Predicting range expansion by thaiv2012


									Biodivers Conserv
DOI 10.1007/s10531-007-9287-y


Predicting range expansion of the map butterfly
in Northern Europe using bioclimatic models

Varpu Mitikka Æ Risto K. Heikkinen Æ Miska Luoto Æ Miguel B. Arau Æ
Kimmo Saarinen Æ Juha Poyry Æ Stefan Fronzek

Received: 22 March 2007 / Accepted: 24 October 2007
Ó Springer Science+Business Media B.V. 2007

Abstract The two main goals of this study are: (i) to examine the range shifts of a
currently northwards expanding species, the map butterfly (Araschnia levana), in relation
to annual variation in weather, and (ii) to test the capability of a bioclimatic envelope
model, based on broad-scale European distribution data, to predict recent distributional
changes (2000–2004) of the species in Finland. A significant relationship between annual
maximum dispersal distance of the species and late summer temperature was detected.
This suggests that the map butterfly has dispersed more actively in warmer rather than
cooler summers, the most notable dispersal events being promoted by periods of excep-
tionally warm weather and southerly winds. The accuracy of the broad-scale bioclimatic
model built for the species with European data using Generalized Additive Models (GAM)
was good based on split-sample evaluation for a single period. However, the model’s
performance was poor when applied to predict range shifts in Finland. Among the many
potential explanations for the poor success of the transferred bioclimatic model, is the fact

V. Mitikka Á R. K. Heikkinen (&) Á J. Poyry
Research Department, Research Programme for Biodiversity, Finnish Environment Institute,
P.O. Box 140, 00251 Helsinki, Finland

V. Mitikka
Metapopulation Research Group, Department of Biological and Environmental Sciences,
University of Helsinki, P.O. Box 65, 00014 Helsinki, Finland

M. Luoto
Thule Institute, University of Oulu, P.O. Box 7300, 90014 Oulu, Finland

M. B. Araujo
Department of Biodiversity and Evolutionary Biology, National Museum of Natural Sciences, CSIC,
      ´     ´
C/ Jose Gutierrez Abascal 2, 28006 Madrid, Spain

K. Saarinen
South Karelia Allergy and Environment Institute, 55330 Tiuruniemi, Finland

S. Fronzek
Research Department, Research Programme for Global Change, Finnish Environment Institute,
P.O. Box 40, 00251 Helsinki, Finland

                                                                                Biodivers Conserv

that bioclimatic envelope models do not generally account for species dispersal. This and
other uncertainties support the view that bioclimatic models should be applied with caution
when they are used to project future range shifts of species.

Keywords Bioclimatic envelope model Á Climate change Á Dispersal jump Á
Model accuracy Á Range shift Á Species distribution modelling

AUC Area under curve of a receiver operating characteristic (ROC) plot
GAM Generalized additive models


Recent studies suggest that many habitats and species are already being affected by climate
change (Parmesan et al. 1999; Walther et al. 2002; Parmesan and Yohe 2003; Parmesan
2006) and that future projected climate change is expected to cause further range shifts of
species (Berry et al. 2002; Midgley et al. 2003; Thomas et al. 2004; Thuiller et al. 2005).
In northern Europe, species are generally predicted to move northwards to track the
changing climate (Bakkenes et al. 2002; Hill et al. 2002, 2003; Skov and Svenning 2004).
   Bioclimatic envelope models have an important role in the assessments of the potential
magnitude and broad patterns of climate change impacts on species distribution (Gitay
et al. 2001; Beaumont and Hughes 2002; Midgley et al. 2002; Pearson and Dawson 2003;
Araujo et al. 2005a). These techniques correlate current distributions of species with cli-
mate variables to then derive a species’ climate envelope, which enables future potential
distributions to be estimated (Bakkenes et al. 2002; Berry et al. 2002; Pearson and Dawson
2003; Thuiller 2003). A fundamental assumption of bioclimatic envelope models is that a
spatial correlation of species distribution vs. climate can be applied to infer spatial shifts in
distribution over time, as the climate changes. The approach is thus based on a realised
species distribution rather than its potential distribution, which may cause uncertainty in
the projections of species future ranges.
   The usefulness of bioclimatic models has recently been questioned on the grounds that
several sources of uncertainty can significantly decrease the accuracy of their predictions.
Sources of uncertainty range from choice of the modelling technique to effects of land
cover and biotic interactions on species distributions, and impacts of scale and species
characteristics on the model performance (Hill et al. 1999; Kadmon et al. 2003; Hampe
2004; Thuiller et al. 2004; Luoto et al. 2005, 2007; Heikkinen et al. 2006; Pearson et al.
2006). Recently, some progress has been made in order to take these uncertainty sources
into account (e.g. Pearson et al. 2004; Lawler et al. 2006; Midgley et al. 2006; Araujo and
New 2007). However, understanding of the capabilities of bioclimatic models for pro-
viding reliable range shift projections in real-life situations is still limited (Araujo and
Guisan 2006; Araujo and Rahbek 2006; Barry and Elith 2006). Improvements in under-
standing are essential if models are to be used for assessing climate change impacts on
biodiversity and for implementing conservation planning strategies under climate change
(Araujo et al. 2005c). This is one of the core elements of the Integrated Project ALARM,
which has the particular objective to test methods of risk assessment for biodiversity

Biodivers Conserv

(Settele et al. 2005). This will be done under different scenarios of future global change,
which include effects of climate and land use changes as well as socio-economic drivers
behind them (Spangenberg 2007).
    We focus here on three critical issues that have been insufficiently studied in the field of
bioclimatic envelope modelling. First, model validation has generally been conducted
using either a resubstitution approach (the data used to calibrate the model are also used to
validate it) or split-sample approach (i.e. dividing the species–environment data randomly
into two subsets, for model calibration and model evaluation (for a review see Araujo et al.
2005a). Both of these approaches are likely to yield overly optimistic estimates of the
model performance in new areas and time periods. Indeed there are very few instances
where bioclimatic models have been tested with independent data, i.e. calibrating the
model with data collected at one point in time and validating the model using data recorded
at another point in time (see also Hill et al. 1999; Araujo et al. 2005a).
    Second, very few studies have performed downscaling of the projections of broad-scale
bioclimatic models to finer spatial resolutions and evaluated the accuracy of the regional
predictions of species distributions (Pearson et al. 2002, 2004; Araujo et al. 2005b;
McPherson et al. 2006), even though finer spatial scales would be better suited for the
purposes of conservation planning.
    Third, none of the studies that validated model projections using independent data,
have, to our knowledge, investigated the possible reasons for the success or failure of
models in predicting the distributions of species in different periods of time. Gathering
such information from empirical studies is important because it increases understanding of
the limitations of bioclimatic models.
    Here we use the European map butterfly (Araschnia levana) as a model system to
examine and model current range shifts. Being a mobile species, not in need of any
particularly rare habitat and with a host-plant widely distributed throughout Europe, this
butterfly species is well-suited for studying potential range shifts under climate change. We
develop a bioclimatic envelope model for the butterfly using broad-scale European dis-
tribution and climate data from the period 1961 to 1995, and validate the model, first, with
a subset of the European data, and then with current species’ distribution and climate data
from Finland (2000–2004).
    Specifically, we addressed the following questions:

(i) How well does the bioclimatic envelope model perform when using a truly
     independent evaluation data collected at a different resolution and time period?
(ii) How suited is the bioclimatic envelope modelling approach for predicting the
      observed range expansion of map butterfly in Finland?
(iii) If the validation results from the European model and downscaled model largely
       disagree, what might the key reasons behind this discrepancy be?
   In addition to these questions we document the history of expansion of the map butterfly
in Finland in relation to the selected climate variables. For mobile generalist species that
are more likely to respond rapidly to changes in climate, such as the map butterfly (cf.
Thomas et al. 2001), strongest periods of range expansion may be associated with one or
two climatically extreme years rather than the mean climate values averaged often over
several years or decades. Therefore we examine whether the years in which the butterfly
has more strongly expanded its range are climatically different from other years, and
whether these potential annual differences in range expansion rate are likely to hamper the
attempts to predict the species response to the changing climate.

                                                                              Biodivers Conserv

Materials and methods

Study species

The European map butterfly (Araschnia levana) has a Palaearctic distribution extending
from Europe to the Russian Far East and Japan (Reinhardt 1972; Kudrna 2002). The range
boundaries in Western Europe reach the Atlantic coast in France, Netherlands and Den-
mark, and in Southern Europe in Spain and northern Greece. The species has actively
expanded its range during the twentieth century, especially during recent decades where
both westward and northward expansions have been recorded (Parmesan 2005). In Finland,
the map butterfly was observed for the first time in 1973 and a rapid northward expansion
has been occurring since the 1980s.
    The map butterfly is a nymphalid species that has two distinct and seasonally poly-
morphic generations per summer (Reinhardt 1972; Fric and Konvicka 2002). The first
(spring) generation is orange with black markings, whereas the second generation
(f. prorsa) is black and white. In central Europe an additional third generation in late
summer is common (Reinhardt 1972; Marttila et al. 1990). In Finland, the second gen-
eration has been regularly observed from 1999 onwards, coinciding with recent warm
    The map butterfly favours semi-open and predominantly moist habitats, like meadows
along riverbanks, pastures surrounded by forest and forest openings (Marttila et al. 1990).
The larval host plant is common nettle (Urtica dioica). Both suitable habitats and host
plants are commonly found all over the European distribution range. The species is mobile
and a rather fast flier, and thus regarded as a relatively good disperser (Marttila et al. 1990;
Fric and Konvicka 2000, 2002; Komonen et al. 2004).

The study area

Two different geographical windows were used in this study. First we applied a 0.5°91°
resolution latitude-longitude grid extending from SW to NE across Europe from 35.5°N,
10.0°W to 71.5°N, 40.0°E, which was used to calibrate the broad-scale butterfly-climate
model. The mean annual temperature (based on grid box averages assuming mean altitude)
in this window for the period of 1961–1995 varies from ca. -6°C in the Fennoscandian
mountains to ca. 19°C in parts of the Mediterranean, and the mean annual precipitation
from ca. 250 mm in Mediterranean regions of Spain and Turkey to ca. 2800–2900 mm in
Scotland and on the western coast of Norway. Second, we used a 10 9 10 km2 resolution
gridded national window for Finland, extending from 59.5°N, 19.5°E to 70.5°N, 32.0°E.
This window was used to evaluate the downscaled model projections for the butterfly.
Finland’s climate becomes more continental away from the coasts and eastwards (Tuhkanen
1984), and rainfall and temperature decrease from the southwestern hemiboreal zone (mean
annual temperature ca. 5°C and mean annual precipitation 600–700 mm during 1961–1995)
to the subarctic region in northernmost Finland (-2°C and 400 mm, respectively).

Species distribution data

The distribution data of the map butterfly for the European window was mainly extracted
from Kudrna (2002). The species was recorded as present in each of the 2298 0.5°N 9 1°E

Biodivers Conserv

grid cells in which it had been recorded after 1950, and the records made exclusively
earlier than that were omitted (Kudrna 2002). However, because the distribution data for
the species were rather sparse in the eastern Europe, we supplemented the data with
distribution notes from the literature and information from specialists for the following
areas: Estonia (Keskula 1992; Tammaru, pers. comm. 2005), Russian Karelia (Gorbach,
pers.comm. 2005; Humala, pers.comm. 2005), the St Petersburg region (Ivanov 1999),
Belarus (Dovgailo et al. 2003), and Ukraine (Popov 2005).
   For documentation on the expansion history and the records of species distribution in
Finland, we used data from the National Butterfly Recording Scheme in Finland (NAFI)
for 1991–2004. The NAFI data is collected using a 10 9 10 km2 grid system from vol-
untary amateur and professional lepidopterists every year based on a uniform questionnaire
for the whole country (Saarinen et al. 2003). The recording intensity (number of recording
days per grid square) is included for the NAFI data (Saarinen 2003). The NAFI data were
complemented with observations collected by the Finnish Lepidopterological Society and
inquiries to individual lepidopterists, in order to gather as detailed an understanding of the
annual range shifts of the butterfly as possible.

Climatological data

Mean monthly temperature and monthly precipitation for the European study area was
extracted from the Climatic Research Unit (CRU) grid-based, interpolated climatological
data set (New et al. 2002; Mitchell et al. 2003). Averages for the time period 1961–1995
were linearly interpolated from the original 0.5° 9 0.5° spatial resolution to the grid cell
size of the butterfly distribution data, 0.5° latitude 9 1° longitude. Interpolated mean
monthly climatological data (temperature and precipitation) for Finland, including data for
all individual years between 1973 and 2004 and data averaged for the period 2000–2004,
were provided by the Finnish Meteorological Institute on a uniform 10 9 10 km2 grid
system matching the national distribution data for the butterfly (Venalainen and Hei-
kinheimo 2002).
    The procedures adopted for climate envelope modelling followed Hill et al. (2003) and
Luoto et al. (2006), focusing on three climate variables (Table 1) known to be important
determinants of the distributions of European butterflies: mean temperature of the coldest
month (MTCO), growing degree days above 5°C (GDD) and water balance (WB). MTCO
is related to the overwintering survival, GDD is regarded as a surrogate for the develop-
mental threshold of the larvae, and WB corresponds to the moisture availability for the
larval host and adult nectar plants (Hill et al. 2003). GDD values were calculated as the
accumulated sum of daily mean temperatures above 5°C (see Luoto et al. 2004). The water
balance (WB) was calculated as the annual sum of the monthly differences between
precipitation and potential evapotranspiration following Holdridge (1967) and Skov and
Svenning (2004). An additional climate variable, which was used in studying the rela-
tionships between climate and the expansion history of the butterfly in Finland, was the
mean temperature of late summer, i.e. July and August (TMPJA; Table 1).

Expansion history in Finland and climate

The expansion history of the map butterfly in Finland was examined as annual changes in
the distribution from 1983 to 2004. The measures of expansion used were: (i) maximum

                                                                                        Biodivers Conserv

Table 1 Climate variables used in bioclimatic envelope modelling for the map butterfly and in analysing
relationships between the annual variation in climate and expansion activity of the species
Variable                              Abbreviation           Definition

Mean temperature of the               MTCO                   The lowest value of the monthly mean
 coldest month                                                 temperature in one year
Growing degree days above             GDD
Water balance                         WB                     WB ¼         ðPi À PETÞ where, Pi = mean
                                                               precipitation in month i PETi = mean
                                                               potential evapotranspiration in month
                                                               i = (58.939Ti)/12[if Ti [ 0°C, else
                                                               PETi = 0] where, Ti = mean temperature
                                                               in month i
Mean temperature of the               TMPJA                  The average of summed monthly mean
 summer months July and                                        temperatures of July and August
Abbreviations are those referred to in the text. The source of the expression for PET = Skov and Svenning

annual dispersal distance, (ii) annual average northward distribution shift, (iii) number of
newly occupied grid cells in each year, and (iv) cumulative number of occupied grid cells
in each year.
   The maximum annual dispersal distance was measured as the distance from grid cells
occupied in the preceding years to the newly occupied 10 9 10 km2 grid cell lying furthest
away from these. The relationship between the maximum dispersal distance and the cli-
mate of the corresponding year was analysed using the distribution records and climate
data for the overall area in which the species had been observed between 1973 and 2004.
Similar analyses of the relationship between the maximum dispersal distance and the
climate were also conducted separately for data from two spatially distinct subpopulations
in Finland, one located in eastern Finland, and the other on the southern coast (Fig. 1c).
Next, we examined the average northward shift of the map butterfly. This was done only
separately for the eastern subregion and southern coastal subpopulation, because there

           a)                           b)                           c)

Fig. 1 Snapshots during the expansion of the map butterfly (Araschnia levana) in Finland in: (a) 1987, (b)
1995 and (c) 2002. Each black dot represents an occupied 10 9 10 km2 grid cell

Biodivers Conserv

were clear differences in the timing of the more active expansion periods in the two
subpopulations. The annual average northward distribution shift in the two sub-regions was
calculated as the average latitude of the 10 northernmost occupied grid cells in each year.
   The influence of the climatically warmer years on the rate of expansion was studied
firstly, by relating the maximum dispersal distance and the annual average northward
distribution shift to two climatic variables (TMPJA and GDD; Table 1) using a linear
regression. Secondly, we arranged the 22 years into two groups according to the annual
values of the two climate variables, i.e. those having (1) higher and (2) lower than average
values of TMPJA or GDD. Because we expected a priori that the butterfly range expansions
would be more pronounced in favourable years, we used one-tailed t-tests (using the
assumption = equal variances not assumed, i.e. the Welch’s t-test) to measure the sig-
nificance between the average maximum dispersal distance and average latitude in the
climatically more favourable years vs. climatically less favourable years. In the analysis of
expansion rates for the two subpopulations, climate variables were based only on data from
the corresponding areas.
   Finally, we calculated the expansion rates for the overall study area in Finland following
Hill et al. (2003) separately for three time periods: (1) 1983–1991; (2) 1992–1998; and (3)
1999–2004, using the formula E = C/Hp. Here, E is the velocity of range expansion, and
C is the slope of line from a regression of the square of the area occupied plotted against
the years (i.e. the square root of area of the cumulative total number of 10-km grid cells
containing species records in each year).

Species–climate envelope modelling with GAM

We used generalized additive models (GAM) for building the climate envelope for the map
butterfly species. Generalized additive models are flexible data-driven non-parametric
extensions of generalized linear models (Hastie and Tibshirani 1986) that allow both linear
and complex additive response curves to be fitted (Wood and Augustin 2002). GAMs were
performed using GRASP (Generalised Regression Analysis and Spatial Prediction) version
3.2 in S-PLUS (version 6.1 for Windows, Insightful Corp.), by applying a logistic link
function for quasi-binomial error distribution (Lehmann et al. 2003). A starting model
including all predictors smoothed with 4 degrees of freedom was fitted first. The variable
dropping, or conversion to linear form, was then determined using Akaike’s Information
Criterion (AIC) (Akaike 1974).
   The validation of the butterfly–climate envelope model was done in two ways. First we
employed the commonly used split-sample approach (Guisan and Zimmermann 2000;
Thuiller 2003; Araujo et al. 2005a). Models were calibrated using a 70% random sample
(1609 grid cells of the 2298 cells) of the European distribution of the species and climate
data for the years 1961–1995, then evaluated by fitting the derived model to the remaining
30% of data (689 grid cells). Second, the model calibrated at the European scale was
projected onto a 10 9 10 km2 grid for Finland using climate data (Venalainen and Hei-
kinheimo 2002) for 2000–2004 and then evaluated with distribution data of the map
butterfly collated (see Species distribution data section) for the same time period. Here, we
excluded all 10 9 10 km2 grid cells with no butterfly recordings or with only one-day
visit. The downscaled model was evaluated using the 829 grid cells which were visited at
least twice during 2000–2004.
   The explanatory power of the butterfly–climate model was evaluated examining the
proportion of explained deviance (D2) of the total deviance of the model (Midgley et al.

                                                                                               Biodivers Conserv

2003; Luoto et al. 2006). The predictive capability of the model was assessed by exam-
ining the AUC (area under curve) of a ROC (receiver operating characteristic) plot and the
Kappa statistics (Fielding and Bell 1997). In order to determine the probability thresholds
at which the predicted values for species occupancy were optimally classified as absence or
presence values, we used the prevalence of the species as the cut-off probability level (Liu
et al. 2005). Thus, as the prevalence of the map butterfly in the European model calibration
data was 0.26, the grid cells with predicted probability of occurrence C0.26 were classified
as occupied (presence) cells, and the remaining cells as unoccupied (absence) cells.


Expansion history

The first observation of the map butterfly in Finland was made in 1973, followed by
10 years without new observations until 1983, when a stable population was discovered in
eastern Finland (Fig. 1). During the 1980s the distribution in eastern Finland remained
fairly local with maximum annual expansion distances of about 10–20 km, and the mean
rate of spread was rather low (Fig. 2a). In the 1990s the rate of spread remained about the
same (Fig. 2b), and the distribution of the species moved both west- and northwards. The
first observation on the southern coast of Finland was made in 1992 (Repo 1993).
   From 1983 to 1998 few observations of the map butterfly were recorded in Finland, with
an average of 1.8 newly occupied 10 9 10 km2 grid squares being detected every year (29
new grid squares occupied in the entire period). However, in the summer of 1999 the
number of grid squares occupied by the species increased notably (to 43) compared to the
summer of 1998. Large numbers of map butterflies (which probably originated from the
Baltic countries) were observed in mid-July 1999 and onwards (Mikkola 2000), especially

                   Sqrt cum. occupied area (km2)




                                                         1985    1990     1995   2000   2005

Fig. 2 Expansion of the map butterfly (Araschnia levana) in Finland during 1983–2004. The expansion is
illustrated as the square root (V) of the area occupied (km) as a function of time (year) and shown separately
for three time periods: (a) the first years of expansion in eastern Finland, (b) expansion after 1992, when the
first observation on the southern coast was made, and (c) expansion since the 1999 invasion of the southern
coast. Expansion rates (E = C/p), calculated from the slopes (C) of the regression lines, for the three periods
are 1.29, 1.47, and 7.50 km/year, respectively

Biodivers Conserv

on the southern coast of Finland. According to NAFI and other lepidopterological records
and Saarinen and Marttila (2000) all the observations on the south coast were of second
generation individuals.
   From the beginning of the 2000s, expansion northwards became more pronounced with
new observations being recorded at the northern edge of the species’ range (Figs. 1c, 2c).
The average annual number of newly occupied grid squares between 2000 and 2004 was
16, and the cumulative number occupied increased to 124. The increased rate of range
expansion starting from 1999 was also clearly detected in expansion rate comparison of the
three time periods based on the approach of Hill et al. (2003). The regression coefficients
(C) of the linear models for the time periods 1–3 in ascending order were: 2.28, 2.60, and
13.30. The calculated expansion rates (E) for the three periods were 1.29, 1.47, and
7.50 km/year, respectively (Fig. 2).

Expansion rate and climatically extreme years

The annual means of GDD and TMPJA (and their standard deviations) during the third time
period used in the expansion rate comparison (1999–2004; GDD = 1112.49 ± 63.29°Cd;
TMPJA = 15.21 ± 0.99°C) were higher than the mean values for the overall expansion
period in Finland (1983–2004; GDD = 1021.15 ± 110.10°Cd; TMPJA = 14.16 ± 1.36°
C). This observation supports the expectation that the rate of expansion of the map but-
terfly might be higher during warmer summers.
   The results of the linear regression between GDD and the maximum annual dispersal
distance for the whole data set showed a statistically non-significant relationship
(F = 0.60; df = 1,20; P = 0.45; R2 = 0.03), whereas the corresponding results for TMPJA
revealed a significant positive relationship (F = 5.40; df = 1,20; P = 0.03; R2 = 0.17).
Similarly, when time periods were grouped into warmer vs. cooler years (with respect to
TMPJA), the average maximum dispersal rates were significantly higher (one-way t-test;
t = -2.664; P = 0.01) in the periods with late summers warmer than average (Table 2,
Fig. 3).
   The examination of the annual dispersal shifts in relation with GDD and TMPJA in the
two subregions revealed different trends. Maximum dispersal distance and annual climate
trends in the eastern region had a greater degree of positive association with range
expansions (particularly in 1994, 2001, and 2003; Fig. 4a) than the population at the
southern coast. The t-tests indicated that the average latitude of species’ occurrences
shifted significantly more notably towards the north in the years with higher GDD than in
years with lower GDD values (one-way t-test; t = -2.438; P = 0.017); the maximum
dispersal distances were also higher in warmer summers than in cooler summers (Table 3).

Table 2 Comparison of the annual dispersal activity of the map butterfly in Finland between years with (1)
lower and (2) higher than average growing degree (GDD) and July–August temperatures (TMPJA) during
                     (1) Lower                   (2) Higher                   t                    P

GDD                  37.50 ± 15.60               47.92 ± 12.90                -0.515               0.307
TMPJA                21.25 ± 7.34                69.50 ± 16.56                -2.664               0.010

Mean (±standard error) values of annual maximum dispersal distance (km) are shown for each grouping
along with the statistical significance of the differences between them based on a one-way t-test

                                                                                                                                                    Biodivers Conserv



                  Maximum dispersal distance






                                                                                                       1                          2
                                                     Grouping of years by July-August temperature

Fig. 3 The mean annual maximum dispersal distance (km) of the map butterfly in Finland in 1983–2004,
by assigning the distances into two categories: years with (1) lower than average July–August temperature
(in 1983–2004) vs. (2) higher than average temperature (in 1983–2004). The difference is statistically
significant (one-way t-test; P = 0.01). Each box shows the median, quartiles, and extreme values within a

Fig. 4 Annual trends in                                                                                    a)
maximum dispersal distance of                                                                    160                                                            20
the map butterfly and July–
August temperature in (a) eastern                                                                                                                               18
Finland and (b) the southern
coastal region. Circles = July–
August temperature;                                                                               80
squares = maximum dispersal
                                                      Maximum dispersal distance per year (km)

distance                                                                                                                                                        14
                                                                                                                                                                     Average July-August temperature (C)


                                                                                                           1984     1988      1992       1996    2000    2004

                                                                                                 160       b)
                                                                                                           1992   1994     1996   1998    2000   2002   2004

Biodivers Conserv

Table 3 Comparison of the annual maximum dispersal distance (km) and annual average northward
distribution shift (mean N-coordinate of the ten northernmost grid cells; m) of the map butterfly in eastern
Finland between years with (1) lower and (2) higher than average growing degree (GDD) and July–August
temperatures (TMPJA) during 1983–2004
                    (1) Lower                     (2) Higher                    t                    P

Maximum dispersal (km)
GDD                    13.89 ± 5.99                  36.16 ± 11.70              -1.694               0.054
TMPJA                  13.64 ± 4.53                  40.46 ± 13.61              -1.870               0.043
Northwards range shift (mean N-coordinate; m)
GDD                 6928141 ± 1883                6947269 ± 7698                -2.413               0.015
TMPJA               6928525 ± 1123                6950363 ± 8885                -2.438               0.017
Mean (± standard error) values are shown for each grouping along with the statistical significance of the
differences between them based on a one-way t-test

In contrast, in the southern coastal subpopulation, the years with higher maximum dis-
persal jumps matched only partly with the years with warmer summers in the years 1992–
2004 (Fig. 4b). In this subregion, no significant differences were detected between the two
range expansion estimates in warmer years and cooler years (Table 4).

Map butterfly–climate envelope models

Accuracy of the GAM model calibrated with the 70% subset of the European 1961–1995
data and evaluated with the remaining 30% of European data was good. The amount of the
deviance (D2) accounted for by the three climate variables was 31%, and the AUC value
from the validation data set (0.875), which is a value classified as providing good model
accuracy by Swets (1988). Visual inspection of the maps of observed and predicted dis-
tributions of the map butterfly for the European window supports the quantitative
assessment provided by AUC (Fig. 5). However, the performance of the butterfly–climate
model was much poorer when the model based on European data was transferred to the
Finnish 10 9 10 km2 climate and butterfly data for 2000–2004. Here, an AUC value of
0.630 was obtained indicating poor model accuracy (see Swets 1988).

Table 4 Comparison of the annual maximum dispersal distance (km) and annual average northward
distribution shift (mean N-coordinate of the ten northernmost grid cells; m) of the map butterfly in the
southern coastal region of Finland between years with (1) lower and (2) higher than average growing degree
(GDD) and July–August temperatures (TMPJA) during 1992–2004
                    (1) Lower                      (2) Higher                       t                P

Maximum dispersal (km)
GDD                    50.00 ± 21.13                  31.67 ± 14.47                 0.716            0.755
TMPJA                  53.57 ± 18.68                  22.00 ± 13.19                 1.389            0.904
Northwards range shift (mean N-coordinate; m)
GDD                 6681881 ± 5890                 6676486 ± 2220                   0.857            0.791
TMPJA               6680552 ± 5271                 6677533 ± 2398                   0.521            0.693
Mean (± standard error) values are shown for each grouping along with the statistical significance of the
differences between them based on a one-way t-test

                                                                                          Biodivers Conserv

    a)                                                     b)

Fig. 5 Distribution of the map butterfly in Europe: (a) observed and (b) projected. Observations are for
species distribution data recorded since 1950 (based on Kudrna 2002, and the supplementary sources for
eastern Europe). Projections are from a bioclimatic model developed using GAM based on climate data for
1961–1995. Data are plotted on a 0.5° latitude 9 1.0° longitude grid

   Based on the Kappa statistics, the matching of the observed occurrences and predicted
presences in the 829 grid cells in Finland in 2000–2004 can be considered moderate
(Kappa = 0.51), according to a classification by Landis and Koch (1977). The maps of the
observed distributions and predicted distributions for Finland show clear spatial discrep-
ancies, most notably because the model predicted presences in many grid cells in SW
Finland where the species has hitherto occurred very rarely (Fig. 6). We also tentatively
fitted the European data based model into the climate data for Finland averaged over the

               a)                                             b)

Fig. 6 Distribution of the map butterfly in Finland in 2000–2004: (a) observed and (b) projected. Black and
grey squares represent presence and absence records in the 1115 10 9 10 km2 grid cells for which
observations are reliable. The projected distribution is based on the downscaled bioclimatic model calibrated
using the European distribution and climate data (Fig. 5)

Biodivers Conserv

years 1970–1979. The outcomes of this exercise suggested that the climate might already
have been suitable for the map butterfly in the 1980s, but for a substantially smaller area
than in 2000–2004, in essence a small coastal zone in southwestern Finland, both on the
mainland and in the archipelago (results not shown). Thus a clear spatial discrepancy also
occurred here, because the first observations of the species were made close to the SE
border of Finland (cf. Fig. 1).


Influence of annual climatic variation on expansion

Our results from eastern Finland show a statistically significant relationship between
annual maximum dispersal distance travelled by expanding map butterflies and tempera-
ture of the late summer. This observation supports the view that the map butterfly (and
potentially other functionally similar species) has dispersed more actively in warmer rather
than cooler summers. Visual inspection indicates that the average distribution limit of the
species has steadily moved northwards from about 2000 onwards, especially in eastern
Finland. We also found that this pattern is consistent with a similar trend in the thermal
sum (GDD) and average late summer temperatures. There are additional reasons for
supporting the assumption that higher summer temperatures in recent years have accel-
erated species expansion in northern Europe (Parmesan 2006). The ranges of many
European butterflies are closely correlated with summer temperature, especially those of
mobile species that have a widespread host plant (Pollard 1988; Pollard and Yates 1993).
Warm summer temperatures of even some few consequent days and certain weather
events, such as the occurrence of warm southerly winds, may also promote northward
dispersal (Mikkola 1986).
   It is likely that the map butterfly has dispersed most actively in Finland during summers
with short duration spells of exceptionally warm weather (even lasting only 2–3 days) in
the late summer. In 1999, field observations suggested that the expansion event from
Estonia (Mikkola 2000; Saarinen and Marttila 2000) was associated with warm south-
easterly air currents on 14–15 July. At the same time, large numbers of other non-resident
butterflies were recorded on the southern coast of Finland, including bath white (Pontia
daplidice), short-tailed blue (Cupido argiades) and Palla’s fritillary (Argynnis laodice).
The lack of complete annual data of such events does not allow the statistical analysis of
this hypothesis. However, our results support the conclusion of Bryant et al. (2002) that
annual and monthly averaged climatic variables do not reveal local, shorter-term climate
variability that is likely to play an important role in driving dispersal processes.
   It has been suggested that in bivoltine species, overwintering as pupae, the second
generation may become more common and occur further north as a consequence of
warming summer temperatures (Virtanen and Neuvonen 1999). The development of sec-
ond and third generations in bivoltine butterfly species is regulated by day length and
temperature, and by genetic variation (Reinhardt 1972; Brakefield and Shreeve 1992). In
the case of the map butterfly, it has been suggested that second generation individuals are
better suited for long-distance dispersal. Fric and Konvicka (2002) found that these but-
terflies were bigger, had larger and less pointed wings and their thorax ratio was higher
compared to first generation butterflies, facilitating an enhanced flight capacity. A mark–
recapture study by the same authors indicated that the second generation individuals had
higher emigration probabilities and shorter residence times than their first generation

                                                                            Biodivers Conserv

counterparts. Thus, consecutive warm summer temperatures of the recent years may partly
explain the accelerated expansion of the species in Finland, where the second generation
has been observed only in warm summers since 1999.

Model performance in predicting distribution

It has been increasingly emphasised that the evaluation of bioclimatic models requires
statistically independent test data collected in other regions or times (Araujo and Guisan
2006). Yet such evaluations are rare (but see Beerling et al. 1995; Hill et al. 1999; Araujo
et al. 2005a; Randin et al. 2006) and this hinders our ability to understand the true
capability of models for predicting climate change impacts (Araujo and Guisan 2006). The
few results available hitherto from such independent validations have yielded contrasting
results. For example, Peterson (2001) reported an excellent predictive ability for 34 bird
species in North America, based on species–climate models built using GARP and a set of
randomly selected states of the United States, and validated using the states omitted from
the model building (validation using different regions; Araujo and Guisan 2006).
   In contrast, the results of Araujo et al. (2005a) showed that accuracies of bioclimatic
envelope models for 116 UK birds were always higher when evaluated by one-time split-
sample than accuracy values derived from fitting the calibrated model to the independent
data recorded ca. 15 years later than the calibration data (validation using different time
periods; Araujo and Guisan 2006). This result supported concerns that predictive accuracy
measured by commonly used split-sample approach may provide an over-optimistic
assessment of model performance when applied into truly independent data (see also
Randin et al. 2006).
   One of the novelties of our study was the attempt to evaluate the bioclimatic envelope
model developed for the map butterfly using the European data with an independent dataset
of climate and species distribution data in Finland, which differed both in terms of the
resolution and time period used. Our results are largely in agreement with those of Araujo´
et al. (2005a). The accuracy of the model calibrated with European data and validated
using split-sample approach was good. However, the performance of the model decreased
drastically when it was transferred to predict the recent observed range shifts in Finland.
Also the visual comparison of the predicted spatial distributions of the map butterfly for the
period 2000–2004 in Finland and the recorded distribution for the same period showed that
the transferred model performed rather poorly.
   There are several sources of uncertainty in bioclimatic envelope models (Kadmon et al.
2003; Hampe 2004; Barry and Elith 2006; Heikkinen et al. 2006), which may contribute to
the poor performance of our transferred models of the map butterfly. Much attention has
been paid to the variation in species range predictions based on the selection of the
modelling method. Recent studies have shown significant differences in present-day pre-
dictions from different modelling techniques which may result in disturbingly dissimilar
future projections (Thuiller 2004; Pearson et al. 2006; Araujo and New 2007). However,
such variability in model predictions is not the most likely cause of the poor success of our
transferred model. This is because a preliminary analysis using three other methods,
classification and regression trees (CTA), neural network (ANN) and generalized linear
models (GLM) provided overall similar projections (results not shown) to the chosen GAM
   More likely explanations of the poor performance of our transferred models might lie
elsewhere. Firstly, there are data deficiencies worth considering. The known distribution of

Biodivers Conserv

the map butterfly is sparse in north-eastern Europe, and this may generate a climatic bias in
our data (corrected only by further extensive sampling from regions E-SE from Finland).
Two recent studies (Kadmon et al. 2003; Thuiller et al. 2004) have shown that insufficient
or biased sampling of the climate range of the species can have a significant effect on the
accuracy of the bioclimatic model predictions.
   Second, there are issues of scale to consider. Our models were calibrated at a coarse
resolution but were then evaluated independently with projections at a finer resolution.
Earlier studies in boreal landscapes (Heikkinen and Birks 1996; Virkkala et al. 2005;
Luoto et al. 2006) and elsewhere (Hill et al. 1999; Pearson et al. 2004; Stefanescu et al.
2004) have shown that species distributions at 10-km resolution (or finer) often reflect the
interplay between habitat availability and climate. Although the map butterfly is consid-
ered to be a generalist species, the spatial distribution of the most suitable habitats—or
those clearly unsuitable—may contribute to the mismatch of the predicted and observed
distribution in Finland. This mismatch was particularly important in certain intensively
managed agricultural landscapes in SW Finland, where the species was projected to thrive
well based on the climate variables alone.
   A third and probably the most important reason for the failure of the transferred map
butterfly model is the non-spatial nature of the projections of the bioclimatic models; they
do not generally take into account dispersal barriers or other migration limitations of the
species (Pearson and Dawson 2003; Hampe 2004). The fact that the species has migrated
over the Baltic Sea only during very favourable weather conditions suggests that the sea
constitutes an effective barrier for dispersal. In contrast, the species has been able to
migrate far more easily (and was recorded first) to south-eastern and eastern parts of
Finland with a land connection to populations in Russian Karelia, Belarus and Estonia.
Thus, from the perspective of butterfly individuals, climatically suitable areas in SW
Finland are located in a remote ‘‘peninsula’’ compared to the areas on the eastern border of
Finland. Similar kinds of potential limitations to dispersal are increasingly being discussed
in the bioclimatic modelling literature (Berry et al. 2002; Pearson and Dawson 2003;
Peterson et al. 2004) and occasionally also integrated into the modelling (Schwartz et al.
2001; Peterson 2003; Midgley et al. 2006), but there is still a need for more research in the
   A potential alternative for modelling distributions, which accounts for the spatial
dependencies of the range shifts, is the use of spatially explicit dynamic models
(Collingham et al. 1996; Hill et al. 1999; Iverson et al. 2004). Detailed examination of the
strengths and limitations of this approach are beyond the scope of this paper. However, the
findings in our study suggest that there might also be limitations in the successful appli-
cation of spatially dynamic models. The most important reason for this conclusion is that it
seems that the magnitude of range shifts of mobile species varies in response to the annual
variation in the weather. Moreover, under favourable weather conditions such species may
be able to undertake long distance jumps in dispersal. When modelling mobile species it is
critical that model’s are able to capture such events (Pearson and Dawson 2005). However,
spatial dynamic models often assume that species migration proceeds as a broad moving
wave, and thus may inevitably underestimate the true dispersal response of the species.


Our results indicate that mobile butterfly species have demonstrated their potential to
migrate to regions that have become climatically suitable due to recent climate warming in

                                                                                       Biodivers Conserv

northern Europe. Moreover, it appears that the rate of dispersal of species such as the map
butterfly is not constant from year to year, but varies according to the annual variation in
climate. In addition, the species appear capable of responding to exceptionally favourable
short-term (warm) weather by exhibiting long-distance jumps in dispersal. Such phe-
nomena can be hard to identify on the basis of climate information averaged over long time
periods and are therefore difficult to predict accurately. The fact that bioclimatic models do
not account for dispersal barriers can also severely hamper their usefulness when they are
used to downscale and transfer models into independent situations. An additional factor
that may cause problems, as suggested by our results, is the possible differences in the
responses to climate warming of separate subpopulations, a phenomenon related to the
local adaptation of the populations under the changing climate (Pearson and Dawson 2003;
Thomas 2005).
   By and large, it seems obvious that spatially accurate predictions of the impacts of
climate warming are rather difficult to achieve. Perhaps the most robust strategy for
obtaining more realistic predictions of the impacts of climate change on species distri-
butions is to use many different approaches, for example a combination of bioclimatic
models, spatially dynamic models and empirical monitoring (Berry et al. 2002; Pearson
and Dawson 2003). Finally, following Araujo et al. (2005a) and Heikkinen et al. (2006) we
argue that more empirical evidence needs to be gathered to improve the knowledge of the
usefulness and limitations of bioclimatic models and their predictions in real-life situations.

Acknowledgements We thank Timothy R. Carter, Ilkka Hanski and Josef Settele for valuable comments
on the manuscript. This research was funded by the EC FP6 Integrated Project ALARM (GOCE-CT-2003-
506675) (Settele et al. 2005) and by the Academy of Finland (project grant 116544). Otakar Kudrna kindly
gave permission to make use of the European butterfly species distribution maps.


Akaike H (1974) A new look at statistical model identification. IEEE Trans Autom Control AU-19:716–722
Araujo MB, Guisan A (2006) Five (or so) challenges for species distribution modelling. J Biogeogr
Araujo MB, New M (2007) Ensemble forecasting of species distributions. Trends Ecol Evol 22:42–47
Araujo MB, Rahbek C (2006) How does climate change affect biodiversity? Science 313:1396–1397
Araujo MB, Pearson RG, Thuiller W, Erhard M (2005a) Validation of species-climate impact models under
   climate change. Global Change Biol 11:1504–1513
Araujo MB, Thuiller W, Williams PH, Reginster I (2005b) Downscaling European species atlas distributions
   to a finer resolution: implications for conservation planning. Global Ecol Biogeogr 14:17–30
Araujo MB, Whittaker RJ, Ladle RJ, Erhard M (2005c) Reducing uncertainty in projections of extinction
   risk from climate change. Global Ecol Biogeogr 14:529–538
Bakkenes M, Alkemade J, Ihle F, Leemans R, Latour J (2002) Assessing the effects of forecasted climate
   change on the diversity and distribution of European higher plants for 2050. Global Change Biol 8:390–
Barry S, Elith J (2006) Error and uncertainty in habitat models. J Appl Ecol 43:413–423
Beaumont LJ, Hughes L (2002) Potential changes in the distributions of latitudinally restricted Australian
   butterfly species in response to climate change. Global Change Biol 8:954–971
Beerling DJ, Huntley B, Bailey JP (1995) Climate and the distribution of Fallopia japonica: use of an
   introduced species to test the predictive capacity of response surfaces. J Veget Sci 6:269–282
Berry P, Dawson T, Harrison P, Pearson R (2002) Modelling potential impacts of climate change on the
   bioclimatic envelope of species in Britain and Ireland. Global Ecol Biogeogr 11:453–462
Brakefield PM, Shreeve TG (1992) Diversity within populations. In: Dennis RLH (ed) The ecology of
   butterflies in Britain. Oxford University Press, Oxford, pp 178–216
Bryant SR, Thomas CD, Bale JS (2002) The influence of thermal ecology on the distribution of three
   nymphalid butterflies. J Appl Ecol 39:43–55

Biodivers Conserv

Collingham YC, Hill MO, Huntley B (1996) The migration of sessile organism: a simulation model with
    measurable parameters. J Veget Sci 7:831–846
Dovgailo KE, Solodovnikov IA, Rubin NI (2003) The butterflies (Diurna, Lepidoptera) of Republic of
    Belarus. CD key and database on the basis of software ‘‘Lysandra’’, Minsk
Fielding A, Bell J (1997) A review of methods for the assessment of prediction errors in conservation
    presence/absence models. Environ Conserv 24:38–49
Fric Z, Konvicka M (2000) Adult population structure and behaviour of two seasonal generations of the
    European Map Butterfly, Araschnia levana, species with seasonal polyphenism (Nymphalidae). Nota
    Lepidopterol 23:2–25
Fric Z, Konvicka M (2002) Generations of the polyphenic butterfly Araschnia levana differ in body design.
    Evol Ecol Res 4:1017–1032
Gitay H, Brown S, Easterling W, Jallow B (2001) Ecosystems and their goods and services. In: McCarthy JJ,
    Canziani OF, Leary NA, Dokken DJ, White KS (eds) IPCC Third Assessment Report. Climate Change
    2001: Impacts, Adaptation, and Vulnerability. Contribution of Working Group II to the Third Assess-
    ment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press,
    Cambridge and New York, pp 248–249
Guisan A, Zimmermann NE (2000) Predictive habitat distribution models in ecology. Ecol Model135:147–
Hampe A (2004) Bioclimate envelope models: what they detect and what they hide. Global Ecol Biogeogr
Hastie TJ, Tibshirani RJ (1986) Generalized additive models. Stat Sci 1:297–318
Heikkinen RK, Birks HJB (1996) Spatial and environmental components of variation in the distribution
    patterns of subarctic plant species at Kevo, N Finland - a case study at the meso-scale level. Ecography
Heikkinen RK, Luoto M, Araujo MB, Virkkala R, Thuiller W, Sykes MT (2006) Methods and uncertainties
    in bioclimatic envelope modelling under climate change. Prog Phys Geogr 30:751–777
Hill JK, Thomas CD, Huntley B (1999) Climate and habitat availability determine 20th century changes in a
    butterfly’s range margin. Proc R Soc Lond Ser B, Biol Sci 266:1197–1206
Hill JK, Thomas CD, Fox R, Telfer MG, Willis SG, Asher J, Huntley B (2002) Responses of butterflies to
    twentieth century climate warming: implications for future ranges. Proc R Soc Lon B 269:2163–2171
Hill JK, Thomas CD, Huntley B (2003) Modeling present and potential future ranges of European butterflies
    using climate response surfaces. In: Boggs C, Watt W, Ehrlich P (eds) Butterflies. Ecology and evo-
    lution of taking flight. The University of Chicago Press, Chicago, pp 149–167
Holdridge LR (1967) Life zone ecology. Tropical Science Center, Jan Hose, Costa Rica
Ivanov AI (1999) Artenverzeichnis der Macrolepidoptera von Sankt-Petersburg und des Sankt-Petersburg
    Gebietes nach Aufsammlungen in den Jahren 1960–1998 (Insecta, Lepidoptera). Atalanta 30:293–356
Iverson LR, Schwartz MW, Prasad AM (2004) How fast and far might tree species migrate in the eastern
    United States due to climate change? Global Ecol Biogeogr 13:209–219
Kadmon R, Farber O, Danin A (2003) A systematic analysis of factors affecting the performance of climatic
    envelope models. Ecol Appl 13:853–867
Keskula T (1992) Distributions maps of Estonian butterflies (Lepidoptera: Hesperionoidea, Papilionoidea).
    Acta Musei Zoologici, vol. 6. Universitatis Tartuensis, Tartu, p 60
Komonen A, Grapputo A, Kaitala V, Kotiaho J, Paivinen J (2004) The role of niche breadth, resource
    availability and range position on the life history of butterflies. Oikos 105:41–54
Kudrna O (2002) The distribution atlas of European butterflies. Oedippus 20:1–342
Landis JR, Koch GC (1977) The measurement of observer agreement for categorical data. Biometrics
Lawler JJ, White D, Neilson RP, Blaustein AR (2006) Predicting climate-induced range shifts: model
    differences and model reliability. Global Change Biol 12:1–17
Lehmann A, Overton JM, Leathwick JR (2003) GRASP: generalized regression analysis and spatial pre-
    diction. Ecol Model 160:165–183
Luoto M, Fronzek S, Zuidhoff FS (2004) Spatial modelling of palsa mires in relation to climate in northern
    Europe. Earth Surf Process Landforms 29:1373–1387
Liu C, Berry PM, Dawson TP, Pearson RG (2005) Selecting thresholds of occurrence in the prediction of
    species distributions. Ecography 28:385–393
Luoto M, Heikkinen RK, Poyry J, Saarinen K (2006) Determinants of the biogeographical distributions of
    butterflies in boreal regions. J Biogeogr 33:1764–1778
Luoto M, Poyry J, Heikkinen RK, Saarinen K (2005) Uncertainty of bioclimate envelope models based on
    geographical distribution of species. Global Ecol Biogeogr 14:575–584

                                                                                        Biodivers Conserv

Luoto M, Virkkala R, Heikkinen RK (2007) The role of land cover in bioclimatic models depends on spatial
    resolution. Global Ecol Biogeogr 16:34–42
                                                                    ¨ ¨                    ¨
Marttila O, Haahtela T, Aarnio H, Ojalainen P (1990) Suomen paivaperhoset. Kirjayhtyma Oy, Helsinki
McPherson JM, Jetz W, Rogers DJ (2006) Using coarse-grained occurrence data to predict species distri-
    butions at finer spatial resolutions – possibilities and limitations. Ecol Model 192:499–522
Midgley GF, Hannah L, Millar D, Rutherford MC, Powrie LW (2002) Assessing the vulnerability of species
    richness to anthropogenic climate change in a biodiversity hotspot. Global Ecol Biogeogr 11:445–451
Midgley GF, Hannah L, Millar D, Thuiller W, Booth A (2003) Developing regional and species-level
    assessments of climate change impacts on biodiversity in the Cape Floristic Region. Biol Conserv
Midgley GF, Hughes GO, Thuiller W, Rebelo AG (2006) Migration rate limitations on climate change-
    induced range shifts in Cape Proteaceae. Divers Distrib 12:555–562
Mikkola K (1986) Direction of insect migration in relation to wind. In: Danthanarayana W (ed) Insect flight:
    dispersal and migration. Springer-Verlag, Berlin, Heidelberg, pp 152–171
                      ¨¨      ¨
Mikkola K (2000) Saa ja hyonteisten vaellukset 1999. Baptria 25:33–43
Mitchell TD, Carter TR, Jones PD, Hulme M, New M (2003) A comprehensive set of high-resolution grids
    of monthly climate for Europe and the globe: the observed record (1901–2000) and 16 scenarios (2001–
    2100). Tyndall Centre Work Pap, vol 55
New M, Lister D, Hulme M, Makin I (2002) A high-resolution data set of surface climate over global land
    areas. Climate Res 21:1–25
Parmesan C (2005) Biotic response: range and abundance changes. In: Lovejoy TE, Hannah L (eds) Climate
    change and biodiversity. Yale University Press, New Haven & London, pp 41–55
Parmesan C (2006) Ecological and evolutionary responses to recent climate change. Annu Rev Ecol Evol
    Syst 37:637–669
Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across natural
    systems. Nature 421:37–42
Parmesan C, Ryrholm N, Stefanescu C, Hill JK, Thomas CD, Descamon H, Huntley B, Kaila L, Kullberg J,
    Tammaru T, Tennent WJ, Thomas JA, Warren M (1999) Poleward shifts in ranges of butterfly species
    associated with regional warming. Nature 399:579–583
Pearson R, Dawson T (2003) Predicting the impacts of climate change on the distribution of species: are
    bioclimatic envelope models useful? Global Ecol Biogeogr 12:361–371
Pearson RG, Dawson TP (2005) Long-distance plant dispersal and habitat fragmentation: identifying
    conservation targets for spatial landscape planning under climate change. Biol Conserv 123:389–401
Pearson RG, Dawson TP, Berry PM, Harrison PA (2002) SPECIES: a spatial evaluation of climate impact
    on the envelope of species. Ecol Model 154:289–300
Pearson RG, Dawson TP, Liu C (2004) Modelling species distributions in Britain: a hierarchical integration
    of climate and land-cover data. Ecography 27:285–298
Pearson RG, Thuiller W, Araujo MB, Martinez-Meyer E, Brotons L, McClean C, Miles L, Segurado P,
    Dawson TE, Lees DC (2006) Model-based uncertainty in species’ range prediction. J Biogeogr
Peterson AT (2001) Predicting species’ geographic distributions based on ecological niche modelling. The
    Condor 103:599–605
Peterson AT (2003) Projected climate change effects on Rocky Mountain and Great Plain birds: generalities
    on biodiversity consequences. Global Change Biol 9:647–655
                    ´                    ´
Peterson AT, Martınez-Meyer E, Gonzalez-Salazar C, Hall PW (2004) Modeled climate change effects on
    distributions of Canadian butterfly species. Can J Zool 82:851–858
Pollard E (1988) Temperature, rainfall, and butterfly numbers. J Appl Ecol 25:819–828
Pollard E, Yates TJ (1993) Monitoring butterflies for ecology and conservation. Chapman & Hall, London
Popov SG (2005) SW Ukranian butterfly database: report 1973–2005, Lepidoptera: Papilionoidea &
    Hesperionoidea, Uzhgorod. Cited 4 Aug 2005
Randin CF, Dirnbock T, Dullinger S, Zimmermann NE, Zappa M, Guisan A (2006) Are niche-based species
    distribution models transferable in space? J Biogeogr 33:1689–1703
Reinhardt R (1972) Der Landkartchenfalter. Die Neue Brehm-Bucherei, Lutherstadt-Wittenberg
Repo S (1993) Records of Finnish macrolepidoptera 1992. Baptria 18:59–65
                                        ¨ ¨
Saarinen K (2003) Valtakunnallisen paivaperhosseurannan vuoden 2002 tulokset. Baptria 28:4–15
                                                        ¨ ¨
Saarinen K, Marttila O (2000) Valtakunnallisen paivaperhosseurannan vuoden 1999 tulokset. Baptria
Saarinen K, Lahti T, Marttila O (2003) Population trends of Finnish butterflies (Lepidoptera: Hesperioidea,
    Papilionoidea) in 1991–2000. Biodivers Conserv 12:2147–2159

Biodivers Conserv

Schwartz MW, Iverson LR, Prasad AM (2001) Predicting the potential future distribution of four tree
    species in Ohio using current habitat availability and climatic forcing. Ecosystems 4:568–581
Settele J, Hammen V, Hulme P, Karlson U, Klotz S, Kotarac M, Kunin W, Marion G, O’Connor M,
    Petanidou T, Peterson K, Potts S, Pritchard H, Pysek P, Rounsevell M, Spangenberg J, Steffan-Dewenter
    I, Sykes M, Vighi M, Zobel M, Kuhn I (2005) ALARM – Assessing LArge-scale environmental Risks
    for biodiversity with tested Methods. GAIA 14:69–72
Skov F, Svenning J-C (2004) Potential impact of climate change on the distribution of forest herbs in
    Europe. Ecography 27:366–380
Spangenberg J (2007) Integrated scenarios for assessing biodiversity risks. Sustain Dev 15(6):343–356
Stefanescu C, Herrando S, Paramo F (2004) Butterfly species richness in the north-west Mediterranean
    Basin: the role of natural and human-induced factors. J Biogeogr 31:905–915
Swets KA (1988) Measuring the accuracy of diagnostic systems. Science 240:1285–1293
Thomas CD (2005) Recent evolutionary effects of climate change. In: Lovejoy TE, Hannah L (eds) Climate
    change and biodiversity. Yale University Press, New Haven and London, pp 75–88
Thomas CD, Bodsworth EJ, Wilson RJ, Simmons AD, Davies ZG, Musche M, Conradt L (2001) Ecological
    and evolutionary processes at expanding range margins. Nature 441:577–581
Thomas CD, Cameron A, Green RE, Bakkenes M, Beaumont LJ, Collingham YC, Erasmus BFN, Ferreira
    de Siquieira M, Grainger A, Hannah L, Hughes L, Huntley B, Van Jaarsveld AS, Midgley GF, Miles L,
    Ortega-Huerta MA, Peterson AT, Phillips OL, Williams SE (2004) Extinction risk from climate change.
    Nature 427:145–148
Thuiller W (2003) BIOMOD – optimizing predictions of species distributions and projecting potential future
    shifts under global change. Global Change Biol 9:1353–1362
Thuiller W (2004) Patterns and uncertainties of species’ range shifts under climate change. Global Change
    Biol 10:2020–2027
Thuiller W, Brotons L, Araujo MB, Lavorel S (2004) Effects of restricting environmental range of data to
    project current and future distributions. Ecography 27:165–172
Thuiller W, Lavorel S, Araujo MB, Sykes MT, Prentice IC (2005) Climate change threats to plant diversity
    in Europe. Proc Natl Acad Sci 102:8245–8250
Tuhkanen S (1984) A circumboreal system of climatic-phytogeographical regions. Acta Bot Fenn 127:1–50
Venalainen A, Heikinheimo M (2002) Meteorological data for agricultural applications. Phys Chem Earth
Virkkala R, Luoto M, Heikkinen RK, Leikola N (2005) Distribution patterns of boreal marshland birds:
    modelling the relationships to land cover and climate. J Biogeogr 32:1957–1970
Virtanen T, Neuvonen S (1999) Climate change and macrolepidopteran diversity in Finland. Chemosphere:
    Global Change Sci 1:439–448
Walther GR, Post E, Convey P, Menzel A, Parmesan C, Beebee TJC, Fromentin J-M, Hoegh-Gulgberg O,
    Bairlein F (2002) Ecological responses to recent climate change. Nature 416:389–395
Wood S, Augustin N (2002) GAMs with integrated model selection using penalized regression splines and
    applications to environmental modelling. Ecol Model 157:157–177


To top