VegZ: Comprehensive Vegetation Data Analysis Package

VegZ is a comprehensive, professional-grade Python package designed specifically for vegetation data analysis and environmental modeling. It provides a complete suite of tools for ecologists, environmental scientists, and researchers working with biodiversity and vegetation data.

1.5.1

A metadata and documentation fix release; no analysis code changed. The declared dependency floors were not installable - VegZ imports QhullError from scipy.spatial, which SciPy only exposed in 1.8, so import VegZ failed on the advertised scipy>=1.7. The floors are now the oldest versions CI actually runs the suite against. See the CHANGELOG.

New in 1.5.0

1.5.0 adds the methods most often reached for after an ordination - PERMDISP/betadisper, multi-factor PERMANOVA (adonis), anova.cca-style permutation tests for constrained ordinations, variance partitioning, forward selection, Baselga turnover / nestedness partitioning, and coverage-based rarefaction and extrapolation of Hill numbers - plus VegData, a container that keeps species, environment, trait and phylogeny tables aligned.

It also lands a full scientific audit of the existing code base, correcting methods that ran without error but returned wrong numbers - among them NMDS (which was silently running metric MDS), the PERMANOVA sum-of-squares decomposition, the ANOSIM R statistic, TWINSPAN, NODF, FEve, FDiv, Moran's I, Geary's C and the ACE richness estimator. Every corrected formula is pinned by a test that checks it against an analytically known answer. See the CHANGELOG.

Complete Feature List

Data Management & Preprocessing

• VegData aligned container (New in v1.5.0) - holds a species matrix with its environmental, trait and phylogeny tables, intersects them once in a stable order, and reports exactly which sites and species each table lost. Misaligned tables are the most common source of a confident, wrong community analysis
• Parse vegetation survey data from multiple formats (CSV, Excel, Turboveg)
• Integration with remote sensing APIs (Landsat, MODIS, Sentinel)
• Darwin Core biodiversity standards compliance
• Species name standardization with fuzzy matching
• Coordinate system transformations
• Multiple data transformation methods (Hellinger, chord, Wisconsin, log, sqrt, standardize)
• Automatic species matrix detection
• Support for heterogeneous data integration
• Online Taxonomic Name Resolution (New in v1.3.0):
  • Validate and update species names against 5 online databases
  • WFO (World Flora Online), POWO (Kew), IPNI, ITIS, GBIF
  • File-based and DataFrame integration
  • Confidence scores and synonym retrieval
• Improved Ecological Terminology (v1.2.0) - Domain-specific language:
  • Use of "sites" instead of generic "samples" for ecological sampling locations
  • Professional ecological nomenclature throughout the package

Data Quality & Validation

• Comprehensive spatial coordinate validation
• Temporal data validation and date parsing
• Geographic outlier detection with country boundary checks
• Coordinate precision assessment
• Invalid coordinate range detection
• Transposed coordinate detection
• Country boundary consistency checks
• Automated quality reporting
• Enhanced Species Name Error Detection (Introduced in v1.1.0):
  • 10+ error categories: incomplete binomial, formatting issues, author citations
  • Hybrid marker detection and validation
  • Infraspecific rank validation (var., subsp., f., cv.)
  • Placeholder name detection (sp., cf., aff., indet.)
  • Invalid character identification
  • Comprehensive error reporting with actionable suggestions
  • Batch processing capabilities for large datasets

Diversity Analysis (15+ Indices)

• Basic indices: Shannon, Simpson (concentration), Gini-Simpson, Simpson inverse, richness, evenness
• Advanced indices: Fisher's alpha, Berger-Parker, McIntosh, Brillouin
• Additional indices: Menhinick, Margalef
• Richness estimators: Chao1, ACE, Jackknife1, Jackknife2
• Hill numbers for multiple diversity orders (q = 0, 0.5, 1, 1.5, 2, etc.)
• Beta diversity analysis (Whittaker, Sørensen, Jaccard methods) - returns a pairwise dissimilarity matrix; whittaker_beta() gives the whole-dataset scalar
• Rarefaction curves - exact Hurlbert expectation with variance, computed in log space so it is stable for large counts
• Species accumulation curves with permutation confidence bands
• Beta diversity partitioning (New in v1.5.0) - Baselga (2010, 2012) separation into turnover (beta_sim / beta_jtu) and nestedness-resultant (beta_sne / beta_jne) components, pairwise and multi-site
• Coverage-based standardization (New in v1.5.0) - Chao & Jost (2012) sample coverage, and diversity compared at equal completeness rather than equal sampling effort
• Hill number rarefaction and extrapolation (New in v1.5.0) - Chao et al. (2014) interpolation/extrapolation for q = 0, 1, 2
• Diversity profiles

Complete Multivariate Analysis Suite

• PCA - Principal Component Analysis with multiple transformations
• CA - Correspondence Analysis with scaling options
• DCA - Detrended Correspondence Analysis with segment control
• CCA - Canonical Correspondence Analysis with constraints
• RDA - Redundancy Analysis for linear relationships
• NMDS - genuine Non-metric Multidimensional Scaling with stress assessment
• PCoA - Principal Coordinates Analysis with Lingoes/Cailliez corrections for negative eigenvalues
• Scientific Method Names (New in v1.2.0) - Professional abbreviated method names:
  • ca_analysis() for Correspondence Analysis
  • dca_analysis() for Detrended Correspondence Analysis
  • cca_analysis() for Canonical Correspondence Analysis
  • rda_analysis() for Redundancy Analysis
  • pcoa_analysis() for Principal Coordinates Analysis
  • Full backward compatibility with existing method names
• Constrained-ordination significance tests (New in v1.5.0) - anova_cca() and anova_rda() in vegan's anova.cca idiom: overall, by axis, by term (sequential) and by margin, with partial ordinations via conditioning
• Variance partitioning (New in v1.5.0) - varpart() across two or three explanatory tables (Peres-Neto et al. 2006) with adjusted R-squared
• Forward selection (New in v1.5.0) - permutation-based, with the Blanchet et al. (2008) double stopping criterion
• Environmental vector fitting (envfit-style) with unit-length direction vectors and permutation p-values
• Procrustes analysis for ordination comparison, with a PROTEST permutation test
• Goodness-of-fit diagnostics
• Multiple ecological distance matrices (Bray-Curtis, Jaccard, Sørensen, Euclidean, Manhattan, Canberra, Chord, Hellinger)

Advanced Clustering Methods

• TWINSPAN - Two-Way Indicator Species Analysis (vegetation classification gold standard)
  • Pseudospecies creation with customizable cut levels
  • Hierarchical divisive classification
  • Indicator species identification
  • Classification tree structure
• Hierarchical clustering with ecological distance matrices
• Comprehensive Elbow Analysis with 5 detection algorithms:
  • Kneedle algorithm (Satopaa et al., 2011) - automatic knee detection
  • Second derivative maximum - curvature-based detection
  • Variance explained threshold - <10% additional variance criterion
  • Distortion jump method (Sugar & James, 2003) - jump detection
  • L-method (Salvador & Chan, 2004) - piecewise linear fitting
• Consensus recommendations with confidence scores
• K-means clustering with multiple initializations
• Fuzzy C-means clustering for gradient boundaries
• DBSCAN for density-based core community detection
• Gaussian Mixture Models for probabilistic clustering
• Clustering validation metrics (silhouette, gap statistic, Calinski-Harabasz, Davies-Bouldin)
• Optimal k determination with multiple methods
• Reproducible by default - every stochastic method accepts random_state

Statistical Analysis

• PERMANOVA - Permutational multivariate analysis of variance
• PERMDISP / betadisper (New in v1.5.0) - Anderson (2006) test of homogeneity of multivariate dispersions, the assumption PERMANOVA depends on; centroid or spatial-median centring, with Holm-adjusted pairwise comparisons
• adonis (New in v1.5.0) - multi-factor PERMANOVA from a model formula: crossed and nested designs, interactions, sequential (Type I) and marginal (Type III) sums of squares, continuous terms, and restricted permutation within strata
• ANOSIM - Analysis of similarities
• MRPP - Multi-response permutation procedures
• Mantel tests and partial Mantel tests for matrix correlation
• Indicator Species Analysis (IndVal) for cluster characterization
• SIMPER - Similarity percentages for group comparisons
• Cophenetic correlation for hierarchical clustering validation

Environmental Modeling

• Generalized Additive Models (GAMs) with multiple smoothers:
  • Spline smoothers
  • LOWESS smoothers
  • Polynomial smoothers
  • Gaussian process smoothers
• Species response curves modeling:
  • Gaussian response curves
  • Skewed Gaussian curves
  • Beta response curves
  • Linear responses
  • Threshold responses
  • Unimodal responses
• Environmental gradient analysis
• Environmental niche modeling

Temporal Analysis

• Phenology modeling with sigmoid, double-sigmoid (Zhang et al. 2003), Gaussian, beta and Weibull curves, fitted from data-driven starting values
• Trend detection - linear, polynomial, spline and LOWESS trends, plus the Mann-Kendall test with Sen's slope and a distribution-free confidence interval
• Time series decomposition (seasonal, trend, residual) - classical, STL and X-11, with a dependency-free fallback when statsmodels is absent
• Seasonal pattern analysis with automatic period detection
• Climate-vegetation response analysis across user-specified lags
• Growth curve fitting - logistic, Gompertz, von Bertalanffy, exponential and power models

Spatial Analysis

• Spatial interpolation methods, with leave-one-out cross-validated RMSE:
  • Inverse Distance Weighting (IDW)
  • Simple kriging with exponential, Gaussian or spherical variograms
  • Radial basis functions (thin-plate spline, multiquadric, Gaussian, linear)
  • Nearest neighbour, linear and cubic
• Landscape metrics calculation:
  • Patch density, mean patch size, patch size coefficient of variation
  • Edge density and largest patch index
  • Landscape shape index and contagion (O'Neill et al. 1988)
  • Shannon, Simpson and evenness indices for landscapes
• Spatial autocorrelation analysis (Moran's I with randomisation-based significance test, Geary's C, empirical variogram)
• Habitat suitability modeling from point occurrences (Random Forest, GLM)

Specialized Methods

• Phylogenetic diversity analysis:
  • Faith's phylogenetic diversity
  • Mean Pairwise Distance (MPD) and Mean Nearest Taxon Distance (MNTD)
  • Net Relatedness Index (NRI)
  • Nearest Taxon Index (NTI)
• Metacommunity analysis:
  • Elements of metacommunity structure
  • Coherence, turnover, and boundary clumping
• Network analysis:
  • Co-occurrence networks from correlation or Jaccard association
  • Betweenness and eigenvector centrality, clustering coefficient
  • Connected components and small-world sigma against a random-graph ensemble
• Nestedness analysis with null models:
  • NODF (Nestedness based on Overlap and Decreasing Fill)
  • Temperature calculator
  • Null model generation and testing

Functional Trait Analysis

• Trait syndrome identification
• Community-weighted means (CWM)
• Functional diversity indices:
  • Functional richness (FRic) - convex hull volume
  • Functional evenness (FEve) - minimum spanning tree based
  • Functional divergence (FDiv)
  • Functional dispersion (FDis)
  • Rao's quadratic entropy
• Functional beta diversity between sites
• Trait-environment relationships
• Functional group identification (hierarchical or k-means)

Machine Learning & Predictive Modeling

• Species Distribution Modeling (SDM):
  • Random Forest models
  • Gradient Boosting models (LightGBM when installed)
  • Logistic regression, including a simplified MaxEnt-style variant
• Classification algorithms for vegetation types
• Regression models for abundance and biomass prediction
• Anomaly detection (Isolation Forest, DBSCAN)
• Dimensionality reduction (PCA, t-SNE) for exploration
• Model validation and performance metrics
• Variable importance assessment

Visualization & Reporting

• Specialized ecological plots:
  • Diversity bar charts and histograms
  • Species accumulation curves
  • Rarefaction plots
• Ordination diagrams with:
  • Site scores plotting, coloured by a continuous or categorical variable
  • Species loading arrows
  • Environmental vector overlays
• Clustering visualizations:
  • Dendrograms with customizable formatting
  • Silhouette plots
  • Comprehensive elbow analysis plots (4-panel layout)
  • Cluster validation plots
• Interactive dashboards using Plotly, with static matplotlib fallbacks when it is not installed
• Automated quality reports with statistical summaries
• Export to HTML (dashboards and reports) and CSV. Plotting functions return matplotlib Figure objects, so fig.savefig(...) covers PNG, PDF and SVG

Quick Analysis Functions

• quick_diversity_analysis() - Instant diversity calculations
• quick_ordination() - Rapid PCA or NMDS analysis
• quick_clustering() - Fast k-means or hierarchical clustering
• quick_elbow_analysis() - Optimal cluster number determination

Quick Start

Installation

pip install VegZ

For extended functionality:

# With spatial analysis support
pip install VegZ[spatial]

# With remote sensing capabilities
pip install VegZ[remote-sensing]

# Complete installation with all features
pip install VegZ[spatial,remote-sensing,fuzzy,interactive]

Basic Usage

import pandas as pd
from VegZ import VegZ

# Initialize VegZ
veg = VegZ()

# Load your vegetation data
data = veg.load_data('vegetation_data.csv')

# Quick diversity analysis
diversity = veg.calculate_diversity(['shannon', 'simpson', 'richness'])

# Multivariate analysis
pca_results = veg.pca_analysis(transform='hellinger')
nmds_results = veg.nmds_analysis(distance_metric='bray_curtis')

# Advanced elbow analysis for optimal clustering
elbow_results = veg.elbow_analysis(
    k_range=range(1, 15),
    methods=['knee_locator', 'derivative', 'variance_explained'],
    plot_results=True
)
optimal_k = elbow_results['recommendations']['consensus']

# Clustering with optimal k
clusters = veg.kmeans_clustering(n_clusters=optimal_k)
indicators = veg.indicator_species_analysis(clusters['cluster_labels'])

# Create visualizations
veg.plot_diversity(diversity, 'shannon')
veg.plot_ordination(pca_results, color_by=clusters['cluster_labels'])

Quick Functions for Immediate Results

from VegZ import quick_diversity_analysis, quick_ordination, quick_elbow_analysis

# Instant analyses
diversity = quick_diversity_analysis(data, species_cols=['sp1', 'sp2', 'sp3'])
ordination = quick_ordination(data, method='pca')
elbow_results = quick_elbow_analysis(data, max_k=10, plot_results=True)

Advanced TWINSPAN Analysis

from VegZ.clustering import VegetationClustering

clustering = VegetationClustering()

# Two-Way Indicator Species Analysis - the gold standard for vegetation classification
twinspan_results = clustering.twinspan(
    species_data,
    cut_levels=[0, 2, 5, 10, 20],
    max_divisions=6,
    min_group_size=5
)

print("Site classification:", twinspan_results['site_classification'])
print("Indicator species:", twinspan_results['classification_tree']['indicator_species'])

Enhanced Species Name Error Detection (Introduced in v1.1.0)

from VegZ.data_management.standardization import SpeciesNameStandardizer

standardizer = SpeciesNameStandardizer()

# Validate individual species names
result = standardizer.validate_species_name("Quercus alba L.")
print(f"Valid: {result['is_valid']}")
print(f"Errors: {result['errors']}")
print(f"Suggestions: {result['suggestions']}")

# Batch validation of species names
import pandas as pd
df = pd.DataFrame({'species': ['Quercus alba', 'quercus sp.', 'Pinus × strobus']})
validated_df = standardizer.batch_validate_names(df['species'].tolist())

# Generate comprehensive error report
report = standardizer.generate_error_report(df, species_column='species')
print(f"Validity rate: {report['summary']['validity_percentage']}%")

Online Taxonomic Name Resolution (New in v1.3.0)

from VegZ import TaxonomicResolver, resolve_species_names

# Quick resolution with default source (World Flora Online)
results = resolve_species_names(['Quercus robur', 'Pinus sylvestris'])

# Using specific source (GBIF)
resolver = TaxonomicResolver(sources='gbif')
results = resolver.resolve_names(['Quercus robur', 'Pinus sylvestris'])

# Multiple sources with fallback
resolver = TaxonomicResolver(
    sources=['wfo', 'powo', 'gbif'],
    use_fallback=True
)
results = resolver.resolve_names(['Quercus robur', 'Pinus sylvestris'])

# Resolve from file
results = resolver.resolve_from_file('species_list.csv')

# Update species names in your data
import pandas as pd
df = pd.read_csv('vegetation_data.csv')
df_updated = resolver.resolve_dataframe(df, species_column='species')

# Export results
resolver.export_results(results, 'resolved_names.xlsx')
resolver.print_summary(results)

Supported databases: WFO (World Flora Online), POWO (Plants of the World Online - Kew), IPNI (International Plant Names Index), ITIS (Integrated Taxonomic Information System), GBIF (Global Biodiversity Information Facility).

Data Format Requirements

VegZ expects data in site-by-species matrix format:

site_id,Species1,Species2,Species3,...
SITE_001,25,18,12,...
SITE_002,32,22,16,...

Environmental data should have matching site IDs:

site_id,latitude,longitude,elevation,soil_ph,temperature,...
SITE_001,44.2619,-72.5806,850,6.2,18.5,...

Target Applications

• Vegetation community classification and mapping
• Biodiversity assessments and monitoring
• Environmental impact studies
• Species distribution modeling
• Ecological restoration planning
• Academic research in plant ecology and environmental science

Requirements

Required:

• Python >= 3.9
• NumPy >= 1.22.4
• Pandas >= 2.2.0
• SciPy >= 1.8.0
• Matplotlib >= 3.5.0
• scikit-learn >= 1.0.0
• Seaborn >= 0.11.0
• Requests >= 2.25.0

These are the oldest versions the suite is run against in CI, not estimates.

Optional (for extended functionality):

import VegZ never requires any of these and never warns about them; each is only needed when you call a feature that uses it.

Extra	Provides	Needed for
VegZ[excel]	openpyxl, xlrd	Reading .xlsx / .xls files
VegZ[spatial]	geopandas, pyproj, shapely, rasterio	Coordinate transforms, geospatial validation
VegZ[fuzzy]	fuzzywuzzy, python-Levenshtein	Faster fuzzy name matching (falls back to difflib)
VegZ[network]	networkx	Advanced co-occurrence network metrics
VegZ[timeseries]	statsmodels	LOWESS smoothing, STL decomposition
VegZ[interactive]	plotly, bokeh	Interactive dashboards
VegZ[remote-sensing]	earthengine-api, geemap, xarray	Remote sensing integration
VegZ[all]	the common subset of the above	Everything except remote sensing

Tested with:

• Python 3.9 - 3.13
• All major operating systems (Windows, macOS, Linux)

Reproducibility

Every stochastic routine accepts a random_state, and VegZ never mutates NumPy's global random state:

from VegZ.statistics import EcologicalStatistics

stats = EcologicalStatistics()
result = stats.permanova(distance_matrix, groups, permutations=999, random_state=42)
# Re-running with the same seed reproduces the p-value exactly.

random_state is available on PERMANOVA, ANOSIM, MRPP, Mantel and partial Mantel tests, indicator species analysis, Procrustes/PROTEST, environmental vector fitting, fuzzy c-means, the gap statistic, null models and species accumulation curves.

Validation

The corrected statistics are pinned by regression tests that check them against analytically known answers rather than against previously recorded output:

• PERMANOVA reproduces the classical one-way ANOVA F exactly on univariate Euclidean distances
• PCoA on Euclidean distances reproduces PCA scores exactly
• CA total inertia x N equals scipy.stats.chi2_contingency's chi-square
• ANOSIM R equals 1 for perfectly separated groups
• NODF equals 100 for a perfectly nested matrix and 0 with no fill gradient
• FEve equals 1 for a perfectly even community
• Geary's C equals 1 under no spatial autocorrelation
• Landscape shape index equals 1 for a square patch
• Rarefaction at full sample size returns the observed richness with zero variance
• TWINSPAN recovers three known vegetation types exactly (adjusted Rand = 1.0)
• PERMDISP's F equals scipy.stats.levene(center='mean') to ten decimal places on univariate Euclidean data
• adonis() reproduces permanova() exactly for one factor, and classical sequential ANOVA sums of squares exactly for two
• anova_rda()'s pseudo-F equals the regression F to ten decimal places on univariate data
• Variance-partitioning fractions sum to exactly 1.0 for two and three tables
• Baselga turnover and nestedness sum to the total beta diversity exactly
• Hill rarefaction at q = 0 equals Hurlbert rarefaction exactly, and extrapolation is continuous with the observed value at the reference size

The suite runs on Python 3.9-3.13 across Linux, macOS and Windows in CI, with additional jobs for the oldest declared dependency floors, a no-optional-dependencies install, and pre-release dependencies.

Run them with:

pytest tests/

Scientific Background

VegZ implements methods from key ecological and statistical literature:

• TWINSPAN: Hill, M.O. (1979) TWINSPAN - A FORTRAN Program for Arranging Multivariate Data
• Elbow Analysis: Multiple algorithms including Satopaa et al. (2011) "Finding a kneedle in a haystack"
• Ordination: Methods from Legendre & Legendre "Numerical Ecology"
• Diversity: Comprehensive indices from Magurran "Measuring Biological Diversity"
• Statistical tests: Anderson (2001) PERMANOVA; Clarke (1993) ANOSIM
• Indicator species: Dufrêne & Legendre (1997) IndVal
• Nestedness: Almeida-Neto et al. (2008) NODF; Atmar & Patterson (1993) temperature
• Functional diversity: Villéger et al. (2008) FRic/FEve/FDiv; Laliberté & Legendre (2010) FDis
• Constrained ordination: ter Braak (1986) CCA
• Landscape metrics: O'Neill et al. (1988) contagion; FRAGSTATS conventions
• Metacommunity structure: Leibold & Mikkelson (2002) elements of metacommunity structure

Contributing

We welcome contributions! Please see the Contributing Guide for details.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

• GitHub Issues: Report bugs or request features
• Documentation: Full user guide and API reference
• Email: For academic collaborations and consulting

Citation

If you use VegZ in your research, please cite:

@software{VegZ,
    author = {Hatim, Mohamed Z.},
    title = {VegZ: A comprehensive Python package for vegetation data analysis and environmental modeling},
    year = {2026},
    version = {1.5.0},
    url = {https://github.com/mhatim99/VegZ}
}

---

VegZ - Empowering ecological research with comprehensive vegetation analysis tools.