An approach to calculate a Species Temperature Index for flora based on open data

Abstract

To study the relation between changes in local plant communities and shifts in temperature, the temperature niches of the species need to be defined. We calculated the Species Temperature Index (STI) as a proxy for this niche. STI is defined as the mean annual temperature within the range of a species. In this paper, a method is described to calculate the STI for the European flora from open data provided by the Global Biodiversity Information Facility (GBIF) and WorldClim global climate data, for 7254 taxa of European vascular plants, bryophytes, algae, and ascomycetes including lichens. The algorithm accounts for incomplete and unbalanced species distribution data. The Community Temperature Index (CTI) is defined as the weighted mean of the STIs of a species assemblage. A 1 × 1 km CTI grid map for the Netherlands is presented as an example of the use of STIs.

Full text

© 2018 Naturalis Biodiversity Center, FLORON & KNBV
You are free to share - to copy, distribute and transmit the work, under the following conditions:
Attribution: You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work).
Non-commercial: You may not use this work for commercial purposes.
No derivative works: You may not alter, transform, or build upon this work.
For any reuse or distribution, you must make clear to others the license terms of this work, which can be found at creativecommons.org/licenses/by-nc-nd/3.0/legalcode. Any of the above conditions can be waived
if you get permission from the copyright holder. Nothing in this license impairs or restricts the author’s moral rights.
Gorteria – Dutch Botanical Archives 40, 2018:
ISSN (online) 2542-8578
INTRODUCTION
Climate change is often regarded as one of the major factors
affecting biodiversity (Mace et al. 2005; Thuiller et al. 2005b, c;
Parmesan 2006). Many species are adapted to a certain tempe-
rature niche (e.g. Dullinger et al. 2009) and climatic change may
act as a selective pressure causing a decline of species that
cannot adapt well to this shift and an increase in species with
high adaptation capabilities and dispersal capacity (Thuiller et
al. 2005a; Devictor et al. 2008; Devictor et al. 2012).
In order to study the relation between changes in local plant
communities and shifts in temperature, the temperature niche
needs to be quantified. A previous study (Thuiller et al. 2005b)
used range maps for a limited number of vascular plants, a
proprietary dataset by Atlas Florae Europaeae. In this paper we
describe the calculation of the Species Temperature Index (STI)
for European vascular plants, bryophytes, algae, and ascomy-
cetes including lichens based on open data. The increasing
number of data published by the Global Biodiversity Information
Facility (GBIF) allows researchers to include many more taxa
in their study. However, species occurrence data in GBIF has
significant data gaps. Therefore, we describe a bootstrapping
method to minimize the effect of biased data.
The Species Temperature Index is defined as the long-term
average temperature within a species range. The Community
Temperature Index (CTI) is defined as the (weighted) average
STI of a species assemblage (Devictor et al. 2008). The CTI
An approach to calculate a Species Temperature Index for flora
based on open data
L. B. Sparrius1, G. G. van den Top2, C. A. M. van Swaay2
Key words
Ascomycotina
Bryophytes
Climate change
GBIF
Open data
Plantae
R
Bootstrapping
STI
WorldClim
Abstract – To study the relation between changes in local plant communities and shifts in temperature, the tempe-
rature niches of the species need to be defined. We calculated the Species Temperature Index (STI) as a proxy for
this niche. STI is defined as the mean annual temperature within the range of a species. In this paper, a method is
described to calculate the STI for the European flora from open data provided by the Global Biodiversity Information
Facility (GBIF) and WorldClim global climate data, for 7254 taxa of European vascular plants, bryophytes, algae, and
ascomycetes including lichens. The algorithm accounts for incomplete and unbalanced species distribution data. The
Community Temperature Index (CTI) is defined as the weighted mean of the STIs of a species assemblage. A 1 × 1 km
CTI grid map for the Netherlands is presented as an example of the use of STIs.
Samenvatting – Om de relatie tussen de veranderingen in lokale plantengemeenschappen en verschuivingen in
de temperatuur te bestuderen is het nodig om de temperatuurniche van de individuele soort te kwantificeren. Wij
berekenden de Species Temperature Index (STI) als een proxy voor deze niche. STI wordt gedefinieerd als de
gemiddelde jaartemperatuur binnen het verspreidingsgebied van een soort. In dit artikel beschrijven we een methode
om de STI voor de Europese flora te berekenen op basis van open data voor 7254 taxa van Europese vaatplanten,
mossen, algen en korstmossen. De methode is ontworpen om te corrigeren voor incomplete of ruimtelijk ongelijk
verdeelde verspreidingsgegevens. De Community Temperature Index (CTI) wordt gedefinieerd als het gewogen
gemiddelde van STI’s voor gemeenschappen van soorten. Een 1 × 1 km CTI rasterkaart voor Nederland wordt
gepresenteerd als voorbeeld van het gebruik van STI’s.
Published on 30 October 2018
SHORT COMMUNICATION
073 – 078
1 FLORON, Toernooiveld 1, 6525 ED Nijmegen; e-mail: sparrius@floron.nl;
2 De Vlinderstichting, Mennonietenweg 10, 6702 AD Wageningen;
e-mail: ggvdtop@gmail.com;
e-mail: chris.vanswaay@vlinderstichting.nl
corresponding author e-mail: sparrius@floron.nl

74 Gorteria – Dutch Botanical Archives: Jaargang 40, 2018
can be used for a trend-analysis if time-series are available,
and for spatial analysis. The STI can be used as a quantita-
tive replacement for the commonly used temperature indicator
values (Ellenberg 1992).
MATERIALS AND METHODS
Study area
Data analysis in this study is restricted to the Western Palaearctic:
Europe, Northern Africa and parts of the Middle East (Fig. 1).
Climate data
We used the WorldClim2 BIO1 dataset (Fick & Hijmans 2017),
which contains the annual average temperature over the period
1970 – 2000 on a 30” grid (Fig. 1).
Species occurrence data
All occurrences of the taxonomical groups Plantae and Asco-
mycota were retrieved from GBIF (GBIF.org 2017). The period
of observations ranges from 1970 – 2017, which was used as
the climate dataset. This resulted in 70.3 million observations
of 67023 taxa. The data shows a highly incomplete and biased
distribution of the study area (Fig 1b), due to factors such as
the lack of monitoring data in remote areas and the fact that
not all available data has been published in GBIF.
Synonyms and accepted taxa
For vascular plants and bryophytes, the Taxonomic Name
Resolution Service v4.0 (Boyle et al. 2013) with data from The
Plant List (2015) was used to identify synonyms and acceptated
taxa based on the “scientificname” field in the GBIF dataset.
For fungi Mycobank (Robert et al. 2005) was used. Occurrence
data of synonyms was merged with records of their accepted
names. In the case of 27010 taxa, the name could not be found
in The Plant List or Mycobank. The “speciesname” field in the
GBIF dataset was then used, ignoring infraspecific taxa and
accepted names according to the GBIF taxonomic backbone.
This procedure was chosen because of a number of incorrect
synonym relations were found in the GBIF taxonomic back-
bone, e.g. Angelica sylvestris L. being treated as a synonym
of Anthriscus sylvestris (L.) Hoffm.
Of the accepted names, only taxa with more than 250 occur-
rences within the study area were selected to avoid errors due
to a lack of data. Handling of large datasets was performed with
the latest versions of EmEditor, MySQL, and QGIS.
STI calculation
In order to correct for the unbalanced data distribution (Fig. 1b),
we used the following bootstrapping approach. At first we super-
imposed a 50 km UTM grid (European Environment Agency
2016) on the observation data and determined for each species
which grid cells are occupied. This step not only makes the data
more workable, but it also removes duplicate observation per
grid cell. This is necessary because the STI is only based on
presence / absence rather than abundance. Next we calculated
the mean annual temperature for each 50 km UTM grid cell.
The normal approach would be to perform a direct calculation
of the mean temperature over the range of each species. Due to
the nature of the data we decided to take a different approach.
A 250 km grid was superimposed on the 50 km UTM grid. Within
each 250 km grid cell we randomly chose one 50 km grid cell
that was occupied by the species of interest. This results in a
more even distribution of 50 km UTM grid cells over the range
of Europe and Northern Africa. These grid cells were used to
calculate the mean annual temperature over the range of the
species by repeating this process a 100 times and averaging
the temperature estimation. The resulting STI is the average
within the standard deviation of these 100 samples. These
steps were performed in R. We published the R-script in a
software repository (Sparrius et al. 2018). For comparison we
also calculated the STI without the bootstrapping approach.
In this case, the same GBIF data was projected on the 50 km
grid cells and the mean temperature of all grid cells together
was calculated.
We applied this STI dataset to the following case:
Plant community temperature index map of the Netherlands
The Plant Community Temperature Index (plant CTI) is here
defined as the average STI of all vascular plants occurring in a
km square, similar to Devictor (2008, 2012). This principle was
applied to distribution data of vascular plants in the Netherlands
(NDFF 2017). The CTI was calculated for each km square with
more than 150 plant species, which we consider as sufficiently
surveyed. No extrapolation to areas with less than 150 species
was performed. We accepted that this selection left out a small
number of grid cells located on the border of the country, grid
cells that fall for a major part in sea or lakes, or are known to
have a low plant diversity, such as extensive inland dunes, dry
heathlands and pine forest.
RESULTS
The STI was calculated for 7254 taxa (Sparrius et al. 2018: data
repository) based on 66 million occurrences. The bootstrapping
method resulted on average in c. 0.27 °C lower STIs (Fig. 2)
than a non-bootstrapped approach. Although only taxa with a
minimum of 250 observations were used, taxa occurring in a
low number of observations (e.g. 1000) tend to have a higher
standard deviation (Fig. 3). The CTI map of the vascular flora
of the Netherlands (Fig. 4) shows a gradient from the warmer
southern and western parts of the country to the colder northeast.
The average temperature indicated by vascular plants is 8.7 °C.
DISCUSSION
Here we show a revised method for the calculation of the STI
over a broad geographical range. This method is better at taking
care of unbalanced observation data than the classical method
without a bootstrapping procedure.
The Species Temperature Index (STI) has proven to be a valuable
tool to study the effect of climate change on birds (Devictor et al.
2008) and butterflies (Devictor et al. 2012). Devictor et al. (2008)
also show that it is a robust parameter in studying the effect on
changes in the Community Temperature Index (CTI), even when
the STI is replaced by its rank number or when the community
data is restricted to presence / absence only. This makes the STI
for plants, as documented in this paper, a valuable additional tool,
which makes it possible to study the effects of climate change
on taxa for which range maps are largely unavailable. The GBIF
occurrence dataset shows a bias towards Western Europe, Medi-
terrean Europe, and Scandinavia, with lower data coverage in
the boreal zone, mountainous areas in Eastern Europe, Russia,
and Northern Africa. Bootstrapping led to lower STIs than the
method without bootstrapping, as all 250 × 250 km regions
were equally sampled. This also becomes visible in Fig 1b, with
many white areas in cooler northeastern part of the study area.

75
Gorteria – Dutch Botanical Archives: 40, 2018
Fig. 1. Maps of (A) the average annual temperature in Europe and (B) the available data of plant diversity for Europe in GBIF (1970 – 2017).
Species diversity
1–300 per 0.1º cell
300–700
700–1000
1000–1300
1300–1616
Mean annual temperature
-14 ºC
-4
6
16
26
A
B

76 Gorteria – Dutch Botanical Archives: Jaargang 40, 2018
Fig. 2. Species Temperature Index (STI) calculated with all available GBIF data versus the bootstrapping approach. Trend line (linear regression) is shown in
red together with its equation. Each data point represents a taxon (n = 7254).
The standard deviation increased with sample size, probably
because common species have a wider distribution area. How-
ever, a low sample size led more frequently to higher standard
deviations. This could be the case for taxa with disjunct distri-
bution, but high standard deviations are more probably caused
by data gaps near the center of the distribution area, or in the
case of poorly known taxa with a wide distribution.
Map
Coastal areas, cities and river valleys show a higher CTI. For
coastal areas, this can be explained by the more temperate
climate near a large water body (Stefan et al. 1998). The tempe-
rature in cities is usually higher than the surrounding areas due
to heating, drought and presence of stone surfaces that easily
heat up in the sun. Western European city centers and can be
up to 2 °C warmer than the surrounding landscape (Santamouris
2007). River valleys are under influence of species dispersal
from upstream areas (Boedeltje et al. 2004; Johansson et al.
1996). In the Netherlands, upstream areas of the Rivers Meuse
and Rhine extend towards the south. This phenomenon is
supported by the fact that small rivers, with a limited catchment
area, do not show up as warm spots on the map. Warm spots
in the river valleys are especially visible in areas with extensive
sand extraction followed by development of new natural areas
with opportunities for pioneer vegetation (Rossenaar et al. 2006).
The average temperature indicated by vascular plants is 8.7 °C.
Meteorological data shows that this was a commonly occurring
annual average before 1980 (KNMI 2018). The average tempe-
rature in the country between 2007 and 2017 was 10.6 °C. This
suggests that the climate in the Netherlands could offer favorable
conditions for many Southern European taxa.
Potential applications
STIs can be used as a more precise replacement for tempe-
rature indicator values (Ellenberg 1992) being a continuous
variable instead of an ordinal scale. STIs can also be used to
predict the effects of climate change on the flora (Bertrand et
al. 2011; Thuiller 2005b), although niche modelling remains
the best solution to predict shifts in species distributions (e.g.
Thuiller 2009). For example, species with a high dispersal
capacity and an STI equal to or slightly higher than the aver-
age temperature within a certain area may be expected there
in future. The STI may also be of use for risk assessments for
invasive species (Thuiller 2005c).
Data quality
As demonstrated in Fig. 3, the accuracy of STIs calculated with
the bootstrap method may improve in future when more data
becomes available. Alternatively, instead of GBIF data, range
maps of e.g. Atlas Florae Europaeae may be used as demon-
strated by Thuiller et al. (2005a), however these maps are not
publicly available and are cover only 20 % of the vascular flora.
Primary data in GBIF is not free of errors either. A small percent-
age of records may have errors in identification or geolocation.
However, this would not dramatically affect the STI when the
sample size is large enough. A minimum sample size of 250
observations was chosen in this study.
With large standard deviations, caused by large distribution
areas, individual STIs must be used with care, but can be
useful in studies where many STIs are combined or related to
other variables.
y = 1.0101x - 0.2786
-10
-5
0
5
10
15
20
25
STI (bootstrapped)
STI (non-bootstrapped)
-5 0510 15 20 25

77
Gorteria – Dutch Botanical Archives: 40, 2018
0
1
2
3
4
5
6
7
8
9
100 1000 10000 100000 1000000
Standard deviation
Number of samples
Fig. 3. The relation between the sample size (number of observations) and standard deviation of a species in the Species Temperature Index (STI) calculations.
Trend line (linear regression) is shown in red.
CONCLUSIONS
The Species Temperature Index for the European flora can be
calculated from open data sources. As more and more data
becomes available, it is likely that STIs can be calculated more
accurately and more species can be included in the calculation.
Acknowledgements – The work presented in this paper has been
made possible by a grant of WWF Netherlands.
REFERENCES
Bertrand R, Lenoir J, Piedallu C, Riofrío-Dillon G, De Ruffray P, Vidal C, Pierrat
J-C, Gégout JC. 2011. Changes in plant community composition lag behind
climate warming in lowland forests. Nature 479: 517– 520.
Boedeltje G, Bakker JP, ten Brinke A, van Groenendael JM, Soesbergen
M. 2004. Dispersal phenology of hydrochorous plants in relation to dis-
charge, seed release time and buoyancy of seeds: the flood pulse concept
supported. J. Ecol. 92: 786 –796; doi: 10.1111/j.0022-0477.2004.00906.x
Boyle B, Hopkins N, Lu Z, Raygoza Garay JA, Mozzherin D, Rees T, Matasci
N, Narro ML, Piel WH, Mckay SJ, Lowry S, Freeland C, Peet RK, Enquist
BJ. 2013. The taxonomic name resolution service: an online tool for auto-
mated standardization of plant names. BMC Bioinformatics 14:16. doi:
10.1186/1471-2105-14-16.
Devictor V, Julliard R, Couvet D., Jiguet F. 2008. Birds are tracking climate
warming, but not fast enough. Proc. Roy. Soc. Biol. Sci. Ser. B 275:
2743 – 2748.
Devictor V, Van Swaay C, Brereton T, Brotons L, Chamberlain D , Heliölä
J, Herrando S, Julliard R, Kuussaari M, Lindström Å, Reif J, Roy DB ,
Schweiger O, Settele J, Stefanescu C, Van Strien A, Van Turnhout C,
Vermouzek Z, WallisDeVries M, Wynhoff I, Jiguet F. 2012. Differences in
the climatic depts of birds and butterflies at a continental scale. Nature
Climate Change 2: 121–124.
Dullinger S, Kleinbauer I, Peterseil J, Smolik M, Essl F. 2009. Niche based
distribution modelling of an invasive alien plant: effects of population status,
propagule pressure and invasion history. Biol. Invas. 11: 2401– 2414; doi:
10.1007/s10530-009-9424-5.
Ellenberg H, Weber HE , Düll R, Wirth V, Werner W, Paulißen D. 1992. Zeiger-
werte von pflanzen in Mitteleuropa. Verlag Erich Goltze KG, Göttingen.
European Environment Agency. 2016. EEA reference grids. See http://eea.
europa.eu/data-and-maps/data/eea-reference-grids.
Fick SE, Hijmans RJ. 2017. WorldClim 2: new 1-km spatial resolution climate
surfaces for global land areas. International Journal of climatology (in
press). doi: 10.1002/joc.5086
GBIF.org. 20th March 2017. GBIF occurrence download http://doi.
org/10.15468/dl.qw9lpv
Johansson ME, Nilsson C, Nilsson E. 1996. Do rivers function as corridors
for plant dispersal? J. Veg. Sci. 7: 593 –598; doi: 10.2307/3236309.
KNMI. 2018. Weerstatistieken, accessed through https://weerstatistieken.nl/
Mace G, Masundire H, Baillie JEM. 2005. Biodiversity. In: Hassan R, Scholes R, Ash
N (eds), Ecosystems and human well-being: Current state and trends: findings
of the condition and trends working group: 77–122. Island Press, Washington.
NDFF. 2017. Vascular plant distribution data download http://www.
verspreidings atlas.nl/.
Parmesan C. 2006. Ecological and evolutionary responses to recent climate
change. Annual Rev. Ecol. Evol. Syst. 37: 637– 669.
Robert V, Stegehuis G, Stalpers J. 2005. The MycoBank engine and related
databases. See: http://www.mycobank.org (accessed April 2017).
Rossenaar AJGA, Odé B, Beringen R. 2006. Natuurontwikkeling en flora
langs de grote rivieren. Levende Natuur 107: 237– 241.
Santamouris M. 2007. Heat island research in Europe: the state of the art.
Advances in Building Energy Research 1: 123 –150.
Sparrius LB, van Strien AJ. 2014. Het berekenen van jaarlijkse trends van
planten op basis van verspreidingsgegevens. Gorteria 37: 31– 40.
Sparrius LB, van der Top G, van Swaay CAM. 2018. R script for species
temperature index with open data (version 1.0). Zenodo. See http://doi.
org/10.5281/zenodo.1155850.
Stefan HG, Fang X, Hondzo M. 1998. Simulated climate change effects on
year-round water temperatures in temperate zone lakes. Climatic Change
40: 547– 576.
The Plant List. 2013. Version 1.1. Published on the Internet; http://www.
theplantlist.org/ [Accessed: 1 June 2017].
Thuiller W, Lavorel S, Araújo MB. 2005a. Niche properties and geographical
extent as predictors of species sensitivity to climate change.

78 Gorteria – Dutch Botanical Archives: Jaargang 40, 2018
CTI (mean) ºC
< 8.5
8.5–8.6
8.6–8.8
8.8–8.9
< 8.9
Fig. 4. Map of the Netherlands with the Community Temperature Index (CTI) of plants expressed as the mean Species Temperature Index (STI) for all plants
found in km squares.
Thuiller W, Lavorel S, Araújo MB, Sykes MT, Prentice IC. 2005b. Climate
change threats to plant diversity in Europe. PNAS 102: 8245 – 8250; doi:
10.1073/pnas.0409902102.
Thuiller W, Richardson DM, Pyšek P, Midgley GF, Hughes GO, Rouget M.
2005c. Niche-based modelling as a tool for predicting the risk of alien plant
invasions at a global scale. Global Change Biology 11: 2234 – 2250; doi:
10.1111/j.1365-2486.2005.001018.x
Thuiller W, Lafourcade B, Engler R, Araujo MB. 2009. BIOMOD – A platform
for ensemble forecasting of species distributions. Ecography 32: 369 – 373.
van Strien AJ, van Swaay CAM, Termaat T. 2013. Opportunistic citizen
science data of animal species produce reliable estimates of distribution
trends if analysed with occupancy models. J. Appl. Ecol. 50: 1450 –1458;
http://dx.doi.org/10.1111/1365-2664.12158.