The ecotope as a pure land unit of the lowest rank was presented as a central concept in landscape ecology [1
]. The ecotope is recognized as an ecologically relatively homogeneous, spatially explicit, functional landscape unit with at least one homogeneous land attribute, and only small variations in other attributes such that gradients within an ecotope cannot be distinguished. In forestry, the term ‘site’ is similarly defined as a geographical location that is homogeneous in its physical environment with respect to climate, topography, soil and vegetation [5
]. Both concepts describe correlative complexes defined by landform, the structure of vegetation forming various cover types, the characteristics of vegetation in the forest stands and the relationship between soil and physiography [2
]. Naveh [8
] proposed that the boundaries of ecotopes could be established pragmatically, based on the subject and the objectives of a particular study, while remaining concretely placed within a certain space and time as actual ecosystems [1
Similar concepts have been adopted for the purpose of analyzing and mapping ecological units at landscape and regional levels. Species assemblage has been used as a categorical variable at the regional scale and in landscape research, based on the assumption that the distribution or pattern of vegetation is influenced by environmental factors [9
]. Over the past several decades, the development of land classifications has been supported by satellite multispectral (MS) imagery and terrain data that can be used to map land classes over large areas. Rasti et al. [12
] provided a technical overview of the state-of-the-art techniques of feature extraction approaches for hyperspectral images. Hyperspectral imaging (HSI) can acquire richer spectral information by sampling the reflective portion of an electromagnetic spectrum, covering a range from the visible to the short-wave infrared region and also the emissive properties of objects in the range of the mid-wave and long-wave infrared regions. HSI technology enables the detection and a high discrimination ability between spectrally similar classes, but its coverage from space is much narrower than the MS ones. In order to obtain a more accurate abundance estimation, Hong et al. [13
] proposed a new method for HS image analysis that models the main spectral variability and variability caused by environmental conditions, the instrumental configurations and the material nonlinear mixing effects separately. To exploit the rich spectral information contained in HS images, they are developing new methods and fusion strategies that will reduce the limitations due to high storage and computational cost [14
Topographic variables are often used as surrogates or predictors for the environmental gradients that affect species, vegetation or ecosystem class distribution [9
]. Pfeffer et al. [18
] showed that, despite the impact of natural and artificial disturbances on the spatial pattern of alpine vegetation, topography was still an important environmental factor. Dirnboeck et al. [19
] stressed the importance of matching topographic data to the scale at which the spatial patterns displayed by mapping units are influenced by biophysical constraints. In calcareous alpine environments, the coarse resolution of the digital elevation model (DEM 50 m) misses most of the microscale variations in site conditions.
Physically heterogeneous landscapes, such as karst landforms developed in limestone and dolomite bedrock, e.g., dolines or sinkholes, where steep slopes alternate with depressions, display special topographical, hydrological and subterranean properties [20
]. Despite a relatively homogeneous parent material, karst relief is characterized by furrows, grykes and different types and combinations of sinkholes. The microforms of a terrain are assumed to be more important factors in soil formation and vegetation patterns than the macro-scale landforms of the surface. In high karst areas, luvisols are typically present at the bottom of sinkholes as a result of the flow and rinsing of water and sediment [21
] from the edges of sinkholes and ridges, where more shallow soil types are prevalent, especially leptosols and shallow cambisols. Significant differences in vegetation diversity and composition occur inside and outside the sinkholes, which indicates their presence has important ecological impacts [22
]. Within forest ecotopes, the soils of steep and rocky surfaces are often shallow and less developed and contain a higher amount of organic matter [23
]. Consequently, growth conditions and forest vegetation can change rapidly, even within very short distances [25
]. Based on vegetation surveys and soil analyses of diverse microtopographic site conditions, it is difficult to determine the boundaries between the smallest homogeneous units and adapt them to research or forest management problems [24
As with ecotopes and their gradient assessments, forest sites can be used to represent the environmental factors that impact the growth and yield of the tree species in a given area. Yield tables and site index concepts were used to estimate future forest stand development and to compare various site conditions within forests [5
]. In forestry, parsimonious models have been established based on assumptions of tree height at certain ages. However, the assumptions for estimating site index were restrictive, as forest stands were assumed to be even-aged, monospecific, fully stocked and free of disturbance [28
]. Curt et al. [29
] have shown that correlations among physiography, climatic indices, soil properties and plant associations are useful site index surrogates. In evaluating techniques for modelling forest site productivity in different ecoregions, site quality has been explained as resulting from different environmental variables [30
], with the validity of the empirical models being restricted to the spatial scale for which they were developed. As shown by [31
], site index models for small areas require a higher resolution to accurately represent the short-distance variability as well as the relevant environmental patterns of subregions.
The main obstacle to such an indirect concept of site index estimation in karst terrain is the high diversity of landforms, which can affect site conditions and produce variable soil conditions within forest ecotopes. In high karst areas, diverse growth and stand conditions are present on small scales, raising the key challenge of determining how to separate environmental gradients to define the differences between sites. However, spatial models can be adopted to determine the differences in gradients within tangible forest ecotopes. Many studies have demonstrated the ability of an airborne LiDAR system to penetrate dense forest canopies and reveal the underlying ground topography in karst landscapes [22
]. The LiDAR system was able to detect a variety of differences in ground elevation and karst topography based on a generated high-resolution DEM with a cell size of 1 m.
The objectives of this work were to assess the influence of topography on the tree growth in karst terrain. The following research questions were addressed: (i) Are the growing conditions of individual trees influenced by landforms and topographic factors? (ii) Is a LiDAR-derived DEM with a 1-m resolution appropriate to detect terrain-related gradients and microsite conditions? (iii) Is the difference in tree growth on different landforms large enough to be indicative for the forest management and silvicultural operations?
To analyze the effect of topography on tree growth, we selected uneven-aged forest stands with diverse microtopographic site conditions in the high karst area of the Leskova dolina Forest Management Unit (FMU). We examined the applicability of a DEM-based spatial model of landforms and topographic factors for forest management and silvicultural treatment. The present study was motivated by the need for a spatial model that can be used to compare site or ecotope gradients under field conditions with a special focus on karst terrain.
The results suggest that the topographic factors influencing the height growth of silver fir trees can be detected in mature stands where older firs are responsive to landform and site differences. Using topography modelling, we classified silver fir trees into groups (Figure 5
) with statistically significant differences in height growth (Figure 8
and Figure 9
The spatial model based on topographic factors and differences in the height growth of dominant silver fir trees (Figure 7
) is not meant to determine growth indices but, rather, illustrates the significance of spatial variability within forest ecotopes. The differences among the heights of dominant silver fir trees were greater than 2 m, differences used in Middle Europe for the purpose of developing yield tables and escalating the differences between estimated growth indices [54
Understanding this variation in the space and growth of trees (Figure 7
) is important to forest management, as differences in height growth might serve as the basis for comparing forest site classes and site conditions within forest communities. Recent studies in forest communities (representing the basic units for the evaluation of site indices) have revealed a great variability in estimations [41
], highlighting the need for the refinement of forest site classifications in selected forest communities and the fact that these communities are defined by the similarities among them, notwithstanding the gradual variations and differences in site conditions. In addition, the problem of spatial variability was highlighted by Skovsgaard and Vanclay [56
] when describing the variability of tree heights and the effects of microsite conditions on the growth of individual trees. The gradient and variation in site conditions, influenced by topography and soil, may dampen or reinforce the natural variability in the size of trees due to genetic variation and silvicultural treatment.
In many forest site classifications, the understory vegetation has been used as an indicator of site productivity. In Central Europe, methods for evaluating the ecological characteristics of phytocenoses or plant communities have been developed to assess environmental factors and their gradients [57
]. In forestry, growth types are determined based on vegetation and environmental factors are estimated based on the ecological characteristics of plant species. Consequently, plants are used as indicators of the environmental conditions at individual sites. Bergès et al. [61
] have shown that different understory vegetation and floristic indices may be relevant to site index and productivity assessment. In the sessile oak (Quercus petraea
) stands of northern France, floristic indices and predictions based on climate, topography and physical and chemical soil characteristics were able to explain the same variations in the site index. In contrast, assessments carried out in the high karst research area have shown that the use of understory vegetation in evaluating the differences in site conditions is ineffective [62
]. Within the prevailing forest communities, it was not possible to detect differences in microsites or understory vegetation patterns using phytocenological classifications and the Ellenberg indicator values for vegetation. Thus, the understory vegetation could not be used as a valid indicator of tree growth evaluation on such a small scale. The gradient of the ecological factors of understory vegetation was too small to have an impact on the heterogeneous conditions and differences in the described vegetation units. Similarly, as in other research on the vegetation of karst terrain [63
], some of the understory vegetation species of high karst areas [62
] are only registered if they are growing either in or on the edge of a sinkhole and were not included in the vegetation inventory elsewhere in the research area. An analysis of ecological factors based on phytoindicator values showed very small differences among the site units in beech forests [64
]. The floristic similarity among sample units within sites correlated very well with the differences in the site productivity, whereas less than 10% of the variance in floristic composition was explained by site productivity.
Limited options for plant-based indicators are presented in Figure 8
. Steam analyses (Figure 2
) indicated that the differences in the height growth of silver fir appeared after the transition of the dominant trees during their later development. According to our assumptions, site differences have a more visible influence on the growth of older trees situated within the three different landform categories (Table 1
). The hypothesis that growing conditions for individual trees are influenced by topographic factors was confirmed not only by the trees’ significant height differences but also by the height growth increment of the dominant trees. Equally important is the finding that silver firs within the three groups of topographic factors did not differ in their average age. In addition, the differences in these trees’ growth can be attributed to site differences modified by topographic factors, even though the forest stands are classified as uneven-aged. Steam analyses confirmed the findings of previous studies [37
], which presupposed at least a 40-year rejuvenation phase for today’s dominant fir trees. In the Leskova dolina FMU, 1500 stems of fir larger than 25 cm dbh were analyzed in the year 1962 to identify the age and periods of suppressed growth. It was estimated that two thirds of all trees regenerated in the period 1814−1850 [37
]. In such stand conditions, it is impossible to assess the growth indices using the principles developed for even-aged stands. Based on the stem and dendrochronological analysis, we can estimate, with 95% confidence limits, that the dominant silver trees in our sample regenerated in the period 1824−1833. After a period of suppressed growth, the height differences between the three groups of dominant trees increased (Figure 8
) and at the end of the analysis reached height differences of 3.9 m (Table 1
, Figure 6
and Figure 9
). These differences are comparable to the results in classical surveys of site indices on karts sites. On five plots of 30 × 30 m, placed on the forest site Omphalodo-Fagetum typicum
, the site index values (top height at 100 years, SI100
) ranged between 22 and 32, and on the site Omphalodo-Fagetum mercurialetosum
was estimated to be between 24 and 28 [47
]. In research of growth and yield characteristics of silver fir in Slovenia, the analysis of site productivity showed that site productivity decreases with altitude and surface stoniness, while it is higher on the concave sites [47
]. In our research area, the differences in altitude were within 50 m, but the influence of different landforms on height growth were clearly shown (Figure 8
and Figure 9
The approach adopted here was similar to that used in the development of individual tree growth models in which the data were gathered from the sampling plots of forest inventories [65
] or from experimental observations [40
]. In individual tree growth models, forest stands are divided into a mosaic of individual trees, thus achieving a much higher resolution when modelling their interactions in the system [40
]. Without historical data on stand dynamics, we were not able to analyze how competition for resources varied with tree development or the differences in the neighboring trees of the selected dominant firs. As shown by Coomes and Allen [66
], competition for light has a strong influence on the growth of small trees, whereas competition for nutrients affects trees of all sizes. D’Amato and Puettmann [67
] noted that the relationship between relative dominance and the growth of dominant trees likely arises from these trees’ competitive advantage in exploiting available resources.
Modelling topographic factors plays an important role in forest ecotopes, where the differences in gradients of ecological factors can be estimated. However, these differences can often not be determined within an area due to the limited numbers of experimental sampling units included in plant indicator or pedological and chemical studies. Topographic characteristics and soil properties may offer insight into the variation in site productivity, but detailed site mapping is often prohibitively expensive for operational use in forest management [56
]. This challenge particularly holds true for ecotopes, in which soil development is likely a consequence of differences in microrelief and the disintegration of limestone parent material [68
Recent studies have shown that the LiDAR system allows for a synoptic view of landforms and topography, which is not possible with field plots alone [69
]. Trevisani [70
] presented the potential for exploiting the geomorphological information from high-resolution digital terrain models derived from airborne LiDAR surveys in complex morphology, which raises new prospects for regional analyses in alpine environments. Evaluations of the gradients of ecological factors based on indicator values of understory plants have proven to be insufficiently reliable in high karst areas (Figure 3
), whereas for the purposes of estimating the ecological factor gradients within forest ecotopes, the growth differences indicated by topographic factors are reliable.