**1. Introduction**

Climate change is expected to have a major impact on long-term community dynamics worldwide. Among other things, the global temperature increase and extreme weather events have a significant impact on the structure of forest ecosystems and also on the dynamics of environmental factors in forest patches and nearby open areas [1–3]. Habitats with different attributes (e.g., vegetation physiognomy and plant species composition) may respond differently to the changes, depending on their sensitivity and resilience. Transition zones could be sensitive habitats, where the structures of two different vegetation types have significant effects on each other, while their strong biotic and abiotic relationships

determine the dynamics of the vegetation [4]. Typical transition zones are forest fragments, which, in many cases, are human-induced [5].

Studies about fragmented habitats have mainly focused on the e ffects of the surrounding open areas around the forest fragments [6–8]. However, there are naturally fragmented vegetation types, such as forest-steppe habitats in Central Hungary. In the case of this sandy forest-steppe vegetation, the fragmented structure has a natural origin. In this habitat, natural drying processes have been observed for decades [9,10], inhibiting the growth of woody vegetation. In addition to these processes, the seeds of tree species in this habitat can only germinate in depressions with ideal topographic conditions where a su fficient amount of water can accumulate. However, these seedlings can rarely develop into even loosely closed tree groups. Thus, the succession dynamics in a sandy-forest-steppe habitat are rather slow due to the vegetation edges, where the strong abiotic di fferences between the grassland and the fragment prevent forest expansion and development [10]. The sandy forest-steppe is a vegetation type that is in danger of total extinction in the near future due to the aridification in the Pannonian region [11]. Thus, improving our knowledge of this type of ecosystem is important for understanding the dynamics of abiotic and biotic factors in transition zones. Therefore, it is important to examine the e ffect of smaller groups of trees and larger forest patches on the surrounding grassland matrix where an edge e ffect is observed [12–14].

Among the environmental factors that characterize a habitat, the microclimate strongly influences the growth and distribution of plant species [15–18]. The most important microclimatic components that are usually examined included the air temperature, air humidity, wind force, and solar radiation [4,16]. In a study by Lin and Lin [19], environmental measurements showed that solar radiation and wind velocity had significant opposite e ffects on air temperature in urban vegetation. The intensity of solar radiation had a positive e ffect, while wind velocity had a negative e ffect. Thus, the cooling e ffect of woody vegetation is highly dependent on these factors. Hence, temperature and humidity are the main variables, from which we can also obtain the vapour pressure deficit [4,20,21]. The vapour pressure of the air in the forest and in the open areas would be similar if the vapour pressure deficit and relative humidity were determined by the air temperature. The di fference in the air temperature causes the di fferences between the relative humidity of open and canopy-covered areas [17,22,23]. The vapour pressure deficit (VPD) can be an important limiting factor in plant growth because conditions with above-threshold values can be considered stressful. Under conditions with above-threshold values, photosynthetic CO2 uptake is strongly limited due to stomatal closure to prevent water loss [5,22,24,25].

These variables are also closely related to each other and are influenced by aspects of the vegetation structure such as the height, canopy cover, species composition, etc. [1,26]. The canopy cover directly influences the below-canopy and nearby microclimate, decreasing the solar radiation energy reaching the surface during the day while retaining outgoing longwave radiation during the night [1,26,27]. In an Atlantic forest fragment, the transient nature of the edge was observed by microclimate measurements, where the air temperature at the edge was higher than in the forest interior, while the relative humidity was lower [27]. This transient nature was also reported in a Douglas-fir forest in the U.S. [13]. Therefore, the forested areas heat up less during daytime and cool down less during night-time. In the UK, in a study of a semi-natural temperate forest, the air temperature in the forest was lower than that of the grassland and decreased from the canopy to the understory on a sunny summer day. The largest di fference in air temperature between these two di fferent areas was approximately 3 ◦C [12]. These microclimatic di fferences between open and canopy-covered areas may be much larger and more relevant, depending on the geographical location and physiognomy of the vegetation studied. Thus, little information is available about the spatio-temporal relationship between the microclimate components, the vegetation structure and ecosystem functioning in fragmented vegetations, although this knowledge is important for understanding the ecological functioning of forest ecosystems.

The edge e ffect is the microclimatic or vegetation structural di fference between the forest edge and the interior of the forest [5]. The size of the forest patch considerably influences the extent of the edge e ffect and the di fference between the forest interior and peripheral areas such as nearby open grassland [4,14]. According to Murcia's review [14], there are three types of edge e ffects on the forest fragments. The first one is the abiotic e ffect, in which a change in the environmental conditions will create a habitat matrix with physical conditions di fferent from the forest (e.g., di fferences in structural complexity and biomass). The two factors influencing the abiotic e ffects include the physiognomy and orientation. The second is the direct biological e ffect, in which there is a change in the abundance and distribution of species in the edge area because of di fferences in species tolerances. This is due to the sudden changes in the physical conditions in the edges, which may increase the tree mortality caused by wind force (wind throw) or by fire. This will lead to a change in species composition. The third is the indirect biological e ffect: alterations in species interactions, such as competition, predation, pollination, etc.

The stages of natural succession modify the edges during the development time; therefore, fluctuations in the distribution of plant species can be observed [28–30]. The current vegetation pattern corresponds to the present environmental conditions. The duration of the study is a very important parameter for such research topics, and also, the scale of sampling is relevant, but fine-scale studies are very rare [31]. Most studies about forest fragments and transition zones have used large-scale resolutions with a minimum of 20–50 m intervals between the measurement points, and the vegetation sampling has usually consisted of random, non-contiguous plots. As a result, information on fine-scale vegetation patterns is lost, and significant microclimatic di fferences may not be detected in the transition zone [5,13,20]. The fine-scale resolution with a sensor network helps to explore small-but-significant di fferences within the vegetation [31].

The main goal of this article was to assess the relationship between the microclimate and the microcoenological structure of the herb layer and to describe the microclimate-modifying e ffect of a grove in a sandy forest-steppe habitat. In three phenological stages of the one-year-long vegetation period, we analysed the air vapour pressure deficit (VPD) and plant species composition of the herb layer in four intersecting transects with di fferent cardinal directions through a group of trees. The air temperature and humidity were measured in the herb layer at a fine spatial and temporal scale with 89 devices during the measurement campaigns, and 350 microcoenological relevés were made per season. We hypothesized that the VPD-modifying e ffect of the grove would gradually decrease in all directions away from the edge. We also assumed that the spatial microclimate patterns did not di ffer from season to season, that only the intensity of the modifying e ffect changed. The response of vegetation to the microclimate pattern was quantified according to the distribution of ecological indicator values. Our third hypothesis was that the coenological and indication structure of the herb layer di ffered only between the grove and the open areas and they were homogeneous in the open grassland.

#### **2. Materials and Methods**

#### *2.1. Study Area*

Our study site (Figure 1) was situated in Central Hungary (Fülöpháza region of the Kiskunság National Park; 46◦5328.18 N 19◦2446.91 E., 107 m a.s.l.). Throughout history, this region has been heavily cultivated, and quicksand was stopped by a fforestation with invasive tree species. These processes destroyed much of the sandy forest-steppe habitat and gave space to both woody and herbaceous invasive species [10,32]. Natural forests have remained in the landscape in relatively small patches, and the sandy grasslands have, in many cases, been regenerated by secondary succession. Thus, the maintenance and restoration of the remaining natural areas is extremely important in this region. This habitat is characterized by a semi-desert climate, well indicated by the xerophytic dominant grass species, which, in the primary grasslands, are *Festuca vaginata* Waldst. & Kit. ex Willd. or *Festuca rupicola* Heu ff., while the secondary grasslands contain mostly *Bromus tectorum* L. and *Secale sylvestre* Host. The dominant tree species in the natural forests are *Populus alba* L. or *Quercus robur* L., while the forest plantations are mostly dominated by *Robinia pseudoacacia* L. or *Pinus nigra*

J.F. Arnold [32]. A grove of poplar (*Populus alba*) and the surrounding grassland were selected for this study. Measurements of microclimate components and vegetation sampling were performed in four intersecting transects (44 m long each) with different cardinal directions, forming, together, a star-shape sampling arrangemen<sup>t</sup> with the group of trees in the middle. The diameter of the tree group was 15 m on average.

**Figure 1.** Study site (**a**) and relief map with the indication of the grove (**b**). Coordinates refer to the Hungarian Unified National Projection System. There were 89 measurement positions (4 × 22 positions in the four transects and the centre as position 12 in all transects). White dashed line: visual tree edge. Black dashed lines: the position and extent of individual shrubs and poplar sprouts.

#### *2.2. Microclimate Measurements*

Air temperature and air humidity were measured in the herb layer with a sensor network for 48 h (1-min resolution) during three measurement campaigns in different phenological stages of the vegetation in 2018 (May, July, and October). The data loggers were placed 20 cm above the soil surface, at the average height of the herbaceous vegetation, along the transects in 2 m intervals (23 measuring positions in each transect). The Crossbow MICA XM2110CA mote (Crossbow Technology Inc., Milpitas, CA, USA), UNI-T UT330B Mini USB Temperature Humidity logger (UNI-TREND Technology Co. Ltd., Guangdong, China), and Voltcraft DL-120TH USB Temperature Humidity logger (Voltcraft, Hirschau, Bavaria, Germany) were used in a sensor network, including 89 dataloggers altogether. The sensors were shielded with a white plastic plate to avoid solar radiation heating. Before the measurements, the sensors were calibrated. We selected precipitation-free measurement periods, but the sky was cloudy during the observation in May. The main changes in the weather were recorded (e.g., clouds' shading and movement). The location of the visual edges of the grove, the positions of bushes and trees in the surrounding area, and the shadow of the grove were also recorded.

#### *2.3. Vegetation Sampling*

Vegetation data were also collected along the transects. Microcoenological relevés were recorded in 0.5 m × 0.5 m contiguous plots in each season in parallel with the micrometeorological measurements. Plant names and indicator values (TZ temperature requirement; WZ moisture requirement elaborated by Zólyomi in Table S1) were used according to FLORA Database 1.2. These ecological indicator values quantify the environmental optimums for plant species based on their occurrences in natural habitats. Lower indicator values mean lower temperature and moisture requirements, while higher values mean higher requirements [33].

#### *2.4. Vapour Pressure Deficit and Duration Curve Method*

The vapour pressure deficit was computed from the relative air humidity (RH) and air temperature (t) according to the formula developed by Bolton [21]:

$$VPD = (100 - RH) \times 6.112 \times e^{\gamma} (17.67 \times t/(t + 234.5)) \tag{1}$$

with *t* in ◦C, *RH* in %, and *VPD* in Pa.

In hydrology, the "flow duration curve" is a widely used method for detecting the rate of occurrence of values for a variable above a certain critical limit (flooding degree in hydrology). With the help of this method, one can identify the duration of a flood, which means the number of days flooded during the current period [34,35]. Since micrometeorological data are large sets of fluctuating time series, similar to the hydrological data, we consider the duration curve to be a promising tool for the analysis of temperature, humidity, or vapour pressure deficit data in plant ecological studies. Our data have a diurnal cycle and may be of any temporal resolution. Our study focused primarily on the derived data, such as the percentage of the VPD values above an appropriate threshold (1.2 or 3.0 kPa) over a 24-h period (exceedance rate) that can indicate the microclimatic conditions of the vegetation. A VPD duration curve (DC) was constructed from a 24-h period of records, from 12:00 to 12:00 in each measuring position. The DC of one variable was created by sorting all the data in descending order. Thus, the rank of the highest value was 1, while that of the smallest was n (number of measurements). The ordered data can be plotted to show the DC, where the relative order (e.g., percentage) on the X-axis reflects the exceedance probability for a particular value of the variable on the Y-axis (e.g., the vapour pressure deficit at one measuring position), indicating the percentage of time a given value was equalled or exceeded over the measurement period. The tendency of the curve shows the relationship between the exceedance probabilities and the examined variable. This graph is called a period-of-record DC. Based on this method, the exceeded values can be easily determined over the measurement period [34,36].

#### *2.5. Data Processing*

Data processing was carried out on the temperature and humidity data recorded at a per-minute frequency; during 2018, altogether, more than 1,200,000 records were processed. A 24-h recording period was used to calculate the VPD and VPD exceedance rates (%) for each measurement position between 12:00 and 12:00. Statistical evaluation was performed in R [37]: VPD calculation, the spline interpolation of spatial plots (akima package) [38], principal coordinate analysis (PCoA; vegan package) [39], boxplots (calculation of average, median, min, and max), and data visualization (ggplot2 package) [40].

The principal coordinate analysis (PCoA) ordinations were made based on microcoenological data and the DCs of VPD values. In the latter case, the ordination was practically performed on the quantiles of the VPD (1% resolution). Spatial maps were generated by spline interpolation to plot the VPD data for the whole study site. The interpolation error can be smaller in this case than that when using other interpolations [41].
