Assessing methods for comparing species diversity from disparate data sources: the case of urban and peri‐urban forests

Abstract

Multi-scale forest inventory and monitoring data are increasingly being used in studies assessing forest diversity, structure, disturbance, and carbon dynamics. Also, local-level urban forest inventories are providing plot data and protocols to study tree diversity and ecosystem services in urban forests worldwide. But, differences in the sampling methods underlying these disparate protocols and data sources is a non-trivial concern in formulating comparative analyses. We assess commonly used methods for comparing tree diversity in peri-urban and urban forests when available data have different sample sizes, plot sizes, and sampling intensities. We present methods for appropriately evaluating species richness, as well as methods for comparing species distributions via community data matrices. Using permanent plot data from the southeastern United States, we present a case study comparing urban and peri-urban forests along a north-south gradient, and assessing species richness and the ecological homogenization hypothesis. Our findings indicate that comparisons of tree species richness among communities, or forest types, are often inconclusive since commonly used sample sizes do not provide precise estimates of the number of species present. While the ecological homogenization hypotheses can be tested under conditions of unequal sampling effort, we suggest robust methods such as PERMANOVA and the Raup-Crick dissimilarity index. A framework for selecting appropriate methods is also discussed. As forests are increasingly being altered by anthropogenic drivers, future studies using disparate data sources must account for differences in measurements and sampling protocols in order to produce results that are both statistically defensible and useful for science-based management.

Urban Forestry in Brazilian Amazonia

“…Meanwhile, there is need to study this parameter at a regional scale. Knowledge of the patterns of urban species diversity and what can influence them contributes to better planning of conservation actions, especially for the population of trees on the streets of the city [15,16]. A study conducted in the city of Manaus found a positive relationship between socioeconomic variables and the valorization of vegetation in an urban area by the residents of the neighborhood [17].…”

Comparative morphometric analysis of lungs of the semifossorial giant pouched rat (Cricetomys gambianus) and the subterranean Nigerian mole rat (Cryptomys foxi)

“…Nearly three decades ago, from statistical-and evolutionary standpoints, Garland and Adolph 66 argued that 'limitations' exist in comparisons between two species or populations of animals where interpretation(s) of adaptation(s) is/are made. Such studies have continued to be performed to characterise, differentiate and explain traits between species and populations, including supposed adaptive features [76][77][78][79] . Here, where we have cautiously termed the significantly higher pulmonary morphometric parameters of C. foxi (compared to the lower ones of C. gambianus) as 'specializations' instead of 'adaptations' .…”

Plant diversity in the diet of Costa Rican primates in contrasting habitats: a meta-analysis

“…Furthermore, contrary to traditional parametric multivariable analysis (e.g. MANOVA), PERMANOVA is unaffected by the correlation between variables and is less sensitive to heterogeneity in dispersion [58]. We ran the PERMANOVA via the function ‘adonis2’ of the R package vegan and we set the matrix of Euclidean distances with 999 permutations to estimate the p -value.…”

“…To determine the influence of the primate species, habitat type, study site, province, forest disturbance level, forest size, forest successional stage, monkey group/population size, study sampling effort (in months), and the interaction between primate species * habitat type on the plant assemblage in diet, we used a permutational multivariable analysis of variance (PERMANOVA), a non-parametric multivariable analysis of variance, as recommended for meta-analyses based in studies with disparate sampling efforts [58]. This method constructs ANOVA-like test statistics from a matrix of resemblances (distances, dissimilarities, similarities) calculated between the sample units, and obtains p -values using random permutations of observations among the groups.…”