Technical ReportPDF Available

Potential Primary Forests Map of Romania (published by Greenpeace CEE Romania; Centre for Econics and Ecosystem Management, Eberswalde University for Sustainable Development; Geography Department, A. I. Cuza University of Iași)

Authors:

Abstract and Figures

Authors and contributions: Freya Kathmann: International Forest Ecosystem Management, Eberswalde University for Sustainable Development Alexandru Ciutea: Ph. D student, Geography Dept., Al. I. Cuza University of Iași, Romania Iovu-Adrian Biriș: Primary forests protection in Romania; Prof. Dr. Pierre L. Ibisch, Professor for Nature Conservation: Romania’s key role in the conservation of European primary forests Valentin Sălăgeanu: Forests and Biodiversity Campaigner, Greenpeace Romania: Why a Primary Forest Potential Map?
Content may be subject to copyright.
1
2
Imprint Publisher:
Greenpeace CEE Romania; Centre for Econics and Ecosystem
Management, Eberswalde University for Sustainable Development;
Geography Department, A. I. Cuza University of Iași,
Date:
November 2017
Authors:
Freya Kathmann: International Forest Ecosystem Management,
Eberswalde University for Sustainable Development
Alexandru Ciutea: Ph. D student, Geography Dept., Al. I. Cuza University
of Iași, Romania
Iovu-Adrian Biriș: Primary forests protection in Romania;
Prof. Dr. Pierre L. Ibisch, Professor for Nature Conservation: Romania’s
key role in the conservation of European primary forests
Valentin Sălăgeanu: Forests and Biodiversity Campaigner, Greenpeace
Romania: Why a Primary Forest Potential Map?
Editors:
Prof. Dr. Pierre L. Ibisch, Professor for Nature Conservation
Lect. Dr. Adrian Ursu, Geography Dept, A. I. Cuza, University of Iași
Photo credits:
Dan Câmpean, Mitja Kobal, Cristi Grecu, Ionuț Brigle
Layout:
Quickdata
Paper:
printed on recycled paper
3
Contents
Primary forests protection in Romania 4
Romania’s key role in the conservation
of European primary forests 11
Why a Primary Forests Potential Map? 13
Results 14
Frequently asked questions on the
Primary Forests Potential Map of Romania 24
Methodology 26
1 | Used data 26
2 | Spectral characteristics of the forests 27
3 | Image preparation 28
3.1 Radiometric correction 28
3.2 Cloud identification 29
4 | Forest area derivation 30
4.1 Segmentation 30
4.2 Supervised classification 30
4.3 Forest stand age classification 31
5 | Logged forest and intact forest separation 34
6 | Global Forest Change Data Exclusion 37
7 | Exclusion of Plantations 38
8 | Exclusion of Roads and Railways 39
9 | Size evaluation 42
10 | Connectivity Assessment 42
Publication bibliography 43
Glossary 50
4
Primary forests protection
in Romania
Iovu - Adrian Biriș
In the last century, the topic of protecting the virgin forests1 has constantly/
continuously been both on the agenda of the environment and forestry
studies’ scientific community and of the environmental organisations from
Romania and the world. Not only the scientific arguments but also the
perseverance in supporting the role and the importance of their protection
have convinced the decision factors/powers from the forestry industry
to give this objective the chance it needed. At first it was a rather vague
idea – a matter of principle; but it has slowly become more and more
achievable once the tools and the procedures for the protection of forests
were implemented.
Between 2000 and 2008, a series of stipulations concerning the
protection of virgin forests existed on paper as declarations without
being taken to an operational level in order to produce concrete
results. However, they were included in official or normative/bounding
documents: e.g. The policy and the development strategy of the forest
sector in Romania between 2001 – 2010 (MAPPM, 2000) , The technical
standards/norms for forest administration, approved by the ministerial
order 1672/2000 (MAPPM, 2000) , The Forest Code (law no. 46/2008).
Following a massive communication campaign – Save the Virgin Forests! –
which received incredible support from the public, the forest legislation for
1. To be consistent with the Romanian legal terminology ”virgin” is used below as an equivalent for
primary.
5
the protection of the virgin forests was gradually approved between 2012
and 2016. Thus, in September 2012, the Forest and Environment Ministry
passed the ministerial order no. 3397/2012 establishing the criteria and
the identification indicators of Romania’s virgin forests. In 2015, law no.
133/2015, which modified and completed law no. 46/2008 – The Forest
Code, stipulated the strict protection of the virgin and quasi-virgin forests
and the creation of the «National Catalogue of Virgin and Quasi-virgin
Forests» (art.26 line 3). Then, in 2016, by the order of the Environment,
Water and Forests Ministry no. 1417/11.07.2016, the creation of the
National Catalogue/Register of Virgin and Quasi-virgin Forests was
approved as an instrument of cataloguing and managing Romania’s virgin
and quasi-virgin forests – just like they are defined in the annex of the
Environment, Water and Forests Ministry’s order no. 3397/2012 which
established the criteria and the identification indicators of Romania’s virgin
and quasi-virgin forests.
Once the ministerial order no. 1417/11.07.2016 was applied/executed,
it revealed a series of deficiencies which complicated the process
of identifying and mapping the virgin forests. Therefore, it has been
replaced by the Environment, Water and Forests Ministry’s order no.
2525/31.12.2016.
Thus, the legal requirements and the tools for the identification and
protection of the virgin and quasi-virgin forests in Romania were created
between 2012-2016. However, the implementation of the legislation
concerning the strict protection of the virgin forests proved to be quite
difficult because the people who were in charge of performing it showed
indifference, lack of involvement and even unwillingness to apply it.
Although the ministerial order no. 3397/2012 offers the administrators
6
the tool to identify the virgin and quasi-virgin forests for their inclusion in
the appropriate forest classification to ensure their strict protection, the
implementation of the legislative requirements was completely ignored by
them. Relatively only small areas have been classified as virgin or quasi-
virgin forests, often as a result of the evidence provided by the public. At
the same time, although the order no. 2525/2016 foresees the obligation
of the administrator / provider of forest services to ‘communicate to the
specialised territorial structure of the central public authority in charge
of forest management the information regarding the placement and the
surface of the identified forest and to stop any forest works in the area’
if ‘he/she identifies in his/her administrative area a forest that meets the
criteria and the indicators specified by the Environment, Water and Forests
Ministry’s order no. 3397/2012’, even this provision has been ignored by
those people in charge.
Although there is a coherent and an adequate legal structure for the
identification and the protection of the virgin forests and both the
state owned and private forests administrations, including the certified
operators, are in possession of all the necessary information required to
establish if a forest meets the criteria and the indicators to be declared
a virgin or quasi-virgin forest, the responsible parties haven’t fulfilled
their obligations required by the law. What is more, some owners or
administrators / forest service providers procrastinate on purpose to
protect them or even ‘force’ forest work so that the forest wouldn’t
correspond to the requirements that would classify them as virgin or
quasi-virgin forests. All these happen under the ‘supervision’ of the
territorial structures that belong to the central public authority who is in
charge of forest management and who watch helplessly or even pretend
not to see/notice when the virgin forest protection legislation is being
ignored.
The information presented above shows without a doubt that the bottom-
up approach regarding the identification and protection of virgin and
quasi-virgin forests, which is the main type of approach suggested by
the law, hasn’t led to the expected results/ hasn’t worked, of course with
some exceptions, because of the lack of interest or lack of involvement of
the people and institutions in charge.
The environmental non-governmental organisations (NGOs) have
demanded a speedy/an urgent implementation of the law regarding
the protection of virgin and quasi-virgin forests; therefore, at the end of
2016, the central public authority responsible with forestry initiated a
governmental decision that would use the Environmental Fund to finance
a study for the identification of Romania’s virgin and quasi-virgin forests in
order to enlist them in the “National Catalogue of Virgin and Quasi-virgin
Forests”. The budget allocated to this project was 2.58 million lei/RON.
This kind of a top-down approach would have accelerated the process if
a general enforcement of the legal requirements regarding the protection
of the virgin and quasi-virgin forests had been implemented in the entire
country. The necessity of a project for the identification and mapping of
these forests, rolled out on a national level, has been acknowledged by
all the parties involved. Unfortunately, this initiative has been abandoned
/ postponed by the new decisional board of the Environment, Water and
Forests Ministry assembled at the beginning of 2017. Thus, no steps
have been taken towards this until the present moment. Any delay in
the implementation of such a project, whether deliberate or caused by
incompetence/irresponsibility/idleness on behalf of the authorities, means
7
the irreversible loss of large areas of virgin forests.
More than a year after the legislation concerning the protection of virgin
and quasi-virgin forests has been adopted, we sadly ascertain that quite a
small surface of forests has been included in the National Catalogue. On
the other hand, concern rises at the alarming rate of their destruction.
The facts presented above clearly illustrate why the mapping of the virgin
and quasi-virgin forests is essential for their protection. The forest surfaces
which are to be protected must be identified and located with precision,
so they can be included in foundation studies and added to the National
Catalogue.
Such an action is especially necessary and urgent as the last time these
forests were catalogued was between 2001-2004, approximately 15
years ago. Meanwhile, Romania’s forests have gone through quit a
tumultuous period: their restitution to the ex-owners, the diversification
of the property’s structure and the fragmentation of forest property, the
roads created to access the virgin forests’ basinets to explore them,
illegal deforestation etc. All of these have affected the virgin forest areas,
which had already been catalogued, to such an extent that they no longer
correspond to the selection criteria anymore.
In fact, the work methodology regarding the identification of the virgin and
quasi-virgin forests (Annex 3, ministerial order no. 2525/2016) mentions
as a preliminary analysis stage a screening / and investigation that must
be performed using satellite information or orthophoto-plans for the
configuration of forest surfaces which have the potential to meet the
identification criteria of virgin and quasi-virgin forests. (Ministerial order no.
8
2525/2016, Annex 3, Stage 2. The preliminary analysis: “At this stage we
use specific office methods to identify the surfaces affected by human
intervention which do not meet the natural criteria and, as such, are to be
excluded from the documentation required for the next stage. The areas
constituted in stage 1 are to be analysed with the help of satellite images
or orthophoto-plans provided by the executor of the foundation studies,
e.g. Google Maps, Bing Maps etc.).
Currently, the main challenge regarding the protection of the virgin
forests is the development of concrete activities that should focus on
the identification and localization of areas which have a high potential of
fostering virgin and quasi-virgin forests throughout the country or in the
Carpathian Mountains. The creation of such a map / geo-spatial database
would allow the focus of resources and efforts of all the parties involved on
a smaller surface which has a higher probability of accommodating forests
with a high degree of naturalness.
The initiative of Greenpeace Romania to create a complete /
comprehensive map of the areas that have high potential to house virgin
and quasi-virgin forests comes as an answer to the urgent and concrete
need regarding their protection, and represents an extremely useful tool
in the identification of those areas / surfaces with forests which appear to
meet the naturalness criteria.
The aim of the study is represented by the creation of a geo-spatial
database/maps of potential primary forests / old-growth forests, which
have a high degree of naturalness and the potential to present a primary
/ virgin or quasi-virgin characteristic, specified in Romania’s Carpathian
Convention.
9
The methodology used for the development of this map / geo-spatial
database has a scientific foundation both from the point of view of the
sets of data used and also the methods used for its processing, analysis
and interpretation. All the data used in the study such as: Sentinel 2
satellite images, the digital rendering of the ground, high-resolution images
from Google Earth, the limits of the area falling under the Carpathian
Convention, comes from open sources, verifiable, official or scientifically
validated at an international level, available on request or freely. The
methods used for the processing, analysis and interpretation of data
are enlisted with all the required information, so they can be verified and
reproduced. From the practical point of view, the methodology used
combines the supervised classification with the assisted one which has
allowed for the delimitation of areas / surfaces occupied by old-growth
forests. They have a high probability of accommodating forests with a high
degree of naturalness / virgin or quasi-virgin.
The result of the study should be followed-up by a ground analysis in
order to analyse the meeting of the tools and criteria for the identification
of virgin and quasi-virgin forests and to exclude the surfaces that
might not qualify. The limitation of the study is mainly caused by the
lack of information from the forest administration regarding the human
interference that happened in the past as well as the forest division layer
required to cut out the identified areas / polygons on the surface lines
or surface parts. The lack of information from the forest administration
has partially been compensated by using some data or algorithms for
the exclusion of the forests which seem not to meet the naturalness
criteria, such as coniferous plantations outside their natural areas, pure
coniferous plantations growing in the subfloor of forests which mix beech
and softwood, the surfaces that have been affected by forest loss in the
10
last 15 years, and the surfaces that include transportation routes / roads /
cableways. However, it is important to specify that the use of the data and
algorithms mentioned before does not guarantee the total exclusion of
areas which do not meet the naturalness criteria from old-growth forests
which present the potential of being primary / virgin or quasi-virgin.
In spite of the limitations already mentioned, which are to be expected
in such a large study aimed to cover the entire area of the Carpathian
Convention, which hasn’t benefitted from the information regarding human
intervention noted by the forest administrations, the obtained results can
be considered quite consistent and especially useful to all the parties that
have the responsibility to implement the legal requirements regarding virgin
forests as well as those interested to get involved in their identification and
protection.
In the last decade, the identification and protection of forests with a high
degree of naturalness has become a regional and European interest.
From the point of view of the Carpathian Region, the Protocol related
to the durable management of forests, which was adopted on the 27th
May 2011 and was part of the Convention regarding the protection and
durable development of the Carpathian Mountains, stipulates in art. 10
line 1 that “each Party is to pass their own legislation at a national level
for the identification and protection of natural forests, especially the
virgin ones from the Carpathians, by assigning natural areas that are
sufficiently protected from the perspective of numbers and surface and by
implementing other specific measures of protection”.
Quite recently, the Wild Europe organisation has initiated and coordinated
the elaboration of a “Strategy for the protection of Europe’s old-growth
forests”, aiming to save the last areas of virgin forests from the European
continent. One of the goals of this strategy is to create a geo-spatial
database / interactive map of forests with a high degree of naturalness.
This platform, known as “The Last European ancient Forest (LEAF)”, will
be the foundation for the development of an Early Warning System for
the identification and the prevention of threats concerning these forests
(https://www.wildeurope.org/index.php/wild-areas/old-growth-forest-
protection-strategy).
By taking concrete actions for the protection of virgin forests such as:
the creation of an adequate legislation, the implementation of practical
measures of protection, the development of a mapping methodology and
of interactive maps, providing compensation payments, Romania has
demonstrated its dedication to accomplish this endeavour. Thus, Romania
could become a best practice example to be adopted by the rest of
Europe.
11
For a long time, in the history of European nature conservation, forests
and especially ancient and primary forests have been seriously neglected.
In many regions, forests had been severely reduced and degraded in
earlier centuries, before they were partly reestablished and/or managed.
It seemed to be a given fact that Europe would have responsibility for
mainly conserving cultural landscapes, severely modified by human
action, reflecting more or less historical land uses and harbouring many
species that would have come to Europe with or after the introduction of
agriculture. The conservation of forests and wilderness areas was an issue
to be delegated to tropical or boreal countries. Indeed, in most European
countries, the ancient and primeval forests – originally and potentially
representing the vastest and continuous ecosystem complex across the
whole continent - have entirely vanished or been reduced to more or less
small remnant patches.
The different ancient and primeval forests help us to understand that
ancientness is somehow relative. Of course, many primeval forests in
Europe are ‘just’ a few thousand years old, and developed in parallel to
humans colonizing and changing the European land. Topography, political
processes and special land tenure patterns contributed to their protection.
Ancient and primeval forests are also a key component of the European
heritage and identity. Therefore, it has been a great achievement that
the serial UNESCO World Natural Heritage property dedicated to the
Primeval Beech Forests of the Carpathians caused a steep increase of
attention given to this ecosystem type. The extensions achieved in 2011
and 2017 that led to the establishment of the largest and most complex
transnational World Heritage property, confirm the growing interest in and
importance of preserving these ‚crown jewels‘ of European forests, now
called Ancient and Primeval Beech Forests of the Carpathians and Other
Regions of Europe stretching over 12 countries (Albania, Austria, Belgium,
Bulgaria, Croatia, Germany, Italy, Romania, Slovakia, Slovenia, Spain,
Ukraine). The Europe-wide screening process in search of appropriate
sites to be included (Ibisch et al. 2017 ), inspired the foundation
of a network that comprises scientists, protected area managers,
organizations and citizens from all over the continent, the European Beech
Forest Network.
The values of ancient and primeval forests, all too often, were recognized
when it was (almost) too late. Today, many of them represent top priority
areas, which receive much attention from conservationists, visitors and
scientists. Beyond being unique and irreplaceable reserves for biodiversity
that has been lost elsewhere, the values of ancient and primeval forests
refer to their multiple cultural and regulating ecosystem services they
provide to humans. Many of the first component parts of the World
Heritage property to be established (in Ukraine, Slovak Republic and
Germany) have already witnessed how the public attention and touristic
relevance can contribute to local sustainable development. Amongst
Romania’s key role in the
conservation of European
primary forests
Pierre L. Ibisch
Centre for Econics and Ecosystem Management at Eberswalde
University for Sustainable Development
European Beech Forest Network
12
others, primeval forests also represent natural laboratories that can inform
our ecosystem management; they are the baseline to our managed
forests, and are likely to provide ‘nature-based solutions‘ to future
problems.
In many parts of Europe we have built our economic richness on the
ruins of the natural ecosystems, and the loss of ecological functionality
is compensated by the ongoing import of goods and services from
others parts of the world, ever more externalizing environmental costs.
Whilst this has to be stopped, and Europe has to learn to live wisely
and sustainably from its own resources, as a region with a relatively high
population density, it has to show to the World that the universal task
of nature conservation must and can be achieved everywhere – without
losing out economically and socially. For natural and historical reasons,
the Carpathians are the core area of the current European ancient
and primeval forests. The growing European Beech Forest Network
appreciates that Romania became part of the joint pan-European
endeavour adding precious and ecologically important Carpathian forests
to the serial World Natural Heritage property. Romania is the country with
the largest extent of old European beech forests; and it is also amongst
the country with the highest loss rates. It seems to be a little unfair that
the Carpathian countries that somehow did not damage their natural
heritage in the past (as much as others did), now have a relatively higher
responsibility for preserving the ancient forests. A business logic could
be that Romania has the right to degrade its ecosystems for boosting
socioeconomic development, as it was done in the past in Germany, the
Netherlands, Belgium and many other countries. Indeed, Romania has the
sovereign right to repeat the same mistakes committed by other nations.
But it has also the opportunity to come up with a new approach and turn
towards a more ecosystem-based model of sustainable development.
Romania has a key role in the conservation of European forests – and a
key opportunity to be Europe’s sustainable forest country No. 1.
On our European continent, those countries that are responsible
for a substantial share of past biodiversity loss and ongoing ‘nature
consumption’ also outside their national territories, and therefore may
have even gained some development advantages, have the duty to
collaborate with others empowering them to develop in a truly sustainable
way without further damaging our continental natural heritage. Clearly,
in Europe we need new strategies and instruments for sharing short-
term socioeconomic burdens related to preserving ecosystems and the
services they provide and thereby investing in a sustainable future. The
fundament for further building a coherent European conservation system
that can also financially support states and regions with special challenges
and responsibilities, will be a good and transparent knowledge of the
distribution and extent of the natural treasures at stake. Therefore, we
were very happy to support the current mapping exercise that will be
another input for creating an informed discourse on the conservation of
ancient and primeval forests in Romania.
13
Why a Primary Forests
Potential Map?
Valentin Sălăgeanu
Greenpeace România
The completion of the National Catalogue of Virgin and Quas-virgin Forest
could be easily achieved if the authority in charge truly accepted thieir
responsibility. The process would begin with the evaluation of the data
existent in forest administrations found either directly in the archives of the
Ministry of Water and Forest, or in each Forest Area, followed by a proper
ground analysis operated by the specially instructed staff of the Forest
Guards, and would end up being validated by the Technical Committee of
Forest Approvals (CTAS), the highest specialist commission.
The process would not have to start anew, since a partial classification
has been already done between 2001 and 2004. The National Research
and Development Forest Institute ‘Marin Drăcea’ (ICAS at that time) has
undertaken, with external financial aids, a project for the identification
of Romania’s primary forests. Almost 220.000 ha have been mapped,
without counting the surfaces smaller than 50 ha.
Today we have the benefit of a national legislation, which creates a
norm for the processes of identification and protection of Romania’s and
Europe’s invaluable natural patrimony, which also belongs to the world, as
proven by the recent inclusion of over 24000 ha of Romanian virgin beech
forests in the world patrimony of UNESCO.
However, the process of saving the virgin forests is being halted by the
lack of interest from the authorities in charge. Unfortunately, the Water and
Forest Ministry’s lethargy justifies the inaction of the other involved parties
when it comes to following the law. This evident lack of responsibility
results in the huge loss of an important surface of one of nature’s last
refuges and an irreparable decline in its biodiversity.
The map of potential virgin forests that has been produced by Greenpeace
Romania in partnership with the Centre for Econics and Ecosystem
Management at Eberswalde University for Sustainable Development from
Germany, and University Alexandru Ioan Cuza from Iași is the first stage
in the protection process of these forests (ministerial order 2525/2016,
Annex 3). The purpose of this first analysis based on satellite images is to
evaluate the state of the forests, in order to eliminate the degraded areas,
which do not correspond to the identification criteria and in order to define
and create an hierarchy for the next stages of documentation and final
ground evaluation.
We have to underline that the methodology developed for the Primary
Forest Potential Map of Romania is perfectible and that, as such, this
exercise can be continued.
We expect this exercise to be considered a call to action by the Ministry
of Water and Forest, for fast and concrete decisions towards saving
Romania’s virgin forests.
14
Results
Potential
Primary Forests
295.779 ha
19%
overlaps
Pin-Matra
forests
11.555 ha
Bucharest
Surface
22.800 ha
Top Protected Areas
of Potential Primary
Forests
are Potential
Primary Forests
in UNESCO
designated
sites (ha)
4%
59%
are found
in SCI
13%
are found in
National Parks
Făgăraș Mountains
61.423 ha
Domogled -
Valea Cernei
15
Brasov
49.601 ha
Hunedoara
24.841 ha
Prahova
14.240 ha
Covasna
10.095 ha
Mehedinti
6.146 ha
Cluj
5.230 ha
Neamt
3.340
ha
Harghita
2.678 ha
Timis
2.284 ha
Bistrita Nasaud
990 ha
Satu
Mare
806 ha
Buzau
949 ha
Arad
1.688
ha
Bacau
2.555 ha
Mures
3.046
ha
Suceava
6.074 ha
Maramures
7.652 ha
Dambovita
7.602
Gorj
12.600
Alba
10.716 ha
Valcea
21.979 ha
Sibiu
21.930
Caras Severin
39.513 ha
Arges
38.918 ha
SJ VR BH
Top Counties with Potential
Primary Forests
16
Potential primary forest map of the Romanian Carpathians
17
Potential primary forest map of the Western Carpathians
18
Potential primary forest map of South-Western Carpathians
19
Potential primary forest map of the western half of the Southern Carpathians
20
Potential primary forest map of the eastern part of the Southern Carpathians
21
Potential primary forest map of the Curvature and south-eastern part of Carpathians
22
Potential primary forest map of the north-eastern part of the Carpathians
23
Connectivity into all Directions of the Potential Primary Forests in the Romanian Carpathians
24
Frequently asked
questions on the
Primary Forests Potential
Map of Romania
1.Why did you focus on the Carpathian Convention area?
Because the protection of primary forests is clearly addressed by the
Sustainable Forestry Protocol of the Carpathian Convention. As the
national legal framework and its enforcement are weakened by political
instability and corruption, it is of high relevance to highlight the multilateral
obligations and call for international support. As a next step, the mapping
and screening will cover all forests.
2.What does this map show?
This map presents a pre-assessment of potential primary forests. It
represents a searching area where potentially primary forests are likely to
be.
These forests represent the remaining parts of potential primary forests in
Romania. Applying a precautionary principle, Greenpeace Romania will
ask the Ministry of Waters and Forests for a moratorium on further logging
these valuable forests.
3.What does it not show?
The presented map does not present an area that is for sure primary,
virgin or quasi virgin forest. There might exist certain additional valuable
ancient forest remnants that have been excluded due to the criteria
described in the methodology. Still, ground-truthing might show that the
area of ancient and primeval forests is even bigger.
4.How close does it come to reality?
An official inventory of virgin forests is lacking. This map is a first approach
towards identification of primary forests, with the application of freely
available data. The second step would then be a field verification of the
sites. We invite everybody to contribute to the verification of the map. This
way a comparison with reality and thus further improvement of the map
can be achieved.
5.How much does the result differ from other/previous mapping
exercises?
In 2005, the PIN-MATRA study identified about 218.500ha of virgin
forest1. About 20% of the potential primary forest polygons intersect with
the PIN-MATRA polygons. There are many reasons that could explain
these differences. Firstly, the PIN-MATRA study was done using a totally
different methodology, based on the field surveys and compared to our
methodology which is based on remote sensing. PIN-MATRA polygons
are based on the limit of the forest units, which include naturally occurred
canopy gaps, while potential primary forest polygons are delineated only
by the crown cover. This difference is visible especially in the high altitude
coniferous forests situated at the contact with the alpine meadow. Also,
some of the forests that were included in PIN-MATRA were logged.
For example, in the Cumpăna and Cumpănița watersheds 38% of the
PIN-MATRA forests were affected by logging activities 2. PIN-MATRA
1. http://www.mmediu.ro/articol/proiect-pin-matra-padurile-virgine-din-romania/2068
2. http://www.greenpeace.org/romania/ro/campanii/paduri/tabara/tabara-padurii-la-final/
25
polygons are shifted from their original position (Figure 1), this being
another potential factor that could increase the differences between the
two studies.
6. How transparent and replicable is the analysis?
A strong emphasis was given to the transparency of the methodology.
Following the flowchart (Annex 1-3) the methodology is replicable. (In the
manual differences in corrections may occur, due to potentially varying
expertise of the operators; but they are of minor importance).
FIGURE 1. Comparison between potential primary forest polygons
and the pin-matra polygons
26
Methodology
The assessment of the potential primary
forests is based on a mixed approach,
automated and manual techniques being
used in the process. It is a completely
transparent procedure, which will be
described step by step indicating potential
shortcomings and challenges.
1 | Used data
The identification of the potentially primary
forests of Romania was made using Sentinel 2
satellite images, which are freely available and
have a 10m spatial resolution for the visible
and near infrared bands. The images were
downloaded from the Copernicus Data Hub1.
Sentinel 2 images are composed from 12
bands, which cover the visible, near-infrared
(NIR) and mid-infrared (SWIR) wavelengths. Four
bands have a spatial resolution of 10m (2, 3, 4,
8), the rest of them having a resolution of 20m
and 60m.
1. https://scihub.copernicus.eu/dhus/#/home
were not available, thus those areas were
excluded from the analysis (11,1% of the study
area).
Sentinel 2 images were used in detriment of
Landsat images, mainly because of the better
Images from 2016 were mostly used, except
for the clouded areas where images from 2015
were used, if available (Figure 2). Also in some
areas, like the extreme west of the Romanian
Carpathians (west of the Apuseni Mountains)
images with a quality suitable for our analysis
FIGURE 2. Pairs of Sentinel 2 images used in the identification of the potentially old
growth forests
27
spatial resolution. For the visible, NIR and
SWIR bands, Landsat 8 images have a spatial
resolution of 30m, which is lower than Sentinel’s
2 10m and 20m resolution. The disadvantage
of using sentinel 2 images is the short temporal
coverage, because the Sentinel 2 mission
started in 2015. Thus, cloud free images that
could be used in the analysis are scarce.
The Carpathian convention area for Romania2
was used as study area. The Convention covers
68913,2 km2 and includes the mountainous
areas of Romania.
SRTM (Shuttle Radar Topography Mission)3
digital elevation model (DEM) with a 30m spatial
resolution was used.
Google earth4 images (CNES, Airbus,
DigitalGlobe) were used for a visual verification
of the results. Google images have a higher
spatial resolution, but recent images are not
widely available.
Global forest change data set up by Hansen et
2. http://www.carpathianconvention.org/
3. https://www2.jpl.nasa.gov/srtm/
4. https://www.google.com/earth/
al. (2013)5 is used in this study in the version
1.3 indicating forest loss from 2000 – 2015.)
From the dataset only the forest loss data was
utilized.
OpenStreetMap (OSM)6 data is a global
voluntary based and user generated
infrastructure map. It was downloaded at. From
this dataset roads and railways were utilized.
The forest layers from the Corine Land Cover
(CLC)7 data from 2012 were used as reference
in the supervised classification of the tree
species.
5. https://earthenginepartners.appspot.com/science-2013-
global-forest/download_v1.3.html
6. https://www.openstreetmap.org
7. https://www.eea.europa.eu/publications/COR0-landcover
2 | Spectral
characteristics of the
forests
Vegetation in general has a low reflectance
in the visible region of the electromagnetic
spectrum. Higher values are specific to the
infrared region, especially near infrared.
The spectral signature of a forest is mostly
dependent on the interaction between the
electromagnetic radiation and the crown cover.
In our case, the electromagnetic radiation
reflected by the forests was analysed using
Sentinel 2 satellite images. Therefore, the level
of processing and the quality of the images
influences the perceived appearance of the
forests, the values being more or less accurate.
For example, after the topographic correction
the reflectance values of the forests situated on
the shaded slopes are more realistic.
Crown cover reflectance depends on a series
of factors, such as species composition,
vegetation period (growing season) or age. In
the case of a satellite image the reflectance
values are also influenced by the gaps in the
crown cover, shadows created by different
28
tree heights, slope exposition and gradient or
atmospheric conditions.
Because of the complex appearance of the
forests, an implementation of a fully automatic
identification algorithm was considered to be
impractical. In addition, old growth / primary
forests that could be used as samples in an
automated classification are few. Considering
these aspects, we chose to use a semi-
automatic algorithm, thus combining visual
interpretation of the satellite images with
automated methods.
FIGURE3. Old growth forests as seen on sentinel 2 false color
infrared composites
3 | Image preparation
3.1 Radiometric correction
In order to reduce classification errors due to
different illumination conditions or atmospheric
effects, Sentinel images were corrected using
a dedicated Sentinel 2 processor called
Sen2Cor, which is capable of atmospheric and
topographic correction.
One of the most important steps was the
topographic correction, which was useful in
minimalizing the shadow effect that appears
due to different illumination conditions (Figure 4).
Because of the strongly fragmented terrain in
the mountain areas, shadow effects can be
very pronounced. Atmospheric correction was
useful in converting the top of the atmosphere
(TOA) reflectance values in bottom of the
atmosphere (BOA) ones. Sentinel 2 images were
downloaded as L1C processing level products.
With the Sen2Cor software the images were
converted to L2A. Only the 10 and 20m bands
were processed.
An SRTM 30m resolution digital elevation model
was used in the process. The lower spatial
resolution of the SRTM (30m) could influence
the quality of the topographic correction
FIGURE 4. Comparison between a L1A and a L2C corrected image
29
process. Thus, using a higher resolution DEM
may have resulted in a more accurate correction
of the pixels belonging to the shadowed slopes.
In addition, the topographic correction algorithm
has its limitations, the shadow effect from the
slope with very high inclination values being
difficult to remove (Figure 5).
3.2 Cloud identification
Clouds are one of the biggest problems of the
multispectral satellite images, in the case of
Sentinel 2 the availability of images with a low
level of nebulosity being limited, because of
the short temporal coverage (2015 and above).
Satellite images with a cloud coverage between
0 and 2% were utilized.
Clouds and cloud shadows were extracted
manually. Sen2Cor software is capable of
automatic identification of clouds and cloud
shadows, but the accuracy is not ideal for our
analysis (Figure 6). From the 68913km2 of the
study area, 4.,1% is covered by clouds.
Because of the cloud coverage, some primary
forests could appear fragmented or shrunk
(Figure 7). Where available, 2015 images were
used for the areas affected by nebulosity on the
2016 images. Thus, the areas analysed using
FIGURE 5. Examples of topographic
correction limitation
older images could contain fresh loggings that
were not charted.
FIGURE 6. Clouds and cloud shadows (a)
automatically identified using Sen2Cor (b)
and manually mapped (c)
FIGURE 7. Errors due to the cloud coverage
30
4 | Forest area derivation
4.1 Segmentation
An object based image analysis was utilized
for the identification of the potentially primary
forests of Romania. The first procedure in an
object based analysis is the segmentation
process. First, the forest cover was extracted, in
order to have an area to work with.
Segmentation can be defined as the object
(polygon) creation based on neighbouring
pixels and their spectral and spatial properties1.
The multi-resolution segmentation algorithm
was used with the bands 2, 3, 4, 5, 8 and 11,
corresponding to the visible, NIR and SWIR.
Multiresolution segmentation is a “bottom-up
segmentation strategy assembling objects to
create larger objects” 2. The best values for
the scale parameter, shape and compactness
were found using an iterative process. The
used values are 80, 0.8, and respectively 0.4.
The scale parameter defines the size and
respectively the number of objects that are
being created, a higher value resulting in larger
1. Kumar Navulur, 2006
2. H E Adam et al, 2016
polygons. The value of the scale parameter
was chosen in order to obtain polygons with
an area as large as possible, but also to best
delineate the forested areas from the other
types of vegetation or loggings, and to separate
the different types of forest stands with different
species composition or age. Large polygons are
ideal for easier visual interpretations that will be
done in the following selection process of the
intact forest polygons.
Because the segmentation process sometimes
doesn’t offer very accurate results, some
polygons had to be manually reshaped. This
operation was done using the “Manual Editing
Toolbar” in eCognition.
4.2 Supervised classification
The forest cover was extracted using a
supervised classification method applied to
FIGURE 8. Segmented Sentinel 2 image
31
the polygons resulted from the segmentation
process. The nearest neighbour classification,
available in eCognition, was used. Firstly,
polygons were divided in two classes,
respectively forest and non-forest. Sample
polygons were selected for both forest and
non-forest classes. In this stage, the forest class
will contain not only old-growth forests, but also
young, planted or degraded forest stands. A
number of six raster images were used in the
classification process, respectively the bands
2, 3, 4, 5, 11 and the Normalized Difference
Vegetation Index (NDVI). NDVI was calculated
using the red and near infrared bands (4, 8),
NDVI being very useful in the classification of
the forest cover. The feature space optimization
tool, available in the eCognition software, was
used for the selection of the classification
parameters.
In some areas manual corrections were
necessary, especially in the alpine areas or
croplands where some polygons were falsely
classified as forest.
It is important to mention that only the crown
cover was extracted. Larger canopy gaps
that could potentially belong to a forest are
excluded. This is the case for the highly
fragmented coniferous forests situated at the
transition zone to alpine meadows (Figure 10).
For example, on the potential primary forest
map the total area of the Museteica forest
(Figure 11) will be 14% smaller when compared
to the total area of the forest units. The size of
the automatically identified gaps is dependent
on the values of the segmentation parameters
that were used previously.
4.3 Forest stand age classification
Young forest stands have different spectral
and textural characteristics than primary and
old-growth forests. Using vegetation indices
derived from Sentinel 2 bands, young forest
FIGURE 9. Supervised classification of the forest cover
FIGURE 10. Manual corrections subsequent to the supervised
classification of the forest cover
32
primary and old-growth forests.
Leaf Area Index (LAI) is defined as half the
developed area of photosynthetically active
elements of the vegetation per unit horizontal
ground area3. It determines the size of the
interface for exchange of energy (including
radiation) and mass between the canopy and
the atmosphere4. As demonstrated in several
studies5 the value of LAI is inversely proportional
3. http://land.copernicus.eu/global/products/lai
4. M. Weiss, F. Baret, 2016
5. R. Pokorný, S. Stojnič, 2012
FIGURE 11. Forest units from the official
management plan and potential old-growth
forest polygons of the Mușeteica forest
to the forest stand age. The highest LAI values
were observed in the youngest forest stands.
As the forest gets older, the LAI values slowly
decreases.6
Observations made on Sentinel 2 satellite
imagery also confirm that young forest stands
and plantations have very high LAI values.
The LAI was calculated with the biophysical
processor integrated in the ESAs SNAP
software.
6. R. Pokorný, S. Stojnič, 2012
FIGURE 12. Young forest stand on Bing imagery (left) and LAI raster (right)
stands can be identified to a certain degree.
Photosynthetic processes are more intense in
young forest stands, than old-growth forests.
Also the young stand’s texture appears
smoother on the satellite images because of
similar tree heights and high tree density with
small canopy gaps. This leads to a very high
reflectance in the infrared bands compared to
33
Coniferous and deciduous species have very
different leaf structures, therefore LAI will have
lower values in the case of coniferous species.
To overcome this limitation, LAI CW (leaf
water content) was used for the identification
of young coniferous forest stands. LAI CW is
more sensitive to coniferous forest stands and
has higher values in comparison to deciduous
species. Thresholds were established by
analysing the LAI and LAI CW values of the
primary and old-growth forests and young
forests charted in the field studies made of
Cumpăna and Mușeteica watersheds in August
20167.
7. http://www.greenpeace.org/romania/ro/campanii/paduri/
camp/
To further filter the remaining forest polygons, a
texture-based classification was implemented.
The roughness of the forest can be quantified
by calculating the slope of the NDVI (Figure 14).
Thus, the mean values of the slope of the NDVI
for each forest polygon were calculated, and
a threshold-based classification was run. The
threshold was set by analysing the values of
the forests charted in the field by Greenpeace
in 2016. It was considered that forest polygons
with lower slope values are specific to young
forest stands and classified as such.
Not all different aged forest stands could be
separated using the above methods, but at least
this is narrowing down the area where primary
forests could be found. Extremely accurate
assessments regarding the forest stand age
can be done in the field studies, or using highly
precise remote sensing data, like Lidar.
Because of the small number of primary and
old-growth forests taken as reference in the
establishment of the threshold that was used
in the age classification and also because
of the complex aspect, forests with similar
characteristics compared to the potentially
primary ones could have been automatically
eliminated. The LAI values and the smoothness
of a forest could differ because of the different
illumination conditions created by different
terrain expositions and inclinations. Atmospheric
FIGURE 13. Comparison between LAI and LAI CW for different tree
species
FIGURE 14. Comparison of the NDVI slope values in the case of a
young and an older forest polygon
34
5 | Logged forest and
intact forest separation
Due to the high complexity of the shape,
texture or spectral characteristics of logged
forests, a visual analysis was considered to be
more practical and more accurate compared
to automated methods. After the automatic
classification of the different aged forests
stands, using LAI and NDVI slope, the forest
polygons that were not previously classified as
young forest were analysed. Logged and intact
forest polygons were manually separated by a
visual interpretation of the satellite images.
conditions could also affect the appearance of
the forests, especially near the clouded areas.
Thus, some potential primary forest polygons
could appear truncated.
FIGURE 15. Forests with a similar appearance compared to the
potentially old-growth forests could have been automatically
eliminated in the process of age classification
FIGURE 16. Canopy disturbances West to Cumpăna watershed
For this analysis, false colour infrared (8, 4, 3)
and true colour (4, 3, 2) Sentinel 2 composites
were used in a side by side view in the
eCognition software.
Canopy disturbances of at least 200m2 were
taken into consideration. Because of the 10m
spatial resolution, a pixel will have an area of
100m2.
Low near-infrared reflectance and a high
reflectance in the visible spectrum is specific to
bare soil blue pixels in the infrared composite
from (Figure 15). Very high infrared reflectance
(pink pixels) is specific to ground vegetation
35
and shrubs or young vegetation. These two
categories of pixels were considered as
disturbances in the forest cover.
Forest polygons containing a mixture of different
kinds of reflectances specific to different aged
forests, ground vegetation and bare soil were
eliminated (Figure 17).
FIGURE 17. Forest thinning in Cumpăna Valley (Greenpeace, 2016)
Clear cuts are relatively easy to identify
especially fresh ones, where the reflectance
in the visible spectrum is high and infrared
reflectance is low (Figure 18). They appear very
bright compared to the forest canopy. Most of
the fresh clear cuts were automatically classified
as non-forest from the beginning. Below is an
example of a clear cut situated on the ridge
between Cumpăna and Cumpănița valley that
were studied in the field.
Forest polygons containing canopy disturbances
with an unnatural shape, pattern, or spatial
distribution were eliminated.
FIGURE 18. Clear cut situated on the ridge between Cumpăna and
Cumpănița Valley
36
For example, the presence of straight lines in the
canopy are a clear indicator of anthropogenic
intervention (Figure 20). These could be skidding
line marks, roads or just linear cuttings.
Some canopy gaps or anthropogenic
structures, like roads, cannot be successfully
identified from Sentinel images, the restrictive
factor being, mainly, the spatial resolution.
Therefore, the potential primary forest polygons
derived from the Sentinel images were imported
in Google Earth and checked using high
resolution satellite imagery. For example, roads
can often be confused with rivers when they are
partially covered by the canopy. It is possible to
distinguish these features with high resolution
in some cases. In addition, the 3D terrain view,
available in the Google Earth software was very
useful in the analysis.
Although Google images have higher spatial
resolution, they do not have a very good
temporal coverage. For this reason, many fresh
cuttings could have been overlooked.
A fully automated methodology would have
been ideal, but was not implemented due to
previously mentioned reasons. For the Făgăraș
Natura 2000 site, where the most potential
primary forests were identified, about 50%
of the excluded forests have been manually
mapped, the other half being automatically
classified as young forest. In other regions,
where potential primary forest polygons are
scarce, the ratio will turn in favour of manually
mapped forests.
In order to best delineate the canopy gaps that
could potentially be forest cuttings, some of the
polygons resulted from the segmentation had to
be reshaped, as shown before.
Although the scope was to identify the
manmade canopy disturbances, some naturally
occurred canopy gaps have also been mapped
and erased from the forest layer. This happened
especially in the high altitude coniferous forests,
at the contact with the alpine meadows, where
canopy gaps are large and have been classified
as non-forest in the forest cover extraction step.
FIGURE 19. Canopy disturbances FIGURE 20. Straight lines in the canopy
37
The accuracy of the canopy disturbance
analysis is dependent on several factors, for
instance the spatial resolution, the quality of
the images or the experience of the interpreter.
As in the case of the forest age classification,
accurate assessment of the canopy
disturbances can only be done in field studies.
FIGURE 21. Analyzed forest polygons visualized in the Ecognition Software
6 | Global Forest Change
Data Exclusion
In order to further exclude potential forest loss
that could have been missed when the satellite
images were visually analysed, a comparison
with a globally assessed and widely applied
dataset was realized by using global forest loss
data (Hansen et al. 2013). A total of 862,54ha
were excluded, which represent 0,28% of
the total area of primary forest polygons that
resulted from the previous section. The Global
Forest Loss Data was projected to WGS 1984
UTM Zone 35 N and polygonised. From the
resulting polygon the value ‘1’ is selected, which
indicates the loss. The resulting loss polygon
was then erased from the potential primary
forest area.
When identifying forest loss, only recent
losses can be perceived as they appear in
different pixel values. It is possible that older
disturbances were present but not captured by
this analysis.
As the minimum cell size of the Hansen data
(600 m²) is bigger than the threshold applied for
the manual extraction, an inconsistency occurs
because bigger gaps are considered natural.
On the one hand it is possible that the forest
loss is underestimated due to the threshold. On
the other hand it is also possible that the area
with natural disturbance patterns is eliminated
especially when it comes to small and diffuse
patterns.
38
7 | Exclusion of
Plantations
Furthermore it is intended to exclude planted
forests. The assessment of the plantations is
based on two assumptions:
1. The plantations consist of pure coniferous
stands.
2. Conifers that occur below 1200 m are
planted.
As in this case it was not differentiated between
northern and southern slopes, a rather
conservative altitude threshold of 1200 m was
selected. Conifers and contour lines in the
northern Fagaras Mountains part are visualized
in Figure 22. Consequently the risk of falsely
excluding primary forest areas was minimized.
The altitude is assessed with contour lines
derived from the SRTM. ‘Contour = 1200’ were
selected and polygonised. With an object based
classification the pure coniferous stands within
potential primary forests were extracted.
In the eCognition software segmentation objects
belonging to the forest class were classified
FIGURE 22. Exclusion of non-natural conifers
39
in coniferous and deciduous species. The
nearest neighbour (NN) classification method
was used, sample objects being selected by
a visual assessment of the satellite images
and also by using Corine Land Cover (CLC)
2012 data as reference. Bands 8 and 11,
corresponding to the NIR and SWIR, were used
in the classification. Coniferous species have a
very low reflectance in the near and short wave
infrared bands compared to the deciduous
species, thus, the NIR and SWIR bands are
ideal for this classification.
For the deciduous forest class, sample objects
were selected from pure decidous and also
mixed forests. For the coniferous forest class
only pure coniferous forest polygons were used
as samples, because the objective was to
identify pure coniferous forest stands.
From the coniferous within primary forest layer
the 1200 m contour polygon was erased. This
resulted in coniferous stands below 1200m,
which were then erased from the potential
primary forest area derived in the previous
section.
Plantations could be underestimated.
Coniferous plantations occurring within their
natural occurrence were not excluded. As only
pure coniferous plantations are considered,
other plantations may occur in the old growth
forest area.
Furthermore inaccuracies the classification may
exist. It is possible that not all coniferous forests
in the classification are pure stands but mixed
stands.
It is also possible that pure stands are qualified
as mixed. An exact species distinction was
not realized as it was not affordable within
the scope of this analysis, due to lacking high
resolution imagery. Furthermore, the threshold
can possibly not fit the reality and the natural
conifers were overestimated or underestimated.
The plantations can also be overestimated, as it
is possible that natural coniferous forests occur
below 1200 m.
8 | Exclusion of Roads
and Railways
A further driving factor of forest loss is
infrastructure, such as roads or railways. These
are thus erased from the potential primary forest
area. In contrast to Ibisch et al. (2016) there is
no buffer drawn around the roads and railways,
as here, the roads are solely regarded as forest
area decline. The impact of roads or railways
on the surrounding forests is not taken into
account. Roads and railways are considered to
be 10 m wide as this equals the minimum cell
size. Thus, generating false accuracy is avoided.
The data was assessed from OpenStreetMap
(OSM) and are available as line features. Small
paths (‘fclass = path’ and ‘fclass =footway’)
are not considered as their surface does not
interrupt the canopy cover in a way that that it
can be observed from satellite images. The non-
vehicle roads were selected and erased from
the roads. In order to reflect the surface a 5 m
buffer for roads and railways was performed
on both sides of the lines. The buffer polygons
were then erased from the potential primary
forest area. In the end a ‘Multipart to Singlepart’
operation is performed and polygons smaller
40
than 200 m², resulting as leftovers from previous
erasing processes, are erased.
OSM data is generated by its several
contributors. Therefore a high accuracy is not
guaranteed. Even though studies reveal that
OSM data reaches high accuracy compared
to proprietary datasets, in Great Britain and
Germany (Haklay 2010; Neis et al. 2012)
these results cannot be directly transferred to
Romania, as the number of active users differ
notably.
While Germany had from January to August
2017 420-600 daily active members, the
number accounts to 13-31 for Romania in
the same time period (Neis, El Loco 2017).
Furthermore, Hecht, Stephens (2014) identify
the data accuracy of Volunteered Geographical
Information (VGI) to be decreasing with distance
to urban areas. This may lead to lower accuracy
in the country of Romania with focus on rural
areas, than assessed by Haklay (2010) or Neis
et al. (2012). Data accuracy is not only related
to positional but also to thematic inaccuracies
(Capineri et al., 2016), e.g. wrongly allocated
classes. Consequently it cannot be ruled out
that the classes footway and path, which were
not erased, are logging roads. A clearly defined
FIGURE 23. Example of roads intersecting potential primary forest polygons near
Lake Vidraru
41
class called logging road does not exist. For the
class path the use of cycles or horses cannot
be excluded as it is described as a general path
without the use of vehicles (OpenStreetMap
Wiki 2017). To exclude major errors a visual
verification with satellite images was performed.
OSM data was chosen as it is open access
data. Assessing proprietary data was not
affordable within the study.
Some potential primary forest polygons could
be crossed by roads and skidding trails that
were not successfully identified on the satellite
images and were not available in the OSM
dataset yet. Analysing satellite images captured
in the cold season could represent one solution,
but this would work only for the deciduous
forests.
Using a detailed road dataset provided by
authorities for the Vidraru forest district, roads
that intersecti potential primary forest polygons
and could not be successfully mapped can
be visualized. From a total of 443 km of roads
existing in the Vidraru area, 17 km (4%) of roads
are intersecting potential primary polygons.
42
9 | Size evaluation
In the end, the potential primary forest polygons
are classified into 4 classes:
0 - 20 ha does not correspond to official
criteria for virgin and quasi virgin forests
20.1 - 0 ha: corresponds to the official
criteria for virgin forests
30.0 - 50 ha corresponds to official criteria
for quasi-virgin forests
> 50 ha corresponds to the criteria for
primary forests applied in the PinMatra (
Biriș, Veen 2005)
10 | Connectivity
Assessment
Well connected forest patches have a higher
ecological functionality (e.g. genetic exchange)
than isolated patches. The connectivity
assessment represents a first step towards a
priority setting for the protection. Nevertheless
it cannot be directly concluded that well
connected patches have a high conservation
priority. Small and isolated patches require the
most urgent protection as they are the most
exposed to external threats.
Connectivity was assessed with Thiessen
Connectivity into all directions (according
to Ibisch et al.(2016)). It is described by the
proportion of a patch to its nearest surrounding
area (Thiessen area).”Each Thiessen polygon
defines an area of influence around its sample
point, so that any location inside the polygon is
closer to that point than any of the other sample
points.” (ESRI 2016)
To include forest patches that were cut by
the study area and thus are outside, a buffer
of 2000 m was applied to the study area
outline. Thiessen area was calculated with the
“Euclidean allocation” tool. The Euclidean output
was then clipped to the study area buffer and
not charted areas were erased. Thiessen Net
area was calculated by erasing OGF 5 from
the Thiessen area. In the end the proportion
of the Thiessen net area and forest patch area
was calculated. For display purposes an inside
buffer of 2000 m was performed in the end. The
resulting values were classified by quantiles into
5 classes (blue indicating well connected, red
indicating poor connected).
43
Publication bibliography
Biriș, Iovu-Adrian; Veen, Peter (2005): Virgin forests in Romania. Inventory
and strategy for sustainable management and protection of virgin forests
in Romania. In
Document ICAS, Bucharest.
ESRI (2016): Thiessen polygons | Definition - Esri Support GIS Dictionary.
Available online at http://support.esri.com/en/other-resources/gis-
dictionary/term/Thiessen%20polygons, updated on 4/28/2016, checked
on 10/9/2017.
Haklay, Mordechai (2010): How Good is Volunteered Geographical
Information? A Comparative Study of OpenStreetMap and Ordnance
Survey Datasets. In
Environ Plann B Plann Des
37 (4), pp. 682–703. DOI:
10.1068/b35097.
Hansen, M. C.; Potapov, P. V.; Moore, R.; Hancher, M.; Turubanova,
S. A.; Tyukavina, A. et al. (2013): High-resolution global maps of 21st-
century forest cover change. In
Science (New York, N.Y.)
342 (6160), pp.
850–853. DOI: 10.1126/science.1244693.
H E Adam et al 2016 IOP Conf. Ser.: Earth Environ. Sci. 37 012061
Hecht, Brent; Stephens, Monica (2014): A Tale of Cities. Urban Biases in
Volunteered Geographic Information. Available online at https://www.aaai.
org/ocs/index.php/ICWSM/ICWSM14/paper/view/8114.
Holtmeier, Friedrich-Karl; Broll, Gabriele (2017): Treelines—Approaches at
Different Scales. In
Sustainability 9 (5)
, p. 808. DOI: 10.3390/su9050808.
Ibisch, Pierre L.; Hoffmann, Monika T.; Kreft, Stefan; Pe’er, Guy; Kati,
Vassiliki; Biber-Freudenberger, Lisa et al. (2016): A global map of roadless
areas and their conservation status. In
Science
354 (6318), pp. 1423–
1427. DOI: 10.1126/science.aaf7166.
Capineri, Cristina; Haklay, Muki; Huang, Haosheng; Antoniou, Vyron;
Kettunen, Juhani; Ostermann, Frank; Purves, Ross (Eds.) (2016):
European handbook of crowdsourced geographic information. London,
London: Ubiquity Press.
Navulur, Kumar (2006): Multispectral Image Analysis Using the Object-
Oriented Paradigm
Neis, Pascal; El Loco, Santos (2017): OSMstats - Statistics of the
free wiki world map. Available online at http://osmstats.neis-one.
org/?item=countries&country=Romania, updated on 9/2017, checked on
9/1/2017.
Neis, Pascal; Zielstra, Dennis; Zipf, Alexander (2012): The Street Network
Evolution of Crowdsourced Maps. OpenStreetMap in Germany 2007–
2011. In
Future Internet
4 (4), pp. 1–21. DOI: 10.3390/fi4010001.
OpenStreetMap Wiki (2017): Key:highway - OpenStreetMap
Wiki. Available online at http://wiki.openstreetmap.org/w/index.
php?title=Key:highway&oldid=1498196, checked on 20.8.17.
Mueller-Wilm U. (2016): Sen2Cor Configuration and User Manual
Pokorný, R.; Stojnič, S.; (2012): Leaf area index of Norway spruce stands
in relation to age and defoliation, Beskydy, 5 (2): 173–180
Richter R., Louis L., Uwe Müller-Wilm (2012): Sentinel-2 MSI – Level 2A
Products Algorithm Theoretical Basis Document
44
Study Area
LAI and LAI CW
derivation
NDVI calculation
Multiresolution
segmentation
Composite bands
Sentinel 2 L1C
SRTM DEM
Google Imagery
Radiometric correction
(Sen2Cor
Sentinel 2 L2A (10m
and 20m bands) Resampling 10m Sentinel 2 layer stack
10m
ANNEX 1. Potential primary forest polygons
derivation algorithm
LEGEND
Input data
Result Process
45
NDVI raster
Slope calculation NDVI slope raster
Supervised
classification (object
based)
Forest cover polygons
Segmentation objects
(polygons)
Manual separation
of logged and intact
forest polygons
Sentinel derived PVF
polygons
Logged forest
Erase logged forests
Old Growth forests
without gaps
Visual validation
of Sentinel derived
PVF polygons and
identification of
remaining logged
forests
Threshold value
classification (mean/
stdev)
Old forest polygons
True color composite
False color Infrared
composite
LAI
LAI CW
46
OSM Railways Buffer Distance = 1 m
OSM Roads
Global Forest Loss
v.1.3
Old Growth forests
without gaps
Coniferous
Classification
Clip conifers to
identified area
Conifers within
identified Area
Erase Contour from
conifers
SRTM DEM
Erase Global Forest
Loss
Contour Countour Lines SelectContour = 1200 1200 m Contour line
Select fclass= path
OR fclass= footway Non vehicle roads
Erase non vehicle
Roads from OSM
Roads
Roads with vehicle
use
Buffer Distance =
1,5 m
Old Gowth Forests
without Global forest
loss
ANNEX 2. Filtering and prioritization of the
potential primary forest polygons
LEGEND
Input data
Result Process
47
Railway Buffer 1 m
Conifers below 1200
m
Erase conifers below
1200 m
Old Growth Forest
without Plantations
Exclude OSM Roads
and Railways
Select Polygons >
200 m²
Old Growth Forest
without Roads and
Railways
Road Buffer 1,5 m
Potential Primary
Forests
48
Old Growth Forests
Study Area
Erase Thiessen Net
Polygon Updated Thiessen Net
Add field:
proportiontopatch
Add field:
For_patch_size
Uncharted Areas
Polygon to Raster
Buffer with distance =
2000 m
Study Area
with Buffer Clip Euclidean Allocation
Polygon clipped
OGF Raster Int OGF Raster as Integer Euclidean Allocation
Clip
Feature
Erase
Feature
ANNEX 3. Process flowchart of the
connectivity evaluation
49
Calculate Field:
For_patch_size =
ThiessenArea –
ThiessenNetArea
Calculate Field:
Proportiontopatch=
ThiessenNetArea/
For_patch_sizeArea
Thiessen Area Add Geometry
Attributes
Updated Euclidean
Allocation
Erase
Raster to Polygon Euclidean Allocation
as Polygon
Dissolve
(field = gridcode)
Corrected Euclidean
Allocation
Euclidean Allocation
Raster
Connectivity into all
directions
50
Glossary
For this study the following definitions have
been used:
DEFORESTATION
The conversion of forest to other land use or the
permanent reduction of the tree canopy cover
below the minimum 10 percent threshold1.
FOREST DEGRADATION
Changes within the forest which negatively
affect the structure or function of the stand or
site, and thereby lower the capacity to sup-ply
products and/or services2.
ILLEGAL LOGGING
Timber is harvested in violation of national laws3.
CANOPY
The forest cover of branches and foliage formed
by tree crowns4.
1. FRA (2015) - Forest Resources Assessment Working Paper
180, FAO, 2015
2. FAO (2001) – Global Forest Resources Assessment 2000
3. European Commission (2003) - The European Union Action
Plan for Forest Law Enforcement, Governance and Trade (EU
FLEGT)
4. FAO (2001) – Global Forest Resources Assessment 2010
VIRGIN FOREST / PRIMARY FOREST5
A primary forest is a forest that has never been
logged and has developed following natural
disturbances and under natural processes,
regardless of its age. It is referred to “direct
human disturbance” as the intentional clearing
of forest by any means (including fire) to
manage or alter them for human use. Also
included as primary, are forests that are used
inconsequentially by indigenous and local
communities living traditional lifestyles relevant
for the conservation and sustainable use of
biological diversity. In much of Europe, primary
forest has a different connotation and refers
to an area of forest land which has probably
been continuously wooded at least throughout
historical times (e.g., the last thousand years). It
has not been completely cleared or converted
to another land use for any period of time.
However traditional human disturbances such
5. Some key characteristics of primary forests are: - they
show natural forest dynamics, such as natural tree species
composition, occurrence of dead wood, natural age structure
and natural regeneration processes; - the area is large enough
to maintain its natural characteristics; - there has been no
known significant human intervention or the last significant
human intervention was long enough ago to have allowed the
natural species composition and processes to have become
reestablished.
as patch felling for shifting cultivation, coppicing,
burning and also, more recently, selective/partial
logging may have occurred, as well as natural
disturbances. The present cover is normally
relatively close to the natural composition and
has arisen (predominantly) through natural
regeneration, but planted stands can also be
found. However, the suggested definition above
would include other forests, such as secondary
forests.6
Naturally regenerated forest of native species,
where there are no clearly visible indications of
human activities and the ecological processes
are not significantly disturbed7.
Virgin forest is a natural woodland where the
tree and shrub species are in various stages
of their life cycle (seedlings, young growth,
advanced growth, maturity and old growth) and
as dead wood (standing and laying) in various
stages of decay, thus resulting in more or less
complex vertical and horizontal structures
6. CBD - https://www.cbd.int/forest/definitions.shtml
7. FRA (2015) - Forest Resources Assessment Working Paper
180, FAO, 2015
51
as a product of a dynamic process, which
enables the natural forest community to exist
continuously and without limit in time.
In virgin forests the dynamics inherent to
living systems are connected to ecological
properties (including longevity) of the dominant
tree species, impact of other organisms (e.g.
outbreak of insects) and to the impact of abiotic
factors related to substrate, climate and to the
complex of topography and water table (e.g.
wind, snow, flooding). Part of this dynamics
is the temporary occurrence of gaps or larger
tree-less stages.
Virgin forests differ within the given
phytogeographic zone, forming specific types of
forest communities with characteristic species
composition, spatial structure, dynamics and
overall diversity due to site conditions related
to the position above sea level and topography,
macroclimate, and nutrient and water availability.
Virgin forests reflect herewith the natural unity
of forest community and abiotic conditions,
fully rooted in their millennia-long continuous
Holocene development8.
8. Biriș și Veen (2005)
OLD-GROWTH FORESTS / PRIMEVAL
FOREST
A former virgin forest which has suffered in the
meantime some anthropogenic influences but
without experiencing a significant degradation of
structure or function.
ANCIENT FOREST
A forest old enough to include a natural diversity
of species and ages, trees that reached
physiological longevity, and other characteristics
that certify the naturalness (standing dead
wood, lying dead wood in various decaying
stages, certain species of plants and animals,
which indicate the health status, and the
naturalness of the forest), from which there is
the possibility of some isolated tree extractions,
but without modifying the forests composition
and structure. This concept especially
emphasizes the continued existence of forest
over time and does not necessarily refer to the
complete absence of human activity.
52
... Therefore, the distribution of these forest patches is much more similar to the location of natural-virgin and natural landscapes returned by the EMI and ECON models, than in case of the ones depicted by the EOC. Ibisch et al. (2017) developed a map whith the potential primary forests of Romania by analyzing high resolution satellite imagery, whereas Schickhofer and Schwarz (2019) assessed the potential primary and old-growth forest areas map of Romania through identifiyng structures of close to nature forest habitats on satellite maps. The finding corresponds with our maps concerning assessing naturalness through the EMI, since in our models, the largest areas covered by natural virgin ecosystems were also located in the southern Carpathians, and mainly in the Fȃgȃraș Mountains. ...
Article
The landscape naturalness may be defined and analysed by various concepts and methods attempting to encapsulate as much as possible the degree of natural conditions over a given territory. The Machado Index (MI) was developed by the Spanish biologist Antonio Machado and uses a qualitative approach to naturalness, being characterized by its versatile application throughout different environments. The outcome of the expert-based evaluation is a score that corresponds to various degrees of naturalness. This research aims to turn the MI into a semi-objective tool, introducing land cover and the ‘neighbouring to natural’ criteria as quantitative components. The MI was applied for assessing the landscape naturalness over Romania, as a case-study, and the Expert Opinion Classification (EOC) method was performed in order to identify the primary benefits and limitations of the MI. Further, the study uses the Principle of Naturalness Spatial Gradient (PNSG) as the basis for conceiving a new approach, named the Edge Contrast method (ECON). The assessment of naturalness consists in the usefulness of a previous Landscape Ecology metric, the Edge Contrast Index (ECI). We finally propose a third method, encompassing the advantages of the prior two, named the Enhanced Machado Index (EMI). The main result of this study is an enhanced method which can be used for assessing the degree of naturalness in a semi-objective manner. A set of three examples taken from different areas of Romania are assessed in a comparative analysis in order to highlight the differences between the EOC, ECON and EMI methods. The EMI results are validated by comparison with different databases.
... În 2017, Greenpeace a publicat un studiu asupra pădurilor potential virgine din România, studiu în cadrul căruia au fost identificate 296.000 ha (Ibisch et al., 2017b). Studiul s-a bazat pe evaluarea unor date ce pot fi accesate liber. ...
Article
Full-text available
DOI: 10.4316/bf.2021.009 Raportul „Pădurile virgine în inima Europei” (2021), autori; Rainer Luick, Albert Reif, Erika Schneider, Manfred Grossmann și Ecaterina Fodor este o analiză detaliată a importanței, situației actuale și viitorului pădurilor seculare din România. Autorii subliniază marea lor simpatie față de România și abordează importanța ultimelor păduri seculare din Carpații românești pentru moștenirea naturală a Europei. De asemenea, sunt descrise detaliat neputința și lipsa de interes din partea instituțiilor statului în protejarea pădurilor seculare și virgine. Investigațiile asupra corupției profunde și imixtiunii criminale în sectoarele silviculturii și exploatării lemnului scot la iveală interferențe șocante între politică, administrație și corporații. Existând informații asupra tăierilor la scară mare în arii protejate, studiul este centrat pe întrebarea de ce instituțiile UE au avut reacții puține în decursul multor ani la aceste probleme. Autorii cer ca protecția ultimelor păduri virgine din centrul Europei să devină o preocupare a organismelor pan-europene. Acest demers este formulat ca un element cheie pentru Strategia pentru Biodiversitate a Europei până în 2030. România ar deveni astfel un test de litmus asupra șanselor de reușită ale strategiei.
... In 2017, Greenpeace published a study it had commissioned on potential virgin forest areas in Romania, which identified some 296,000 hectares (Ibisch et al. 2017b). The study is based on an evaluation of current and freely available data records: The authors emphasised that they saw the results as search areas with a high probability of virgin and very near-natural old-growth forests (corresponding to the Romanian category of quasi-virgin forests) within the delimited polygons. ...
Book
Full-text available
The report "Virgin forests at the heart of Europe“ by Rainer Luick, Albert Reif, Erika Schneider, Manfred Grossmann & Ecaterina Fodor (2021) is a comprehensive analysis of the importance, situation and future of virgin and old-growth forests in Romania. The authors emphasize their great sympathy with the country and address on many aspects the importance of the remaining virgin forests in the Romanian Carpathians for the European natural heritage. The widespread extensive powerlessness and the lack of interest of state institutions to protect virgin and old forests are also meticulously portrayed. The investigations on a profound corruption and a criminal mixture in the forestry and timber sector that has developed in the interplay of politics, administration and corporate actors is shocking. Knowing about the large-scale loggings in protected areas, the study also picks out as a central theme and question why the EU institutions have shown very few reactions for many years. The authors demand that it must be a pan-European concern to protect the last coherent virgin forests in the more central parts of Europe. This is also formulated as a key element in the EU's Biodiversity Strategy 2030. Romania will be a litmus test as to whether this will succeed.
... In 2017, Greenpeace published a study it had commissioned on potential virgin forest areas in Romania, which identified some 296,000 hectares (Ibisch et al. 2017b). The study is based on an evaluation of current and freely available data records: The authors emphasised that they saw the results as search areas with a high probability of virgin and very near-natural old-growth forests (corresponding to the Romanian category of quasi-virgin forests) within the delimited polygons. ...
Book
Full-text available
The report (virgin forests at the heart of Europe) provides an overview of the distribution and situation of the last remaining large-scale virgin forests in Central Europe, with a particular focus on Romania. Most people usually associate images of destruction of forests with tropical rain forests. But this also takes place right here on our doorsteps. We in Europe share a global responsibility to protect our unique, irreplaceable natural heritage. These Carpathian forests are some of the last remaining wildernesses, and a precious archive of biodiversity, history, of impressive images and beauty. As consumers, processors and sellers of timber and wood-based products we all have responsibility to stop the pressures placed on these forests, and have the duty to protect this natural heritage for future generations.
Article
Full-text available
Scales in treeline research depend on the objectives and must match the underlying natural processes. Factors and processes at one scale may not be as important at another scale. In the global view, the number of factors influencing climatic treeline position can be reduced to the effects of heat deficiency. Emphasis, however, should be laid on differentiation of the treeline by their regionally and locally varying physiognomy, diversity, spatial and temporal features, and heterogeneity. An assessment of the relative importance of the factors shaping regional/local treeline physiognomy, spatial patterns, and dynamics should have priority. This can be achieved only by syndisciplinary research. Such studies are indispensable for assessing treeline response to climate change at the regional and landscape scales.
Article
Full-text available
Roads fragment landscapes and trigger human colonization and degradation of ecosystems, to the detriment of biodiversity and ecosystem functions. The planet’s remaining large and ecologically important tracts of roadless areas sustain key refugia for biodiversity and provide globally relevant ecosystem services. Applying a 1-kilometer buffer to all roads, we present a global map of roadless areas and an assessment of their status, quality, and extent of coverage by protected areas. About 80% of Earth’s terrestrial surface remains roadless, but this area is fragmented into ~600,000 patches, more than half of which are <1 square kilometer and only 7% of which are larger than 100 square kilometers. Global protection of ecologically valuable roadless areas is inadequate. International recognition and protection of roadless areas is urgently needed to halt their continued loss. Read more on www.roadless.online ...
Article
Full-text available
Quantification of global forest change has been lacking despite the recognized importance of forest ecosystem services. In this study, Earth observation satellite data were used to map global forest loss (2.3 million square kilometers) and gain (0.8 million square kilometers) from 2000 to 2012 at a spatial resolution of 30 meters. The tropics were the only climate domain to exhibit a trend, with forest loss increasing by 2101 square kilometers per year. Brazil's well-documented reduction in deforestation was offset by increasing forest loss in Indonesia, Malaysia, Paraguay, Bolivia, Zambia, Angola, and elsewhere. Intensive forestry practiced within subtropical forests resulted in the highest rates of forest change globally. Boreal forest loss due largely to fire and forestry was second to that in the tropics in absolute and proportional terms. These results depict a globally consistent and locally relevant record of forest change.
Article
Full-text available
The OpenStreetMap (OSM) project is a prime example in the field of Volunteered Geographic Information (VGI). Worldwide, several hundred thousand people are currently contributing information to the “free” geodatabase. However, the data contributions show a geographically heterogeneous pattern around the globe. Germany counts as one of the most active countries in OSM; thus, the German street network has undergone an extensive development in recent years. The question that remains is this: How does the street network perform in a relative comparison with a commercial dataset? By means of a variety of studies, we show that the difference between the OSM street network for car navigation in Germany and a comparable proprietary dataset was only 9% in June 2011. The results of our analysis regarding the entire street network showed that OSM even exceeds the information provided by the proprietary dataset by 27%. Further analyses show on what scale errors can be reckoned with in the topology of the street network, and the completeness of turn restrictions and street name information. In addition to the analyses conducted over the past few years, projections have additionally been made about the point in time by which the OSM dataset for Germany can be considered “complete” in relative comparison to a commercial dataset.
Book
Bringing a fresh new perspective to remote sensing, object-based image analysis is a paradigm shift from the traditional pixel-based approach. Featuring various practical examples to provide understanding of this new modus operandi, Multispectral Image Analysis Using the Object-Oriented Paradigm reviews the current image analysis methods and demonstrates advantages to improve information extraction from imagery. This reference describes traditional image analysis techniques, introduces object-oriented technology, and discusses the benefits of object-based versus pixel-based classification. It examines the creation of object primitives using image segmentation approaches and the use of various techniques for object classification. The author covers image enhancement methods, how to use ancillary data to constrain image segmentation, and concepts of semantic grouping of objects. He concludes by addressing accuracy assessment approaches. The accompanying two CD-ROMs present sample data that enable the use of different approaches to problem solving. Integrating remote sensing techniques and GIS analysis, Multispectral Image Analysis Using the Object-Oriented Paradigm distills new tools to extract information from remotely sensed data. © 2007 by Taylor & Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business.
Article
Geotagged tweets, Foursquare check-ins and other forms of volunteered geographic information (VGI) play a critical role in numerous studies and a large range of intelligent technologies. We show that three of the most commonly used sources of VGI - Twitter, Flickr, and Foursquare - are biased towards urban perspectives at the expense of rural ones. Utilizing a geostatistics-based approach, we demonstrate that, on a per capita basis, these important VGI datasets have more users, more information, and higher quality information within metropolitan areas than outside of them. VGI is a subset of user-generated content (UGC) and we discuss how our results suggest that urban biases might exist in non-geographically referenced UGC as well. Finally, because Foursquare is exclusively made up of VGI, we argue that Foursquare (and possibly other location-based social networks) has fundamentally failed to appeal to rural populations. Copyright © 2014, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
Article
Within the framework of Web 2.0 mapping applications, the most striking example of a geographical application is the OpenStreetMap (OSM) project. OSM aims to create a free digital map of the world and is implemented through the engagement of participants in a mode similar to software development in Open Source projects. The information is collected by many participants, collated on a central database, and distributed in multiple digital formats through the World Wide Web. This type of information was termed ‘Volunteered Geographical Information’ (VGI) by Goodchild, 2007. However, to date there has been no systematic analysis of the quality of VGI. This study aims to fill this gap by analysing OSM information. The examination focuses on analysis of its quality through a comparison with Ordnance Survey (OS) datasets. The analysis focuses on London and England, since OSM started in London in August 2004 and therefore the study of these geographies provides the best understanding of the achievements and difficulties of VGI. The analysis shows that OSM information can be fairly accurate: on average within about 6 m of the position recorded by the OS, and with approximately 80% overlap of motorway objects between the two datasets. In the space of four years, OSM has captured about 29% of the area of England, of which approximately 24% are digitised lines without a complete set of attributes. The paper concludes with a discussion of the implications of the findings to the study of VGI as well as suggesting future research directions.