Article

Where are Europe’s last primary forests?

Abstract

Aim Primary forests have high conservation value but are rare in Europe due to historic land use. Yet many primary forest patches remain unmapped, and it is unclear to what extent they are effectively protected. Our aim was to (1) compile the most comprehensive European‐scale map of currently known primary forests, (2) analyse the spatial determinants characterizing their location and (3) locate areas where so far unmapped primary forests likely occur. Location Europe. Methods We aggregated data from a literature review, online questionnaires and 32 datasets of primary forests. We used boosted regression trees to explore which biophysical, socio‐economic and forest‐related variables explain the current distribution of primary forests. Finally, we predicted and mapped the relative likelihood of primary forest occurrence at a 1‐km resolution across Europe. Results Data on primary forests were frequently incomplete or inconsistent among countries. Known primary forests covered 1.4 Mha in 32 countries (0.7% of Europe’s forest area). Most of these forests were protected (89%), but only 46% of them strictly. Primary forests mostly occurred in mountain and boreal areas and were unevenly distributed across countries, biogeographical regions and forest types. Unmapped primary forests likely occur in the least accessible and populated areas, where forests cover a greater share of land, but wood demand historically has been low. Main conclusions Despite their outstanding conservation value, primary forests are rare and their current distribution is the result of centuries of land use and forest management. The conservation outlook for primary forests is uncertain as many are not strictly protected and most are small and fragmented, making them prone to extinction debt and human disturbance. Predicting where unmapped primary forests likely occur could guide conservation efforts, especially in Eastern Europe where large areas of primary forest still exist but are being lost at an alarming pace.
Diversit y and Distributions. 2 0 18 ; 1–14 .  wileyonlinelibrary.com/journal/ddi  
|
 1
© 2018 John Wi ley & Sons Ltd
Received:20September2017 
|
  Accepted:19April2018
DOI : 10.1111 /ddi .127 78
BIODIVERSITY REVIEW
Where are Europe’s last primary forests?
Francesco Maria Sabatini1| Sabina Burrascano2| William S. Keeton3|
Christian Levers1| Marcus Lindner4| Florian Pötzschner1| Pieter Johannes Verkerk5|
Jürgen Bauhus6| Erik Buchwald7| Oleh Chaskovsky8| Nicolas Debaive9|
Ferenc Horváth10| Matteo Garbarino11| Nikolaos Grigoriadis12| Fabio Lombardi13|
Inês Marques Duarte14 | Peter Meyer15| Rein Midteng16| Stjepan Mikac17|
Martin Mikoláš18| Renzo Motta11| Gintautas Mozgeris19| Leónia Nunes14,20 |
Momchil Panayotov21| Peter Ódor10 | Alejandro Ruete22 | Bojan Simovski23|
Jonas Stillhard24| Miroslav Svoboda18| Jerzy Szwagrzyk25| Olli-Pekka Tikkanen26|
Roman Volosyanchuk27| Tomas Vrska28| Tzvetan Zlatanov29| Tobias Kuemmerle1
1Geography Department, Humboldt-Universität zu Berlin, Berlin, Germany
2Department of Environmental Biology, Sapienza, Università di Roma, Rome, Italy
3RubensteinSchoolofEnvironmentandNaturalResources,Universit yofVermont ,Burlington,Vermont,USA
4European Forest Institute, Bonn, Germany
5European Forest Institute, Joensuu, Finland
6Faculty of Environment and Natural Re sources, University of Freiburg , Freiburg, Germany
7TheDanishNatureA gency,Randbøl,Denmark
8InstituteofForestManagement,NationalUniversityofForestryandWoodTechnology,Lviv,Ukraine
9Réser ves Naturelles de France, Dijon Cede x, France
10InstituteofEcologyandBotany,MTACentreforEcologicalResearch,Vácrátót,Hungar y
11Depar tmentofAgricultural,ForestandFoodSciences(DISAFA),UniversityofTorino,Grugliasco,Italy
12ForestResearchInstituteofThessaloniki,Vassilika,Greece
13Depar tmentofAgraria,MediterraneanUniversit yofReggioCalabria,ReggioCalabria,Italy
14CentreforAppliedEcology“ProfessorBaetaNeves”(CEABN),InBio,SchoolofAgriculture,UniversityofLisbon,Lisbon,Portugal
15Northwest German Forest Research Institute, Göttingen, Germany
16AsplanViak,Sandvika,Nor way
17Facult y of Forestry, Depa rtme nt of Fores t Ecolog y and Silv iculture, University of Zagreb, Zagreb, Croatia
18FacultyofForestryandWoodSciences,CzechUniversityofLifeSciencesPrague,Praha-Suchdol,CzechRepublic
19InstituteofForestManagementandWoodScience,AleksandrasStulginskisUniversity,Akademija,Lithuania
20CITABCentreoftheResearchandTechnologyofAgro-EnvironmentalandBiologicalScience,UniversityofTrás-os-MontesandAltoDouro,VilaReal,Portugal
21Dendro logy Depart ment, University of Fore stry, Sofia, Bul garia
22GreenswayAB,Uppsala,Sweden
23FacultyofForestryinSkopje,DepartmentofBotanyandDendrology,Ss.CyrilandMethodiusUniversityinSkopje,Skopje,RepublicofMacedonia
24WSL Swiss Fe deral I nstitu te of Fores t, Snow and Land scape Research, Birmensdor f, Switzerland
25FacultyofForestry,InstituteofForestEcologyandSilviculture,UniversityofAgriculture,Kraków,Poland
26School of Forest S ciences, University of E astern Finlan d, Joens uu, Finland
27WWFDanube-CarpathianProgrammeUkraine,Lviv,Ukraine
2 
|
   SABATINI eT Al.
28Forest Ecology D epar tment , Silva Tarouc a Research Institute, Brno, Czech Republic
29InstituteofBiodiversit yandEcosystemResearch,BulgarianAcademyofSciences,Sofia,Bulgaria
Correspondence
Francesco Maria Sabatini, Geography
Department, Humboldt-Universität zu
Berlin, Berlin, Germany.
Email: sabatinf@hu-berlin.de
Present addresses
Francesco Maria Sabatini , Instit ut für
Biologie - Martin-Luther-Universität Halle-
Wittenberg, Halle, G ermany and German
Centre for Integrative Biodiversity Research
(iDiv)-Halle-Jena-Leipzig,Leipzig,Germany.
Funding information
H2020MarieSkłodowska-CurieActions,
Grant/AwardNumber:658876
Editor: Fr anz Essl
Abstract
Aim:PrimaryforestshavehighconservationvaluebutarerareinEuropeduetohis-
toric land use. Yet many primary forest patches remain unmapped, and it is unclear to
whatextenttheyareeffectivelyprotected.Ouraimwasto(1)compilethemostcom-
prehensiveEuropean-scalemapofcurrentlyknownprimaryforests,(2) analysethe
spatial determinantscharacterizing theirlocationand (3) locate areaswhere sofar
unmappedprimaryforestslikelyoccur.
Location: Europe.
Methods: We aggregated data from a literature review, online questionnaires and 32
datasets of primary forests. We used boosted regression trees to explore which bio-
physical, socio- economic and forest- related variables explain the current distribution
ofprimaryforests. Finally,wepredictedand mapped therelativelikelihood ofpri-
maryforestoccurrenceata1-kmresolutionacrossEurope.
Results: Data on primary forests were frequently incomplete or inconsistent among
countries.Knownprimaryforestscovered1.4Mhain32countries(0.7%ofEurope’s
forest are a). Most of these fo rests were protecte d (89%), but only 46% of them
strictly.Primaryforestsmostlyoccurredinmountainandborealareasandwereun-
evenly distributed across countries, biogeographical regions and forest types.
Unmapped primaryforestslikelyoccurintheleastaccessibleandpopulatedareas,
where forests cover a greater share of land, but wood demand historically has been
low.
Main conclusions: Despite their outstanding conservation value, primary forests are
rare and their current distribution is the result of centuries of land use and forest
management.Theconservationoutlookforprimaryforestsisuncertainasmanyare
notstrictlyprotectedandmostaresmallandfragmented,makingthempronetoex-
tinctiondebt andhuman disturbance.Predictingwhereunmappedprimary forests
likely occur coul d guide conservation efforts, espe cially in Eastern Europe w here
large areas of primary forest still exist but are being lost at an alarming pace.
KEYWORDS
boosted regression trees, forest naturalness, land-use change, old-growth forest, primary
forest, spatial determinants, sustainable forest management, virgin forest
1 | INTRODUCTION
Primaryforestsare becomingrareasforestlandgloballyiscleared
foragricultureorputunderactivemanagement(Mackeyetal.,2015;
Potapov eta l., 2017). Primar y forests, acc ording to the Food an d
AgriculturalOrganization(FAO),refertonaturallyregeneratedfor-
ests of native species where there are no clearly visible indications of
human ac tivities an d the ecologic al processes a re not signific antly dis-
turbed(FAO,2015).Giventheirirreplaceabilityanduniquequalities,
protectingprimaryforestsisaglobalconcern(Mackeyetal.,2015).
Notonlyareprimaryforestscherishedfortheirwildnature(Navarro
&Pereira,2012),andrepresentasocialperceptionofuntouchedna-
ture (Schn itzler, 2014), but they are al so ecologic ally impor tant in
regions wh ere forests a re highly fragm ented (Vandeker khove, De
Keersmaeker,Menke,Meyer,&Verschelde,2009).Primaryforests,
for instance, ser ve as refuges or sources of propagules for rare or
endangered species, especially for forest species sensitive to human
disturbance (Paillet etal., 2015; Peterken, 1996). Furthermore,
    
|
 3
SABATINI eT A l.
primar y forests serve as a model for understanding natural distur-
bance andsuccessional dynamics (Král, McMahon,Janík,Adam, &
Vrška,2014;Kuuluvainen&Aakala,2011;Leibundgut,1959),espe-
cially in the face of climate change, and provide baselines for the
delivery of ecosystem services under unmanaged conditions, includ-
ingcarbonstocksandsequestration(Burrascano,Keeton,Sabatini,
& Blasi, 2013; Ha rmon, Ferrel l, & Franklin, 1990). F inally, primary
forests help us to evaluate human impacts on forest ecosystems and
to understand the potential and limitations of close- to- nature for-
estmanagement(Bauhus,Puettmann,&Messier,2009;EEA,2014;
Kuuluvainen&Aakala,2011).
In Europe, as in other human- dominated regions, historical de-
forestation and forest exploitation came close to eliminating primary
forest s (Kaplan, K rumhardt , & Zimmerman n, 2009; Pot apov etal.,
2017). Europe’s for ests are now mainl y composed of semin atural
forests,whileforestsundisturbedbymanaccountforonly4%ofthe
total(FORESTEUROPE,2015).Eventhislittleshareofundisturbed
forest is heavily fragmented as virtually no intact forest landscapes
>500km2 exist outside European Russia and boreal northern Europe
(Potapov etal., 2017). Finally, although some Eastern European
countries may still contain relatively large areas of primary forests
(Frank etal., 2007; Kulakowski etal., 2017), these remain often
unmapped and unprotected and are being lost at an alarming rate
(Chylarecki&Selva,2016;Knornetal.,2013;Mikolášetal.,2017).
Seminatural forests cannot be easily restored to a primary status
(Ford&Keeton,2017).Intheabsenceofanthropogenicdisturbance,
forest s slowly recover the natural disturbance dynamics and develop
thosestructuralfeatures(e.g.,deadwood,largelive treesandpres-
ence of can opy openings of va rious sizes) that ar e typical fo r the
old-growth phases of primary forests, although this process takes
decades(Burrascanoetal.,2013;Pailletetal.,2015;Vandekerkhove
etal., 20 09). The ongoing p rocess of agricu ltural intens ification i n
productive areas, which co- occurs with deintensification or even
abandonment of marginal areas, may offer important conserva-
tion opportunities (Jepsen etal., 2015; Navarro & Pereira, 2012;
Schnitzler,2014). In many Western European countries, satisfying
wood demands increasingly relies on imports, while forests located
in remote areas are today being managed much less intensively than
inthepast(Burrascanoetal.,2016;Navarro &Pereira,2012). As a
result of these economic changes, as well as of changing manage-
ment priorities, the proportion of European forests in the older-
age classes is increasing, although wide regional differences exist
(FORES T EUROPE, 2015). Effo rts devoted at i dentifyi ng and pro-
tecting primar y forests should also include late- successional forests,
especially given that in many European regions, these forests rep-
resent the most natural forests still existing in the landscape. Late-
successional forests play an important role in terms of biodiversity
conservation, ecological functioning and provisioning of ecosystem
services.
Forthepurposeofthisstudy,weusetheterm“primary forests,”
to include all forests having a high naturalness, without implying
that these forest s were never cleared nor disturbed by man, which
is in line wit h the FAO definition of pr imary fore sts (Buchwal d,
2005; FAO, 2015). Research on the s tructure and dy namics of
primaryforestsinEuropehasalongtradition(Leibundgut,1959).
For instance, strictly protected forest areas were in the focus
of two large collaborative efforts to coordinate, harmonize and
link research onforest reserves (Diaci, 1999;Frank etal., 2007;
Parviainen,2000).Agrowingbodyofknowledgehasaccumulated
eversince(Burrascanoetal.,2013;EEA,2014;Keetonetal.,2010;
Kuuluvai nen & Aakala, 2 011),in cluding data o n the most iconi c
primar y forests, su ch as Białowieża in Po land, Uholka-S hyrokyi
LuhinUkraine,ŽofínintheCzechRepublicandIzvoareleNereiin
Romania(Bernadzki,Bolibok,Brzeziecki,Za̧jaczkowski,&Zybura,
1998; Hobi, Commarmot, & Bugmann, 2015; Král etal., 2014;
Veen etal., 2010). Nevertheless, only afew countrieshave sys-
tematicallyinventoriedremaining primaryforestfragments(e.g.,
Adam&Vrška,2009)asidefromforestreservesandinternation-
ally recognized primary forest patches. Large regional gaps thus
remain, especially in those countries where the political resistance
to the designation of additional strict reserves is hindering ef-
fort s to identify r emaining prim ary forest ( Mackey etal., 2015).
Furthermore, transboundary efforts for mapping and protecting
primaryforestsarerareandconfinedtospecificecoregions(e.g.,
theCarpathians, thegreenbelt of Fennoscandia)orforest types
(e.g.,UNESCOnetworkofprimevalbeechforests).
Despite these past efforts for consolidating and harmonizing in-
formationat thecontinental scale (Diaci, 1999; Franketal., 2007;
Parv iainen, 200 0), no up-to -date and sp atially detai led European-
wide database and map of primar y forests are currently available
(GarcíaFeced,Berglund,&Strnad,2015).Asaresult,systematicre-
search to quantify the extent of primar y forests in Europe, to assess
whether primary forests are adequately protected or to understand
what determines their spatialdistributionismissing.Amap of the
primar y forests of Europe is thus highly needed, to ensure that pri-
mary forestsreceiveadequaterecognitionandprotection(Mackey
etal.,2015)andasastartingpointforasystematicgapanalysisthat
highlights those biogeographical regions or forest types for which
primar y forests are absent or underrepresented. Such a map is in-
creasingly needed in the light of international commitments, such
astheEuropeanBiodiversityStrategy(Target3b-Action12,which
calls for Member States to ensure the preservation of wilderness
areas)ortheEU’sGreenInfrastructureStrategy,toensurethatpri-
mary forests and the ecosystem services they provide can be pro-
tected. Finally, analysing the determinants of the spatial distribution
of primar y forests could help to identify the socio- economic driv-
ers (e.g.,bioenergy production) behind thethreats facedby these
forests(e.g., illegalloggingandanthropogenic wildfires),as wellas
candidate sites for restoration initiatives, for instance where land-
use pressure and opportunity- cost of restoration are decreasing
(Navarro&Pereira,2012;Schnitzler,2014).
In this paper, we addressed the following questions:
1. What is the currently known distribution of primary forests
across Europe, biogeographical regions, forest types and pro-
tection levels?
4 
|
   SABATINI eT Al.
TABLE1 Descriptionofpredictorsusedtomodelthelikelihoodoffindingpreviouslyunmappedprimaryforests.Foreachpredictor,wereportedthemeasurementunit,resolution(Res),data
source,thesignoftheexpectedrelationshipwiththelikelihoodofoccurrenceofprimaryforests(+positive,−negative),anddataformat.(R—raster,V—vector).Onlyunderlinedvariableswere
retained in the final model
Class Predictor Description Mes. Unit Res Source Expected relationship Format
Climate Growing degree days
(GDD)
Number of days per year having a mean
temperature >5°C
Days 30 arc s Hijmans,Cameron,Parra,Jones,
andJarvis(2005)
R
Mean annual
temperature
Long- term mean annual temperature °C 30 arc s Hijmansetal.(2005) − R
Water availability Priestly–Taylorcoefficient:difference
between precipitation and potential
evapotranspiration
Ratio 30 arc s TrabuccoandZomer(2010) + R
Soil Crop suitability Maximum suitability value across 16 crops 0–1 30 arc s Zabel,Putzenlechner,and
Mauser(2014)
R
Topography Elevation Elevation a.s.l. m 30 arc s NASA,(2006) + R
Slope %30 arc s NASA,(2006) + R
Aspect 30 arc s NASA ,(2006) ± R
Ruggedness Terrain ruggedness expressing relief
energy
m 30 arc s NASA,(2006) + R
Solar radiation Potentialannualdirectincidentradiation log(MJ/cm2
*year)
1km McCuneandKeon(2002) R
Forest conditions Forest cover Percentageofforestedarea % 1km Kempeneers,Sedano,Seebach,
Strobl,andSan-Miguel-Ayanz
(2011)
+R
Forest core area Percentageofforestedareaclassifiedas
core
% 1km SoilleandVogt(2009) + R
Growing stock m3/ha 1km Gallaunetal.(2010) + R
Net annual increment Annualabovegroundbiomassincrement Ton d.m.* per
ha*yea r
1km Buset to, Barredo, and San-
Miguel-Ayanz(2014)
R
Biogeographical region Dummy BfN,(2003) V
Socio- economic Population density Landscan dataset n/km230 arc s OakRidgeNationalLaboratory,
(2005)
R
Travel time to the
nearest city
Estimated travel time to the nearest city of
50,000 or more people in the year 2000
Min 1km Nelson(2008) +R
Harvest intensity
(2000–2015 average)
Percentageofnetannualincrement
harvestedin(20 00-2015average)
%Country FORESTEUROPE(2015) − V
Historical legacies Forested cover (1850) Amountoflandsuitableforagriculturestill
forested in 1850
%Country Kaplanetal.(2009) + V
Wood demand (1828) Historic al wood demand reconstructed for
the year 1828
TgC 0.5° McGrathetal.(2015) − R
    
|
 5
SABATINI eT A l.
2. Which biophysical, socio-economic historical land-use factors de-
termine the extant pattern of primar y forests?
3. What are theareas with the highestlikelihoodoffindingprevi-
ously unmapped primary forests?
2 | METHODS
2.1 | Primary forest database
Toproduce the first map of known European primary forests, we
adoptedFAOdefinitionofprimaryforests(FAO,2015).Wefollowed
theframeworkproposedbyBuchwald(2005),accordingtowhichthe
term primary forest comprises all those forests previously indicated
asprimeval,virgin,near-virgin,old-growthandlong-untouched(i.e.,
classesn10ton5 inBuchwald,2005—seeSupportingInformation
Appendix S1—for definitions). Here, we embrace a positivist per-
spective implying that empirical evidence can be used to infer
whether forests have been impacted by human activities within the
last two centuries.
Based on this set of conceptual definitions, we conducted a liter-
ature revie w and collected al l the studies pub lished betwee n January
2000 and January 2017, reporting basic information on primary for-
ests in Europe, excluding Russia. We limited our review to papers
published after 20 00, to avoid including those forests that, although
being reported as primar y in older papers, may have meanwhile lost
their primary status due to human disturbance. We identified rele-
vantpublicationsintheISIWebofKnowle dgeusingthesea rchter m
“(primary OR virgin OR old-growth OR primeval) AND forest*” in
thetitlefield.Weconservativelyavoidedothertermssuchas“un-
managed”(=notunderactivemanagement),“natural”(=stockedwith
naturallyregeneratednativetrees)or“ancient”(=neverclearedfor
agriculture).AlthoughwidelyusedintheEuropeanliterature,these
concepts represent necessary but not sufficient conditions for con-
sidering a forest as primary for our paper.
The initial search was then refined using geographical and sub-
jectareasasfilters(seeSupportingInformationAppendixS2forde-
tails).Thispreliminarylistofpaperswasthensupplementedwiththe
literatureintheirownreferencelistsaswellaswithstudiesknown
to the authors. For all papers, we extracted the location and basic
information on the primary forests described. In addition, we sent
out a questionnaire to scientists and experts on primary forests to
collec t information on (1) exi sting maps an d databases of p rimary
forestsintheir country,(2) primary forestsnotyetincluded in ex-
isting ma ps and databa ses, and (3) cont acts of add itional exp erts.
In total, we contacted 134 forest exper ts from 33 European coun-
tries (Supporting Information Table S1). After finding a suitable
dataset or map, we invited the data owner to join our informal re-
search networkandsharethedatasetintheirpossession.Toavoid
terminological inconsistencies, the inclusion of a countr y dataset
was conditional on the establishment of an explicit equivalence be-
tweencountry-specificdefinitionsandthe definitionframeworkof
Buchwald(2005).
We integrated all data into a geodatabase, where each primary
forest patch was reported either as a polygon or as a point location.
Our minimum mapping unit was two ha. For each forest, we gath-
ered a set of basic descriptors, including name, location, naturalness
level(followingthebroaddefinitionsreportedinBuchwald,2005—
Supporting Information Appendix S1), extent and dominant tree
species. We assigned each stand to a broad forest type, based on
thestand’sdominanttreespecies,elevationandbiogeographicalre-
gion(BfN,2003;EEA,2006).Wederivedtheprotectionstatusand
IUCN category of each forest patch based on the World Database of
ProtectedArea (UNEP-WCMC & IUCN, 2017). Adetaileddescrip-
tion of the database architecture and each dataset is in Supporting
Infor mat ion(Su ppor t ingInf ormatio nTab les S2, S3a ndA ppen dixS3).
2.2 | Biophysical and socio- economic location
characteristics of the mapped forests
Based on the variables that were previously used as spatial deter-
minants of harvest intensit y and wood production across Europe
(Levers etal.,2014;Verkerk etal.,2015),weidentified a setof19
biophysical (including climate, soil, topography and forest condi-
tions),socio-economicandhistoricalland-use variables that could
explainprimary forest distribution (Table1).Most predictorswere
availableasrasterlayerswitharesolutionof1×1kmorfiner,with
the exception of three variables that either had a 0.5° resolution, or
were available at the country level. We reprojected all predictors to
theLambertazimuthalequal-areaprojection.Wecheckedforcollin-
earity and excluded collinear predictors when an individual variable
returnedavarianceinflationfactor(VIF)>10(Dormannetal.,2013)
orreturnedaPearson’sr>0.7withanothervariable(inthiscase,the
variablehavingthehighestVIFwasexcluded;Table1).
2.3 | Relative likelihood of the occurrence of
undetected primary forests
Weconvertedthemapofprimaryforeststoa1-kmpresence–ab-
sencerasterandusedboostedregression trees(BRTs)toexplore
the relationships between our set of predictors and the occurrence
of primar y forests. I n this way, we estimated the r elative likeli-
hood that a grid cell cont ained a primary forest patch, although
we recognize that the relatively coarse scale of most predictors
mayweakentheperformance ofourmodel.Werelied on model-
lingas,toourknowledge,noreliable workflow existsthatallows
differentiating primary from nonprimary forest using remote sens-
ing data only.
BRTs are nonparametric models based on decision trees in a boost-
ingframework.Theyhavetheadvantageofnotrequiringpriorassump-
tions and being relatively robust against overfitting, missing data, and
collinearity. Therefore, BRTs represent a flexible approach for uncover-
ing nonlinear relationships and interactions among predictors. BRTs are
increasingly used for attaining system understanding, hypothesis test-
ing and statisticalinferences (Dormann etal., 2013; Elith,Leathwick,
&Hastie, 2008). Our BRTwasparameterized using a learning rateof
6 
|
   SABATINI eT Al.
0.02,atreecomplexityof5andabagfractionof0.7(Elithetal.,2008).
We used the gbm.step routine provided by the dismopackage(Hijmans,
Phillips, Leathwick, & Elith, 2011) in r (R Development Core Team,
2017)todeterminetheoptimalnumberoftrees.Weranalltheanaly-
sesaftermaskingnonforestareas(Gallaunetal.,2010).
Asthedataonprimaryforestpresencewerespatiallyclustered
andthismayleadtoinaccuratemodels(Phillipsetal.,2009),weused
aspatialfilteringapproachtorarefytheavailabledataona5×5-km
grid. To account for the bias in our dataset due to some countries not
reporting any or ver y few data, we also created a map of sampling
effor t (1:h igh sampling e ffort, 0 : low samplin g effort ; Support ing
Information FigureS1).Wethen stratified theselection of 37,060
pseudo-absencepoints(i.e.,tentimesthenumberofpresencesafter
therarefaction)basedonthedistributionofpresencepointsinthe
mapofsamplingeffort(Kramer-Schadtetal.,2013).Toaccountfor
remaining spatial bias, we used the pwdSample function in the dism o
packagetopaireachtestpresencesitewiththeclosesttestpseudo-
absence site prior to evaluating the performance of our model, thus
removingtheremainingspatialsortingbias(Hijmans,2012).Wealso
testedforspatialautocorrelationinmodelresidualsusingMoran’sI.
Weusedthereceiver-operatingcharacteristiccurves(ROC)and
theareaunderthecurve(AUC)toevaluatepredictionperformance
based on 10-fold cro ss-validat ion. As AUC is only ran k-bas ed, we
also calculated Pearson’s correlation between the observed pres-
ence\pseudo-absence and the likelihood predictedf rom the BRT
model (Phillips etal., 2009). Finally,weused the true-positive and
true- negative rates, to calculate model accuracy and precision when
usingdifferentlikelihoodthresholdsfordiscriminatingbetweenpre-
dicted primary forest occurrence vs. absence. We used the thresh-
oldreturningthehighest accuracy tocreate amapofthe1×1-km
forested grid cell potentially containing one or more patches of pri-
mary forest. The relative importance of predictors was evaluated
according to the number of times that a variable was selected for
splitting, weighted by the squared improvement to the model as a
resultofeachsplitandaveragedoveralltrees(Elithetal.,2008).For
those predictors with a relative importance above that expected by
chance (100%/numb er of predicto rs), we produced pa rtial depen -
dencyplotsconstrainedbetweenthe2.5and97.5percentilesofthe
predic torrangeandsmoothedusingaLOESSinterpolation(spanpa-
rameter=0.2)toenhanceinterpretability.
FIGURE1 Distribution of primary
forest patches retrieved for Europe by
forest t ype. The map of biogeographical
regionsinthebackgroundfollowsBfN
(2003).ForesttypesfollowEEA(2006):
FT1—borealforest,FT2—hemiborealand
nemoralconiferous-mixedforest,FT3—
alpineconiferous,FT4-5—mesophytic
deciduousandacidophilusforest,FT6—
beechforest,FT7—mountainousbeech
forest,FT8—thermophilusdeciduous
forest,FT9—broadleavedevergreen
forest,FT10—coniferousMediterranean
forest,FT11-12—mireandswampforests
andfloodplainforest,FT13—nonriverine
alder,birchoraspen,NA-NC—nodata/
unclassified
    
|
 7
SABATINI eT A l.
3 | RESULTS
Our database covered 1.4 Mha of primary forest in 32 European
countries(Figure1).Thisdatabasewascomposedof32regional
dataset s (Supporting Information Table S3) that we integrated
with data on additional 254 primary forest patches, described
in94 studies or reports retrieved through theliterature review
(Suppor ting Inform ation Table S4). A list of th e data sources is
in Supporting Information Appendix S4. Most of the primary
forests for which data were available were located in northern
Europe, especially Finland (0.9 Mha), and Eastern Europe (0.2
Mha), especially Ukraine, Bulgaria and Romania (Supporting
InformationTableS5).Thecountries havingthehighestpropor-
tion of primaryforestwere Finland (2.9%of national territory),
Switzerland, Lithuania, Slovenia and Bulgaria (eachabout 0.5%;
Figure2). T hese rankin gs, however, are heavil y affected by t he
availability of data and disregard the contribution of countries for
which we could not retrieve adequate data. We found complete
inventoriesonlyforthreecountries(CzechRepublic,Slovakiaand
Hungary)andpartialorincompleteinventoriesforadditionalfour
countries, but either limited to specific mountain ranges (e.g.,
Carpathians—Romania,Ukraine)orprotectedareas(France,Italy;
Figure2). Countriesforwhichwewerenotabletoretrievedata
on primar y forests were L atvia, Belarus, Moldova and Ireland. For
Sweden,Austria, the UK, BosniaandHerzegovina,Montenegro
and Serbia, we only found scat tered information, that is ver y
fewrecordsintheliterature,butno(orverylimited)spatialdata-
sets deriving fromlocal inventories (Figure2).Nevertheless,we
cannot exclude that additional data may exist for these or other
countries that we did not manage to retrieve, especially for coun-
tries expected to host wide stretches of primar y forest, such as
Sweden.
Primar y forests occurred mostly in the boreal (1 Mha, 1% of
thatbiogeographicalregion)andthealpineregions(0.4Mha,0.6%).
The Macaronesian region also had a high relative proportion of pri-
maryforests,allofitlocatedintheLaurisilvaofMadeira(15,100ha,
1.5%; Supporting Information Table S6). The mapped primary for-
est patches were, on average, very small: The median size was only
24ha,andonly4.3%ofthepatcheswerelargerthan1,000ha.Most
(89.1%)oftheprimaryforestinourdatasetwasprotected,butonly
46%wascurrentlyunderstrictprotection(IUCNcategoryI),withan
additional 24%being includedinnationalparks(IUCN category II;
Figure3,SupportingInformationTableS5).
With regardtotheforesttypes(FTs,sensu EEA ,2006),boreal
forest(FT1)accountedforthehighestshareofthemappedprimary
forests(1.09 Mha),followed bymountain beech forest (FT7—0.15
Mha) and, to a mi nor extent, alpi ne coniferous fores t (FT3—0.07
Mha;Figure1,Supporting InformationFigure S2).According tothe
definitions reported in Buchwald (2005—Supporting Information
Appendix S1), most of the primary forests in our dataset were
near-virgin (n7—1.20 Mha), while old-growth (n6—0.15 Mha) or
long-untouchedstands(n5—0.11Mha)accountedonly for a minor
fraction(10%)ofthe cumulative area we mapped. However,when
considering the number of polygons rather than the area, the highest
FIGURE2 Contry-wise completeness of primary forest data
andproportionofprimaryforestunderstrictprotection(IUCN
categoryI),includedinprotectedareashavingotherIUCN
categories or unprotected. The size of the pie is propor tional to the
logarithm of the total primar y forest extent mapped in a country.
The pie fractions only represent the data currently available and
they should not be directly compared across countries, as dat a
quality and availability differ. Furthermore, for some countries,
onlyinventoriesofprimaryforestlocatedeitherinside(e.g.,Italy,
FinlandandFrance)oroutside(Norway)protectedareaswere
available
FIGURE3 AreaofEuropeanprimaryforestacrossIUCN
categories.I—strictnaturereservesorwildernessareas;II—national
parks;III—naturalmonumentsorfeatures;IV—habitat/species
managementareas;V—protectedlandscapes;andVI—protected
area with sustainable use of natural resources. When a patch
of primar y forest was protected under multiple levels, we only
considered the strictest category
8 
|
   SABATINI eT Al.
share of the forest patches we mapped were classified as old- growth
forestsandbelongedtotheboreal(FT1),alpineconiferous(FT3)and
mountainbeech (FT7) foresttypes(Supporting InformationFigure
S3).
The boosted regression tree modelling provided insights into the
relative importance of our predictors in determining the spatial pat-
terns ofknown primaryforests.The BRT modelfitted2,050 trees
andreturned a relatively high cross-validated AUC andcorrelation
(mean ± SD range 0.86±0.005 and 0.63±0.008, respectively).
When evaluating the model performance on the test data selected
tocontrolforspatialsortingbias(Hijmans,2012),theAUCand the
correlationwerelower(0.70and0.33,respectively),indicatingthat
the model performance was affected by the spatial dependency of
thetrainingdata.Thehighestmodelaccuracy(0.64)wasobserved
forathresholdcorrespondingtothe90thpercentileoftheprobabil-
itydistribution(SupportingInformationTableS7,FigureS4).
Biophysical, socio- economic and historical variables all played
arole in determining thelikelihood of primaryforest occurrence
(Figure4). Primary forests were more likely found in areas with
higher ruggedness and water availabilit y. Socio- economic factors
had the highest relative importance among the selected variables,
with accessibility and population density selected in 12.6% and
12.2% ofallmodelruns.Primaryforestsoccurredmorelikely far-
ther away from major towns and where population density was
lower. Both historical variables we used were important predictors:
The likelih ood of occurrence of p rimary fore st decreased fo r in-
creasing historical levels of wood demand up to a certain threshold,
above which it increased again. The amount of land suitable for ag-
riculture still forested in 1850, instead, showed a reverse U- shaped
relationship. Finally, our model also highlighted differences across
biogeographical regions: The likelihood of occurrence of primary
forestswashigherthanaverageforthealpine,BlackSeaandboreal
regions.
The areas w ith the highe st primar y forest likeli hood (Figure5 )
were along the northern Finnish–Russian border, in the Finnish–
Swedish border and in mountain ranges, especially the Carpathians,
the eastern Alps, the Dinaric Mountains and, to a lesser extent,
the highest parts of thePyrenees. Areas with lowprimary forests
FIGURE4 Partialdependency
plots(PDPs)showingtherelationship
between spatial determinants and
therelativelikelihoodofoccurrence
of primar y forest patches in a given
1×1-kmpixel.Theverticalaxisofthe
PDPsshowsfittedvaluesforeach
observationalongthevariable’sdata
range(horizontalaxis).X-axesare
equipped with rug plots that visualize the
distribution of the respective data space.
Numbers in parentheses represent the
relative importance of a given variable.
Biogeographicalregions:ALP=alpine,
BLK=BlackSea,CON=continental,
MED=Mediterranean,PAN=Pannonian,
ATL=Atlantic,BOR=boreal)
0 200 600 1000 1400
0.00.4 0.
8
Travel time (12.8%)
Fitted values
02468
0.00.4 0.
8
Pop. density (log) (12.2%)
40 50 60 70 80 90 100
0.00.4 0.8
Water availability (alpha) (11.5%)
Fitted values
0e+00 2e+11 4e+11 6e+11
0.00.4 0.8
Wood demand − 1828 (8.2%)
20 40 60 80
0.00.4 0.8
Forest cover − 1850 (8%)
Fitted values
0 100 200 300 400
0.00.4 0.8
Ruggedness (6.8%)
0.00.4 0.8
Biogeographical reg. (6%)
Fitted values
ALP
BLK
CON
MED
PAN
ATL
BOR
    
|
 9
SABATINI eT A l.
likelihoodweretheAtlantic region, the BritannicArchipelago, the
Middle European lowlands, the Pannonian plain and the hemibo-
realBalticregion.Areaswithpredictedandobservedprimary for-
est (Supporting Information Figure S5) matched in those regions
wherewehadahighsamplingsize(northernFinland,Slovakianand
Ukrainian Carpathians, Balkan mountains). On the contrary, our
model predicted the occurrence of scattered and isolated primar y
forest patches in southern Finland, in the continental lowlands or
inthe westernMediterraneanareas weakly.Only 38% of thearea
predicted to host primary forest was included in protected areas, of
whichonly 5.6% was understrictprotection(i.e., IUCN category I;
SupportingInformationFigureS6).
4 | DISCUSSION
Our study produced the most comprehensive spatially explicit data-
setonknownprimaryforestsinEuropecurrently available.Known
primar y forests covered approximately 1.4 Mha in 32 European
countries,whichrepresent 0.25%ofterrestrialEuropeand0.7%of
FIGURE5 AreaswiththehighestlikelihoodofoccurrenceofprimaryforestinEuropeata1×1kmresolution.Thetop-ranking5%pixels
werehighlightedinpurpleandthe90–95thpercentileinblue.ForestsarereportedingreyandfollowGallaunetal.(2010)
(a) (b) (c)
(d)
(e)
10 
|
   SABATINI eT Al.
Europe’sforestareaexcludingRussia.Thismeansthatwemanaged
tomapaboutone-fifthofthe7.3Mhaofforestestimatedtobe“un-
disturbed by man”inEurope (FORESTEUROPE, 2015). Wefound a
general increase in the number of primar y forest patches from the
west to the east and from the south to the nor th. Most of the pri-
mary fo rests in our dat aset were locat ed in Finland (0.9 Mha), in
theCarpathians(0.16Mha)andintheBalkans(0.08Mha),although
some important data gaps exist.
For many countries, we noted a mismatch between the total area
of primar y forest included in our map and the estimates reported in
FORESTEUROPE(2015),possiblybecausethesewerebasedonthe
data not inherently designed for mapping primar y forest, such as ex-
trapolationfromforestinventories(Italy,Norway)orremotesensing
datanotverifiedinthefield(e.g.,Romania,FORESTEUROPE,2015).
The area of primary forest we mapped for Finland is three times
largerthanpreviousestimates(FORESTEUROPE,2015).Itpossibly
depends on the fac t that we considered as primary forest s not only
old-growthstandsolderthan160–200years(asinFORESTEUROPE,
2015),butalsothoseprimaryforestscomposedofamosaicofsuc-
cessionalphasesoccurringintheextremenorthofFinland(Bernier
etal.,2017;Kuuluvainen &Aakala,2011;Potapovetal., 2017).On
the contrary, the amount of primary forest area mapped for Sweden
and the Carpathians is far lower than current estimates. For Sweden,
wemappedonly0.03Mhaofprimaryforest,whichrepresents<2%
ofthecurrentestimation(2.4MhainFORESTEUROPE,2015).Given
that Sweden is expected to host the widest continuous stretches
ofprimaryforestoftheEuropeancontinent(Parviainen,1999),this
represents the most severe data gap of our dataset. Similarly, for the
Carpathians,wemappedca.30%ofthe0.44Mhaofprimaryforest
currentlyestimatedtoexist(FORESTEUROPE,2015).Thedatawe
aggregated for the C arpathians mostly derived from sur veys coordi-
natedwithin theframework ofthe UNEP—CarpathianConvention.
Not only are these inventories still incomplete in countries such as
UkraineandRomania,butthey alsoprioritizethose forests having
the highest naturalness levels. Therefore, a considerable share of
forest of lower naturalness levels, but still qualif ying as primary, may
remainunmappedintheCarpathians(Kulakowskietal.,2017).
The low share of primar y forest in Western Europe was ex-
pected considering the historically high population density, and long
history of land use, especially in the Mediterranean (Jepsenetal.,
2015).Species-richMediterraneanforesttypes(i.e., FT8, FT9 and
FT10)wereparticularlyscarceinourmap (SupportingInformation
Figures S2andS3).Mediterraneanforestsshow fundamentallydif-
ferent structural characteristics from temperate mesic forests, due
to the high- drought stress Mediterranean forests experience during
thesummerandduetofiredisturbance(Karavanietal.,2018).The
role of wildfires in shaping the structure of Mediterranean primary
forests is particularly complex as today most wildfires are human-
induced (Ganteaume etal., 2013; Vacchiano, Garbarino, Lingua,
&Motta, 2017). These conditions mayhinderthe developmentof
structural features typically associated with old- growth stages, such
as deadwood or large trees (Burrascano etal., 2013; Kulakowski
etal.,2017).Asthesefeaturesarecommonlyusedonthegroundfor
identifyingprimaryforests(atleastintheirlate-successionalstages),
significant portions of Mediterranean primary forest may remain
overlooked.
Primary forest disproportionatelyoccurred in remote, scarcely
populated areas, mostly in rugged mountain areas or at high latitudes
(i.e., on land with low agricultural productivity or low profitability
forforestryoperations).Thismakesintuitivelysense,asaccessibility
andthedistancefrommarketsorothercentresofdemandisoneof
the main drivers of land- use allocation. Indeed, in remote and unfa-
vourable areas such as northern Fennoscandia and the Carpathians
mountains, land- use histor y has been shorter and less intense than
inthe restof Europe (Jepsen etal., 2015;Kulakowski etal.,2017),
makingthepersistenceofprimar yforestsmorelikely.Thisfindingis
alsoconsistentwithpreviousworkinFennoscandia(Kuuluvainen&
Aakala,2011),aswellaswiththeknownbiasinprotectedareadis-
tributiontowardshigherelevationandmoreremotelocations(Joppa
&Pfaff,2009).Interestingly,accessibilityandpopulationdensityare
also important spatial determinants for explaining the patterns of
wood productionand harvesting intensity in Europe (Levers etal.,
2014;Verkerketal.,2015).Thecorrelationbetweenprimar yforest
and water availabilit y probably reflects the same pattern, as a direct
effec t of water availabi lity on the like lihood of find ing patches of
primar y forests i s unlikely and water av ailability i s usually high in
mountain and boreal regions. Finally, our model predicted an unex-
pectedlyhighlikelihoodofoccurrence ofprimaryforestinthe rug-
gedportions of thePyrenees and the Alps.The Pyreneesand the
Alpshavealongerhistoryoflanduseandhigher historicalratesof
forest management intensity than other European mountain ranges,
which our models could not account for.
Although difficult to map at high spatial resolution, historical
land-usepressuresplayedakeyroleinourmodeltoexplainpresent-
dayprimary forestdistribution.Primaryforests,forinstance,hada
lowerlikelihoodofoccurringinthoseregionswithhigherhistorical
wood dema nd (Figure4), but only up t o a threshold, af ter which
the likelihood increased unexpectedly. We believe this relation-
ship derives by the occurrence of several primary forest patches in
some historical mining areas, where a historical high wood demand
co- occurred with a high historical forest cover, such as the Upper
SilesianProvince(Poland).Thehistoricalvariablesweused,however,
did not fully capture the role of historical events and contingencies.
For instance, the occurrence of some primary forest patches may de-
pend on the short distance from major historical political boundaries
asinthecaseoftheBieszczadyregion(SEPoland)ortheRhodope
mountains(betweenGreeceandBulgaria).The peripherallocation
oftheseregionsand/orthelackofeffectivemeansfortimbertrans-
portation left considerable areas of primary forest well into the
20th century. These areas could have followed a trajectory similar
to other peripheral areas where primary forests were extensively
cut in the last centur y if political upheavals, including the establish-
mentoftheIronCurtain,hadnot occurred(Keetonetal.,2010).In
addition to major historical events, peculiar local episodes could
also explain the presence of some primar y forest patches, such as
Fonte Novello, a 50- ha old- growth stand in Gran Sasso National
    
|
 11
SABATINI eT A l.
Park (centralItaly), whichislocatedattheboundary between two
municipalities. Ownership of this forest remnant has been contented
between the two municipalities since their formal establishment at
thebeginningof 19th centuryandremainsunresolvedas of today.
This dispute coupled with the deep economic depression of this
mountain area saved the stand from being exploited for timber and
degradationuntilitsrecent“rediscover y”andprotection.Otherem-
blematic examples include primar y forests that were set aside centu-
riesagoashuntinggrounds,suchasinBiałowieża(lowlandPoland),
Biograd ska Gora Nation al Park (Montenegr o)o r Central Bohem ia
(CzechRepublic).
The result of an unprecedented international collaboration, our
dataset should be considered as a necessary first step towards a
more complete inventory. Important limitations include high vari-
ability in data quantity and quality across countries. Variabilit y may
derivefromadifferentinterpretationofFAOdefinitionofprimary
forests. Although authoritative and widely accepted internation-
ally,FAOdefinitionisconceptual, ratherthan operational, which
mayresultininconsistenciesinreportingamongcountries(Bernier
etal.,2017).Formanycountries,nocompleteinventoryexists,and
data der ive from the kn owledge of loca l expert s or from par tial
inventories with relatively narrow breadth, focussing on either
forest inside(e.g.,France, ItalyandFinland)oroutsideprotected
areas (e.g.,Norway)or specific regions(e.g., theTranscarpathian
region of Uk raine, the French P yrenees). In som e countries, we
found only incomplete data although extensive forestr y statistics
and databases are generally available for these countries. This was
either because we did not manage to engage loc al researchers
inhelpinguslocate,extract andharmonizeexistingdata (Latvia,
Sweden) or because relevant data are kept strictly confidential
by public authorities, possibly to avoid conflicts with private for-
estowners(e.g.,Austria).Whilefillingthese knowledgegaps is a
priorit y to achieve a more accurate description of primary forest
distribution in Europe, the good- quality datasets we retrieved for
neighbouringcountrieswithsimilarecologicalconditions(Norway
and Finland in the case of Sweden, or Switzerland and Slovenia in
thecaseofAustria)grantrobustnesstoourstatisticalresults.For
other countries with abundant forest resources and presumably
alsoarelativelyhighfractionofprimaryforest(e.g.,manyBalkan
countriesandBelarus),datawereunavailable,atleastintheinter-
national scientific literature or in digitized forms. In this case, we
advocate a higher commitment from the international communit y
to suppor t local research institutions and NGOs in the collection
or digitization of data on primar y forests. Few data also exist for
thosecountrieswithlowforestcover(e.g.,<10%)andinwhichsig-
nif ic antareasofprimaryforestareunlikelytooccurduetohisto ric
clearing or biophysical factors, such as the British Isles, Moldova
or Cypr us.
Granting adequate protection to European primary forests
should be aconservation priority (Mackeyetal.,2015),especially
given the recent concerns about commercial exploitation of old-
growthforestsinEasternEurope(Chylarecki&Selva,2016;Knorn
etal., 2013).The majority (89%) ofprimary forest inour dataset
is currently under some form of protection; nevertheless, its fu-
tureprotectionre mainsuncertain.Ahighfrac tionofprimaryforest
(54%)iscurrently outside strictlyprotectedareas,andbroaddif-
ferences exist among European countries in the management re-
strictionappliedinotherprotectedareas(Diaci,1999;Parviainen,
2000;Verkerk,Zanchi,&Lindner,2014).Insomecountries,some
forest management activities (e.g., salvage logging) are allowed
even in protec ted areas, representing a threat to primary forests
(Thornetal., 2018).Anotherconcernisthe small averagesize of
primar y forest patches. Even if protected, a small patch of forest
may not be large enough to host the full range of ecological pro-
cesses,andbiodiversitymaysufferfromextinctiondebt(Peterken,
1996). When large patch es of primar y forest do not ex ist, main-
taining existing patches in a large matrix of natural or seminatural
forests should be the priority. This is necessar y both to buffer the
effects of direct and indirect anthropogenic disturbance on primary
forestsandbecausethesepatchescouldfunctionas“strongholds”
for the recovery and recolonization of many specialist species in
the surroundingforest(Vandekerkhove etal., 2009). Our map of
primary forest in Europe can therefore inform efforts aiming at
preserving wilderness areas, in line with the requirements of the
European Biodiversity Strategy and EU’s Green Infrastructure
Strategy. Given the current low share of primar y forests, their res-
torationshouldbeaprioritythroughoutEurope(Navarro&Pereira,
2012;Schnitzler,2014).Ourmapcouldbeusedtoprioritizethose
regions and forest t ypes for possible restoration efforts. For in-
stance,ourworkhighlightedareas,suchasthemostruggedparts
oftheAlpsandthePyrenees,whereland-usepressureisrelatively
low and primary forests could potentially occur, thus suggesting
that the opportunity costs of restoring primary forests and associ-
ated ecosystem processes and biodiversity in these areas may be
lower than elsewhere.
ACKNOWLEDGEMENTS
FORESTS & CO was funded by the European Union under the Marie
Sklodowska-CurieGrantAgreementNo.658876.M.MikolášandM.
Svoboda were supported by the Czech University of Life Sciences,
Prague(CIGANo.20184304)andbytheinstitutionalprojectMSMT
CZ.02.1.01/0.0/0.0/16_019/0000803. Additional funding sources
forindividualdatasetsare in SupportingInformationAppendixS3.
WethankA.Dušan,H.Kirchmeir,A.Kraut,R.Pisek,O.Schwendtner,
J.Vysoky,D.VallauriandthestaffofWWFRomania,especiallyC.
Bucur and R. Melu, for providing valuable data and suggestions. We
also than k S. Varis and S. Zudin for t echnical supp ort. This work
would not have been possible without all those who responded to
the questionnaires and those who collected all the data presented
here.
DATA ACCESSIBILITY
The data on primar y forests here presented were collected
within the F&CO- NET initiative and remain property of the
12 
|
   SABATINI eT Al.
institutions, organizations or person who created or collected
them. The custodian of each dataset, that is person who owns
or represents the contributed data, is listed in Supporting
Information Table S3. F&CO- NET is available on request for dis-
closing data to individuals or groups of individuals for research
or application purposes. Request s will be considered by the
F&CO-NETcoordinators(F.M.Sabatini,T.Kuemmerle)anddata
released after receiving the approval from the respective cus-
todians. A ll derived data a re available upon r equest from th e
corresponding author.
ORCID
Francesco Maria Sabatini http://orcid.org/0000-0002-7202-7697
Christian Levers http://orcid.org/0000-0003-4810-9024
Inês Marques Duarte http://orcid.org/0000-0002-1524-5487
Leónia Nunes http://orcid.org/0000-0002-2617-0468
Peter Ódor http://orcid.org/0000-0003-1729-8897
Alejandro Ruete http://orcid.org/0000-0001-7681-2812
Tobias Kuemmerle https://orcid.org/0000-0002-9775-142X
REFERENCES
Adam,D.,&Vrška,T.(2009).Importantlocalitiesofold-growthforests.
InT.Hrnčiarová,P.Mackovčin&I.Zvara(Ed.),Landscape atlas of the
Czech Republic (pp. 209). Průhonice, Czech Republic: Ministry of
EnvironmentandSilvaTaroucaResearchInstitute,Prague.
Bauhus, J., Puettmann, K., & Messier, C. (2009). Silviculture for old-
growth attributes. Forest Ecology and Management, 258, 525–537.
https://doi.org/10.1016/j.foreco.2009.01.053
Bernadzki, E.,Bolibok, L.,Brzeziecki,B.,Za̧ jaczkowski,J.,&Zybura, H.
(1998).Compositionaldynamics ofnaturalforestsintheBialowieza
NationalPark,northeasternPoland.Journal of Vegetation Science, 9,
229–238.https://doi.org/10.2307/3237122
Bernier,P.,Paré,D.,Stinson,G.,Bridge,S.,Kishchuk,B.,Lemprière,T.,…
Vasbinder,W.(2017).Movingbeyondtheconceptof“primaryforest”
as a metric of forest environment quality. Ecological Applications, 27,
349–354.https://doi.org/10.1002/eap.1477
BfN (2003). Map of natural vegetation of Europe. Bundesamt fur
Naturschutz, Deutschland. Retrieved from http://www.bfn.de/.
National data included.
Buchwald,E.(2005).Ahierarchicalterminologyformoreorlessnatural
forest s in relation to sust ainable management and biodiversity con-
servation. Proceedings: Third expert meeting on harmonizing forest-re-
lated definitions for use by various stakeholders.FoodandAgriculture
OrganizationoftheUnitedNations,Rome,17–19January2005.
Burra scano, S., Chy trý, M., Kuem merle, T., Giarr izzo, E., Luys saert, S .,
Sabatini,F.M.,&Blasi,C.(2016).CurrentEuropeanpoliciesareun-
likelytojointlyfostercarbonsequestrationandprotectbiodiversity.
Biological Conservation, 201, 370–376. https://doi.or g/10.1016/j.
biocon.2016.08.005
Burrascano, S., Keeton, W. S., Sabatini, F. M., & Blasi, C. (2013).
Commonality and variability in the structural attributes of moist
temperate old-growth forests: A global review. Forest Ecology
and Management, 291, 458–479. https://doi.org/10.1016/j.
foreco.2012.11.020
Busetto, L., Barredo, J., & San-Miguel-Ayanz, J. (2014). Developing
a Spatially-Explicit Pan-European Dataset of Forest Biomass
Increment. In 22nd European Biomass Conference and Exhibition,
pp.41–46,https://doi.org/10.5071/22ndEUBCE2014-1AO.8.1
Chylarecki,P.,&Selva,N.(2016).Ancientforest:Spareitfromclearance.
Nature, 530,419–419.https://doi.org/10.1038/530419b
Diaci, J. (1999). Vir gin forest s and fores t reser ves in Centra l and East
European countries-History, present status and future development.
Proceedings of the invited lecturers’ reports presented at the COST
E4 management committee and working groups meeting in Ljubljana,
Slovenia. Ljubljana
Dormann,C.F.,Elith,J.,Bacher,S.,Buchmann,C.,Carl,G.,Carré,G.,…
Leitão,P.J.(2013). Collinearity:Areview ofmethodstodealwithit
and a simulation study evaluating their performance. Ecography, 36,
27–46 . ht tp s://doi.o rg /10.1111/j.160 0- 05 87.20 12.073 48 .x
EEA(2006).Europeanforesttypes.Categoriesandtypesforsustainable
forest managementreportingand policy.EEATechnicalReport No
9/2006.EEA ,Copenhagen.
EEA(2014).Developing a forest naturalness indicator for Europe. Concept
and methodology for a high nature value (HNV) forest indicator. EEA
Technical re port No 13/2014, Luxem bourg: Pub lications O ffice of
the European Union.
Elith,J.,Leathwick,J.R.,&Hastie,T.(2008).Aworkingguidetoboosted
regression trees. Journal of Animal Ecology, 77, 802–813. https://doi.
org/10.1111/j.1365-2656.2008.01390.x
FAO(2015).GlobalForestResourcesAssessment2015.Termsanddefi-
nitions . In: Forest resourc es Assessment Work ing Paper 180,p.36.FAO,
Rome.
Ford, S.E., & Keeton, W.S. (2017). Enhanced carbon storage through
management for old-growth characteristics in northern hardwood-
conifer forests. Ecosphere, 8, e01721. ht tps://doi.org/10.1002/
ecs2.1721
FORESTEUROPE(2015).StateofEurope’sForests2015.
Frank, G ., Parviaine n, J., Vandekerhove, K ., Latham, J., S chuck, A., &
Little,D.(2007).COST Action E27. Protected Forest Areas in Europe-
analysis and harmonisation (PROFOR): results, conclusions and rec-
ommendations. Federal Research and Training Centre for Forests,
NaturalHazardsandLandscape(BFW).
Gallaun, H., Zanchi, G., Nabuur s, G.-J., Hengeveld, G., Schardt, M., &
Verkerk, P. J. (2010). EU-wide m aps of growing stoc k and above-
ground biomass in forests based on remote sensing and field mea-
surements. Forest Ecology and Management, 260, 252–261. https://
doi.org/10.1016/j.foreco.2009.10.011
Ganteaume, A., Camia, A., Jappiot, M., San-Miguel-Ayanz, J., Long-
Fournel,M.,&Lampin,C.(2013).Areviewofthemaindrivingfactors
of forest f ire ignition over Europe. Environmental management, 51,
651–662.https://doi.org/10.1007/s00267-012-9961-z
GarcíaFeced,C.,Berglund, H., &Strnad, M.(2015).Scopingdocument:
information related to European old growth forests. ETC/BD rep ort
totheEEA .
Harmon,M. E.,Ferrell,W.K.,&Franklin,J.F.(1990). Effectsoncarbon
storag e of conversion of old- growth fo rests to young fo rests. Science,
247,699–702.https://doi.org/10.1126/science.247.4943.699
Hijmans , R. J. (2012). Cross-vali dation of specie s distributio n models:
Removing spatial sorting bias and calibration with a null model.
Ecology, 93,679–688.https://doi.org/10.1890/11-0826.1
Hijmans,R.J.,Cameron,S.E.,Parra,J.L.,Jones,P.G.,&Jarvis,A.(2005).
Very high re solution interpolated climate sur faces for global land
areas. International Journal of Climatology, 25,1965–1978.https://doi.
org/10.1002/(ISSN)1097-0088
Hijmans,R.J.,Phillips,S.,Leathwick,J.,&Elith,J.(2011).Package ‘dismo’.
Retrieved from http://cran.r-project.org/web/packages/dismo/
index.html.
Hobi, M. L ., Comma rmot, B., & B ugmann, H . (2015). Patte rn and pro-
cess in the largest primeval beech forest of Europe (Ukrainian
Carpathians).Journal of Vegetation Science, 26, 323–336. https://doi.
org /10.1111/j vs.12234
    
|
 13
SABATINI eT A l.
Jepsen,M. R., Kuemmerle, T.,Müller,D.,Erb,K .,Verburg, P.H.,Haberl,
H.,…Reenberg,A.(2015).TransitionsinEuropeanland-management
regimes b etween 1800 and 2010. Land Use Policy, 49, 53–64. https://
doi.org/10.1016/j.landusepol.2015.07.003
Joppa, L.,&Pfaff,A .(2009).Highandfar:Biasesin thelocationofpro-
tected areas. PLoS ONE, 4, e8273. https://doi.org/10.1371/journal.
pone.0008273
Kaplan , J. O., Krumhard t, K. M., & Zimm ermann, N. (2 009). The pre-
historic and preindustrial deforestation of Europe. Quaternary
Science Reviews, 28, 3016–3034. https://doi.org/10.1016/j.
quascirev.2009.09.028
Karavani, A., Boer, M. M., Baudena, M., Colinas, C., Díaz-Sierra, R.,
Pemán,J.,…RescodeDios,V.(2018).Fire-induceddeforestationin
drought-proneMediterraneanforests:Driversand unknownsfrom
leaves to communities. Ecological Monographs, 88,141–169.https://
doi.org/10.10 02/ecm.1285
Keeton,W.S.,Chernyavskyy,M.,Gratzer,G.,Main-Knorn,M.,Shpylchak,
M., & Bihun, Y. (2010). Structural characteristics and aboveground
biomass of old- growth sp ruce- fir stands in the eastern Carpathian
mountains, Ukraine. Plant Biosystems, 144, 148–159. https://doi.
org /10.10 80/11263500903560512
Kempeneers,P.,Sedano,F.,Seebach,L.,Strobl,P.,&San-Miguel-Ayanz,J.
(2011).Datafusionofdif ferentspatialresolutionremotesensingim-
ages applied to forest- type mapping. IEEE Transactions on Geoscience
and Remote Sensing, 49, 4977–4986. https://doi.org/10.1109/
TGRS.2011.2158548
Knorn,J.,Kuemmerle,T.,Radeloff,V.C.,Keeton,W.S.,Gancz,V.,Biriş,
I.A., … Hostert, P.(2013).Continuedlossoftemperateold-growth
forest s in the Romanian Carpathians despite an increasing protected
areanetwork.Environmental Conservation, 40,182–193.https://doi.
org/10.1017/S0376892912000355
Král, K.,McMahon, S. M., Janík,D.,Adam,D.,& Vrška,T.(2014).Patch
mosaic of developmental stages in central European natural forests
along vegetation gradient. Forest Ecology and Management, 330, 17–
28. https://doi.org/10.1016/j.foreco.2014.06.034
Kramer-Schadt,S.,Niedballa,J.,Pilgrim, J.D.,Schröder,B.,Lindenborn,
J.,Reinfelder,V.,…Augeri,D.M.(2013).Theimportanceofcorrect-
ing for sampling bias in MaxEnt species distribution models. Diversity
and Distributions, 19,1366–1379.https://doi.org/10.1111/ddi.12096
K u l a k o w s k i , D . ,S e i d l , R . , H o l e k s a , J . , K u u l u v a i n e n , T . , N a g e l , T. A . , P a n a y o t o v , 
M.,…Whitlock, C. (2017). A walkonthewild side: Disturbancedy-
namics and the conservation and management of European mountain
forest ecosystems. Forest Ecology and Management, 388, 12 0 –1 3 1 .
https://doi.org/10.1016/j.foreco.2016.07.037
Kuuluvainen, T., & Aakala, T. (2011). Natural forest dynamics in boreal
Fennoscandia:Areviewandclassification.Silva Fennica, 45, 823–841.
Leibundgut, H. (1959). Über Zweck und Methodik der Struktur-und
Zuwachsanalyse von Urwäldern. Schweiz. Zeitschrift für Forstwesen
110(3): 111–124
Levers, C.,Verkerk, P.J.,Müller,D.,Verburg,P.H.,Butsic,V.,Leitão,P.
J., … Kuemm erle, T. (2014). Drivers of fore st harvesti ng intensity
patterns in Europe. Forest Ecology and Management, 315, 160–172.
https://doi.org/10.1016/j.foreco.2013.12.030
Mackey, B., DellaSala, D. A., Kormos, C., Lindenmayer, D., Kumpel,
N., Zimm erman, B. , … Watson, J. E. M . (2015). Policy o ptions for
the world’s primar y forests in multilateral environmental agree-
ments. Conservation Letters, 8, 139–147. https://doi.org/10.1111/
conl.12120
McCune , B., & Keon, D. (20 02). Equations for p otential annu al direct
incident radiation and heat load. Journal of Vegetation Science, 13,
603–606. https://doi.org/10.1111/j.1654-1103.2002.tb02087.x
McGrath,M.,Luyssaert,S.,Meyfroidt,P.,Kaplan,J.,Bürgi,M.,Chen,Y.,
…Naudts, K. (2015). Reconstructing European forest management
from 1600 to 2010. Biogeosciences, 12, 4291–4316. https://doi.
org /10.5194/bg-12- 4291-2015
Mikoláš, M., Tejkal, M., Kuemmerle, T., Griffiths, P., Svoboda, M.,
Hlásny, T., … Morriss ey,R . C. (2017). Fore st manageme nt impact s
oncapercaillie(Tetrao urogallus)habitatdistributionandconnectiv-
ity in the Carpathians. Landscape Ecology, 32,163–179.https://doi.
org/10.1007/s10980-016-0433-3
NASA(2006).ShuttleRadarTopographyMission.Retrievedfromhttp://
www.jpl.nasa.gov/srtm
Navarro, L . M., & Pereira, H. M. (2012). Rewilding abandoned land-
scapes in Europe. Ecosystems, 15,900–912.https://doi.org/10.1007/
s10021-012-9558-7
Nelson,A.(2008).Traveltimetomajorcities:AglobalmapofAccessibilit y.
Ispra, Italy: Global Environment Monitoring Unit - Joint Research
Centre of the European Commission.
OakRidgeNationalLaboratory(2005).Highresolutionglobalpopulation
dataset. Retrieved from: http://web.ornl.gov/sci/landscan/
Paillet , Y.,Pe rnot, C., B oulanger, V., Debaive, N ., Fuhr, M., Gilg, O., &
Gosselin, F. (2015). Quantifying the recovery of old-growth at-
tributes in forest reserves: A first reference for France. Forest
Ecology and Management, 346, 51–64. http s://doi.org /10.1016/j.
foreco.2015.02.037
Parviainen, J. (1999). Strict forest reserves in Europe–efforts to en-
hance biodiversity and strengthen research related to natural for-
estsinEurope.InJ.Diaci(Ed.),Proceedings of the COST Conference:
Virgin forests and forest reserves in Central and East European
countries. History, present status and future development (pp. 145–
171). Ljubljana: Department of Forestry and Renewable Forest
Resources - Biotechnical Faculty.
Parviainen, J. (Ed.)(2000). COST Action E4. Forest reserves research net-
work. Luxembourg: European Commission.
Peterken, G. F. (1996). Natural woodland: Ecology and conservation in
northern temperate regions. Cambridge, UK: Cambridge University
Press.
Phillips,S.J.,Dudík,M.,Elith,J.,Graham,C.H.,Lehmann,A.,Leathwick,
J., & Ferrier, S. (2009). Sample selection bias and presence-only
distribution models: Implications for background and pseudo-
absence data. Ecological Applications, 19, 181–197. https://doi.
org/10.1890/07-2153.1
Po t apov,P. ,Ha n sen, M .C., L aesta d ius, L . ,Turu b anov a ,S. ,Ya ros h enko ,
A., Th ies, C., … Es ipova, E. (2017). T he last front iers of wilde r-
ness: Tracking loss of intact forest landscapes from 2000 to
2013. Science Advances, 3, e160 0821. ht tp s://doi.o rg/10.1126/
sci ad v.16 00 82 1
RDevelopmentCoreTeam(2017).R: A language and environment for sta-
tistical computing. R Foundation for Statistical Computing.
Schnit zler, A. (2014). Towards a new Eur opean wilde rness: Embr acing
unmanaged forest growth and the decolonisation of nature.
Landscape and Urban Planning, 126, 74–80. https://doi.org /10.1016/j.
landurbplan.2014.02.011
Soille, P., & Vogt, P. (2009). Morphological segmentation of binary
patterns. Pattern Recognition Letters, 30, 456–459. https://doi.
org /10.1016/j.patrec. 20 08.10.015
Thorn , S., Bässler, C., Bra ndl, R., Bu rton, P. J., Cahall , R., Campb ell, J.
L.,…Müller,J.(2018).Impactsofsalvageloggingon biodiversity:A
meta- analysis. Journal of Applied Ecology, 55, 279–289. ht tps ://doi.
org /10.1111/1365-2664.12945
Trabucco,A.,&Zomer,R.J.(2010).GlobalSoilWaterBalanceGeospatial
Database.CGIARConsortiumforSpatialInformation.In,Published
online, available from the CGIAR-CSI GeoPor tal. Retrieved from
http://www.cgiar-csi.org.
UNEP-WCMC&IUCN(2017).Protected Planet: The World Database on
Protected Areas (WDPA). Retrieved from www.protectedplanet.net.
Vacchiano, G., Garbarino, M., Lingua, E., & Motta, R. (2017). Forest
dynamicsanddisturbanceregimes in the ItalianApennines. Forest
Ecology and Management, 388, 57–66. https://doi.org/10.1016/j.
foreco.2016.10.033
14 
|
   SABATINI eT Al.
Vandekerkhove, K., De Keersmaeker, L., Menke, N., Meyer, P., &
Verschelde,P.(2009).Whennaturetakesoverfromman:Deadwood
accumulation in previously managed oakan dbe echwoodlan dsin
North- western and Central Europe. Forest Ecology and Management,
258,425–435.https://doi.org/10.1016/j.foreco.2009.01.055
Veen, P.,Fanta, J.,Raev, I.,Biriş, I.-A., de Smidt, J., & Maes, B.(2010).
Virgin forests in Romania and Bulgaria: Results of two national in-
ventory projects and their implications for protection. Biodiversity
and Conservation, 19, 1805–1819. https://doi.org/10.1007/
s10531-010-9804-2
Verkerk, P. J., Levers, C., Kuemmerle, T., Lindner, M., Valbuena, R.,
Verburg, P. H., & Zudin, S. (2015). Mapping wood production in
European forests. Forest Ecology and Management, 357, 228–238.
https://doi.org/10.1016/j.foreco.2015.08.007
Verkerk, P. J., Zanchi, G., & Lindner, M. (2014). Trade-offs be-
tween forest protection and wood supply in Europe.
Environmental Management, 53, 1085. ht tps ://doi.org/10 .1007/
s00267-014-0265-3
Zabel , F., P utzenlechn er,B ., & Mauser, W. (2014). Global agr icultural
land resources–a high re solution suitability evaluation and its per-
spectives until 2100 under climate change conditions. PLoS ONE, 9,
e107522. htt ps://doi.org /10.1371 /journal.pone.0107522
BIOSKETCH
Francesco Maria Sabatini is interested in the processes and de-
terminants underlying the distribution of plant biodiversity in
forests. Most of his research has been on old- growth and virgin
forests, regarded as reference systems for understanding base-
lines and processes related to natural disturbance regimes, forest
dynamic s, tree demography and carbon cycling.
Author contribuions: F.M.S. and T.K. conceived the ideas; S.B.,
E.B.,O.C.,N.D.,H.F.,M.G., N.G.,F.L.,I.M.D.,P.M.,R.M., S.M.,
M.M.,R.M.,G.M.,L.N.,M.P.P.Ó.,A.R.,B.S.,J.St.,M.S.,J.Sz.,O.-
P.T.,R.V.,T.V.andT.M.Z.collectedthedata;F.M.S.,C.L.,P.J.V.and
F.P.analysedthedata;F.M.S.,T.K.,P.J.V.,W.K.,M.L.andJ.B.led
the writing; all authors provided major input on the manuscript.
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Suppor ting Information section at the end of the article.
How to cite this article:SabatiniFM,BurrascanoS,KeetonWS,
etal.WhereareEurope’slastprimaryforests?Divers Distrib.
201 8; 0 0 :1–14 . ht tp s://doi.o rg /10.1111/ddi.12778
... ng ows of vital substances are in direct and dynamic connection. Consequently, it is essential to study and interpret these processes in an integrated conceptual framework. Furthermore, oakdominated mixed forests, characteristic of the Pannonian Ecoregion, are of particular interest, because referential old-growth stands are almost totally missing (Sabatini et. al. 2018(Sabatini et. al. , 2021. ...
... The main Natura 2000 habitat types are 9130 -Asperula-Fagetum beech forests, 91G0 -Pannonic woods with Quercus petraea and Carpinus betulus, 91H0 -Pannonian woods with Quercus pubescens, 91M0 -Pannonian-Balkanic Turkey oak-sessile oak forests (CEC 1992) The core area has been set aside for preservation in 1991, though several previous conservation acts had already secured spontaneous natural development of the VSFR stands (Aszalós et al 2017). Most of the stands are classi ed/considered as 'long untouched' forests according to Sabatini et al (2018). ...
Preprint
Full-text available
The long-term study of old-growth forest stands helps to understand the small-scale natural forest dynamic processes driven by the population biology of trees. The sampled trees can be considered from the viewpoint of their life and death cycle from establishment through to growing and death to full degradation. We provide a methodological framework for tracking and interpreting tree history in interaction with resurvey sampling events. The history of living and dead trees is consecutive and closely related to each other due to mortality events. The evanescent rate of disintegrated and decomposed dead trees is also a substantial dynamic characteristic of stands like regeneration, maturation, and mortality rates. They all form the intrinsic stand dynamic profile based on the whole life-dead-decomposed cycle of trees. This framework fills a knowledge gap in the European forest dynamic studies, where most of the research approaches point the focus on forest structure and pattern of developmental stages, and separately on living population dynamic features and deadwood structure and decay processes. In our pilot study, we performed resurvey of 53 Permanent Sampling Plots of a long untouched forest at Várhegy Strict Forest Reserve, Hungary. We estimated annual rates of ingrowth regeneration, maturation, mortality (13.2, 1.5, and 10.0 tree/hectare/year respectively), and 2.4 tree/hectare/year evanescence of disintegrated and decomposed dead trees. Apparent trends that accumulation of dead trees increases considerably – especially in oak-hornbeam stands, while 67% of annual regeneration ingrowth density is produced by shrubs, and only 9% by upper canopy trees.
... This anthropogenic effect will erase the inherited topography and reset the palimpsest of the microtopography of soils. Among the cultivable lands in western Europe, it is very difficult, except for mountain areas, to find areas that have not experienced past agricultural activity (Sabatini et al. 2018). In contrast, in a forest context characterized by a dense and continuous vegetation cover, the soil surface is protected/stabilized and therefore the conservation of its landforms is better. ...
... This work also reveals that the past agricultural use in the now-forested lands of northwest Europe can still be characterized owing to this very gradual acidification of the soil. Lowland temperate forests with a primaeval character no longer exist in Europe (Peterken, 1996) and most European forests have been cultivated in a more or less distant past (Sabatini et al. 2018). Thus, soil pH profiles may be an indicator of afforestation oldness in many forested areas. ...
Chapter
Full-text available
The interpretation of soil archives, although complex, is nevertheless essential for understanding the impact of soils on present‐day environments and deciphering the environmental history of a site. Soil archives are local, although within a continuous soilscape, they allow the study of spatial structures. This chapter presents some key elements for understanding the archiving process through to the reconstruction of a fragment of paleoenvironmental history. It describes the biological, physical and anthropological archiving elements of pedological processes. Geomorphological processes also affect the soilscape and can produce excellent soil and sediment archives or destroy them. The combined study of the different soil archives (plant remains, soil organic matter, chemical parameters, macro‐ and micromorphological features, microtopography, archaeology) allows an increasingly less lacunary picture to emerge of the environmental history and its legacies in the soil; an environmental story whose heritage, while fading, still structures today's terrestrial ecosystems.
... Due to their carbon storage capacity, they play a key role in the global carbon cycle [2,3], helping to conserve and protect soil resources, and last but not least, enhancing the beauty of the landscape [4]. Picea abies L., known as Norwegian spruce, is one of the most important species of conifers but also one of the dominant species in the forests of the Carpathian Mountains, which is one of the largest forest-covered regions of the continent [5,6]. In addition, it is considered one of the most economically important species in Central and Northern Europe [7], extending to more than 30 million hectares [8]. ...
Article
Full-text available
The aim of this research was to investigate the influence of an electric field and gamma radiation upon the germination of spruce seeds. In order to carry out the research, spruce seeds from different provenances have been subjected to different treatments: electric field (EF) with 10 V, 30 V, and 50 V voltages and intensity of E = 266V/m, exposure time of 15 and 35 min, and gamma (G) radiation with several treatments (1 Gy-31 min, 1.5 Gy-46 min, 2 Gy-62 min, and 6 Gy-186 min). Under the influence of EF, the best results upon seed germination (80.83%) were recorded when seeds were treated with 30 V for 15 min, for all provenances investigated. Regarding gamma radiation, the highest germination percentage (87.50%) was achieved in T5G when seeds were subjected to 6 Gy for 186 min. It was also considered the interaction between seeds origin and the different EF and G treatments applied to the seeds to induce germination and further seedlings’ development. The results obtained after seeds were exposed to gamma radiation came out on top compared to electric field treatments, both for the germination and seedlings’ height.
... Most of the primary beech and beech-dominated forest remnants are located in the Carpathians-Slovakia, Ukraine and Romania-and in the mountain ranges of the western Balkans-Slovenia, Bosnia and Herzegovina and Albania [10]. According to Korpel' [11,12], the primary beech forests of eastern Slovakia belong to the best-preserved primary forests, not only in Slovakia, but also throughout Europe. ...
Article
Full-text available
The study investigates the links and interactions between soil properties, soil microorganisms and the structure of a primary beech forest. The study was performed in the reserve Havešová (Bukovské vrchy Mts., Slovakia). On 40 sampling plots, soil samples from the O-horizon and from the first 10 cm of the organo-mineral horizons were taken to analyze the physico-chemical and biological properties. Moreover, stand structural characteristics (volume of trees, additive stand density index, coefficient of homogeneity, tree influence potential, development stage indices, etc.) were measured and calculated. In general, we did not observe any strong effects of forest structure on the topsoil characteristics. The effect of stand structure was more reflected in the physico-chemical properties than in the biological attributes. We found that the P and K content in the forest floor increased at plots with a higher volume or density of trees per plot. Moreover, a positive correlation was found also between the K content and tree influence potential. The development stages expressed by the indexes based on the diameter structure were reflected especially by the soil reaction in the A-horizon. Within functional groups of microorganisms based on the Biolog assay, significant differences were found, especially in the utilization of D-cellobiose, which positively correlated with the presence of the optimum stage index. The effect of soil physico-chemical properties on biological indicators was more pronounced than the effect of stand structure.
... In Europe, forests are now mainly seminatural, with undisturbed forests accounting for just 4% of all forest area [24]. Further, only 0.7% of European forests are undisturbed forests composed of native species [21], and just 46% of these forests are strictly protected [25]. ...
Article
Full-text available
The protection and conservation of old-growth forests (OGFs) are becoming a global concern due to their irreplaceability and high biodiversity. Nonetheless, there has been little research into the identification and characterization of OGFs of the oldest tree species in Mediterranean areas. We used forest inventory data, low-density airborne laser scanning (ALS) metrics, and geostatistical analysis to estimate old-growth indices (OGIs) as indicators of old-growth forest conditions. We selected a pilot area in European black pine (Pinus nigra subsp. salzmannii) ecosystems where the oldest known living trees in the Iberian Peninsula are found. A total of 756 inventory plots were established to characterize standard live tree and stand attributes. We estimated several structural attributes that discriminate old growth from younger age classes and calculated different types of OGI for each plot. The best OGI was based on mean tree diameter, standard deviation of tree diameter, and stand density of large trees (diameter > 50 cm). This index is useful for assessing old-growthness at different successional stages (young and OGFs) in Mediterranean black pine forests. Our results confirm that the estimation of OGIs based on a combination of forest inventory data, geostatistical analysis, and ALS is useful for identifying OGFs.
... Our study was conducted in Białowieża Primeval Forest (approximately 1500 km 2 ), one of the best preserved forests in Europe (Jaroszewicz et al., 2019;Sabatini et al., 2018). This forest massif stretches over the border between Poland and Belarus, in the eastern part of the Central European Lowland (52 o 41 ′ N; 23 o 49 ′ E) and covers the flat plain with a mean altitude of 170 m a.s.l. ...
Article
Full-text available
Mushrooms play an important role in the maintenance of ecosystem processes and delivering ecosystem services, including food supply. They are also an important source of income for many people worldwide. Thus, understanding which environmental factors influence mushroom productivity is a high practical and scientific priority. We monitored the production of mushrooms in temperate mixed deciduous forest in Białowieża Primeval Forest in eastern Poland for two yielding seasons. The research plots were set under similar environmental conditions (topography, geology, soil type) but differed by tree species composition and tree species richness. The main factor explaining mushroom production (close to 35% of the variation explained by the model) was the species richness of mushrooms. In turn, the species richness of mushrooms was mainly explained by soil properties (pH and C/N ratio) and stand characteristics (including tree species richness and wood increment) for ectomycor-rhizal mushrooms and by soil pH for saprotrophic mushrooms. Higher precipitation in 2021 resulted in higher mushroom production than in 2020, while low levels of precipitation in 2020 resulted in stronger effect of ambient temperature. The differences in mushroom yield between years varied highly among plots. They were explained by stand characteristics, and in the case of saprotrophic mushrooms by tree richness and their own species richness. Our results suggest that promoting mushroom species richness is fundamental for increasing mushroom yield and should be taken into account in forest management.
... Unmanaged forests or forests managed for biodiversity are more attractive to people 18 . Only 0.7% of Europe's forest area is classified as primary forest 19 , indicating that most European Table 1. Attributes and levels used in choice experiments to investigate public forest preferences in Poland and Norway. ...
Article
Full-text available
Recreation is a crucial contribution of nature to people, relevant for forest ecosystems. Large carnivores (LCs) are important components of forests, however, their contribution to forest recreational value has not yet been evaluated. Given the current expansion of LC populations, the ongoing forest conservation debate, and the increasing use of nature for recreational purposes, this is a timely study. We used discrete choice experiments and willingness-to-travel to determine people’ preferences for both forest structural characteristics and presence of four LC species in Poland (N = 1097 respondents) and Norway (N = 1005). In both countries, two-thirds of the respondents (termed ‘wildness-positive’) perceived LCs as contributing positively to forest recreational value and preferred to visit old forests with trees of different species and ages and presence of dead wood (i.e. natural forests). Respondents with negative preferences towards LCs preferred more intensively managed forest (‘wildness-negative’); their preferences were stronger than in wildness-positive respondents and in Norway. Preferences towards wild nature were highly polarized and there were hardly neutral people. Our results showed a strong link between preferences for LC presence and forest structure, and reflected the dualism of human-nature relationships. This study highlights the need to consider the contribution of forests and LCs to human recreation services in ecosystem management policies.
Article
Full-text available
Ecological knowledge on long-term forest dynamics and development has been primarily derived from the study of old-growth forests. Centuries of forest management have decreased the extent of temperate old-growth forests in Europe and altered managed forests. Disentangling the effects of past human disturbances and climate on current species composition is crucial for understanding the long-term development of forests under global change. In this study, we investigated disturbance and recruitment dynamics in two forests in the Western Pyrenees (Spain) with contrasting management history: an old-growth forest and a long-untouched forest, both dominated by the two shade-tolerant species Fagus sylvatica (European beech) and Abies alba (Silver fir). We used dendroecological methods in seven plots to analyse forest structure, growth patterns and disturbance histories in these forests. We benchmarked these data with the dynamic vegetation model ForClim to examine the effects of natural and human-induced disturbances on forest development, structure and species composition. Disturbance regimes differed between the study forests, but none showed evidence of stand replacing disturbances, either natural or human induced. Low disturbance rates and continuous recruitment of beech and fir dominated the old-growth forest over the last 400 years. In contrast, the long-untouched forest was intensively disturbed in 1700-1780, probably by logging, with lower natural disturbance rates thereafter. Beech and fir recruitment preferentially occurred after more intense disturbances, despite the high shade tolerance of both beech and fir. Higher fir abundance in the long-untouched forest than in the old-growth forest appeared to be related to its human-induced disturbances. ForClim closely simulated forest potential natural vegetation with a dominance of beech over fir, but overestimated the presence of less shade-tolerant species. Previously observed local fir decline may result from natural forest successional processes after logging. Within ~200 years after logging cessation, some long-untouched forest structural attributes converged towards old-growth forest, but legacy effects still affected species composition and structure. Natural disturbance regimes in beech-fir forests of the Western Pyrenees induce temporal fluctuations between beech and fir abundance, with a natural tendency for beech dominance in advanced developmental stages with low disturbance rates.
Book
Full-text available
Towards a coherent protected area network – Priorities of protecting biodiversity in Finland (In Finnish with an English abstract) The objective of the Towards a coherent protected area network (KOKASU) project was to collect data for the national definition and implementation of the EU Biodiversity Strategy, which sets the target of protecting 30% of land and sea areas. The project gathered data produced by research and study projects and various processes and tapped a wide range of GIS data sources on the Finnish protected area network and its development needs. A key objective was investigating the current conservation status of the main biotopes and threatened species in land and sea areas and identifying gaps in the protection of biodiversity, in other words ecological features that are underrepresented in the Finnish network of protected areas, considering their threatened status or conservation needs identified by other means. In addition, an effort was made to outline regional priorities for protecting additional areas with the aim of improving the coherence and connectivity of the existing protected area network. The following main habitats were identified in the project: 1) forests, 2) mires, 3) rocky habitats, 4) coastal habitats, 5) fells, 6) traditional rural biotopes, 7) inland waters and shores, and 8) the Baltic Sea. Major variations regarding the need to protect different biotypes and species and conservation methods were found between these biotopes and species. For each habitat covered by the KOKASU report, gaps were found in Finland's current network of protected areas. A large number of threatened biotypes and species is found outside protected areas, and the report proposes urgently complementing the network of protected areas with them and, where possible, also improving connectivity between concentrations of biotopes and species. The report additionally identifies significant needs for ecological management and restoration of species and biotopes. Attention should also be paid to connections between habitats. For example, measures taken in the catchment areas of inland waters, especially those aiming to protect forest and mire biodiversity, also play a key role in improving the state of these waters. Southern and Central Finland are key areas for efforts to develop Finland's terrestrial protected area network. The areas with the greatest conservation needs associated with forests, mires, rocky habitats and scree, inland waters and shores as well as traditional rural biotopes are found to the south of Forest Lapland (in hemiboreal, southern boreal and middle boreal zones, and for the part of mires, in southern parts of the north boreal zone). Coastal and marine habitats in need of protection are found in the Gulf of Finland, the Åland Islands, the Archipelago Sea, Kvarken and the Bay of Bothnia. A significant part of the conservation needs in both land and sea habitats are associated with areas in private ownership. Consequently, incentives and different policy instruments for protecting private land and water areas are an important area of development. More detailed conservation needs related to certain better known biotopes or species could be identified regionally, however taking into consideration the bias of biotope and species inventory data, in other words their focus on protected areas. The best data are produced by methods that address the complementary nature and connectivity of sites, including the Zonation analyses carried out for the Baltic Sea in the KOKASU project. Due to bias and gaps in the data, presenting specific protection needs in hectares is difficult. This is why the KOKASU report refrains from setting conservation objectives in hectares. Instead, the report proposes measures for each habitat type with the aim of developing the Finnish protected area network and improving the status of biodiversity. The report also highlights some of the most significant information gaps. Keywords: Baltic Sea, biodiversity strategy, coastal habitats, fell habitats, forests, habitat type, inland waters, mires, protected areas, rocky habitats, seminatural grasslands, shores, species.
Article
Full-text available
Over the past 15 years, 3 million hectares of forests have been converted into shrublands or grasslands in the Mediterranean countries of the European Union. Fire and drought are the main drivers underlying this deforestation. Here we present a conceptual framework for the process of fire-induced deforestation based on the interactive effects of fire and drought across three hierarchical scales: resistance in individuals, resilience in populations, and transitions to a new state. At the individual plant level, we review the traits that confer structural and physiological resistance, as well as allow for resprouting capacity: deforestation can be initiated when established individuals succumb to fire. After individuals perish, the second step toward deforestation requires a limited resilience from the population, that is, a reduced ability of that species to regenerate after fire. If individuals die after fire and the population fails to recover, then a transition to a new state will occur. We document trade-offs between drought survival and fire survival, as embolism resistance is negatively correlated with fire tolerance in conifers and leaf shedding or drought deciduousness, a process that decreases water consumption at the peak of the dry season, temporally increases crown flammability. Propagule availability and establishment control resilience after mortality, but different hypotheses make contrasting predictions on the drivers of post-fire establishment. Mycorrhizae play an additional role in modulating the response by favoring recovery through amelioration of the nutritional and water status of resprouts and new germinants. So far, resprouter species such as oaks have provided a buffer against deforestation in forests dominated by obligate seeder trees, when present in high enough density in the understory. While diversifying stands with resprouters is often reported as advantageous for building resilience, important knowledge gaps exist on how floristic composition interacts with stand flammability and on the “resprouter exhaustion syndrome,” a condition where pre-fire drought stress, or short fire return intervals, seriously restrict post-fire resprouting. Additional attention should be paid to the onset of novel fire environments in previously fire-free environments, such as high altitude forests, and management actions need to accommodate this complexity to sustain Mediterranean forests under a changing climate.
Article
Full-text available
Logging to "salvage" economic returns from forests affected by natural disturbances has become increasingly prevalent globally. Despite potential negative effects on biodiversity, salvage logging is often conducted, even in areas otherwise excluded from logging and reserved for nature conservation, inter alia because strategic priorities for post-disturbance management are widely lacking. A review of the existing literature revealed that most studies investigating the effects of salvage logging on biodiversity have been conducted less than 5 years following natural disturbances, and focused on non-saproxylic organisms. A meta-analysis across 24 species groups revealed that salvage logging significantly decreases numbers of species of eight taxonomic groups. Richness of dead wood dependent taxa (i.e. saproxylic organisms) decreased more strongly than richness of non-saproxylic taxa. In contrast, taxonomic groups typically associated with open habitats increased in the number of species after salvage logging. By analysing 134 original species abundance matrices, we demonstrate that salvage logging significantly alters community composition in 7 of 17 species groups, particularly affecting saproxylic assemblages. Synthesis and applications. Our results suggest that salvage logging is not consistent with the management objectives of protected areas. Substantial changes, such as the retention of dead wood in naturally disturbed forests, are needed to support biodiversity. Future research should investigate the amount and spatio-temporal distribution of retained dead wood needed to maintain all components of biodiversity.
Article
Full-text available
Forest management practices emphasizing stand structural complexity are of interest across the northern forest region of the United States because of their potential to enhance carbon storage. Our research is part of a long-term study evaluating silvicultural treatments that promote late-successional for- est characteristics in northern hardwood-conifer forests. We are testing the hypothesis that aboveground biomass development (carbon storage) is greater in structural complexity enhancement (SCE) treatments when compared to conventional uneven-aged treatments. Structural complexity enhancement treatments were compared against selection systems (single-tree and group) modified to retain elevated structure. Manipulations and controls were replicated across 2-ha treatment units at two study areas in Vermont, United States. Data on aboveground biomass pools (live trees, standing dead, and downed wood) were collected pre- and post-treatment, then again a decade later. Species group-specific allometric equations were used to estimate live and standing dead biomass, and downed log biomass was estimated volumetri- cally. We used the Forest Vegetation Simulator to project “no-treatment” baselines specific to treatment units, allowing measured carbon responses to be normalized against differences in site characteristics affecting tree growth and pre-treatment stand structure. Results indicate that biomass development and carbon storage 10 yr post-treatment were greatest in SCE treatments compared to conventional treatments, with the greatest increases in coarse woody material (CWM) pools. Structural complexity enhancement treatments contained 12.67 Mg/ha carbon in CWM compared to 6.62 Mg/ha in conventional treatments and 8.84 Mg/ha in areas with no treatment. Percentage differences between post-treatment carbon and simulated/projected baseline values indicate that carbon pool values in SCE treatments returned closest to pre-harvest or untreated levels over conventional treatments. Total carbon storage in SCE aboveground pools was 15.90% less than that projected for no-treatment compared to 44.94% less in conventionally treated areas. Results from classification and regression tree models indicated treatment as the strongest predictor of aboveground C storage followed by site-specific variables, suggesting a strong influence of both on carbon pools. Structural enhancement treatments have the potential to increase carbon storage in managed northern hardwoods. They offer an alternative for sustainable management integrating carbon, associated climate change mitigation benefits, and late-successional forest structure and habitat.
Article
Full-text available
An intact forest landscape (IFL) is a seamless mosaic of forest and naturally treeless ecosystems with no remotely detected signs of human activity and a minimum area of 500 km². IFLs are critical for stabilizing terrestrial carbon storage, harboring biodiversity, regulating hydrological regimes, and providing other ecosystem functions. Although the remaining IFLs comprise only 20% of tropical forest area, they account for 40% of the total aboveground tropical forest carbon. We show that global IFL extent has been reduced by 7.2% since the year 2000. An increasing rate of global IFL area reduction was found, largely driven by the tripling of IFL tropical forest loss in 2011–2013 compared to that in 2001–2003. Industrial logging, agricultural expansion, fire, and mining/resource extraction were the primary causes of IFL area reduction. Protected areas (International Union for Conservation of Nature categories I to III) were found to have a positive effect in slowing the reduction of IFL area from timber harvesting but were less effective in limiting agricultural expansion. The certification of logging concessions under responsible management had a negligible impact on slowing IFL fragmentation in the Congo Basin. Fragmentation of IFLs by logging and establishment of roads and other infrastructure initiates a cascade of changes that lead to landscape transformation and loss of conservation values. Given that only 12% of the global IFL area is protected, our results illustrate the need for planning and investment in carbon sequestration and biodiversity conservation efforts that target the most valuable remaining forests, as identified using the IFL approach.
Article
Full-text available
The United Nations Food and Agriculture Organization (FAO) has been reporting country-level area in primary forests in its Global Forest Resource Assessment (FRA) since 2005. The FAO definition of a primary forest (naturally regenerated forest of native species where there are no clearly visible indications of human activities and the ecological processes are not significantly disturbed) is generally accepted as authoritative and is being used in policy making. However, problems with this definition undermine our capacity to obtain globally-coherent estimates. In addition, the current reporting on primary forests fails to consider the complementarity of non-primary forests towards the maintenance of ecosystem services. These issues undermine the appropriate tracking of changes in primary and non-primary forests, and the assessment of impacts of such changes on ecosystem services. We present the case for an operational reconsideration of the primary forest concept and discuss how alternatives or supplements might be developed. This article is protected by copyright. All rights reserved.