised the manuscript.
This study was possible due to financial support from CONICET, Grant from PICT2018-04228. We thank all the colleagues who collaborated with field sampling and discussion of ideas during the last few years.