Download ZONAL FOREST TYPES, CLIMATIC VARIABLES AND EFFECT OF

ZONAL FOREST TYPES, CLIMATIC VARIABLES AND EFFECT OF CHANGES
FOR HUNGARY
LÁSZLÓ GÁLHIDY1, BÁLINT CZÚCZ2, FRANCK TORRE3
1
Hungarian Academy of Sciences, Production Biology Research Group, H–9400 Sopron, Bajcsy-Zs. u.
4,Hungary; [email protected], 2Corvinus University, Budapest, H–1118 Budapest, Villanyi út 29–43,
Hungary; [email protected], 3Institut Méditerranéen d'Ecologie et de Paléoécologie, F–13545 Aix-enProvence cedex 04, France; [email protected]
GÁLHIDY, L., CZÚCZ, B., TORRE, F.: Zonal forest types, Climatic variables and effect of changes for
Hungary. Lesn. Čas. – Forestry Journal, 52(1–2): 99–105, 2006, 4 fig., ref. 13. Original paper. ISSN
0323–1046
Distribution of the zonal forest types is critically influenced by climatic conditions, and climate
change probably will affect their future distribution in most regions including South-East Europe. Our
aim was to study the relationship between climatic variables and zonal forest types in Hungary based on
their current distribution. We used two national databases for Hungary (forest inventory data, 2001 and
climatic data, 1961–1990). We used Classification and Regression Tree (CART) analysis to select the
most important climatic variables that explain the recent distribution of zonal forest types. Probability of
presence was determined for the whole area of Hungary for beech (Fagus sylvatica) and oak (Quercus
petraea) dominated zonal forest types. Future predictions were also taken for beech-dominated forests,
using different regional climatic scenarios. In certain regions probability of presence will decrease, and
the maintenance of present forest types might become difficult or impossible.
Keywords: climate change, CART, Fagus sylvatica, Quercus petraea
1. Introduction
In general, distribution of vegetation is influenced by many factors such as climate, soil,
drainage, human and animal activities. Large-scale climatic factors are responsible for the
distribution of zonal communities, and they are commonly used to predict their current and
future potential distribution or species turnover in response to global change (MALCOLM and
PITELKA 2000, THUILLER 2003). In this paper we focus on the relationships between climatic
variables and forest types.
Forests are especially vulnerable to climate change, due to the longevity of trees. Through its
impact on forest growth, climate change will affect both long-term wood supply and carbon
sequestration in trees, forest soils and wood products (KRAMER 2001). Modelling of the
distribution of tree species and forest types is a powerful tool to measure spatial changes.
Different zonal forest types can characterize forested landscapes in Hungary. Forests are
defined here as "zonal" if influenced primarily by the climatic conditions. They can generally
be found on flat surfaces and on slight slopes where the effect of edaphic factors is less
pronounced (MÁTYÁS and CZIMBER 2004). Previous works in Hungary aimed to classify the
climate from the aspect of forestry (JÁRÓ 1972). This approach used "test species" and the
border between forest types dominated by test species was defined by the 2 pm. (local time)
air humidity of July. This classification is still widely accepted in Hungary and used as a
determinant of applied species and forestry methods. Recent studies used modelling as a
method to find relationships between forest types and zonal climate (MÁTYÁS and CZIMBER
2004).
In our work we aim to predict the present and future potential distribution of zonal forest
types, taking the Global Climate Change (GCC) into consideration. The consequences of
changes in potential distribution of zonal forests is a challenge for foresters and
conservationists as well, addressing a set of tasks concerning reforestation of abandoned
agricultural lands and maintaining managed forest.
Our questions in this work were the following: (1) Which climatic variables can describe best
the present distribution of the two most important zonal forest tree species (Fagus sylvatica,
Quercus petraea) in Hungary? (2) What are the climatic probabilities for these species to be
present in the different parts of Hungary? (3) How would these potential distributions shift in
the future based on the predicted change in climatic variables?
2. Materials and methods
2.1. Study area
Hungary is a relatively flat country covering 93.036 km2 in the Carpathian basin. The highest elevation is
1014 m (Mt. Kékes), and more than half of the country is less than 200 m above sea level. Mountain ranges
which divide the country along a southwest-northeast line (Transdanubian and Northern Mountains), the
Transdanubian Hills and the floodplains of the Danube and the Tisza rivers can be highlighted as the most
characteristic regions.
The climate of Hungary can be characterized continental, with longer warm season (PÉCZELY 1979) or nemoralsemihumid (RÖHRIG et al. 1991). The four seasons are distinct, thermal amplitude between winter and summer is
relatively high. Higher proportion of precipitation falls during the summer. Both temperature and precipitation is
highly variable monthly and annually as well. Mediterranean and Atlantic climatic influences are more
pronounced on the southwestern and on the western parts of the country.
The potential vegetation of the country is generally deciduous broadleaved forest. On the lower mountains and
hills dry, oak-dominated (Q. cerris, Q. pubescens) forest are still present. More mesic sites can be characterized
by oak-hornbeam (Q. petraea, Carpinus betulus) forests with hornbeam as an admixed species. At the highest
elevations and in the southwestern part of the country – beech (Fagus sylvatica) dominated forests can be found.
The plains used to be covered by vast areas of dry forest-steppe vegetation accompanied by alluvial forests.
These are now almost completely converted to agricultural land (JÁRAINÉ-KOMLÓDI et al. 1995). About 18
percent of Hungary is covered by forests and plantations, of which less than half is considered to have natural
species composition. According to forest inventory data (MÁTYÁS 1992) the ratio of native tree species is only
29% on the Great Plain, the rest of the stands are formed by exotic species and plantations, such as black locust
(Robinia pseudo-acacia) and Euramerican poplars – the latter two alone cover more than half of the forest area.
2.2. Data sets
Climatic data were taken from the corrected climatic surface maps for Hungary based on the
observations 1961–1990 (source: National Weather Service [OMSz]). We have chosen 20
climatic variables as possible predictors of zonal forest type distributions, namely: mean
temperatures (annual, summer, winter, May, June, July August), precipitation (annual,
summer, winter, May, June, July August), potential evapotranspiration (April, July), relative
air humidity (April, July), number of summer days (Tmax > 25 °C) and number of heat days
(Tmax > 30 °C). This set of climatic variables reflects the fact, that in Hungary it is the summer
drought, which is most limiting for zonal vegetation. As a derived variable, we also used the
so-called Ellenberg-index (July temperature/annual precipitation), which has been a classic
measure to delineate the areas suitable for beech in Germany (RÖHRIG 1991). The original
resolution of the data surfaces has been coerced into a coarser grid ("forestry grid",
1517×1896 m) in order to be able to join it to the forest inventory data.
Forest stand data were derived from the Forest Inventory database of the Hungarian State
Forest Service. We used data from 2001. The database contains detailed information on all the
forest compartments of the country, and compartment data were joined to the forestry grid. A
preliminary selection was carried out in order to eliminate stands on steep slopes, sites with
shallow soil or water surplus. Compartments with non-native species and stands of native
species planted outside their natural range were also left out.
2.3. Statistical analysis
We used classification and regression tree analysis (CART, see BREIMAN et al. 1984) to
examine the relationship between the climatic conditions and the presence of the studied
species. The computations were performed with the R software environment (VENABLES &
SMITH 2005), using the rpart module for tree development (RIPLEY 2005). CART models are
hierarchical classification trees, created by repeatedly splitting the learning database into two
subsets, according to simple rules based on a single explanatory variable in each step. The
splitting rule is chosen in each step so that the two groups would differ from each other the
most, becoming as homogenous as possible themselves. The length of the tree is controlled by
pruning it over an independent subset of the full database. A major advantage of CART
models is that they provide a hierarchical, easy-to-interpret view of the relationship between
the independent and the dependent variables, making it possible to identify which climatic
variables account most for the presence of a species. CART technique accepts correlated data
of several types as independent variables, and can represent their interactive effects in an
easily readable way (IVERSON et al. 1999).
3. Results
According to the CART analysis the most important climatic variables were different in the
case of the two different dominant tree species in describing their present distribution. For
beech Ellenberg-index was the most important variable, which split the data set on more than
one level (Figure 1a).
a)
b)
Figure 1. The main predictor of a) beech (Fagus sylvatica) and b) sessile oak (Quercus petraea) dominated
forests and probabilities of presence, modelled by CART.
On the CART tree four final twigs can be seen, which represent different probability values.
In case of sessile oak the annual temperature was the main predictor, which is followed by the
annual precipitation on one branch and the summer temperature on the other one (Figure 1b).
The probabilities of presences are not as differentiated for the branches as in the case of
beech.
On the maps created by using the CART information, probabilities of presence are spatially
displayed based on the climatic data for the whole country. For beech (Figure 2a and b) and
sessile oak (Figure 3a and b) predicted (potential) and observed (actual) occurrences can be
compared.
a)
b)
Figure 2. a) Potential distribution map of beech (Fagus sylvatica) dominated forests. The darker colours
represent higher probabilities. b) Observed occurrences of beech (Fagus sylvatica) dominated forests. Red dots
represent the "forestry grid" cells with beech forest occurrences according to the inventory data.
a)
b)
Figure 3. a) Potential distribution map of sessile oak (Quercus petraea) dominated. The darker colours represent
higher probabilities. b) Observed occurrences of beech (Quercus petraea) dominated forests. Red dots represent
the "forestry grid" cells with beech forest occurrences according to the inventory data.
In the case of the beech we tried to model the effects of GCC on a local scale. For this we
used the regional climate scenarios from MIKA 2000 (summarized in Table 1)
Table 1. Regional scenarios of climate change for Hungary (based on IPCC data, MIKA 2000).
Scenario
I.
II.
III.
IV.
Global temperature increase
+0,5 °C
+1 °C
+2 °C
+4 °C
1. Summer temperature change, Hungary
+1,0 °C
+1,3 °C
+2,0 °C
+4,0 °C
2. Winter temperature change, Hungary
+0,8 °C
-1,7 °C
+3,0 °C
+6,0 °C
3. Annual precipitation change, Hungary
-40 mm
-66 mm
near 0
+40–400 mm
The presumed future distribution of beech in case of the different scenarios can be followed
on Figure 4. It can be assumed, that along with GCC the potential distribution shifts
dramatically in some regions of Hungary, especially in the western part of the country.
However there are also regions, especially on the north, where changes seem to be less
apparent. The first two scenarios predicting increase in temperature and decrease in
precipitation generally show a decrease in the potential distribution of beech (see Figure 4b
and c), while the last scenario assuming the highest increase in temperature shows a possible
expansion in the potential distribution of beech (Figure 4d), although the exact picture
depends much on the precise amount and monthly distribution of the precipitation increase,
which is still poorly known.
a)
b)
c)
d)
Figure 4. Potential distribution of beech (Fagus sylvatica) dominated forests, using different climate scenarios.
a) Present (no change) b) summer temperature: +1 °C, annual precipitation: -40 mm, c) +1.3 °C, annual
precipitation: -66 mm, d) +4 °C, annual precipitation: +220 mm (See also Table 1).
4. Discussion
Using climatic variables in our CART analysis we found that different variables played an
important role in determining the distribution of beech and sessile oak respectively.
Beech, which is a typical European tree species preferring milder climate (MAJER 1968), was
predicted with Ellenberg-index best (Figure 1a). This index represents the relation between
summer (July) temperature and annual precipitation. On those sites where July temperature is
relatively high, annual precipitation has to be also high to have beech forest, which suggests
that absence of severe drought must play an important role in the distribution pattern of beech.
Regions that meet this criteria are predicted on the highest mountains and also in the
southwestern part of the country which is most influenced by Atlantic climate (Figure 2a).
The observed (real) occurrences generally support our predictions, with some exceptions.
These exceptions – for instance the lack of beech stands in some regions in the southwestern
part of Hungary – are probably due to special (sandy) soil conditions (see Figure 2b).
However, variables concerning soil conditions were not implemented into the model.
The area of sessile oak spreads into more continental areas of Europe (MAJER 1968), it
demands less precipitation and can withstand more extreme temperatures as well. According
to CART analysis we found that occurrence of sessile oak is predicted primarily on the cooler
and at the same time drier parts of the country – especially at medium elevations of the
Northern Mountains (Figure 3b). At higher elevations sessile oak becomes rare, which is
primarily caused by the dominance of the very competitive beech. Although human
intervention altered the species composition of beech forests, they were in general not
replaced by stands of other species such as sessile oak. The rarity of sessile oak in drier
regions is caused by complex reasons; the climatic conditions probably play a lesser role. The
interacting role of soil characteristics and human land use (agriculture and flood-control)
should be mentioned as primary reasons. The highly transformed landscape of the Great Plain
obviously influenced the model, since there were very few data on forests with natural species
composition in this region. In the easternmost part of Hungary, where sessile oak is predicted
with higher probability, but is completely absent according to forest inventory database. The
most obvious explanation is the presence of sandy soils.
Based on climatic scenarios we could predict the future distribution of beech-dominated
forest. It is probable, that in case of higher temperature and lower precipitation, the potential
distribution of beech will decrease significantly, particularly in some areas of Transdanubia
where beech occurs on hills and on lower mountains, what represents the borderline of its
range. The predicted drier climate in Europe threatens not only beech but also sessile oak
(RAFTOYANNIS and RADOGLOU 2002).
In evaluating our results some limitations of the model should be taken into consideration.
First, the spatial resolution of the data is relatively coarse, consequently in some locations
forest types could be characterized by less relevant climatic variables. Second, local factors
can override the effect of large-scale effect of climatic variables. Third, it is hard to separate
the climatic and human influences behind the absence and presence of forest or a certain
forest type. Forest types thus could be either under- or over represented in the database.
5. Conclusions
Mapping of potential distribution of zonal forests can be carried out based on CART analysis.
However, implementing more variables into model is necessary to improve its applicability at
finer scales. We hope that our results can support the efforts in solving the problems of
forestry and forest conservation in the face of Global Climatic Change.
6. Acknowledgements
The support of the National Research and Development Program (NKFP), the National
Research Fund (OTKA) and of the Hungarian Academy of Sciences (supported research
group on Production Biology) is gratefully acknowledged. Special thanks to Prof. Cs. Mátyás
for his useful comments.
References
1. IVERSON, L.R., PRASAD, A., SCHWARTZ, M.W., 1999: Modelling potential future individual tree-species
distributions in the eastern United States under a climate change scenario: a case study with Pinus virginiana.
Ecological Modelling, 115, 77–93. – 2. JÁRAINÉ-KOMLÓDI, M. (ed.), 1995: Vegetation of Hungary. Pannon
Encyklopédia (in Hung.), Dunakanyar 2000, Budapest, p. 158–205. – 3. JÁRÓ, Z., 1972: System of siteevaluation in forestry. In DANSZKY, I. (ed.): Forest Management, p. 47–256, (in Hung.). – 4. KRAMER, K.
MOHREN, G.M.J (eds.), 2001: Long-term effects of climate change on carbon budgets of forests in Europe.
Alterra, Green World Research, Wageningen (internal report) p. 11. – 5. MÁTYÁS, CS., 1992: Conservation
problems of forest ecosystems on the Great Hungarian Plain. Hung. Agricultural Rev. 1, p. 33–38. – 6. MÁTYÁS,
CS., CZIMBER, K., 2004: Climatic sensibility of zonal closed forest limit in Hungary. In MÁTYÁS, CS. – VIG, P.:
Forest and Climate, IV. Sopron, 35–44 (Hungarian with English summary). – 7. MIKA, J., 2000: Climatic
scenarios for Hungary. In Forests and Climate III. Kossuth Egyetemi Kiadó, Debrecen. 9–23. – 8. PÉCZELY, GY.,
1979: Climatology Nemzeti Tankönyvkiadó, Budapest, p. 258–259. (In Hung). – 9. MÁJER, A., 1968: Forest
associations of Hungary. Akadémia Kiadó, Budapest (in Hung.). – 10. RAFTOYANNIS, Y., RADOGLOU, K., 2002:
Physiological responses of beech and sessile oak in a natural mixed stand during a dry summer. Annals of
Botany 89, p. 723–730. – 11. RIPLEY, B., 2005: The rpart package – Version 3.1–23. http:// cran.rproject.org/doc/packages/rpart.pdf. – 12. RÖHRIG, E. ULRICH, B. (eds.), 1991: Ecosystems of the World 7.
Temperate deciduous forests. Elsevier, 11 p. – 13. VENABLES, V.N., SMITH, D.M., 2005: An introduction to R –
Version 2.1.1. http://cran.r-project.org/doc/manuals/R-intro.pdf