Flooding drives tropical dry forest tree community assembly in southeast Brazil
Data files
Oct 17, 2023 version files 303.61 KB
-
agbabd.csv
-
ave_diam.csv
-
field.csv
-
nmds.csv
-
README.md
-
soil_analysis.csv
-
soil_pca.csv
-
venn.csv
Abstract
In this study, we characterized and compared vegetation types associated with geomorphological units susceptible to distinct flooding levels. Differences in vegetation are related to landform variations. We aimed to (i) characterize the vegetation structure and quantify community compositional differences among landforms and (ii) compare landforms soil characteristics and how these correlate with the tree vegetation. The study area is located in the Brazilian Caatinga Domain, near the Verde Grande River, a tributary of São Francisco River (coordinates 14º 54’ 38’’ S 43º 42’ 53’’ W). We allocated six plots in each landform sampled from wettest to driest sites: (i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest). We conducted multivariate analyses (non-metric multidimensional scaling and Principal component analyses) to determine if the five sampled environments formed distinct floristic and environmental groups across the flooding gradient. We used ANOVA and Tukey post-hoc tests to assess soil variable differences among vegetation types (where relationships between edaphic variations and tree communities' floristic-structural patterns are the purpose of each analysis). A total of 1422 individuals, 26 families, 70 genera, and 89 species were recorded. The NMDS revealed two distinct floristic groups: one group is associated with landforms with assumed higher flood frequency (RF, RWF, WF) and one with less frequently flooded landforms (OFF and UF). The RF, OFF, and UF landforms contained exclusive species (that only occurred in the plots of a particular landform). The species Geofroea spinosa (Fabaceae) was responsible for 70% of the total biomass recorded in the landforms RWF and WF. The soil analysis showed a gradient of soil acidity and fertility related to water saturation, whereby the most frequently flooded plots had the highest acidity values and highest fertility. We found that flood-related conditions significantly influence tree community structure and species distribution in this floodplain in the Brazilian Caatinga Domain.
README: FLOODING DRIVES TROPICAL DRY FOREST TREE COMMUNITY ASSEMBLY IN SOUTHEAST BRAZIL
Cite this dataset
Madeira, Denise et al. (2023). Flooding drives tropical dry forest tree community assembly in southeast Brazil [Dataset]. Dryad. https://doi.org/10.5061/dryad.rjdfn2zjq
Abstract
In this study, we characterized and compared vegetation types associated with landform units susceptible to distinct flooding levels, from tropical dry forests fragments located in the Brazilian Caatinga Domain. Each response variable was measured during plot sampling in sites.
Methods
Our study was conducted near the Verde Grande River, in northern Minas Gerais, near Bahia state in Brazil. We allocated six plots in each woody vegetation type sampled from wettest to driest sites where differences in vegetation are related to landform variations. More specifically, sites are located in flood-prone areas of the Verde Grande River that can be differentiated into five vegetation types based on landforms flood frequency from wettest to driest: (i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest) In each plot (400 m²), the circumference at breast height (1.3 m above the soil) of every living tree that met the inclusion criterion (minimum circumference: 10 cm) was recorded. For trees with multiple stems, we included those of which the square root of the sum of squares of each stem's circumference met the inclusion criterion. Tree identification was performed by a specialist in the field. Family names were standardized based on The Angiosperm Phylogeny Group (APG IV 2016), and species names were based on Reflora database (reflora.jbrj.gov.br/reflora/herbarioVirtual). For the soil data, we collected superficial samples of 0.5 liters to 10 cm of depth in three points in each plot (beginning, middle and end). Each sample was stored in a plastic bag and later sent to the Federal University of Lavras Laboratory of Soil Analysis
Usage notes
These datasets were collected in the landscape surrounding the Verde Grande River, São Francisco basin, northern Minas Gerais state, Brazil (2014).
Details for each dataset are provided in README file.
Datasets included:
- Tree species identified in the plots;
- Phytosociological and NMDS analysis (abundance matrix);
- Venn diagram analysis data ( packages venn diagram (Chen 2011) and tidyverse (Wickham et al. 2019) used in the R environment;
- Soil variables for ANOVA and Tukey tests for differences in the original soil variables among the vegetation types per plot;
- Soil variables for Principal Component Analysis (PCA) analysis;
- Aboveground biomass (AGB) (t ha-¹) per tree, per plot and per vegetation type;
- Average diameter (cm) per tree, per plot and per vegetation type;
In addition, a list of species with their respective families, abundances, aboveground biomass (AGB) (t ha-¹), canopy cover (CC) (%) and taxonomic classification table information (data for taxonomic and variation in taxonomic distinctness analysis) are available in the Supporting information of the manuscript. (Madeira et al. 2022 - Nordic Journal of Botany - 10.1111/njb.03913)
SHARING/ACCESS INFORMATION
- Links to publications that cite or use the data: Madeira, Denise et al. (2023). Flooding drives tropical dry forest tree community assembly in southeast Brazil. Oikos.
- Links to other publicly accessible locations of the data: None
- Links/relationships to ancillary data sets: None
- Was data derived from another source? No
DATA & FILE OVERVIEW
- File List: A. field.csv B. nmds.csv C. venn.csv D. soil_analysis.csv E. agbabd.csv F. ave_diam.csv
- Relationship between files, if important: None
- Additional related data collected that was not included in the current data package: None
- Are there multiple versions of the dataset? No A. If yes, name of file(s) that was updated: NA i. Why was the file updated? NA ii. When was the file updated? NA
Tree data
DATA-SPECIFIC INFORMATION FOR: field.csv
- Number of variables: 7
- Number of cases/rows: 1423
- Variable List:
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- ind = tree marked with numbered aluminium plates.
- family, genus, species;
- wd = wood density (Chave et al. (2014)(for R BIOMASS package (Rejou-Mechain et al. 2017);
- dq = quadractic diameter(trees with multiple stems) (cm);
- agb = tree individual above ground biomass (ton. ha -¹);
- yr = year of sampling
- Missing data codes: None
- Specialized formats or other abbreviations used: None <br>
Phytosociological and NMDS analysis (abundance matrix)
DATA-SPECIFIC INFORMATION FOR: nmds.csv
- Number of variables: 92 (abundance matrix)
- Number of cases/rows: 30
- Variable List:
- plot_code = plots;
- landform = There are 5 different landforms in the study area (i) Marginal Dike, Upper terrace, Lower Terrace, Lower Plain and Upper Plain;
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- species.
- Missing data codes: None
- Specialized formats or other abbreviations used: None
===========================================================
Venn diagram analysis (R package venn diagram (Chen 2011)
DATA-SPECIFIC INFORMATION FOR: venn.csv
- Number of variables: 6
- Number of cases/rows: 1423
- Variable List:
- plot_code = plots;
- landform = There are 5 different landforms in the study area (i) Marginal Dike, Upper terrace, Lower Terrace, Lower Plain, and Upper Plain
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- family;
- species
- dq = for presence /absence use only
- Missing data codes: None
- Specialized formats or other abbreviations used: None
===========================================================================================================
Soil variables for ANOVA and Tukey tests for differences in the original soil variables among the vegetation types per plot
DATA-SPECIFIC INFORMATION FOR: soil_analysis.csv
- Number of variables: 19
- Number of cases/rows: 31
- Variable List:
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- plot_code = plots;
- pH in water;
- phosphorus (P) (mg/cm³);
- potassium (K) (mg/cm³);
- calcium (Ca) (mg/dm³);
- magnesium (Mg) (mg/dm³);
- aluminum (Al) (mg/dm³);
- potential acidity (H+Al);
- sum of bases (SB)(cmolc/dm³);
- effective Cation Exchange Capacity (CEC)(cmolc/dm³);
- Cation Exchange Capacity at pH 7.0 (CECpH7)(cmolc/dm³);
- base saturation (BS) (%);
- aluminum saturation (AS)(%);
- organic matter (OM) (dag/kg);
- remnant phosphorus (rem P)(mg/L);
- clay (dag/kg);
- silt (dag/kg);
- sand (dag/kg);
- Missing data codes: None
- Specialized formats or other abbreviations used: None
=====================================================================================================
Soil variables for Principal Component Analysis (PCA) (R_Rescaled)
DATA-SPECIFIC INFORMATION FOR: soil_pca.csv
- Number of variables: 7
- Number of cases/rows: 31
- Variable List:
- plot_code = plots;
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- pH in water;
- phosphorus (P) (mg/cm³);
- aluminum (Al) (mg/dm³);
- sum of bases (SB)(cmolc/dm³);
- organic matter (OM) (dag/kg);
- Missing data codes: None
- Specialized formats or other abbreviations used: None
===================================================================================================
Aboveground biomass (AGB) per plot and per tree (t ha-¹)
DATA-SPECIFIC INFORMATION FOR: agbabd.csv
- Number of variables: 5
- Number of cases/rows: 31
- Variable List:
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- plot_code = plots;
- agb = above ground biomass(t ha-¹);
- abund = tree abundance per plot;
- agb_tree = average agb per tree in the plot (t ha-¹);
- Missing data codes: None
- Specialized formats or other abbreviations used: None
======================================================================================================
Average diameter (cm)
DATA-SPECIFIC INFORMATION FOR: ave_diam.csv
- Number of variables: 4
- Number of cases/rows: 31
- Variable List:
- vg_type = ((i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest);
- plot_code = plots;
- ave_diam = average diameter (cm);
- abund = tree abundance per plot;
- Missing data codes: None
- Specialized formats or other abbreviations used: None
Funding:
This research was supported by:
Fundação de Amparo à Pesquisa do Estado de Minas Gerais FAPEMIG.
Conselho Nacional de Desenvolvimento Científico e Tecnológico – CNPq
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
<br>
Methods
Study site and sampling
Our study was conducted near the Verde Grande River, in northern Minas Gerais, near the state of Bahia in Brazil (coordinates 14º 54’ 38’’ S 43º 42’ 53’’ W). The Verde Grande River (557 km of extension), which belongs to the São Francisco river basin (area of 641.000 km²), begins in the village of Alto Belo in the Bocaiúva municipality, northern Minas Gerais, and flows into the São Francisco River in Malhada, Bahia state (Inema). According to Köppen’s climate classification, the region falls into the Aw category (hot and humid tropical climate), experiencing hot summers and dry winters with mild temperatures. Seasonality is another important factor in the basin’s climate, with well-defined rainy (October to March) and dry seasons (April to September) (ANA). The average total annual precipitation is 785 mm, with values ranging from 700 mm in certain parts of the basin to over 1,300 mm at the headwaters (ANA).
The study sites belong to the Caatinga biogeographic domain, in a region where seasonally dry tropical forests predominate. More specifically, the sites are located in flood-prone areas of the Verde Grande River that can be differentiated into five vegetation types based on landforms flood frequency from wettest to driest: (i) Marginal Dike (RF – Riparian Forest), (ii) Upper terrace (RWF – Riparian Wetland Forest), (iii) Lower Terrace (WF – Wetland Forest), (iv) Lower Plain (OFF – Occasionally Flooded Forest), and (v) Upper Plain (UF - Unflooded Forest) (Figure 1).
The “Marginal Dike” can be described as a sedimentary ridge along the riverbanks. The bank’s width varies between half and four times the channel’s width, while the bank’s height varies from a few meters to more than 10 meters and tends to be topographically higher closest to the river. Soils are sandier than in other floodplain sites and its tree-dominated vegetation is distributed in forest fragments. The marginal lagoons are flat, often water-saturated, and flood-prone. In these environments, the soils are usually clay-rich, and the vegetation is dominated by flood-adapted herbaceous plants (trees are rare in this environment) (Christofoletti 1980, Lobo and Joly 2000, Silva et al. 2012).
The terraces represent former floodplains and can be subdivided into a lower terrace and an upper terrace according to flood frequency. The “Lower Terrace” is under the considerable influence of paleochannel flooding due to lateral overflow of the main bed and/or rainwater accumulation during the rainy season. The “Upper Terrace” is located above the floodplain level and, therefore, less susceptible to flooding (Christofoletti 1980, Pereira 2013).
Floodplains are located near shoreline ponds or in depressions and vary in susceptibility to flooding depending on topographical position or proximity to the main bed or shoreline pond. The “Lower Plain” is located in gentle depressions along the floodplain, generally close to the river and/or the marginal lagoons, while the “Upper Plain” is located at higher elevations further from the main riverbed or marginal lagoons and thus is not affected by flooding regimes.
Data collection
Woody vegetation on each of the five landform types was sampled in 2014. Three plots of 400 m² (20x20 m) were allocated on each side of the river, a total of six plots per site. Our classification of flooding regime of each site was based on local knowledge and observable landscape signs. For example, flooded forest sites are located in riparian habitats or depressions and have evidence of sedimentation and short-lived ponds of standing water. The occasionally flooded sites are characterized by shallow depressions and are flooded at about once every 30 years. Finally, sites that are never flooded are typically 500-1900 m from watercourses, with no evidence of temporary ponds.
In each plot, the circumference at breast height (1.3 m above the soil) of every living tree that met the inclusion criterion (minimum circumference: 10 cm) was recorded. For trees with multiple stems, we included those of which the square root of the sum of squares of each stem's circumference met the inclusion criterion (Scolforo and Mello 1997, Souza et al. 2021). Tree identification was performed by a specialist in the field or by comparisons with herbarium specimens. Family names were standardized based on the Angiosperm Phylogeny Group (APG IV 2016), and species names were based on Reflora database (reflora.jbrj.gov.br/reflora/herbarioVirtual).
For the soil data, we collected superficial samples of 0.5 liters to 10 cm of depth at three points in each plot (beginning, middle, and end). Each sample was stored in a plastic bag and later sent to the Federal University of Lavras Laboratory of Soil Analysis. The following variables were analyzed: proportions of clay, silt, and sand; pH in water, potassium (K), phosphorus (P), calcium (Ca), magnesium (Mg), aluminum (Al), organic matter (OM); potential acidity (H+Al); sum of bases (SB), base saturation (BS); aluminum saturation (AS); effective Cation Exchange Capacity (CEC); Cation Exchange Capacity at pH 7.0 (CECpH7) and remnant phosphorus (rem P).
Data analysis
For the tree data, we calculated the Shannon-Wiener (H’) and Pielou equability (J) indices (Durigan, 2003). For taxonomic distinctness analysis (Clarke and Warwick, 2001), six taxonomic categories (subclass, superorder, order, family, genus, and species) of trees were used (Supporting information). The value of Δ+ (average taxonomic distinctness - average taxonomic path length between two randomly chosen species) and Λ+ (variation in taxonomic distinctness, which reflects the unevenness in the taxonomic tree of a given species’ list and represents the variance of these pair-wise path lengths) were calculated for each patch and plotted against the corresponding number of species. The index was calculated by the ‘taxa2dist’ and ‘taxondive’ functions of the R package ‘vegan’ (Oksanen et al. 2019). To assess whether the five sampled environments across the flooding gradient formed distinguishable floristic groups, we employed non-metric multidimensional scaling (NMDS) (Gotelli and Ellison, 2011) using abundance data (Bray-Curtis distance) and the function "metaMDS" from the vegan package (Oksanen et al. 2020). We also evaluated floristic specificity and species sharing among the environments. For this, we used the packages venn diagram (Chen 2011) and tidyverse (Wickham et al. 2019) in the R environment (<www.r-project.org>).
We compared the vegetation of the five environments in terms of abundance, above-ground woody biomass (AGB), and canopy cover (CC). We calculated each species' AGB with the pantropical modified equation of Chave et al. (2014), using the biomass package (Rejou-Mechain et al. 2017). This allometric equation estimates AGB based on stem diameter, a bioclimatic stressor variable E (derived from the WorldClim database and based on plot geographic coordinates, aiming to compensate for the lack of height information through the relationship of this variable with macro-scale climatic patterns), and mean wood density values. When available, we used species-level values of wood density, otherwise (if species-level information was not available), we used genera- or family-level means. These values were obtained from the Global Wood Density database (Zanne 2009).
We calculated the proportion of family, species, and abundance with the highest biomass in each vegetation type. In each landform, we summed the total AGB per species, starting with the most dominant species and arranged them in descending order, to add up to 70% of each vegetation type total biomass. For the structural analysis of the forests, we calculated the average diameter and average biomass/stem/plot in the landforms. On a microscale, the vegetation changes sharply between environments. To investigate these changes, we analyzed how environmental harshness relates to the flooding regime and how the vegetation types differ from each other. In this analysis, we only considered the data on exclusive species (that only occurred in the plots of a particular landform). To determine the abundance, richness, and AGB of these exclusive species for each vegetation type we calculated the percentage of each of these variables (abundance, richness, and AGB) attained by the exclusive species relative to the rest of the vegetation type community in the landform.
We assessed whether the environments differ significantly in terms of each one of the edaphic variables. We used ANOVA and Tukey tests for differences in the original soil variables among the vegetation types. We employed Principal Component Analysis (PCA) in the R environment with the soil variables collected in the 30 plots to assess soil variations among the landforms. Due to intercorrelation, we removed Ca, Mg, and Cation Exchange Capacity (CEC) from the analysis.