The author thanks the two anonymous reviewers for the precious help in the improvement of the earlier manuscript. The author sincerely thanks: Dr. Bernard Defaut, for the positive exchange of opinions, for the suggestions, for the correction of the French abstract, and finally, for the encouragement; Mr. Chris Matera, for the correction of the English manuscript.

Introduction

Phytoclimatic maps as tools for ecological regionalization, analysis and modelling

1Many different methods exist to approach climate-vegetation relationships. The usefulness of each method depends on the intentions of the particular study. In the present paper, the analysis, assessment and planning of landscape and ecosystems, and their regionalization are emphasized. For instance, in such fields it is useful to define land units with clearly defined boundaries at different scales. These units should also have some temporal stability to be helpful with diachronic analysis. Furthermore, sharing of information and data needs standardized and reproducible methods. Vegetation is widely diffused and is a synthetic indicator of ecological conditions (Emberger 1939, Ozenda 1994). Climate determines broad type and distribution of vegetation and ecosystems (Walter and Box, 1976). Consequently, a climate classification related to vegetation distribution (i.e. a phytoclimatic one) should be appropriate as a basis for small-scale ecological regionalization (e.g. Bailey 1987; Klijn and De Haes, 1994), to characterize and compare different ecosystems, even where natural vegetation is not available as an environmental indicator due to lack or excessive disturbance.

Defaut’s System of phytoclimatic stages

2Concepts about climate classification are historically linked to those concerned with vegetation classification. As stated by Perera (1968), “climatic classification grew out of a desire to make sense of the vegetation pattern of the earth”. Conversely, some vegetation classification systems were strongly influenced by climate concepts (De Laubenfels, 1975). Numerous methods to relate and map climate and vegetation have been proposed at this point (Box 1981). Summarizing can recognize two main approaches. One defines phyto- or bio-climates relevant for vegetation (or life in general) by using thresholds related to distribution of zonal vegetation types (e.g. Köppen 1884; Köppen and Geiger 1936; Holdridge 1947, 1967; Bagnouls and Gaussen 1957; Troll and Paffen 1963, 1964; Walter and Box 1976; Meurk 1984; Ren et al. 2008; Rivas-Martinez 1997) while the other defines vegetation/biomes regions by grouping species with similar climatic/ecological requirements (or life/growth/phenological forms) (e.g. Box 1981; Haxeltine and Prentice 1986). In the last few years, more quantitative models have been elaborated, based on clustering and species distribution modeling approaches (e.g. Zhang et al. 2017). This paper presents a quantitative, high-resolution, phytoclimatic map of Europe (hereinafter “PME”) based on the “System of phytoclimatic stages” (hereinafter “DSPS”) devised by Defaut (1996). Parameters considered for classification by DSPS are aridity/humidity, mean annual temperature, mean temperature of warmest month and thermal continentality. DSPS relies on commonly collected and available climatic parameters, i.e. monthly mean air temperature and precipitation. Emberger’s works (1955, 1964, in Defaut 1996) inspired DSPS (Defaut 1991, 1992, 1996, 2004). Unlike the former, which is limited in its use to the Mediterranean zone or at least a zone with only one climatic regime (Morat, 1969), the latter was conceived for the entire Palearctic Domain. Defaut, like Emberger, did not draw a priori of the phytoclimatic stages on his climagram (Figure 1), but, instead, devised his system directly from vegetation, as characterized by phytosociology. DSPS is not fixed but is open to improvement. Indeed, the limits of the vegetation stages on the climagram can and will have to be modified in light of the progress of phytosociological and climatic knowledge. It may be necessary to recall here that Defaut traced the phytoclimatic stage borders by plotting the vegetation (i.e. whenever possible, the highest rank phytosociological syntaxa) of numerous meteorological stations on the Qn2 - T / tc diagram. The outline drawing was then simplified to account for stationary corrections (Defaut, 1996, 2015). For instance, the Subhumid hot and temperate stages (SH2-3) are characterized by the presence of Quercetalia ilicis Br.-Bl. ex Molinier 1934 communities; the Subxeric temperate stage (SX3) by the Quercetalia pubescenti-petraeae Klika 1933; the Colline or axeric temperate stage (AX3) by the Quercetalia roboris Tx. 1931 and Carpino betuli - Fagenalia sylvaticae Rameau in J.-M.Royer et al. 2006; the Boreo-Montane or Axeric cool stage (AX4) by the Fagenalia sylvaticae Rameau in Royer et al. 2006 and Cephalanthero rubrae - Fagenalia sylvaticae Rameau in J.-M.Royer et al. 2006; the Boreo-Subalpine or Axeric cold stage (AX5) by the Piceetalia excelsae Pawłowski et al. 1928; the Arctic-Alpine or Axeric very cold stage (AX6) by the Caricetalia curvulae Br.-Bl. in Br.-Bl. et Jenny 1926 and Carici rupestris-Kobresietea bellardii Ohba 1974; and so on. The general framework is further enriched and detailed by continentality / oceanicity influence, as explained below, in the section “Determination of phytoclimatic stage”. In the climagram (Figure 1) are shown (strongly simplified) the major syntaxa in their phytoclimatic stages. Unfortunately, phytosociology is not used or developed in several countries. Therefore, some parts of the climagram were completed by using physiognomic vegetation types. Defaut (1996, 2015, 2017) gave a detailed description of each phytoclimatic stage and related vegetation. DSPS distinguishes three large groups of phytoclimatic stages. A first group includes stages that experience a dry period (xeric bioclimates): in this case, the stages are mainly under the control of aridity-humidity, and secondarily under the control of temperature or continentality. A second group involves stages without a dry period (axeric bioclimates). Between these two groups, there is a third, transitional, formed by stages that experience a period of subaridity, (subxeric, or also subaxeric bioclimates, according to Defaut, 2004, 2015). These two latter groups are mainly under the control of temperature and to a lesser extent under the control of continentality.

Materials and methods

Study area

3The PME covers all of Europe (Figure 2 left). Customary eastern European boundaries were slightly enlarged to explore trends of phytoclimates and their transition toward Asia.

Data

4PME is based on data from scattered climatological stations, other than Europe, in Siberia, Middle East, Africa, Macaronesian islands and Greenland (Figure 2 right). Climatological data from 1113 stations were used. Main data sources were Mühr (2010) and Defaut, (1996). From Defaut’s work, original climate data was taken, nonclassified phytoclimatic stages, to ensure homogeneity of elaboration and results. Other data sources were AEMET, ARPAE; MET OFFICE; MeteoSwiss, 2016; Perosino and Zaccara, 2006; Pogodaiklimat; Zaninović et al (2008); Wikipedia. Furthermore, 15 fictitious stations were added on some mountain tops, mostly at average snowline (Iceland, Scandinavian Alps, Svalbard, Franz Josef Land, Novaya Zemlya, Maritime Alps, Caucasus, Pico de Europa, central and southern Apennines), where climatological stations were lacking. Climatological stations are unevenly distributed in the study area. On the other hand, it is important to have more stations in complex terrain areas (where climate is highly diversified; e.g. Mediterranean borders) than in large plains (where climate changes very slowly across landscape; e.g. east European plains). Central Southern Europe, from southwestern Germany to central Italy hosts the highest density of stations. The lowest density of stations is found in Boreal Eastern Europe. Appendix 1 shows climatological data sources.

Determination of phytoclimatic stage

5In this study, the method devised by Defaut (1996) to determinate the phytoclimate of each climatological station was followed. Climatological data were plotted as points on a climagram (i.e. a diagram in which each axis is correlated with a climatic parameter, see Figure 1), according to values of their climatic parameters. Climatic parameters of Defaut’s system are the following: 1) average annual temperature T (in °C), or average monthly temperature of the warmest month tc (in °C) for the phytoclimates with T < 4.5°C (±0.5°C), 2) thermal continentality tc-tf (in °C), where tf is the average monthly temperature of the coldest month, 3) aridity index Qn2 (Defaut 1991), calculated as following:

where P is the average annual rainfall and P’ the cumulative rainfall of the driest three consecutive months (“dry” sensu Gaussen 1957, in Defaut 1996) where dryness is inversely proportional to the ratio between annual rainfall and average temperature. In the climagram, the x axis shows the values of Qn2, the y axis shows the values of T and tc. In particular, the boundaries between stages 1, 2, 3, 4 and 5 refer to the middle and lower part of the Y axis, which corresponds to the annual temperature T (and which is measured from +5 to + 30 ° C), while the boundaries between stages 5 and 6, and between stages 6 and 7 refer to the upper part of the Y axis, which corresponds to tc (and which is measured from -20 to + 20 ° C) because when summer temperatures are sufficiently low, trees (stage 6) disappear or permanent ice appears (stage 7). Position of a point on the climagram, plus its thermal continentality class gives the phytoclimatic stage. Tables 1, 2, 3 show classes of the three parameters with the corresponding codes, names and ranges of values. Range of aridity/humidity depends on the position on the climagram. For the sake of consistency and simplicity, original codes of axeric phytoclimatic stages (i.e. C, BM, BS, AA and N) were named together “AX”, Axeric. Another difference between the present study and Defaut (1996) involves the distinction and the grouping of substages (e.g. the Thermo-Atlantic or the Medio-European). Indeed, in the present paper, phytoclimatic classes are qualified and named exclusively by their parameter classes, and all have “stage” rank. For instance, a climatological station with parameters of classes AX (Axeric), 3 (Temperate) and c (suboceanic), will have the formula AX3c and will be the “Axeric, temperate, suboceanic stage”.

Table 1: Classes of aridity/humidity.

Code	Name
E	Eremic
HA	Hyperarid
A	Arid
SA	Semiarid
SH	Subhumid
SX	Subxeric (or Subaxeric)
AX	Axeric

Source: Defaut 1996.

Table 2: Classes of temperature.

Code	Name	Range
1	Very hot	23.0°C (±2) ≤T
2	Hot	16.5°C (±1) < T <23.0°C (±2)
3	Temperate	10.0°C (±1) < T <16.5°C (±1)
4	Cool	4.5°C (±0.5) < T <10.0°C (±1)
5	Cold	10.5°C < tc et T < 4.5°C (±0.5)
6	Very cold	2.0°C (±2) < tc < 10.5°C
7	Nival	tc ≤2.0° C (±2

Source: Defaut 1996. Legend: “T” is the mean annual temperature in ° C; “tc” is the mean temperature of the warmest month in ° C.

Table 3: Classes of thermal continentality.

Code	Name	Range
a	Hyperoceanic	9°C ≥ tc-tf
b	Oceanic	16°C > tc-tf > 9°C
c	Suboceanic (= Subcontinental)	22°C > tc-tf > 16°C
d	Continental	42°C > tc-tf > 22°C
e	Hypercontinental	tc-tf ≥ 42°C

Source: Defaut 1996. Legend: “tc” is the mean temperature of the warmest month in ° C; “tf” is the mean temperature of the coldest month in ° C.

GIS processing

6Development of phytoclimatic map were carried out using software QGIS (2.18.0), SAGA (2.1.2) and GRASS (7.0.5). Figure 3 shows the GIS workflow.

7Classified climatological stations were implemented in a geodatabase. Data were coded according to classes of their three phytoclimatic parameters, i.e. temperature, aridity/humidity and thermal continentality. For each parameter, a model was created by universal kriging (Figure 4). The Digital Elevation Model (DEM) used for the kriging was the Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010), courtesy of the U.S. Geological Survey. The resolution chosen to develop the map was 15 arc-second with RMSE (Root Mean Square Error) range between 29 and 32 meters. Resulting patterns of aridity, temperature and continentality differ clearly from one another (Figure 4). Whereas temperature shows a latitudinal gradient and continentality a longitudinal one, aridity/humidity is an intermediate one. Although aridity and continentality are related to temperature, their contrasting patterns is reassuring regarding the appropriateness of choosing them as basic variables.

8By overlaying the three models, the final phytoclimatic map was obtained, with a cellsize of around 1.1 Km. A post-processing of such a map was necessary to compare PME to the Map of the Natural Vegetation of Europe (Bohn et al, 2000/2003) (hereinafter MNVE), due to the different grain pattern of the two maps. Indeed, the MNVE has smoother and larger patches. Therefore, the more fine-grained pattern of PME was simplified and smoothed by using a resampling filter. The choice of such a method was the result of a trial and error process. The different appearance of the full resolution and resampled map is illustrated by a detail showing the Alpine chain (Figure 5).

Comparison of the PME to the Map of the Natural Vegetation of Europe (MNVE)

9MPE should be predictive of (climatic) potential vegetation. Therefore, it was chosen to compare such a map to MNVE by the Kappa statistics, a method widely used in assessing agreement between different maps. MNVE is very useful because it presents vegetation types of the whole continent in a homogeneous way. Furthermore, although it is based on physiognomy, it contains a large amount of phytosociological information, useful here because DSPS is principally based on phytosociology. Use of vegetation maps to validate phytoclimatic maps faces at least three major issues. First, although phytoclimatic classification should be predictive of potential vegetation, it should not be expected to match perfectly. Indeed, vegetation of a given place is an expression of the overall environment (and history), and not only the climate. Second, Defaut devised his phytoclimatic system using distribution of potential vegetation. Therefore, to validate a map based on such a system using vegetation could lead to circular reasoning (e.g. Webb, 1968; De Laubenfels, 1975). Third, basis map levels of generalization affect the results.

10Regarding the first issue, omitting azonal vegetation from the comparison should solve at least part of the problem. As stated by Defaut (1996), edaphic, microclimatic, or other causes, can mimic macroclimate factors (e.g. heat or moisture) and allow even zonal vegetation to occur “out of place” (i.e extrazonality). Defaut speaks about such local effects as “site corrections”. In such places there is no correspondence between phytoclimate and vegetation. Regarding the second issue, it should be noted that DSPS and MNVE are based on different systems of vegetation classification. Indeed, while the former relies on phytosociological taxa (at the highest rank possible), the higher hierarchical levels of the latter follow a physiognomic-ecological classification (Bohn et al, 2000/2003).

11Keeping in mind Ozenda’s statement (1994), different approaches to vegetation description should not lead to strikingly different results, therefore a comparison between PME and MNVE should not be too arbitrary and the risk of circular reasoning should be (at least largely) avoided. Note that comparing the present map with a vegetation classification system different from that used by Defaut could be a positive test of the robustness of such a phytoclimatic system. Furthermore, in the original Defaut study, climatological data were almost completely lacking in some areas of Europe, such as the eastern Alps, Italy and the Balkan Peninsula. Therefore, a comparison between the phytoclimatic map and MNVE explores climate-vegetation relationship in an area larger than originally analyzed by Defaut. This is a further test of such a phytoclimatic system.

12It should be remembered that PME is larger than MNVE. Thus, only the area covered by the MNVE has been considered for comparison. For the Kappa analysis, the data sets that are compared must distinguish the same classes (Metzger et al. 2005). Therefore, both PME and MNVE were reclassified in 17 classes (see Appendix 2). First, MNVE’s mapping units (except azonal mapping units) were assigned to phytosociological syntaxa according to information presented in the accompanying MNVE documentation. Phytosociological nomenclature has been updated according to Mucina et al. (2016). The vegetation types obtained were grouped in 17 major phytoclimatic classes (see Appendix 2) according to Defaut (1996). It was not possible to classify all zonal mapping units (and thus every phytoclimatic stage), because of the lack of information about some vegetation types or regions. In any case, a large part of Europe (89,59%) was classified. To homogenize PME to MNVE classes, the phytoclimatic stages of the former were grouped. In general, there was a major distinction between oceanic sensu lato group of phytoclimatic stages and continental ones (see Appendix 2). Other than the Kappa stastistic, the distribution of MNVE vegetation formations in the phytoclimatic stages were also analysed. To minimize negative effects caused by the third issue, a more generalized phytoclimatic map was developed by post-processing the full-resolution version, as explained above.

Results

Comparison to MNVE

13The Kappa statistic for PME and MNVE is 0.554673. Values between 0.55 and 0.7 are generally considered good (e.g. Monserud and Leemans 1992; Haxeltine and Prentice 1996; Metzger et al. 2005).

Distribution of European phytoclimatic stages

14PME shows distribution and pattern of phytoclimatic stages in Europe, as determined mainly by latitude, distance from the ocean and sea, and elevation (among other factors). In addition, anthropization (e.g. through lack of vegetation and urbanization) can affect phytoclimatic stages in some areas (Defaut, 1996). Figure 6 illustrates the PME. This image is intended for general orientation only, and the reader should refer to the shapefile annexed.

15Unlike most climate classifications, Defaut’s hierarchy put moisture first, not heat (Box 2016). Thanks to this characteristic of the system, it is easy to visualize the major phytoclimate groups and consequently the broad continental zones, as shown in Figure 7. Furthermore, the true Mediterranean mountains that experience summer drought in the south, are grouped with lowland Mediterranean climates and not with central or northern mountains. In detail, Figure 7 (left) illustrates Xeric phytoclimates (A, SA, SH in Table 1): which form a continuous band across SW and SE Europe. Continentality subdivide Xeric phytoclimates into a SW subgroup that is oceanic to suboceanic, from a SE subgroup that is suboceanic to continental. Figure 7 (middle) illustrates Subxeric (SX) phytoclimates, transitional between xeric and axeric. Such phytoclimates experience summer subaridity; and form a belt between xeric and axeric phytoclimates. It is possibile to subdivide them into a Western subgroup, oceanic to suboceanic, and into an Eastern subgroup, continental. Caucasus region host an outlier of Subxeric phytoclimate. Figure 7 (right) illustrates Axeric (AX) phytoclimates: except for some outliers on the east coast of the Black sea and on the Central Southern reliefs, they occupy the central and northern part of the continent and range from the warmer and hyperoceanic stages of Western Europe, to the colder and continental Northern and Eastern Europe.

Coverage of European phytoclimatic stages

16Fifty different phytoclimatic stages were found in Europe through this study. The aridity/humidity factor ranges from arid to axeric; the temperature factor from hot to nival and continentality from hyperoceanic to continental. The following stages are completely lacking in the study area: eremic and hyperarid classes of aridity/humidity; very hot temperature class and hypercontinental continentality class. Figure 8 bar chart illustrates area, in percent, of each phytoclimatic stage. After resampling process to create the generalized version of the map, three phytoclimatic stages were missed, which are SX3a, AX6a and AX7a. These stages were very small, and percent area covered by the stages after the resampling process remained almost unchanged.

17The most widespread ten stages cover three-quarters (74,8%) of Europe, and the most extensive singular stage, the “Axeric cold continental” (AX5d) covers almost a quarter of the continent. Xeric climates cover 20.7%, subxeric 16.9% and axeric 62.5% of the study area.

Composition of vegetation in phytoclimatic stages

18The bar chart in Figure 9 illustrates the vegetation composition of each phytoclimatic stage. Vegetation units considered here are the main formations of MNVE, as listed in Table 4. As PME extent is larger (toward the East) than MNVE, only the same area as MNVE is considered in the graph, and dominant formations gradually replace one another from an arid and hot pole to a moist and cold one. Although such a gradient is somewhat broken because of each group of climates (i.e. Xeric, Subxeric and Axeric) it is further subdivided by gradients of temperature and continentality, and the main sequence of formation seems clear. Hot and temperate Xeric phytoclimatic stages are clearly separated in two groups by continentality: first, oceanic and suboceanic stages, which are dominated by Mediterranean sclerophyllous forest and scrub (formation “J”); and second, continental stages, which are dominated by Deserts (“O” formation) and Steppes (“M” formation). In detail, the Mediterranean forest and scrub formation is dominant in arid and semiarid oceanic to suboceanic stages, and hot to cool temperature levels (A2b, SA2b, Sa2c, SA3b, SA3c, SA4c). This formation dominates also in subhumid hot, temperate, oceanic and suboceanic stages (SH2b, SH2c, SH3b, SH3c). In the subhumid temperate suboceanic stage, dominance of Mediterranean vegetation is not so absolute, and is associated with “Thermophilous mixed deciduous broadleaved forests” (formation “G”). Subhumid, cool oceanic (SH4b) stage presents the reverse situation, with a dominance of thermophilous vegetation associated with Mediterranean vegetation. Desert prevails in arid and semiarid temperate stages (A3d, A4d and SA3d), while steppe dominates in semiarid cool (SA4d) and subhumid temperate and cool stages (SH3d and especially SH4d). The subhumid cool suboceanic stage (SH4c) shows an intermediate situation, with presence of “G”, Thermophilous mixed deciduous broadleaved forests, other than “J” and “M” formations, Mediterranean sclerophyllous forest and scrub and Steppes, respectively. Regarding subxeric stages, it is noticeable that Mediterranean vegetation is present (but not important) only in temperate/cool, oceanic/suboceanic ones (SX3b, SX3c, SX4b, SX4c), whereas forest steppe and steppe (formations “L” and “M”) compare in continental temperate to cold stages. Thermophylous forests are present in stages temperate to cool, irrespective of their continentality. “Mesophytic deciduous broadleaved forests and mixed coniferous-broadleaved forests” (formation “F”) are important in subxeric stages, up to cold ones (SX5c, SX5d). In cold stages (SX5b, SX5c, SX5d) there are also “Subarctic, boreal and nemoral-montane open woodlands as well as subalpine and Oro-Mediterranean vegetation” and “Mesophytic and hygromesophytic coniferous and mixed broadleaved-coniferous forests” (formations “C” and “D”). The former formation becomes dominant in very cold stages (SX6c and SX6d) and is associated with “Arctic tundra and alpine vegetation” (formation “B”). Axeric temperate/cool and oceanic/suboceanic stages (AX3b, AX3c, AX4b and AX4c) are largely dominated by mesophytic deciduous and mixed deciduous-conifer forests (formation “F”). The presence of this formation diminishes in importance in hyperoceanic, continental, and cold stages. In hyperoceanic phytoclimatic stages (AX3a, AX4a and AX5a) become important (and dominant in AX4a and AX5a), the “Atlantic dwarf shrub heaths” and “Mires” (formations “E” and “S”). Continental, temperate and cool stages are dominated by azonal “Vegetation of floodplains, estuaries and fresh-water polders and other moist or wet sites” with the presence of thermophylous vegetation (AX3d) or “Mesophytic and hygromesophytic coniferous and mixed broadleaved-coniferous forests” (AX4d). This latter formation is dominant in the cold continental, axeric stage (AX5d) and it remains important in cold oceanic/suboceanic stages (AX5b and AX5c) associated with “Subarctic, boreal and nemoral-montane open woodlands as well as subalpine and Oro-Mediterranean vegetation”. Very cold stages are dominated by “Arctic tundras and alpine vegetation” (formation B). From oceanic to continental stages the dominant formation diminishes and “Mesophytic and hygromesophytic coniferous and mixed broadleaved-coniferous forests” (formation D) increases. Nival stages were considered here altogether. Glaciers cover more than half the area, followed by the presence of “Arctic tundra and alpine vegetation” and by “Polar deserts and subnival-nival vegetation of high mountains” (formation A). Note that the map of the natural vegetation of Europe does not show glaciers southwards of Scandinavia.

Table 4: List of the main formations (Bohn et al 2000/2003).

Zonal vegetation (primarily climatically conditioned)

A Polar deserts and subnival-nival vegetation of high mountains

B Arctic tundras and alpine vegetation

C Subarctic, boreal and nemoral-montane open woodlands as well as subalpine and Oro-Mediterranean vegetation

D Mesophytic and hygromesophytic coniferous and mixed broadleaved-coniferous forests

E Atlantic dwarf shrub heaths

F Mesophytic deciduous broadleaved forests and mixed coniferous-broadleaved forests

G Thermophilous mixed deciduous broadleaved forests

H Hygro-thermophilous mixed broadleaved deciduous forests

J Mediterranean sclerophyllous forest and scrub

K Xerophytic coniferous forest, woodlands and scrub

L Forest steppes (meadow steppes alternating with deciduous broadleaved forests) and dry grasslands alternating with xerophytic scrub

M Steppes

N Oroxerophytic vegetation (thorn-cushion communities, tomillares, mountain steppes, open scrub)

O Deserts

Azonal vegetation (determined by specific soil properties and water balances)

P Coastal vegetation and inland halophytic vegetation

R Tall reed vegetation and tall sedge swamps, aquatic vegetation

S Mires

T Swamp and fen forests

U Vegetation of floodplains, estuaries and fresh-water polders and other moist or wet sites

Discussion and conclusions

19PME shows phytoclimate distribution at three different scales. The first one is zonal or continental. Phytoclimates result from interplay of positional and circulational factors (Box 2016), namely, latitude, distance from the Atlantic Ocean and global air mass circulation. At this scale, the phytoclimates pattern is useful to define boundaries of zonal ecological land units. For instance, Axeric phytoclimates, distinguished by their temperature degrees represent Polar/Subpolar, Boreal and Nemoral (or Humid Temperate) ecozones, according to classification by Schultz (1995). Subxeric phytoclimates, distinguished by their continentality degree, can be considered merely as two transitional zones (zonoecotonal, sensu Walter and Box, 1976) between Nemoral/Boreal and Mediterranean/Steppe zones, or as two distinct zones (e.g. Thermonemoral and Forest-steppe, according to Ozenda, 1994 and Ozenda and Borel, 2000). Finally, Xeric phytoclimates, distinguished by continentality, represent Mediterranean (or summer drought subtropics, Schultz, 1995) and arid temperate ecozones (Figure 7).

20The second scale is a subcontinental or regional one, where phytoclimates pattern seems determined by major landforms. For instance, distance from the sea, rain-shadows and relative position and elevation of the plain and mountains lead to a complex pattern of phytoclimatic stages in the Po plain (Fig. 5). In such an apparently homogeneous area, phytoclimatic stages range from temperate suboceanic Axeric and Subxeric (AX3c and SX3c) stages at the Alps’ foothills, to temperate continental, Subxeric and Subhumid stages (SX3d and SH3d) in the centre of the plain, to a Subhumid temperate suboceanic stage (SH3c) in the eastern and southern plain. The difference between the drier central sector of southern Alps, centered on Subxeric Lake Garda and Adige River valley, and the moister western and eastern Axeric sectors is also noticeable. Finally, on the third scale, phytoclimatic map units follow the pattern of elevation. In this case, the map shows broad altitudinal belts. Naturally, the three scales of phytoclimatic units are intermixed and are often not easy to distinguish. As shown above, PME shows good agreement with MNVE. Despite this fact, as expected, PME fails to predict potential vegetation perfectly. While this fact is self-evident in areas with azonal vegetation, it is less obvious in areas with zonal vegetation types. Disagreement between climate map and vegetation map are due to many causes, some implied with the nature itself of climate and vegetation, some related to the scale of generalization and mapping mode (e.g. type of interpolation), some due to climatic factors not considered by the map, and lastly, some caused by the influence of other environmental factors, e.g. landform, soil, disturbance, history, etc. It is possible to summarize some aspects of the comparison between PME and MNVE as follows: 1. PME correlates better with sub-formations than formations of MNVE. Perhaps, this fact might depend on the method used to group the vegetation units; 2. Vegetation composition of warmer Xeric phytoclimates looks very similar. This fact could be due to the detailed subdivision of xeric stages and Mediterranean vegetation by DSPS, whereas MNVE groups Mediterranean vegetation types in a unique formation; 3. It is possible that on highly degraded lands, PME shows a drier stage than indicated by potential vegetation because of anthropogenetically altered interactions on the climate-vegetation-soil system. This fact has been already discussed by Stewart (in Daget 1977) and by Daget (1977) about the Emberger bioclimatic system, and by Defaut (1996) regarding his own system; 4. As described above, DSPS was applied in this study in an elementary form. Perhaps following Defaut more strictly in grouping the elementary phytoclimates would result in a somewhat different correlation with MNVE. Another issue is regarding the uneven density and distribution of stations. Whereas this is not a problem in flat homogeneous regions, concerns are raised for mountainous areas. A very dense network of station covers the central part of Europe, from South Germany to the Alps and the Italian peninsula (Figure 2 right). Likely, this area is the most sensitive in the study area, because of its geographical position between ocean, continent, the Mediterranean and because of the complicated interplay of high mountains, seas and low basins. In the other regions of the study area, climatic changes seem more gradual, so the lesser density of stations should not overly affect the map. In any case, future further efforts could improve the PME by including more stations. Theoretically, in the section of the PME with a denser network of stations, it would be possible to increase the scale of the map, though this increase should realistically also take into account local factors that affect climate at local scales. Some concerns might arise about the PME and DSPS due to being based on Potential Natural Vegetation. Summarizing, this concept was criticized (e.g. Chiarucci et al 2010) principally because it is deterministic and static, whereas real world vegetation is dynamic and stochastic. Nevertheless, as suggested in several papers (e.g. Somodi et al. 2012), PNV concept is still useful, for instance as a baseline for planning and comparison between different scenarios and is still widely used (e.g. Liang et al. 2012; Reidl et al. 2013). Of course, PME possesses the same weaknesses of all sharp boundary maps. In conclusion, while keeping in mind the results of the map comparisons, it is felt that PME, thanks also to its reliability and relative simplicity could be a useful tool in ecological analysis and environment assessment, as well as in climate change studies, and for educational purposes.