You are on page 1of 14

Ecological Indicators 94 (2018) 128–141

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Original Articles

A spatially explicit patch model of habitat quality, integrating spatio- T


structural indicators

Barbara Riedler , Stefan Lang
University of Salzburg, Interfaculty Department of Geoinformatics – Z_GIS, Schillerstrasse 30, 5020 Salzburg, Austria

A R T I C L E I N F O A B S T R A C T

Keywords: Habitat quality – referring to the structure of a habitat – is crucial for the habitat functions and biodiversity-
Habitat quality assessment related ecosystem services. Its effective assessment and monitoring is important to meet not only scientific
Riparian forest monitoring interest, but also practical conservation regulations. We herein present the approach of using a composite
Geon indicator of riparian forest quality, both quantitatively (habitat quality index HQI [0|1]) and categorically
Cluster analysis
(quality types QT1…4), integrating a set of spatio-structural indicators derived from earth observation data. We
Habitats directive
apply the geon concept to aggregate multiple sets of statistically profound and conceptually meaningful
EO-based indicators
indicators to the well-established patch concept. The main difference is that instead of using a priori units (e.g.
existing patch delineations), new patches (geons) are derived that directly represent the phenomenon of habitat
quality. Patch boundaries correspond to the gradients imposed by the multivariate behavior of the underlying
indicators. The approach provides a truly spatially explicit assessment, while at the same time retaining the
potential for decomposability on the level of individual indicators. The example of a riparian forest is chosen, as
riparian zones are complex ecosystems with a high biodiversity that are highly threatened. In our case, riparian
forest habitat quality was assessed using four indicators: (I) tree species composition, (II) horizontal forest
structure, (III) vertical forest structure, and (IV) water regime. The distribution of habitat quality (HQI)
highlights hot- and cold-spots, where conservation measures may be needed. Cluster analysis reveals four types
of patches that are characterized by a specific behavior of the aggregated indicators (QT1 = fair composition,
QT2 = (old growth) forest plantations, QT3 = characteristic tree species, QT4 = forest gaps). This categorization
enables the prevailing or lacking aspects of quality to be determined based on the decomposability of the index.
In addition, the resulting patches were evaluated using landscape metrics. The findings achieved on a
statistically significant level show that patches with high HQI scores are better connected and form large
patches with a characteristic tree species composition. In contrast, areas with low HQI values are characterized
by a non-favorable tree species composition and the existence of clear-cut areas or access roads. A comparison
with an assessment using a traditional composite indicator approach reveals the sensitivity of the different sets of
indicators and assessment methods. The presented habitat quality index can be considered as suitable for the
assessment and monitoring of riparian forest quality, supporting spatially explicit conservation measures and the
evaluation of applied measures.

1. Introduction within the European network of Natura 2000 habitats – an ambitious


initiative prescribing mandatory monitoring and regular reporting of
1.1. Assessing habitat quality in riparian forests habitat quality and of the conservation status (European Commission,
2005).
Habitat quality can be assessed based on the specific structure and Due to their linear structure, representing small shares of the land
functions of habitats necessary for their long-term maintenance and surface only, riparian zones are among the most precious, yet highly
survival of typical species (HabDir Art. 1; Riedler et al., 2015). Effective threatened European ecosystems (Tockner and Stanford, 2002), with
assessment and monitoring are essential for the general conservation natural riparian habitats being one of the most anthropogenically
purpose, and even legally binding for protected sites within the disturbed habitats in Europe (Dimopoulos and Zogaris, 2008). Hu-
Habitats Directive. In Europe, most precious habitats are protected man-induced pressures include alterations of natural hydrology and


Corresponding author.
E-mail address: barbara.riedler@sbg.ac.at (B. Riedler).

http://dx.doi.org/10.1016/j.ecolind.2017.04.027
Received 8 August 2016; Received in revised form 23 January 2017; Accepted 10 April 2017
Available online 09 May 2017
1470-160X/ © 2017 Published by Elsevier Ltd.
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 1. Characteristics of geons and geon boundaries depending on the variability of the underlying indicators.

river geomorphology through river regulation and isolation from river EO technology in this domain.
channels, land use change in the form of agriculture, forestry and urban Measuring and monitoring such individual indicators – like the
development, the invasion of alien species, and climate change species composition – is essential for assessing habitat quality. At the
(Decamps et al., 2008; Dimopoulos and Zogaris, 2008; Tockner and same time, capturing the complexity of habitat quality and representing
Stanford, 2002). These impacts can lead to fragmentation of riparian cumulative effects in the sense of systemic behavior (Capra and Luisi,
forests, reduced natural dispersal, changes in plant species commu- 2014) remains a challenge. In its complexity, habitat quality can be
nities, decreased species diversity, and structural simplification considered as a latent phenomenon, which cannot be observed and
(Dimopoulos and Zogaris, 2008). At the same time, riparian zones measured directly (Byrne, 1998). Still it may be represented by a set of
provide multiple and diverse important ecosystem services (Clerici integrated indicators (Müller et al., 2000) that are conceptually linked
et al., 2011; Clerici et al., 2013; NRC, 2002), such as food resources for to the phenomena (Lang et al., 2014). Composite indicators are a
animal and plant species, moving corridors, services for the regulatory powerful tool for measuring such multi-dimensional concepts (OECD,
function of water quality like water storage, nutrient and organic 2004). They are highly suitable for translating complexity in policy-
matter cycling, the maintenance of water quality through water relevant information and are applied in the environmental domain
filtering and cooling of streams by shading, flood protection, the (OECD, 2003). We consider the fact that they are calculated on pre-
regulation of local climatic conditions, and finally leisure and recrea- defined (e.g. administrative) boundaries to be one of the drawbacks of
tional potential (Azim, 2006; Döringer and Tockner, 2008; Dwire and composite indicators. This research therefore presents a new approach
Lowrance, 2006; Hughes, 1997; Naiman et al., 2005; NRC, 2002; for assessing habitat quality, through the integration of various remote
Wantzen and Junk, 2008). Therefore, riparian zones are considered sensing based indicators in a spatially explicit manner.
among the most complex ecological systems that represent biodiversity
hotspots worldwide (Naiman et al., 2005). The flooding regime creates
1.2. Approaching complex spatial phenomena using geons
dynamic geomorphic landforms and a vegetation distribution with a
high structural and compositional diversity (Döringer and Tockner,
Geons can be defined as “homogenous spatial units in the sense of a
2008; Ward et al., 2002). Riparian forests are characterized by diverse
region, yet semi-automatically delineated with expert knowledge
and rich but distinct vegetation patterns, dominated by hygrophilous
incorporated, scaled and of uniform response to a phenomenon under
species, adaptive to mechanical stress from flooding and to fluctuations
space related policy concern” (Lang et al., 2014). They are constructed
in water levels (Dimopoulos and Zogaris, 2008). Thus, due to their high
by regionalizing a multiple indicator set. Geon boundaries correspond
value and critical status, the protection of these habitats is a major
to steps along gradients imposed by the multivariate behavior of the
challenge (Decamps et al., 2008) that strongly depends on effective
underlying indicators (Fig. 1). Thus, these boundaries are delineated
management and monitoring (Gordon et al., 2004).
drawing on the idea of region building (Berry, 1967), which assumes
The use of indicators to support forest monitoring has often been
that changing cumulative behavior causes a boundary. The resulting
proposed both on a national level, for instance the Austrian Forest
homogenous geons have minimized inner variance, but gradients
Biodiversity Index (Geburek et al., 2010) or the Biodiversity Indicators
towards neighboring regions. Geons therefore explicitly represent the
for UK Managed Forests (Humphrey and Watts, 2004), and on an EU
spatial distribution of the examined phenomenon (Kienberger et al.,
level, such as the Forest Biodiversity Indicators of the European
2009; Lang et al., 2014).
Environmental Agency (EEA, 2012; for an overview, see Schuck and
From an ecological point of view, the geon concept resembles the
Rois, 2004).
patch concept, but is considered to be more general regarding the
To facilitate seamless, area-wide coverage and the option of regular,
phenomena under investigation and comprises multi-dimensional data
cost-efficient updates (Duro et al., 2007; Vanden Borre et al., 2011), the
integration. In congruence with the patch concept (Forman and
use of Earth observation (EO) data and derived indicators have been
Godron, 1986), geons aim to delineate homogenous areas. Patches
suggested (Noss, 1999; Bock et al., 2005; Gillespie et al., 2008).
have been defined as “non-linear surface areas differing from its
Previous studies have applied habitat classifications to a variety of
surroundings in appearance (Forman and Godron, 1986) or nature
vegetation types using different EO data, and assessed quality indicators
(Turner et al., 2001)”, usually described by vegetation composition
based on the classification, composition, and age of resident species
(Forman, 1995). Following the origins of landscape ecology coinciding
(e.g. Förster et al., 2008; Haest et al., 2010; Delalieux et al., 2012;
with the availability of aerial photographs (Troll, 1939), the delineation
Neumann et al., 2015). While the operational use of such remote
of patches was traditionally carried out by visually interpreting the
sensing based information is inherently limited by image resolution and
imagery. Recently, the concept has been extended to (semi-)automated
the aerial view (Nagendra et al., 2013), mapping activities such as those
delineation of habitat patches, based on the classification and composi-
initiated by the European Copernicus program have boosted the use of
tion of species, using high resolution remote sensing imagery (see e.g.

129
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Förster et al., 2008; Haest et al., 2010). The approach presented herein 13°00′22”; Fig. 3). It is located in a densely populated area of the alpine
provides a solution to integrate further quality indicators for the foreland at the Austrian‐German border along the river Salzach (Riedler
delineation of patches, in addition to visual (or spectral) properties et al., 2015; Strasser et al., 2014), between the city of Salzburg in the
and species distribution. south and the urban centres Oberndorf/Laufen in the north. The
In parallel, we provide a spatially explicit assessment of habitat vegetation forms the dominant Natura 2000 habitat types (Annex 1 of
quality through avoiding the use of a priori defined patches. As the Habitats Directive: Council Directive 92/43/EEC, 1992) of alluvial
calculating indices based on pre-defined boundaries often neglects the forest (91E0*) and riparian mixed forest (91F0) (Strasser et al., 2014),
true spatial distribution of indicator values (Lang et al., 2014), we try to respectively the EUNIS habitat types (Davies et al., 2004) of riparian
circumvent such aggregation problems through the use of geons. The and gallery woodland (G1.1) and mixed riparian floodplain and gallery
modifiable area unit problem (MAUP) describes aggregation effects of woodland (G1.2) (Riedler et al., 2015). The river Salzach was regulated
both scale and zoning (Openshaw, 1984). Here we mainly try to avoid in the early 19th century in the course of demarcating the borderline
the effect of zoning, where different boundaries and compositions of between Austria and Bavaria. Thus, its riparian zone nowadays can be
areal units can lead to variations in results, depending on the grouping considered semi-natural at best. Today, remainders of the then widely
of attributes that are associated with these units. This may evoke the spread riparian forest still face logging and hunting, despite its
effect of ecological fallacy (Gotway and Young, 2002), where a bias is designation as a Natura 2000 site of community interest. For flood
introduced due to the aggregation of individuals and the varying mitigation purposes, parts of the riverbed were recently widened up,
distribution of individuals within zones (Gelfand, 2010). and steep banks were regenerated, which caused important secondary
Composite indicators are formed when “individual indicators are effects for the water regime and the local ecological conditions in
compiled into a single index, on the basis of an underlying model of the general. As a strategic roadmap to the future, a political decision is
multi-dimensional concept that is being measured” (OECD, 2004). This being implemented to re-naturalise the area and to convert it to a
practice reduces the number of separate indicators, while maintaining nature park.
the underlying information (Saisana and Tarantola, 2002). The multiple
sets of indicators relevant for aggregation should thereby not only be 2.2. Indicator selection
conceptually meaningful, but also statistically profound and robust
during all stages of the construction of a composite indicator – from Based on criteria for expert-based field assessments (Ellmauer,
data selection to weighting and aggregation (Nardo et al., 2008). In line 2005; BfN, 2010) and the possibility to derive indicators from EO data
with this, the geon approach uses this framework to describe habitat (Riedler et al., 2015), the following four indicators were used in this
quality, but additionally considers the issue of pre-defined boundaries. study (Table 1):
Another advantage of the geon approach is the flexible treatment of
the generated units with respect to (1) labelling, both quantitative – 2.2.1. Tree species composition
using quantitative index values − and qualitative – using nominal The degree of naturalness of forest systems, which plays a central
categories; (2) their decomposability in terms of the influence of the role in the assessment of forest habitats, can be described through its
contributing indicators; (3) assessing their spatio-structural character- tree species composition (Ellmauer, 2005; Geburek et al., 2010).
istics, for example, using landscape metrics. Thereby, characteristic tree species as potential natural vegetation
(Ellenberg, 2009) facilitate the occurrence of associated plant and
1.3. Aims and ambition animal species and are considered to have a positive impact on the
habitat quality. In contrast, allochthonous tree species can have a
Varieties of indicators, usually used in expert-based field assess- negative impact on the potential natural vegetation by competing for
ments, exist for the assessment of riparian forests and their integrity light, nutrients, water or space (Richardson, 1998).
(Ellmauer, 2005; BfN, 2010). In the context of this paper, four
indicators that are relevant for the study site of the Salzachauen, and 2.2.2. Horizontal forest structures
were previously discussed by Riedler et al. (2015), were selected This comprises small streams and clearings, which are typical
(Table 1). We herein use them to construct integrated geons as a features in alluvial forests and essential for characteristic herbaceous
spatially explicit composite indicator representing habitat quality of layers (Ellmauer, 2005), and other elements of spatial heterogeneity
riparian forests (Fig. 2). often caused by natural disturbances (Ward et al., 2002), which in turn
Thereby, the geon approach incorporates (1) multiple indicators can trigger species richness (Schnitzler, 1994). Spectral heterogeneity
solely derived from EO data to describe riparian forest quality; (2) the of a satellite image is assumed to reflect the occurrence of local
assessment of habitat quality with a composite indicator that integrates structural richness and habitat heterogeneity and has been linked to
these statistically profound indicators into a single index; (3) the species richness (Palmer et al., 2002; Rocchini et al., 2004).
simultaneous delineation of spatially explicit areas of homogenous
habitat quality (geons) with respective categorization. We thereby 2.2.3. Vertical forest structure
facilitate decision-support by providing an easy to understand overview A mix of tree age classes and species is important for forest ecology
map of habitat quality, which enables the detection of spatially explicit and biodiversity (Lowman and Rinker, 2004; Noss, 1999). It has been
hot- and cold-spots where conservation measures may be directed. associated with species diversity and richness (e.g. Goetz et al., 2007;
Through the categorization of qualitative categories that are differen- Bergen et al., 2007) and was proposed for the assessment of the
tiated by a characteristic behavior of the underlying indicators, the ecological integrity of forests (Geburek et al., 2010; Noss, 1999).
evaluation of patches of these clusters using landscape metrics and the Canopy roughness primarily approximates forest structure, whereby a
combination with the quantitative habitat quality assessment provide complex mixed-aged structure has a positive impact on biodiversity
critical information for targeted conservation measures. (Noss, 1999), and to some degree the diversity of tree species
composition. It additionally accounts for the openness of forest patches,
2. Methods which increases light levels below the canopy, in turn favoring more
below-canopy species and general structural diversity (Eycott et al.,
2.1. Study area 2006). Large old-living trees provide essential habitat for both plant
and animal species, but are especially important for different bird
This study was conducted in the Natura 2000 site Salzachauen species (Coote et al., 2013; Berg et al., 1994). Furthermore, it is likely
(Salzburg, Austria; UL: N 47°56′12”/E 12°56′24”; LR: N 47°52′17”/E that old trees will become dead wood in near future, which again

130
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Table 1
Individual indicator processing.

1
Acquired in June 2012.
2
Acquired in April 2006.

increases biodiversity, as dead wood provides niches for certain species two different EO data sources – a very high resolution WorldView-2
(Anderson, 1991). satellite image and LiDAR data (Table 1). The data was processed in a
way that a final set of four statistically profound and robust indicators
could be used for regionalization (Table 1; Fig. 5).
2.2.4. Water regime
For the indicator Tree species, an initial tree species classification
Hydrological conditions and flooding are a prime indicator for
with an overall accuracy of 86.93% and a Kappa value of 0.84 was used
riparian forest habitat assessment (Döring and Tockner, 2008;
(Strasser et al., 2014). In accordance with the definitions of two
Ellmauer, 2005 Ellmauer, 2005). The water regime influences geo-
different classification Schemes – Natura 2000 (Annex 1 of the Habitats
morphic and soil patterns, as well as vegetation and general species
Directive: Council Directive 92/43/EEC, 1992) and EUNIS – level 3
richness (Ellmauer, 2005; Hughes, 1997; Schnitzler, 1994). Terrain
(Davies et al., 2004) – and site-specific management strategies, expert-
roughness thereby describes the micro-topographic relief of the forest
based values emphasizing the importance of specific tree species for the
ground, which designates areas where water is likely to accumulate
overall habitat quality were assigned to each tree species (Table 1). A
resulting in specific soil conditions and plant species distribution
similar expert-based procedure was applied for the sub-indicator Old
(Unger and Muzika, 2008). It also accounts for existing perennial
trees, where it was assumed that higher trees have an over-proportional
streams and riverbeds, which are only activated during floods and
value for habitat quality, not only providing essential habitats, but also
heavy rains that can provide valuable (temporary) habitat for specific
having a higher change of becoming dead wood in near future (see 2.1).
species (Tockner et al., 1999).
To render all indicators comparable, the continuous data sets were
resampled to the same cell size (2 m × 2 m) and were normalized using
2.3. Indicator processing linear min-max normalization (OECD, 2004). The value range was set
to [0 | 255], as further processing in the form of segmentation using
Based on the general workflow of constructing a meaningful and Trimble Geospatial’s eCognition requires 8-bit resolution raster data.
statistically sound composite index proposed by Nardo et al. (2008) and Descriptive statistics were used for detecting missing data and describ-
complemented with steps necessary for the regionalization process ing data distributions. Indicators with a skewness > 1 and kurtosis >
using the geon concept (Kienberger and Hagenlocher, 2014; Lang et al., 3.5 that should be handled before constructing the final index
2014), the following workflow was applied (Fig. 4): (Groeneveld and Meeden, 1984) were log transformed (ln10).
Four indicators describing riparian forest quality were derived from

131
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 2. Assessing habitat quality based on four indicators (I–IV) of riparian forest structure. For HQI (habitat quality index) and QT1…4 (habitat quality type 1–4) see 2.2 and 3.2 for
further explanation.

For Vertical forest structure, the two sub-indicators Canopy roughness process using a similar way of representing the distance and position of
and Old trees were aggregated before constructing the final index to each vector in the feature space (Kienberger et al., 2009). Again, equal
avoid over-representation during regionalization. In accordance with weights were assigned to the four input indicators. For better inter-
the procedure of creating the final index, an equal weighted sum pretability of results, the final habitat quality index was normalized to a
approach was applied. As a prerequisite for this aggregation step, the value range of [0|1], where 0 reflects poor habitat quality and 1 the
sub-indicators were statistically tested to detect possible multi-colli- best possible quality.
nearity using the Spearman’s rank correlation coefficient (ρ). Based on
HQIp = w1 v1 + w2 v2+…+wn vn (1)
thresholds proposed by Nardo et al. (2008), indicators with ρ < I 0.9 I
were considered unproblematic for further aggregation. The same p = patch
procedure was applied to the final set of four indicators before v1-n = observed indicator values
aggregation to the final index. w1-n = weigthing of the correponding indicator.
To identify homogenous categories, a hierarchical agglomerative
cluster analysis based on factor tables was conducted using Philcarto
2.4. Regionalization, habitat quality and categorization
5.0. To describe the characteristics of the resulting clusters based on the
underlying indicators, descriptive statistics and quantitative measures
For regionalization, a final set of four harmonized indicators was
used (Fig. 5). For the delineation of geons, we regionalized the of landscape metrics were calculated. Differences between the different
clusters were statistically tested using a Chi2-test for several indepen-
indicators in a multi-dimensional indicator space (Lang et al., 2014)
using image segmentation routines. We worked in an object-based dent samples, and a Mann–Whitney U test (MWU) for two independent
samples.
image analysis (OBIA) software environment, Trimble Geospatial’s
eCognition 9. The implemented multi-resolution segmentation algo-
rithm employs a growing seed function, merging pixels to larger objects 2.5. Validation
by adhering to pre-determined homogenous criteria (Baatz and Schäpe,
2000), which depend on color (pixel values) and form (compactness The habitat quality index values (HQI) were compared with existing
and smoothness) (Trimble, 2014). Here we used the multidimensional assessments of habitat quality in the study site, namely the RFI_S
indicator set instead of multispectral imagery, where all indicators were (Riparian Forest composite Indicator: focus on structure) computed by
equally weighted. Shape and compactness of the objects were para- Riedler et al. (2015). A similar set of EO-based indicators was used for
metrized with 0.5 to achieve rather compact objects. The segmentation the calculation of the RFI_S, but they were aggregated on pre-existing
scale (scale parameter) was set to 270, as determined by the estimate habitat patches using different weighting and aggregation methods.
scale parameter (ESP)-2 tool by testing the statistical behavior of This reference map only provides forest quality information for regions
variance and the rate-of-change curve (Drăguţ et al., 2014). that were classified as EUNIS-3 habitat types [riparian and gallery
The computation of HQI follows a composite indicator approach woodland (G1.1) and mixed riparian floodplain and gallery woodland
(Nardo et al., 2008). A weighted vector magnitude (Eq. (1)) was used (G1.2)]. To account for this diverging patch delineation, a point-based
for the aggregation of the HQI, as it closely reflects the segmentation comparison was applied, where 500 points were randomly distributed

132
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 3. Study area. The designated Natura 2000 site Salzachauen represents the Austrian part of the riparian forest along the river Salzach north of the city of Salzburg.

over the study site. The correlation between the quality indices was The spatial distribution of the habitat quality index (HQI) in the
computed using Spearman’s rank correlation coefficient (ρ); the differ- Salzachauen is shown in Fig. 6a. It provides an easy to grasp overview
ence between sample points lying inside and outside riparian forest of forest quality, which can be especially helpful when trying to localize
habitats was tested using Mann-Whitney-U tests (MWU). areas of low quality, where conservation measures to improve forest
The delineation of geons was visually compared against (a) EUNIS-3 quality may be needed. Besides the true representation of the spatially
modeling based on tree species composition (Strasser et al., 2014; varying phenomenon of habitat quality, an additional advantage of the
Paccagnel, 2013), and (b) a manual delineation of habitat types based use of geons is the clear discrimination of units. Gradients occurring in
on the same satellite image with a manual assignment of EUNIS-3 the spatial variability of the underlying indicators are emphasized and
classes (Riedler et al., 2013a,b). represented by clear boundaries. This leads to a smoothing of the
internal variance and emphasizes the heterogeneity between them in
the sense of generalization. The polygon delineation has an aggregating
3. Results and discussion
(and thereby generalizing) effect that overcomes the limited interpret-
ability of highly varying spatial indicators. Moreover, polygons can be
3.1. Habitat quality index
adapted to the (mapping) scale of interest and better integrated or
compared with existing delineations. This is not a contradiction to the
Descriptive statistics reveal that two indicators, namely Old trees
current trend in spatially explicit continuous habitat mapping
and Terrain roughness, have skewness and kurtosis values that require
(Neumann et al., 2015), but rather to be seen as complementary. While
handling before constructing an index (). Even after applying a log
this trend builds on traditional habitat quality classifications (Delalieux
transformation, skewness values are still above the proposed threshold
et al., 2012; Förster et al., 2008; Haest et al., 2010), it follows the
(Old trees: skewness 1.523, kurtosis 1.884; Terrain roughness: skewness
technical progression of moving towards finer resolution in both spatial
1.536, kurtosis 2.387), which can be explained by the nature of the
and spectral terms. Still, the focal (O’Neill et al., 1986) or reporting
indicators, as most areas neither have old trees nor streams or a strong
level (Hagenlocher et al., 2014) for such aggregated phenomena of
micro-topographic relief. Being aware that this distribution is not ideal,
concern is not always the theoretically or technically best possible,
but given the high conceptual importance of the indicators (e.g.
which itself is ideally the explanatory one.
Ellmauer, 2005; BfN, 2010), they were not rejected for constructing
the final habitat quality index. Since a multi-collinearity analysis of the
indicators reveals that no correlation values reach a critical level, all
indicators were considered for further aggregation steps.

133
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 4. Workflow for assessing habitat quality with a spatially explicit composite indicator.

134
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 5. Final set of four indicators used for regionalization.

3.2. Habitat quality types clusters (mean proximity QT1 and QT2: MWU, p < 0.001, QT4: MWU,
p < 0.004; Table 2), potentially enhancing the survival of meta-
To identify suitable and spatially targeted measures, knowledge of populations. In line with these factors, these patches exhibit a sig-
the underlying indicators that are responsible for the habitat quality is nificantly higher habitat quality than patches from other clusters
required. A cluster analysis was conducted to localize areas within the (MWU, p < 0.001, Fig. 8).
landscape that are similar regarding the behavior of the underlying QT4 Forest gaps – characterized by a low vertical forest structure and
indicators. Based on the tree diagram, four clusters (habitat quality a non-favorable tree species composition, revealed by a significantly
types QT1…4) jointly explaining 87% of the variance were identified lower mean of the corresponding indicators than all other clusters (Tree
(Fig. 6b). species composition and Vertical forest structure: MWU, p < 0.001;
These clusters can be characterized by a specific combination of Fig. 7). They are mainly patches without or with only a few trees,
behavior of the underlying indicators (Fig. 7), and can be combined which are often clear cuttings or forest roads. The latter is in line with
with habitat quality and information about area, form and connectivity the finding of a significantly higher mean shape index (QT1 and QT2:
(Table 2). Although the latter information was not used in the cluster MWU, p < 0.032, QT3: MWU, p < 0.009; Table 2), representing the
analysis, a highly significant difference of both habitat quality and all elongated form of forest roads. Some of the patches of this cluster
landscape metrics, except for Mean Patch Edge, is visible between the include rivers and small water bodies, areas with no trees, or young
clusters (Table 2). monoculture forest plantations, which have a similar age structure and
QT1 Fair composition – characterized by a fair composition of thus a low vertical forest structure. Together with the lowest mean
indicator values with a slightly positive variance of the indicators patch size (QT1 and QT2: MWU, p < 0.056, QT3: MWU, p < 0.001;
Horizontal forest structure and Water regime to the mean indicator values Table 2), QT4 patches have a significantly lower habitat quality than
(Fig. 6b). They Patches of this cluster are often located at the edge of other clusters (QT1 and QT2: MWU, p < 0.001, QT3: MWU,
water bodies, indicated by high terrain roughness, but also patches with p < 0.001; Table 2; Fig. 8).
small clearings resulting in higher horizontal forest structure. These For forest management, looking at specific patches with different
patches are relatively big, but due to their spatial distribution in the habitat quality in combination with their indicator characteristics is
landscape have the lowest connectivity (mean proximity: QT2 and QT3: essential to identify measures that help conserve or improve forest
MWU, p < 0.001, QT4: MWU, p < 0.018; Table 2). In line with the quality. Here, areas with the highest habitat quality can primarily be
balance of indicator values, they reveal a medium habitat quality described by a high characteristic tree species composition, represented
QT2 (Old growth) forest plantations – characterized by a diverse in QT3 Characteristic tree species (Fig. 8). This finding emphasizes the
vertical structure, but at the same time a non-favorable tree species importance of tree species composition for the habitat quality, which is
composition, indicated by a strong positive, respectively negative in line with literature, where it has been suggested as one of the key
variance of the according indicators to the mean indicator values aspects to assess riparian forest quality (Ellmauer, 2005; BfN, 2010).
(Fig. 6b). They are patches with (old) forest plantations, where the In contrast, patches with the lowest habitat quality all belong either
regularity of planting is clearly visible, but also patches with old trees to QT4 Forest gaps, which contains no riparian forest, but rather clear
and small clearings, resulting in a high variance of the digital height cuts, forest roads or monoculture plantations, or to QT2 (Old growth)
model. In line with the visual exploration, they are relatively low- forest plantations, often characterized by allochthonous tree species
quality, medium-sized patches with rather low connectivity (Table 2; (Fig. 8). Targeted conservation measures in these areas can thus be
Fig. 8). easily identified and would primarily target reforestation with char-
QT3 Characteristic tree species – characterized by a favorable tree acteristic tree species, potentially after the logging of allochthonous
species composition with significantly higher Tree species composition trees.
values than all other clusters (MWU, p < 0.001; Fig. 7). Patches of this
cluster thus resemble mostly natural riparian forest, but also some areas 3.3. Validation
of reforestation with Salix alba and Alnus sp. They have a significantly
higher mean patch size (QT1 and QT3: MWU, p < 0.001, QT2: MWU, For habitat quality as a latent phenomenon, evaluation of the results
p < 0.009; Table 2) and are better connected than patches of all other is not trivial, as such phenomena correspond to hidden complex

135
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 6. (a) Habitat quality index (HQI) with normalized values ranging from 0 (poor) to 1 (good) and (b) Quality types derived from cluster analysis. The cluster profile shows the distinct
behavior of the underlying indicators.

136
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 7. Cluster characteristics. Mean indicator values per cluster are shown.

variables (Lang et al., 2014) that cannot be measured directly in the When considering all sample points, a comparison with the con-
field. One approach is to use an indirect assessment with evidence data tinuous quality index for riparian forests (RFI_S) (Riedler et al., 2015)
related to that phenomenon. Hence, we chose a strategy to system- reveals a significant, medium to strong positive correlation (Table 3).
atically compare the habitat quality index with an already existing When taking into account the boundaries of the reference classification,
assessment. sample points located within riparian forest based on EUNIS-3 defini-

Table 2
HQI and landscape metrics per habitat quality type. Mean values and significance of statistical difference per cluster (Chi2) are shown.

Number of Mean habitat Mean Patch Size Mean Patch Edge Mean Shape Index – Mean Proximity (within
patches quality (ha) (m) MSI 100 m)

QT1 18 0.47 6.55 1956 2.25 3.83


Fair composition
QT2 22 0.22 4.00 1662 2.38 49.15
(Old growth) forest plantations
QT3 39 0.74 6.87 1955 2.26 684.07
Characteristic tree patches
QT4 61 0.13 2.95 1734 2.94 218.85
Forest gaps
2
Chi p < 0.001 p < 0.001 p < 0.236 p < 0.001 p < 0.001

137
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 8. Relation between habitat quality type (QT1...QT4) and habitat quality index (HQI). Examples for each existing combination including the HQI-values are shown.

Table 3 Even though both assessments use a similar set of indicators


Comparison of HQI with an existing habitat quality assessment (RFI_S; Riedler et al., aggregated to a composite indicator for describing habitat quality, the
2015). Mean values and correlations (Spearmans correlation coefficient) are shown.
weak correlation of the results reveals the sensitivity to the differences
Sample points Sample points Sample points
in the used methods. We assume that the aggregation to different
all inside EUNIS 3 outside EUNIS boundaries – a priori defined boundaries based on a semi-automated
habitats 3 habitats habitat delineation (Riedler et al., 2015; Strasser et al., 2014) versus
patches (geons) that directly represent habitat quality – is mainly
Number 500 226 274
Mean HQI 0.522 0.418 0.389
responsible for the differences in results. Additionally, the used sets of
Habitat quality index indicators differ: For the computation of the RFI_S, tree species were
Mean RFI_S 0.216 0.478 x only differentiated in characteristic and allochthonous tree species and
Riparian Forest the Shape Index used in Riedler et al. (2015) as a quality indicator can
composite Indicator:
only be applied when using a priori defined patches. The other
focus on structure
(Riedler et al., 2015) indicators were used in both assessments, but they were calculated
Spearman correlation ρ = 0.419, ρ = 0.027, x per pre-defined patch as one mean value per patch only. This can
p < 0.001 p < 0.684 potentially obscure the true spatial distribution of indicator values,
leading to the zoning problem, as described in Openshaw (1984), and to
ecological fallacy (Gelfland, 2010; Gotway and Young, 2002 Gotway
tions have significantly higher habitat quality index values than points
and Young, 2002). Different weighting schemes (expert-based weight-
outside these patches (MWU, p < 0.001). This is in line with the
ing computing RFI_S versus equal weighting calculating HQI) are
findings of the present study, that areas with lowest habitat quality are
assumed to have an additional impact.
either clear-cut areas or forest plantations (Table 2; Fig. 8), both of
The visual comparison of homogenous habitat quality units with
which are not included in the definitions of G1.1 or G1.2. In contrast,
modeled and manually delineated forest patches provides examples for
no correlation between the habitat quality indices can be found
both similar (Fig. 9a) and very diverging delineations of boundaries
(Table 3) when only considering points placed within EUNIS-3 riparian
(Fig. 9b). With the underlying aggregated indicators, the generated
forest (G1.1 or G1.2).
geons do not necessarily reflect patches in the sense of areas differing

138
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Fig. 9. Comparison of delineations of (1) geons (2) habitat modeling and (3) manual patch delineation of EUNIS 3 habitats. Examples of (a) patches with similar and (b) patches with
diverging delineations are shown.

from their surroundings in appearance (Forman and Godron, 1986) or blems of a priori defined boundaries are overcome.
biological objects in the sense of phytosociological units. This becomes As proposed by Riedler et al. (2015), the assessment of habitat
visible in Fig. 9b, where the reference delineations either solely depend quality following a defined conceptual framework with a given set of
on tree species composition (Strasser et al., 2014), or on visual cues of quality indicators, but with certain flexibility, allows for transferability
VHR satellite imagery (Riedler et al., 2013a,b). As different underlying to other riparian forest sites. Thereby, site-specific adaptations and the
concepts are used – phytosociological units, difference in appearance, integration of in-situ data can be taken into account. Especially the
or objects of the same habitat quality – the delineations of boundaries latter aspect is considered essential when trying to operationalize such
also diverge in the majority of cases. Areas that appear visually diverse an approach, as EO data offer a wide range of advantages, but also have
(e.g. plantations of different and visually easily distinguishable tree limitations. EO data cannot or can only partly be used to assess all
species) are not differentiated in the geon delineation if they exhibit relevant indicators for riparian forest quality, such as detailed species
homogenous habitat quality (Fig. 9b). In contrast, patches with similar information on understory vegetation, impact of game or distribution of
geon boundaries (Fig. 9a) reflect areas with a habitat quality largely endangered species. The presented geon approach offers the flexibility
dependent on Tree species composition. to integrate such important field data not only in the calculation of the
habitat quality index, but already for the segmentation of habitat
quality units (geons).
4. Conclusion and outlook The presented assessment of habitat quality reveals a status quo,
which can be of great interest for decision makers. The approach could
The presented habitat quality index provides a spatially explicit also be applied for monitoring purposes, when considering that
representation of riparian forest quality on landscape level that especially protected areas require regular monitoring, for instance,
supports the simultaneous mapping of several quality aspects to mandatory six-year reporting in Natura 2000 sites (Habitats Directive,
account for ecosystem services and other integrated assessments. Art 17: Council Directive 92/43/EEC, 1992). It is considered to bear
Qualitative categories, defined by similar patterns of underlying great potential for this task, as indicators are to a high degree derived
indicators, in combination with the continuous habitat quality informa- from EO data that enable regular updates at relatively low costs (Duro
tion and spatial characteristics, provide valuable information for policy- et al., 2007; Vanden Borre et al., 2011) and can complement traditional
related decisions. Thus, results of this study may contribute to main- field-based methods. A multi-temporal geon-based assessment provides
taining or improving these habitats by identifying suitable and spatially comparable updates against a specific policy background (Lang et al.,
targeted conservation measures. 2014), whereby not only the re-assessment of habitat quality, but also
The assessment is based on an integrative, yet decomposable the effectiveness of intervention measures can be evaluated. Taking all
conceptual framework that reveals a truly spatially explicit distribution these advantages and challenges into account, we consider the pre-
of latent phenomena. The advantage of this approach is that inherent sented methodology highly suited to enrich the assessment and
information about habitat quality is used directly in the delineation monitoring of riparian forest quality on a landscape level.
process and thus objects with uniform response to the underlying
phenomenon (here: habitat quality) are derived. Consequently, pro-

139
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Acknowledgements in a plantation forest managed by clearfelling. J. Appl. Ecol. 43, 1160–1171.


Förster, M., Frick, A., Walentowski, H., Kleinschmit, B., 2008. Approaches to utilising
QuickBird data for the monitoring of Natura 2000 habitats. Commun. Ecol. 9,
This research was conducted within the project MS.MONINA (www. 155–168.
ms-monina.eu) funded by the European Community’s Seventh Forman, R.T.T., Godron, M., 1986. Landscape Ecology. John Wiley & Sons, New York.
Forman, R.T.T., 1995. Land Mosaics: The Ecology of Landscapes and Regions. Cambridge
Framework Programme (FP7/2007–2013) under grant agreement No. University Press, Cambridge.
263479. We want to thank Thomas Strasser and Geoff Smith for their Geburek, T., Milasowsky, N., Frank, G., Konrad, H., Schadauer, K., 2010. The Austrian
crucial data input, as well as Lena Pernkopf, Michael Hagenlocher and biodiversity index: all in one. Ecol. Indic. 10, 753–761.
Gelfand, A.E., 2010. Misaligned spatial data: the change of support problem. In: Gelfand,
Marion Borderon for their support. A.E., Diggle, J.P., Fuentes, M., Guttorp, P. (Eds.), Handbook of Spatial Statistics. CRC
Press, pp. 517–540.
References Gillespie, T.W., Foody, G.M., Rocchini, D., Giorgi, A.P., Saatchi, S., 2008. Measuring and
modelling biodiversity from space. Prog. Phys. Geogr. 32, 203–221.
Goetz, S., Steinberg, D., Dubayah, R., Blair, B., 2007. Laser remote sensing of canopy
Anderson, L.H.H., 1991. Bryophytes and decaying wood a comparison between a habitat heterogeneity as a predictor of bird species richness in an eastern temperate
managed and a natural forest. Holarct. Ecol. 14, 121–130. forest, USA. Remote Sens. Environ. 108, 254–263.
Azim, U.M., 2006. Structural and Functional Roles of Riparian Management Areas in Gordon, N.D., Macmahon, T.A., Finlayson, B., Gippel, C.J., Nathan, R.J., 2004. Stream
Maintaining Stream Values in the Acadian Forest, Technical Bulletin 922. National Hydrology—An Introduction for Ecologists, 2nd edition. John Wiley & Sons, New
Council for Air and Stream Improvement. York.
Baatz, M., Schäpe, A., 2000. Multiresolution Segmentation: An Optimization Approach Gotway, C.A., Young, L.J., 2002. Combining incompatible spatial data. J. Am. Stat. Assoc.
for High Quality Multi-scale Image Segmentation. In: Strobl, J., Blaschke, T., 97, 632–648.
Griesebner, G. (Eds.), Angewandte Geographische Informationsverarbeitung XII. Groeneveld, R.A., Meeden, G., 1984. Measuring skewness and kurtosis. J. R. Stat. Soc. D:
Wichmann Verlag, pp. 12–21. Stat. 33 (4), 391–399.
Berg, A., Ehnström, B., Gustafsson, L., Hallingbäck, T., Jonsell, M., Weslien, J., 1994. Haest, B., Thoonen, G., Borre, J.V., Spanhove, T., Delalieux, S., Bertels, L., Kooistra, L.,
Threatened plant, animal, and fungus species in Swedish forests: distribution and Mücher, C.A., Scheunders, P., 2010. An object-based approach to quantity and
habitat associations. Conserv. Biol. 8, 718–731. quality assessment of heathland habitats in the framework of Natura 2000 using
Bergen, K.M., Gilboy, A.M., Brown, D.G., 2007. Multi-dimensional vegetation structure in hyperspectral airborne AHS images. In: Proc Third Int. Conf. Asp. Geogr. Object-
modelling avian habitatl. Ecol. Inform. 2, 9–22. Based Image Anal. Gent, Belgium.
Berry, B.J.L., 1967. Grouping and regionalizing: an approach to the problem using Hagenlocher, M., Kienberger, S., Lang, S., Blaschke, T., 2014. Implications of spatial
multivariate analysis. In: Garrison, W.L., Marble, D.F. (Eds.), Quantitative scales and reporting units for the spatial modelling of vulnerability to vector-
Geography, Northwestern Study in Geography, pp. 219–251. bornediseases. In: Vogler, R., Car, A., Strobl, J., Griesebner, G. (Eds.), GI_Forum 2014.
Bock, M., Rossner, G., Wissen, M., Remm, K., Langanke, T., Lang, S., Klug, H., Blaschke, Geospatial Innovation for Society, Wichmann Verlag, pp. 197–206.
T., Vrscaj, B., 2005. Spatial indicators for nature conservation from European to local Hughes, F.M.R., 1997. Floodplain biogeomorphology. Prog. Phys. Geogr. 501–529.
scale. Ecol. Indic. 5, 322–338. Humphrey, J.W., Watts, K., 2004. Biodiversity indicators for UK managed forests:
Byrne, B.M., 1998. Structural Equation Modeling with LISREL, PRELIS, and SIMPLIS. development and implementation at different spatial scales. Marchetti, M. (Ed.),
Lawrence Erlbaum Associates, Inc. Monitoring and Indicators of Forest Biodiversity in Europe ?From Ideas to
Capra, F., Luisi, P.L., 2014. The Systems View of Life. A Unifying Vision. Cambridge Operationality. EFI Proceedings, 51 37–48.
University Press. Kienberger, S., Hagenlocher, M., 2014. Spatial-explicit modelling of social vulnerability
Clerici, N., Weissteiner, C.J., Paracchini, M.L., Strobl, P., 2011. Riparian Zones: Where to malaria in East Africa. Int. J. Health Geogr. 13 (29), 108–120.
Green and Blue Networks Meet. European Community, D.G. Joint Research Centre. Kienberger, S., Lang, S., Zeil, P., 2009. Spatial vulnerability unit—expert-based spatial
Clerici, N., Weissteiner, C.J., Paracchini, M.L., Boschetti, L., Baraldi, A., Strobl, P., 2013. modeling of socio-economic vulnerability in the Salzach catchment, Austria. Nat.
Pan-European distribution modelling of stream riparian zones based on multi-source Hazard Earth Syt. 9, 767–778.
earth observation data. Ecol. Indic. 24, 211–222. Lang, S., Kienberger, S., Tiede, D., Hagenlocher, M., Pernkopf, L., 2014. Geons-domain-
Coote, L., Dietzsch, A.C., Wilson, M.W., Graham, C.T., Fuller, L., Walsh, A.T., Irwin, S., specific regionalization of space. Cartogr. Geogr. Inf. 41 (3), 214–226.
Kelly, D.L., Mitchell, F.J.G., Kelly, T.C., ÓHalloran, J., 2013. Testing indicators of Lowman, N., Rinker, H., 2004. Forest Canopies, 2nd edition. Elsevier.
biodiversity for plantation forests. Ecol. Indic. 32, 107–115. NRC, 2002. Riparian Areas. Functions and Strategies for Management. Committee on
Council Directive 92/43/EEC, 1992. Council directive on the conservation of natural Riparian Zone Functioning and Strategies for Management. Water Science and
habitats and of wild fauna and flora. http://europa.eu.int/comm/environment/ Technology Board, National Research Council, National Academy Press.
nature/habdir.htm. Nagendra, H., Lucas, R., Honrado, J.P., Jongman, R.H.G., Tarantino, C., Adamo, M.,
Döringer, M., Tockner, K., 2008. Morphology and dynamics of riparian zones. In: Arizpe, Mairota, P., 2013. Remote sensing for conservation monitoring: assessing protected
D., Mendes, A., Rabaca, J.E. (Eds.), Sustainable Riparian Zones: A Management areas, habitat extent, habitat condition, species diversity and threats. Ecol. Indic. 33,
Guide, Ripidurable, pp. 23–64. 45–59.
Davies, C.E., Moss, D., Hill, M.O., 2004, EUNIS Habitat Classification Revised 2004. Naiman, D.J., Decamps, H., McClain, M.M., 2005. Riparia—Ecology, Conservation and
Decamps, H., Naiman, R.J., McClain, M.M., 2008. Riparian systems as zones of pervasive Management of Streamside Communities. Elsevier.
anthropogenic stress. In: Arizpe, D., Mendes, A., Rabaca, J.E. (Eds.), Sustainable Nardo, M., Saisana, M., Saltelli, A., Tarantola, S., Hoffman, A., Giovannini, E., 2008.
Riparian Zones: A Management Guide, Ripidurable, pp. 114–121. Handbook On Constructing Composite Indicators: Methodology And User Guide. EC
Delalieux, S., Somers, B., Haest, B., Spanhove, T., Vanden Borre, J., Mücher, C.A., 2012. Joint Research Centre & OECD Statistics Directorate and the Directorate for Science,
Heathland conservation status mapping through integration of hyperspectral mixture Technology and Industry.
analysis and decision tree classifiers. Remote Sens. Environ. 126, 222–231. Neumann, C., Weiss, G., Schmidtlein, S., Itzerott, S., Lausch, A., Doktor, D., Brell, M.,
Dimopoulos, P., Zogaris, S., 2008. Vegetation and flora of riparian zones. In: Arizpe, D., 2015. Gradient-based assessment of habitat quality for spectral ecosystem
Mendes, A., Rabaca, J.E. (Eds.), Sustainable Riparian Zones: A Management Guide, monitoring. Remote Sens. 7, 2871–2898.
Ripidurable, pp. 66–83. Noss, R.F., 1999. Assessing and monitoring forest biodiversity: a suggested framework
Drăguţ, L., Csillik, O., Eisank, C., Tiede, D., 2014. Automated parameterisation for multi- and indicators. For. Ecol. Manage. 115, 135–146.
scale image segmentation on multiple layers. ISPRS J. Photogramm. Remote Sens. 88, O’Neill, R.V., De Angelis, D.L., Allen, T.F.H., Waide, J.B., 1986. A Hierarchical Concept of
119–127. Ecosystems, Monographs in Population Biology 23. Princeton University Press.
Duro, D., Coops, N.C., Wulder, M.A., Han, T., 2007. Development of a large area OECD, 2003. OECD Environmental Indicators −Development, Measurement and Use.
biodiversity monitoring system driven by remote sensing. Prog. Phys. Geogr. 31, OECD Environment Directorate Environmental Performance and Information
235–260. Division, Paris.
Dwire, K.A., Lowrance, R.R., 2006. Riparian ecosystems and buffers—multiscale OECD, 2004. The OECD-JRC Handbook on Practices for Developing Composite
structure, function and management: introduction. J.Am. Water Resourc. Ass. Indicators. OECD Committee on Statistics, Paris.
42, 1–4. Openshaw, S., 1984. The modifiable areal unit problem. Concepts and Techniques in
EEA, 2012, Streamlining European biodiversity indicators 2020: Building a future on Modern Geography. Geo Books, Norwich.
lessons learnt from the SEBI 2010 process, Technical report No 11/2012. Paccagnel, K., 2013. Objektbasierte Klassifikation von Baumarten im Natura 2000
Ellenberg, H., 2009. Vegetation Ecology of Central Europe, 4th ed. Cambridge Univ. Pr., Schutzgebiet Salzachauen auf Basis einer hochaufgelösten WorldView-2
Cambridge. Satellitenbildaufnahme. Master Thesis. University of Salzburg.
Ellmauer, T., 2005, Entwicklung von Kriterien, Indikatoren und Schwellenwerten zur Palmer, M.W., Earls, P., Hoagland, B.W., White, P.S., Wohlgemuth, T., 2002. Quantitative
Beurteilung des Erhaltungszustandes der Natura 2000-Schutzgüter. Band 3: tools for perfecting species lists. Environmetrics 13, 121–137.
Lebensraumtypen des Anhangs I der Fauna-Flora-Habitat-Richtlinie. Im Auftrag der Richardson, D.M., 1998. Forest trees as invasive aliens. Conserv. Biol. 12, 18–26.
neun österreichischen Bundesländer des Bundesministerium f. Land-und Riedler, B., Strasser, T., Pernkopf, L., Tiede, D., Lang, S., 2013a. Habitat quality analysis
Forstwirtschaft, Umwelt und Wasserwirtschaft und der Umweltbundesamt GmbH. of Natura 2000 riparian forests. In: Lang, S., Pernkopf, L. (Eds.), Ecosystem and
European Commission, 2007. Note to the Habitats Committee, Assessment, Monitoring Biodiversity Monitoring: Best Practice in Europe and Globally, Conference
and Reporting of Conservation Status—Preparing the 2001–2007 Report Under Proceedings GI-Forum 2013. Salzburg. pp. 540–543.
Article 17 of the Habitats Directive. DocHab-04-03/03 Rev.3. European Commission, Riedler, B., Pernkopf, L., Strasser, T., Lang, S., 2013b. EO-based habitat structure
Brussels. assessment in a trans-boundary Natura 2000 site. In: Conference Proceedings
Eycott, A.E., Watkinson, A.R., Dolman, P.M., 2006. Ecological patterns of plant diversity 5thSymposium for Research in Protected Areas. Mittersill. pp. 661–664.

140
B. Riedler, S. Lang Ecological Indicators 94 (2018) 128–141

Riedler, B., Pernkopf, L., Strasser, T., Lang, S., Smith, G., 2015. A composite indicators for Tockner, K., Schiemer, F., Baumgartner, C., Kum, G., Weigand, E., Zweimüller, I., Ward,
assesing habitat quality of riparain forests derived from Earth observation data. Int. J. J.V., 1999. The Danube restoration project: species diversity patterns across
Appl. Earth Observ. Geoinf. 37, 114–123. connectivity gradients in the floodplain system. Regul. Rivers 14, 245–258.
Rocchini, D., Chiarucci, A., Loiselle, S.A., 2004. Testing the spectral variation hypothesis Trimble, 2014. eCognition Developer 9.0 Reference Book. Trimble Germany GmbH.
by using satellite multispectral images. Acta Oecol. 26, 117–120. Troll, C., 1939. Luftbildplan und ökologische Bodenforschung. Zeitschrift der Gesellschaft
Saisana, M., Tarantola, S., 2002. State-of-the-Art Report on Current Methodologies and für Erdkunde 241–298.
Practices for Composite Indicator Development, EUR 20408 EN. European Turner, M.G., Gardner, R.H., ÓNeill, R.V., 2001. Landscape Ecology in Theory and
Commission, Joint Research Centre, Italy. Practice—Pattern and Process. Springer.
Schnitzler, A., 1994. European alluvial hardwood forests of large floodplains. J. Biogeogr. Unger, I.M., Muzika, R.-M., 2008. Influence of microtopograhy on soil chemestry and
21, 605–623. understory riparian vegetation. In: Jacobs, D.F., Michler, C.H. (Eds.), Proceedings,
Schuck, A., Rois, M., 2004. Forest biodiversity indicators—a contribute to an eea core set 16th Central Hardwood Forest Conference.
of biodiversity indicators. Marchetti, M. (Ed.), Monitoring and Indicators of Forest Vanden Borre, J., Paelinck, D., Mücher, C.A., Kooistra, L., Haest, B., De Blust, G., Schmidt,
Biodiversity in Europe? From Ideas to Operationality, EFI Proceedings, 51 37–48. A.M., 2011. Integrating remote sensing in Natura 2000 habitat monitoring: prospects
Strasser, T., Lang, S., Riedler, B., Pernkopf, L., Paccagnel, K., 2014. Multiscale object on the way forward. J. Nat. Conserv. 19, 116–125.
feature library for habitat quality monitoring in riparian forests. Geosci. Remote Sens. Wantzen, K.M., Junk, W.J., 2008. Riparian wetlands. In: Erik, S., Brian, F. (Eds.),
Lett. IEEE 559–563. Encyclopedia of Ecology. Academic Press, pp. 3035–3044.
Tockner, K., Stanford, J.A., 2002. Riverine flood plains: present state and future trends. Ward, J.V., Tockner, K., Arscott, D.B., Claret, C., 2002. Riverine landscape diversity.
Environ. Conserv. 29, 308–330. Freshw. Biol. 47, 517–539.

141

You might also like