Species abundance distributions should underpin ordinal cover-abundance transformations

Prepared by Megan J. McNellie, Josh Dorrough and Ian Oliver

Around the world, ecologists are surveying plants in fixed area plots and estimating the cover and abundance of plant species. Photo credit: Fenner School of Environment and Society, The Australian National University.

Our analyses of cover-abundance data are a timely contribution to support the ever-growing volume of plot data that are being compiled into databases. These consolidated data open opportunities for ecologists to test ideas and analyses using a global dataset. Within these datasets, we found that most species abundance data have been recorded in a variant of the classical Braun-Blanquet cover-abundance scale. To extend the use of these ordinal data, for example to quantify vegetation structure or calculate summed cover, we needed to convert ordinal data into continuous (0-100%) data.

We examined the right-skewed distribution in continuous cover data to investigate if this skew might inform an approach to transforming ordinal data. We were especially interested in looking deeper into the patterns of summing cover for different growth forms. We were able to use data from New South Wales (Australia), and check our models using an entirely independent dataset from the West Virginia Natural Heritage Program (USA) extracted from VegBank.

We clearly show that it is necessary to account for right-skewed distributions when transforming Braun-Blanquet cover-abundance scores to a 0-100% continuous form. We also show that previous approaches to transforming ordinal data to the class midpoints will overestimate summed cover.

Different growth forms require different transformation values. Iron bark gum trees, tree ferns, grass trees and palms are scattered throughout the forest in Hat Head National Park, New South Wales. Photo credit: John Spencer/Officer of Environment and Heritage.

We tailored different transformation values for six different growth forms – trees, shrubs, grasses (and grass-like), forbs, ferns and the ‘remaining others’. We show that different growth forms require different transformation values. By applying a tailored transformation to growth forms, we were able to curtail the overestimation of summed cover.

We hope the results of these analyses are useful for others who explore the ecological attributes of plot data, especially in light of the substantial effort undertaken by Bruelheide et al. (2019) to assemble a staggering 23 million records of plant species from around the globe. They calculated that 66% of these records have cover-abundance estimated in an ordinal scale.


Bruelheide, H., Dengler, J., Jiménez-Alfaro, B., Purschke, O. … Zverev, A. (2019). sPlot – a new tool for global vegetation analyses. Journal of Vegetation Science. (https://doi.org/10.1111/jvs.12710)

Different growth forms require different transformation values. Classic outback landscape near Byngnano Ranges, New South Wales. Photo credit: John Spencer/Officer of Environment and Heritage.

This is a plain language summary for the paper of McNellie et al. published in the Applied Vegetation Science (https://doi.org/10.1111/avsc.12437).