Plants Vote: European Indicator Values Win When Every Species Counts Equally

For half a century, ecologists across Europe have relied on a deceptively simple trick to read the environment through plants. Instead of hauling pH meters and temperature loggers to every site, they assign each plant species an expert-judged score for how much soil acidity, moisture, nutrients, or warmth it tolerates, then average those scores across all species found in a vegetation plot. The resulting mean ecological indicator value, or EIV, serves as a cheap and fast proxy for conditions that would otherwise require laborious field measurement. A new study, published in the journal Vegetation Classification and Survey, has now put this method to one of its most rigorous tests to date, and the results overturn two long-standing assumptions about how the averaging should be done.

The research, conducted around the village of Preda in the Swiss canton of Grisons within the regional nature park Parc Ela, used three datasets of vegetation plots paired with hard measurements of soil pH or mean annual near-surface air temperature. The plots spanned elevations from 1,738 to 2,636 metres above sea level, covering forests, grasslands, heathlands, mires, and screes on both limestone and siliceous bedrock. Vegetation was recorded in precisely delimited square plots of one and ten square metres, with all vascular plants listed and their percentage cover estimated separately for the tree, shrub, and herb layers. Soil samples were taken from the top ten centimetres, air-dried, sieved, and measured with a pH electrode, while miniature temperature loggers recorded conditions at thirty-minute intervals both below and above the soil surface for a full year.

Against this measured ground truth, the team calculated mean indicator values using four competing EIV systems that all cover the Alps: the classic Ellenberg system from Central Europe, the Landolt system tailored to Switzerland, the new pan-European Ecological Indicator Values for Europe (EIVE), and the Ellenberg-type indicator values published by Tichý and colleagues in 2023. Each system was combined with four weighting approaches: a simple unweighted average, a mean weighted by species cover, one weighted by the square root of cover, and, where the system allowed it, a weighting by inverse niche width. This produced forty-two combinations of dataset, system, and weighting, each of which was correlated with the measured environmental variable using Pearson and Spearman correlation coefficients, then compared with linear mixed effects models.

The clearest finding concerned cover-weighting, the intuitive idea that a species covering more ground should count more heavily in the average. The data said otherwise. Unweighted means, which treat every species equally regardless of how abundant it is, achieved a mean correlation of 0.78 with measured conditions, slightly better than square-root cover-weighting at 0.75 and clearly better than full cover-weighting at 0.68. When the comparison was restricted to the two systems offering niche width information, unweighted and inverse-niche-width-weighted means tied at 0.79, with full cover-weighting trailing at 0.65. In other words, the more influence cover was given, the worse the predictions became, a result that contradicts the widespread belief that weighted means are more informative, particularly in species-poor communities.

The explanation the authors propose is a statistical principle known as the wisdom of the crowd. Every species indicator value carries an estimation error, because even the best expert sometimes overestimates or underestimates a species’ ecological preferences. When many species enter the average, these errors tend to cancel out, making the mean more reliable than any single judgement. Cover-weighting effectively reduces the number of species that meaningfully contribute to the calculation, since dominant species dominate the mean while rare but highly indicative species are drowned out. A sprawling beech tree can hardly cover less than one percent, while a delicate flax species like Linum catharticum may never exceed fifty percent cover, yet its mere presence can carry strong ecological signal. Counting presence rather than abundance lets the full community cast its vote.

The same crowd principle may explain the second major finding: the performance of the EIV systems themselves. EIVE achieved the highest mean correlation at 0.76, significantly better than the Tichý system at 0.73, while the differences to Ellenberg at 0.75 and Landolt at 0.71 were not statistically significant. This is remarkable because the two new European systems were never intended to replace regional ones. The authors of the Tichý system had explicitly assumed that regional systems would provide more accurate estimates within their own definition areas, since the European dataset averages values across different regions. The Preda data suggest the opposite: pooling many independent expert assessments produces better species-level indicator values than relying on fewer, conceptually similar sources.

The numbers behind this argument are striking. EIVE version 1.0 drew on 28 source systems for soil reaction and 23 for temperature, whereas the Tichý system used only seven for each dimension, and those sources were deliberately chosen for their conceptual similarity to the Ellenberg tradition, making them hardly independent from one another. By including every available system, however eccentric, EIVE maximised the diversity and independence of opinion that the wisdom of the crowd requires. The study’s authors had hypothesised before the analysis that EIVE would outperform its rivals, based on earlier work showing that EIVE’s temperature values correlated most strongly with species’ climatic niches derived from distribution data, and the new field calibration largely confirmed that expectation.

Practical advantages reinforce the statistical ones. In the study’s datasets, EIVE provided definitive indicator values for 100 percent of the 371 valid taxa recorded, while the older systems left between 12.1 and 40.4 percent of taxa without values, either because species were considered too indifferent to score, were entirely missing, or were only covered at a different taxonomic level. Only EIVE systematically assigns values to subspecies, species, and aggregates alike, allowing direct application without tedious and imperfect manual matching. For large-scale analyses, and for regions lacking a well-established regional system, the authors conclude that EIVE is currently the best choice in Europe. Where a strong regional system exists, the choice between regional and European values can depend on whether comparisons are intended to be regional or international.

The study also delivered calibration equations that translate mean EIVE values into physical units, finding relationships between mean reaction values and soil pH, and between mean temperature values and mean annual near-surface air temperature, that were close to linear. Correlations were strongest in the ten-square-metre pH dataset, likely because larger plots contain more species and thus more reliable means, and weaker for temperature, where the sampled gradient covered only about twenty percent of the European range compared with roughly half for soil reaction. The authors caution that their evidence comes from a narrow geographic region with three limited datasets, and they recommend similar calibration studies in other regions and for the remaining niche dimensions of moisture, nitrogen, and light. They also suggest that improved niche width measures in future editions of EIVE, or novel approaches such as multiplying the response curves of all co-occurring species, might eventually push predictive power beyond what simple unweighted averaging can achieve.

For a method that has quietly underpinned decades of vegetation science, from tracking climate-driven vegetation change to comparing management regimes and reconstructing past environments, these findings arrive as both a correction and a validation. The correction is that weighting by abundance, long considered the more sophisticated option, systematically weakens the signal. The validation is that averaging the judgement of many independent experts, whether human contributors to a continental database or the plants of an entire plot, remains one of ecology’s most powerful shortcuts. In the alpine meadows of Preda, as perhaps across Europe, the crowd knows best, and every species gets exactly one vote.

Subject of Research: Calibration and comparison of plant ecological indicator value systems and averaging methods against measured soil pH and temperature

Article Title: Mean ecological indicator values: use EIVE but no cover-weighting

Article References: Ostrowski, G., Aicher, S., Mankiewicz, A., Chusova, O., Dembicz, I., Widmer, S., & Dengler, J. (2025). Mean ecological indicator values: use EIVE but no cover-weighting. Vegetation Classification and Survey, 6, 57-67. https://doi.org/10.3897/vcs.134800

Image Credits: AI Generated

DOI: 10.3897/VCS.134800

Keywords: ecological indicator values, EIVE, Ellenberg indicator values, vegetation ecology, soil pH, bioindication, cover-weighting, wisdom of the crowd, Swiss Alps, niche width, plant communities, environmental calibration