Fine-resolution patterns of plant species richness

across European forests

Martin Veea1, Milan Chytr1, Jan Divek1,2 & data contributors*
1Department

of Botany and Zoology, 2Department of Geography, Faculty of Science, Masaryk University, Brno, Czech Republic.
Corresponding author: M. Veea (martinvec@seznam.cz)

Introduction
Up to now, most studies examining species richness of vascular plants have used coarse-resolution (i.e. spatially highly generalized) data based on atlases or inventories of large areas, which considerably limits our
understanding of species-richness patterns and underlying factors. In this study, we analysed vascular plant species richness in forest vegetation across major part of Europe using a large set of spatially referenced
vegetation plots (phytosociological relevs).

Aims
Our aims were: (1) to map species richness of vascular plants in European forests; (2) to examine environmental
factors determining the species-richness pattern; (3) to create a fine-resolution predictive grid map of vascular plant
species richness in European forests

Material
We used data on species richness from 27 national or regional vegetation-plot databases gathered in
the European Vegetation Archive. In total, we obtained 98 363 spatially referenced forest-vegetation plots with
sampling sizes of 1001000 m2. However, only 2355% of these plots appeared to be suitable for analyses, depending
on the filtering criteria applied. We considered the following climatic, topographical, soil-geological and landscapestructure factors which might affect species richness at the continental scale:
Variable
Mean annual temperature
Annual precipitation
Mean annual potential evapotranspiration
Mean annual actual evapotranspiration
Altitude
Terrain ruggedness index
Topsoil pH measured in water

Description, data source

Annual temperature range

Raster data ~ 1 1 km, Hijmans et al. (2005)
Precipitation seasonality
Raster data ~ 1 1 km, Zomer et al. (2008)
Raster data ~ 1 1 km, Trabucco et al. (2010)
Derived from digital elevation model ~ 90 90 m, Jarvis et al. (2008)
Based on USGS (1996), raster data ~ 1 1 km, Riley et al. (1999)
Raster ~ 4.3 4.3 km, Wieder et al. (2014)
Area of occupancy within a buffer, vector data 1:1 000 000,
Main geological formations
Calcareous substrate
derived from CEC (2004), Panagos et al. (2012)
Acid magmatic and metamorphic rocks, Basic magmatic and metamorphic rocks, Limestone and other calcareous
rocks, Loess, Sand, Organic substrate, Sedimentary rocks, Other non-consolidated sediments

Species richness of vascular plants in European forests in the final data set (49 433 vegetation plots)

Land cover categories

Area of occupancy within a buffer, vector data 1:100 000, CEC (2000)
Forests, Grassland, Arable land, Intensively cultivated land (i.e. arable land, rice fields, vineyards, fruit plantations, olive groves, annual crops associated with
permanent crops), Mosaic (semi-opened) land (i.e. pastures, natural grasslands, complex cultivation patterns, cultivated land with significant areas of natural vegetation,
agro-forestry areas), Shrub-land (i.e. moors and heathlands, sclerophyllous vegetation and transitional woodland-scrub)

Methods
Data processing followed these steps:
(1) the numbers of species in each plot were counted; (2) plots older than 1970 were removed;
(3) plot sizes between 100 and 400 m2 were selected; (4) the number of species in each plot
was rescaled to 400 m2 according to species-area curve; (5) average number of species was
claculated in plots with identical geographical coordinates; (6) a circular buffer zone of an area
of approx. 25 km2 was created around each plot; (7) plots with location imprecision higher than
2821 m (the radius of the buffer zone) were removed; (8) buffer zones occupied by more than
50% of sea were removed; (9) average value for spatially continuous variables within each
buffer zone was claculated; (10) the area of occupancy for categorical spatially discontinuous
variables within each buffer zone was calculated.
Following this procedure we obtained the final dataset containing 49 433 records.
Subsequently we calculated a predictive model using the Random Forest method. Results were
evaluated by visual inspection of the maps and histograms of residuals and by calculating
Global Morans I of residuals. Finally we created a predictive map of vascular plant species
richness of European forests with a 5 5 km resolution.
Precipitation seasonality
Forests
Shrub land
Annual precipitation
Mean annual PET
Grassland
Mosaic land
Mean annual AET
Temperature annual range
Terrain ruggedness index
Altitude
Mean annual temperature
Other non-consolidated sediments
Sedimentary rocks
Intensively cultivated land
Arable land
Topsoil pH in water
Sand
Acid magmatic and metamorphic rocks
Calcareous substrate
Limestone and other calc. rocks
Loess
Basic magmatic and metamorphic rocks
Organic substrates

Terrain ruggedness index

Mean annual PET
Mean annual AET
Annual precipitation
Temperature annual range
Altitude
Precipitation seasonality
Forests
Calcareous substrate
Mosaic land
Limestone and other calc. rocks
Grassland
Shrub land
Mean annual temperature
Topsoil pH in water
Intensively cultivated land
Arable land
Acid magmatic and metamorphic rocks
Sedimentary rocks
Other non-consolidated sediments
Sand
Loess
Basic magmatic and metamorphic rocks
Organic substrates

20 40 60 80

120

Mean decrease in accuracy

Results

200

400

600

800

Mean decrease in node impurity (/1000)

Predictive grid map of vascular plant species richness in European forests.

The map is based on UTM grid cells, each of them spanning 5 5 km

(species richness is predicted for an area of 400 m2).

Random Forest Result

No. of cases
No. of explanatory variables
No. of regression trees
No. of variables tried at each split
Variation explained (%)
Mean of squared residuals
Morans I of residuals

49 433
24
500
8
51.7
102.9122
-0.04, p < 0.001

The random forest model explained 51.7% of variation in species richness. Residuals had almost normal distribution and they did not show any obvious pattern when we
plotted them to the map. The relative importance of explanatory variables was expressed by two different measures called Mean decrease in accuracy and Mean decrease in
node impurity. The higher the value of the first measure, the lower the chance that the effect of the selected variable on species richness is random. The second measure
shows how good predictor of the species richness the selected variable is. Variables having high values of both measures can be considered as the most important predictors.
In our case, these are both variables related to precipitation, area of forests in plots surroundings, mean potential and actual evapotranspiration. The most reliable (i.e. least random) predictor was precipitation
seasonality. The best predictor according to the second measure was terrain ruggedness index.
The map suggests that the richest forests occur in mountainous regions with high proportion of calcareous bedrock and relatively high values of actual evapotranspiration. European forest diversity hotspots include
the Limestone Alps, Western Alps and Jura Mountains, Southern and Northern Dinarides, and Western Carpathians in Slovakia. Other hotspots are located at the south-eastern foothills of the Romanian Carpathians, on
the Transylvanian Plateau, or in the north-eastern part of Poland. In contrast, species-poor forests predominate in north-western Europe, e.g. on the British Isles and in the Netherlands and adjacent lowland areas from
France to north-western Poland.
These are just preliminary results which we are going to improve in the near future by adding more vegetation plots and by cross-validation testing.

Emiliano Agrillo, Pierangela Angelini, Fabio Attorre, Christian Berg, Idoia Biurrun, Henry Brisse, Laura Casella, Panayotis Dimopoulos, Jrg Ewald, na FitzPatrick, Itziar Garca-Mijangos, Stephan Hennekens, Adrian Indreica, Ute Jandt, Florian Jansen, Zygmunt Kcki, Martin Kleikamp,
Vitaliy Kolomiychuk, Daniel Krstonoi, Flavia Landucci, Jonathan Lenoir, Vassiliy Martynenko, Dana Michalcov, Viktor Onyshchenko, Hristo Pedashenko, Valerijus Raomaviius, John Rodwell, Patrice de Ruffray, Gunnar Seidler, Joop Schamine, Jens-Christian Svenning, Grzegorz
Swacha, Jozef ibk, Urban ilc, eljko kvorc, Ioannis Tsiripidis, Pavel Dan Turtureanu, Domas Uogintas, Milan Valachovi, Kiril Vassilev, Roberto Venanzoni, Lynda Weekes, Wolfgang Willner & Thomas Wohlgemuth
Acknowledgements: We thank Ilona Knollov for her help with the data preparation, Frantiek Kuda and Petr Vybral for help with GIS and all the botanists whose vegetation plots we could use for analysing species richness. This study was supported by the Czech Science Foundation
(project no. 14-36079G, Centre of Excellence PLADIAS).
References: CEC, 2000. CORINE land cover technical guide Addendum 2000. Copenhagen, European Environment Agency; CEC, 2004. The European Soil Database distribution version 2.0, European Commission and the European Soil Bureau Network (CD-ROM) EUR 19945 EN.; Hijmans R.J., Cameron S.E., Parra J.L., Jones P.G., Jarvis A., 2005. Very high resolution interpolated climate surfaces for global land
areas. International Journal of Climatology 25, 1965-1978; Jarvis A., Reuter H.I., Nelson A., Guevara E., 2008. Hole-filled SRTM for the globe Version 4. Available from the CGIAR-CSI SRTM 90m Database: http://srtm.csi.cgiar.org/; Panagos P., Van Liedekerke M., Jones A., Montanarella L., 2012. European Soil Data Centre: Response to European policy support and public data requirements. Land Use Policy, 29
(2), 329-338; Riley S.J., DeGloria S.D., Elliot R., 1999. A terrain ruggedness index that quantifies topographic heterogeneity. Intermountain Journal of Sciences 5(1-4), 23-27; Trabucco A., Zomer R.J., 2010. Global Soil Water Balance Geospatial Database. CGIAR Consortium for Spatial Information; US Geological Survey (USGS), 1996. GTOPO30. Sioux Falls, SD: United States Geological Survey Center for Earth
Resources Observation and Science (EROS); Wieder W.R., Boehnert J., Bonan G.B., Langseth M., 2014. Regridded Harmonized World Soil Database v1.2. Data set. Available on-line [http://daac.ornl.gov] from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, Tennessee, USA; Zomer R.J., Trabucco A., Bossio D.A., van Straaten O., Verchot L.V., 2008. Climate Change Mitigation: A
Spatial Analysis of Global Land Suitability for Clean Development Mechanism Afforestation and Reforestation. Agric. Ecosyst. Environm. 126, 67-80.