Recalculated woody crown hydraulic network properties and reduced principal component analyses of Caatinga and Cerrado trees

This dataset provides source records, recalculated supplementary tables, reduced principal component analyses (PCAs), and additional network-size-adjusted analyses of woody crown hydraulic networks (WCHNs) from Caatinga and Cerrado vegetation in Brazil. It accompanies a critical review of the WCHN framework but is documented for independent understanding, verification, and reuse. Source data and recalculated properties The source spreadsheets comprise 14 individuals of Cenostigma pyramidale, ten additional Caatinga species represented by one WCHN each, and 15 Cerrado species represented by five WCHNs per species, totaling 99 networks. They preserve the original WCHN diagrams, decomposition and topology records, equations, and calculated property values. The supplementary tables provide recalculated symmetry (SYM), complexity (COM), and vulnerability (VUL) using revised equations. SYM is defined as the inverse coefficient of variation of initial-to-final node (IN–FN) topological distances, SYM = 1/CV = μ/σ. Complexity is defined as COM = ΣNO × CV, yielding the algebraic identity COM × SYM = ΣNO when both properties are defined. Revised VUL includes all nodes except the initial node in the attackable set, eliminating an artifact associated with changes in the number of final nodes. The dataset includes recalculated comparisons between Caatinga and Cerrado, comparisons among Cerrado phenological groups, and additional reduced-set PERMANOVAs excluding the algebraically derived properties COM and SHA. Original reduced principal component analyses Separate reduced PCAs were performed for Caatinga (24 WCHNs, 11 species) and Cerrado (74 complete WCHNs, 15 species), using five properties: navigability (NAV), revised vulnerability (VUL), laterality (LAT), basitony (BAS), and potential topological plasticity (PTP). Analyses used untransformed data and correlation matrices. The first two principal components explained 91.1% of the total standardized variance in Caatinga and 89.0% in Cerrado. In both vegetation types, PC1 contrasted VUL with NAV, BAS, and PTP, while PC2 was primarily associated with LAT. A sensitivity analysis evaluated the single Cerrado network with undefined BAS. New in version 1.1.0: analyses adjusted for network size This version adds exploratory analyses designed to examine how much of the observed covariance among WCHN properties persists after statistically controlling for network size (ΣNO). Each of the five properties included in the reduced PCAs was regressed separately against ΣNO using ordinary least squares. The resulting residuals were standardized, and new PCAs were calculated from their correlation matrices. Additional analyses used the natural logarithm of ΣNO to evaluate sensitivity to the functional form of size adjustment. The first two components of the linearly size-adjusted PCAs explained 85.9% of residual standardized variance in Caatinga and 84.8% in Cerrado. These percentages describe variance remaining after adjustment and restandardization and are not directly comparable with the proportions of original variance explained by the unadjusted PCAs. Six partial Pearson correlations among NAV, BAS, PTP, and VUL were also calculated separately for each biome, controlling for ΣNO. Multiple comparisons were adjusted using the Holm procedure. In Caatinga, positive associations among NAV, BAS, and PTP persisted after linear size adjustment, whereas none of the three correlations involving VUL remained significant after Holm correction. In Cerrado, all six tested associations remained statistically detectable after Holm correction, including negative associations between NAV and VUL (r = −0.586) and between BAS and VUL (r = −0.539). Additional sensitivity analyses examined logarithmic size adjustment and sampling structure, including analyses restricted to the 14 C. pyramidale individuals and exploratory species-adjusted analyses for Cerrado. Scope, limitations, and reproducibility The new analyses distinguish associations related to network magnitude from covariance that persists after linear adjustment. However, residualization does not eliminate all possible nonlinear size effects or mathematical dependencies among WCHN properties. The analyses are exploratory and do not establish causal relationships. The adjusted analyses use the numerical precision stored in the original reduced-PCA workbook. Species-level replication is uneven, especially in Caatinga, and inferential results should be interpreted accordingly. Version 1.1.0 preserves the original source spreadsheets, recalculated tables, and unadjusted PCA results. It adds a reproducible Python script, an analytical workbook containing adjusted PCA and partial-correlation results, machine-readable numerical outputs, adjusted PCA figures in PNG and SVG formats, and expanded methodological documentation. The dataset assesses the topological component of the WCHN framework. It does not directly measure hydraulic resistance, embolism vulnerability, hydraulic safety margins, or causal effects on attained tree height. The original analyses and table numbering are reported in Prado and Trovão (2023), Ecological Modelling, 481, 110345. https://doi.org/10.1016/j.ecolmodel.2023.110345 Related stochastic WCHN data and computational experiments are available separately at https://doi.org/10.5281/zenodo.21839645.