Mirel Beloiu ,Dimitris Poursnidis ,Antonis Tskirkis ,Nektrios Chrysoulkis ,Smuel Hoffmnn ,Petros Lymerkis,d ,Antonis Brnis,Dvid Kienle,Crl Beierkuhnlein,e,f

a Department of Biogeography,University of Bayreuth,Universit¨atsstra?e 30,95447,Bayreuth,Germany

b Remote Sensing Lab,Institute of Applied and Computational Mathematics,Foundation for Research and Technology Hellas,100 N.Plastira Str.,Vassilika Vouton,Heraklion,70013,Greece

c Samaria National Park,Management Body,Old National Road Chanion-Kissamou,Agioi Apostoloi,Chania,Crete,Greece

d Natural History Museum of Crete,University of Crete,Knossou Av.,71409,Irakleio,Crete,Greece

e GIB Department of Geography,University of Bayreuth,95447,Bayreuth,Germany

f BayCEER Bayreuth Center of Ecology and Environmental Research,University of Bayreuth,95448,Bayreuth,Germany

Keywords:Aerial imagery Protected area Continental island Mediterranean region High mountains Temperature Precipitation Climate change Forest dynamics

ABSTRACT Background: The recent rise in temperature and shifting precipitation regimes threaten ecosystems around the globe to different degrees.Treelines are expected to respond to climate warming by shifting to higher elevations,but it is unclear whether they can track temperature changes.Here,we integrated high-resolution aerial imagery with local climatic and topographic characteristics to study the treeline dynamic from 1945 to 2015 on the semiarid Mediterranean island of Crete,Greece.Results: During the study period,the mean annual temperature at the treeline increased by 0.81 °C,while the average precipitation decreased by 170 mm.The treeline is characterized by a diffuse form,with trees growing on steep limestone slopes (>50°) and shallow soils.Moreover,the treeline elevation decreases with increasing distance from the coast and with aspect (south >north).Yet,we found no shift in the treeline over the past 70 years,despite an increase in temperature in all four study sites.However,the treeline elevation correlated strongly with topographic exposure to wind (R2=0.74, p <0.001).Therefore,the temporal lag in treeline response to warming could be explained by a combination of topographic and microclimatic factors,such as the absence of a shelter effect and a decrease in moisture.Conclusion:Although there was no treeline shift over the last 70 years,climate change has already started shifting the treeline altitudinal optimum.Consequently,the lack of climate-mediated migration at the treeline should raise concerns about the threats posed by warming,such as drought damages,and wildfire,especially in the Mediterranean region.Therefore,conservation management should discuss options and needs to support adaptive management.

1.Background

Over the past century,mean annual temperatures have risen globally(Seneviratne et al.,2014),with pronounced warming trends and rapid biodiversity changes occurring at high elevations and latitudes(Ohmura,2012;Garcia et al.,2014;Lamprecht et al.,2018;Ripple et al.,2020;Vitasse et al.,2021).The Mediterranean region of Europe is particularly sensitive to global warming as it is located in a transition zone between the semi-arid regime of north Africa and the temperate humid regime of central Europe (Giorgi and Lionello,2008).The Mediterranean region was already affected by severe heat waves and forest fires in the summer of 2021 (Copernicus Land Monitoring Service,2021).The European Mediterranean biome is a hotspot of endemic species(Myers et al.,2000)and is home to more tree species than central-northern Europe(Svenning and Skov,2005).Particularly mountain regions and islands in this hotspot are of great importance for biodiversity as endemism increases with elevation and geographical isolation (Steinbauer et al.,2013,2016).Climate change is expected to shift the species elevational optimum(Vitasse et al.,2021),yet it is not clear if trees can track these changes(Hof et al.,2011).Studies showed that the observed tree species shift,is smaller than the expected or predicted shift(Chen et al.,2011;Zhu et al.,2012;Vitasse et al.,2021).Unlike animals,trees cannot migrate by themselves.They have a long lifespan,and their dispersal and establishment at the treeline is rather limited (Dullinger et al.,2004;Neuschulz et al.,2018),therefore it is unclear whether trees can track recent temperature changes.If trees do not adapt to these changes,consequential damage from drought,wildfires,high evapotranspiration and low soil water can be expected(Ven¨al¨ainen et al.,2014;Coulthard et al.,2017;Trotsiuk et al.,2021).

Despite the large numbers of studies on treeline dynamics,continental islands are underrepresented in treeline studies;instead,most studies focus on mainland areas (Bogaert et al.,2011;Mathisen et al.,2014;Suwal et al.,2016;Zindros et al.,2020).A fundamental difference between treelines on islands and the continents is the higher degree of ecological isolation resulting in a high proportion of endemism and less ecological niche occupancy toward the summits of islands (Steinbauer et al.,2016).On the one hand,given their legacy of isolation,edaphic and geological conditions,continental Mediterranean islands are expected to host less adapted tree species compared to mainland areas(Irl et al.,2016).On the other hand,these conditions qualify for strict protection which is also expected to facilitate the expansion of forest cover(Leberger et al.,2020).The subalpine and alpine zone of the Mediterranean and the continental island of Crete are centers of endemism and thus of substantial importance for the Mediterranean biodiversity hotspot (Myers et al.,2000;Vogiatzakis et al.,2003,2016;Kazakis et al.,2007;Spanos et al.,2008).However,an upward shift of treelines can also lead to the reduction of habitats for endemic alpine species(Kidane et al.,2019).Therefore,such processes and possible dynamics need to be monitored precisely.

The upper treeline that is separating the montane and the alpine elevation zone (“alpine treeline”) is a fundamental ecological and conspicuous physiognomic boundary along the elevation gradient in high mountains.The treeline ecotone can range from rather abrupt forest margins to“islands”(group of trees),and to diffuse transition zones including in some cases a zone with characteristic shrub life forms(“krummholz”)(K?rner and Paulsen,2004;Holtmeier,2009;Harsch and Bader,2011).Causes and processes resulting in a sharp transition between forest and open ecosystems are multifold and still under debate(Jobbágy and Jackson,2000;Holtmeier,2009;K?rner,2012;Irl et al.,2016).In many cases,the treeline ecotone has not a clear boundary but it is characterized by a gradual decrease in tree height and density as elevation increases(Harsch and Bader,2011).Obviously,the advantage of“being a tree”comes at the treeline to an end.

The treeline elevations worldwide increase from oceanic islands followed by continental (shelf) islands,to continents (mainland areas) (Irl et al.,2016;Karger et al.,2019).The latitudinal gradient of treeline elevation demonstrates the dependence of treeline elevations on climate.As the establishment of trees beyond the alpine and polar treelines is mainly caused by temperature-driven growing conditions(K?rner,2012;Irl et al.,2016;Karger et al.,2019),climate warming is expected to shift treelines to higher elevations and latitudes,particularly in the Northern Hemisphere (Gatti et al.,2019;Lu et al.,2021).The minimum growth temperature is currently the most prominent explanation of global treeline patterns (K?rner,2012):sapling survival and regeneration are limited since they are atmospherically coupled and touch higher boundaries of air masses under colder conditions.Recent studies have shown that precipitation has a comparable or stronger effect than temperature (Moyes et al.,2015;Sigdel et al.,2018;Rees et al.,2020;Lu et al.,2021).This can be explained by the fact that the treeline isotherm is shifting upslope or poleward with warming,meaning that tree growth rates and tree establishment might be enhanced within this limit if sufficient precipitation and moisture are available (Camarero et al.,2021;K?rner,2021).

In addition to climate,the effects of local site conditions,such as the shelter effect should also be considered (Holtmeier and Broll,2007;McIntire et al.,2016).Disturbance regimes (Jentsch and Beierkuhnlein,2003) and topographical characteristics can influence the local conditions,leading to local treeline dynamic patterns(Salzer et al.,2014;Vitali et al.,2018).Moreover,microhabitats e.g.rocky outcrops with consequences on isolation and snow cover are modifying zonal conditions for a given elevation(Batllori et al.,2009;Scherrer and K?rner,2010;McIntire et al.,2016;Cudlín et al.,2017).Hence,different treeline structure is a result of the combined effect of topography and human influence,however,there is no doubt about the essential influence of long-term climatic conditions on the height of the treeline.Therefore,treelines are seen as a sensor for climatic changes(Suwal et al.,2016;Gatti et al.,2019).

Earth observation and precise geospatial information are key for understanding current patterns of important ecosystem boundaries,which is the precondition to monitor and analyze future changes.Aerial imagery,multispectral satellite data,and airborne-based lidar measurements allow us to identify individual trees and can be utilized to investigate tree growth in inaccessible and remote areas over time,improving the ability to study and understand their dynamics (Chen et al.,2015;Bolton et al.,2018;Hoffmann et al.,2018b).However,these are usually limited in their spatial and temporal coverage,e.g.Landsat MSS data availability begins from the 1970s but spatially too coarse (60 m) to accurately identify trees.Historical aerial imagery from 1945 provides a unique and powerful data source on the land cover that enables long-time studies.Nevertheless,historical aerial imageries at high resolution(1.83-2.74 m) are still rarely used in forest cover changes (Nita et al.,2018;Rendenieks et al.,2020).Nonetheless,they are especially useful when trying to understand past conditions from periods before satellite imagery was available.

In this study,we used both historical imagery from 1945 and new imagery (2008 and 2015) to identify treeline shifts in remote,inaccessible,and protected areas of the steep mountains of Crete over a period of 70-year of anthropogenic climate change.We combined high-resolution aerial imagery with local climatic and topographic characteristics to investigate the temporal dynamics of realized treeline elevation.The study area is established in the Lefka Ori mountains(known also as White Mountains) of the renowned Samaria National Park on Crete,a Mediterranean and continental island of Greece (Fig.1).We specifically address the following questions:1)Did the realized treeline change over time?2) To which degree is the recent continental island treeline influenced by distance to the coast and wind exposure? 3) Is there an asymmetric position of treelines related to aspect?To answer these questions,we mapped the realized treeline elevation over time by using aerial images from 1945,2008,and 2015;all images were ortho-rectified and capable of change detection at a fine scale.We then compared the realized treeline elevation between the years on four study locations with distinct topography.

2.Methods

2.1.Study area

The current area of the Samaria National Park(58,454 ha)is a multidesignated protected area.In 1962 Samaria was declared a National Park.Since then,it was additionally designated as a Man and Biosphere Reserve by UNESCO,and a Diploma of Protected Areas by the Council of Europe.It also includes two sites that belong to EU’s Natura 2000 protected area network and is also an Important Bird Area of Greece,alongside other numerous national designations.The National Park’s area ranges in elevation from sea level up to over 2,400 m a.s.l.(highest peak is at 2,454 m a.s.l.)with over 50 of its peaks exceeding the 2,000 m a.s.l.range.It is dissected by nine main gorges,with a predominant north to south direction,the longest of which is the Gorge of Samaria,with a length of 13 km.The protected areas of Samaria host a unique diversity of priority species in the European context(Hoffmann et al.,2018a).Until the mid-1950’s the area included uncontrolled pastoral pressure,logging,seasonal fires for the establishment of pasture lands,and small-scale agricultural activities.The main activities today (anthropogenic pressures) are tourism and semi-extensive,free-range pastoralism,however,grazing is limited within the Samaria National Park(Spanos et al.,2008).The abandonment of these lands led to the natural reforestation of old pastoral and agricultural plots with pines and cypresses (Pinus brutia,Cupressus sempervirens) (Papanastasis,2004).In contrast to the broadleaved evergreens of the area (e.g.,Quercus coccifera,Pistacia lentiscus),these species are resistant to grazing pressures and thus form the main forest communities.Of the two conifer species present,cypresses are the ones that form the timberline (Spanos et al.,2008).Some of the other prominent species of the alpine areas of the Lefka Ori mountains in the limits of Samaria National Park are the shrubsBerberis cretica(up to 1 m height),Prunus prostrata(～30 cm height),andSatureja spinosa(～12 cm height).The Samaria National Park and its location in Crete and Greece,along with the regions selected for this study,can be seen in Fig.1.

In total,four study sites were selected,in the N,W,S,and E areas of the Samaria National Park,and all entirely within its borders.A few potential study areas had to be excluded from the comparative survey due to the poor quality of some historical aerial photos from 1945.Reasons for this exclusion were strong shading due to extreme inclinations or artifacts due to damage to the initial aerial photo slides.Mediterranean mountain ecosystems have been exposed to millennia of human pressures and land-use changes resulting in soil erosion and degradation (Shakesby,2011;Riva et al.,2017).The investigation includes a variety of anthropogenic pressures,with currently low human pressure areas in northern,western,and southern study sites (N,S,and W)and more pronounced human influences on the eastern study site(E).Tree location was mapped over a 5-km transect along the treeline,with a mean distance between study sites of>6 km.The selected study sites are shown in Fig.1.

The studied treeline elevations ranged from 1,270 to 1,884 m a.s.l.with the lowest treeline elevation on the northern(N)site and the highest treeline elevation on the southern(S)site(Table 1).The northern study site (N) is the most remote concerning road infrastructure and human activities,both historically and presently,with a treeline ranging from 1,250 to 1,550 m a.s.l..The southern study site (S) is characterized by very steep topography and is distant to human activities and pressures,both historically and presently.Dense canopy closure appears on areas below the treeline,signifying forest-line thickening and the treeline appears relatively spread in the 1,500-1,900 m a.s.l.range.In contrast,the eastern study site (E) is the region closest to roads and human settlements,on both study periods,with partial canopy closure on some areas below the treeline,which appears predominantly in the 1,200-1,550 ma.s.l.range.The western study site(W)is close to the Samaria Gorge(the core of the Samaria National Park).At the W site,the treeline is more concentrated in the 1,450-1,850 m a.s.l.range.

Table 1 Tree number for each study site(N=north,S=south,E=east,W=west)per year(1945,2008,2015).The individual trees were mapped on the aerial images and their elevation was calculated based on the Digital Elevation Model(DEM).

Fig.1.(a)Treeline in the four study sites(N=north,E=east,W=west,S=south)of the Samaria National Park in Greece.(b)View of the treeline from the southern slopes of the Samaria Gorge.(c) Tree growing on steep slopes of the Lefka Ori mountains (White Mountains).

2.2.Climate data

We used monthly data provided by the National Oceanic and Atmospheric Administration (NOAA) for the Souda meteorological station(148 m,at a distance of 10 km from the study area),Crete,Greece(NOAA/NCDC,2019).The mean annual temperature (°C) and the average precipitation (mm) were calculated between 1979 and 2020.Climate station data for a longer period and a higher elevation were not available for the study region.However,to account for the climatic conditions at the treeline,we extracted the mean annual temperature(°C)and the average precipitation(mm)from the CHELSA(Climatologies at high resolution for the earth’s land surface areas) timeseries dataset(1979-2013).The Chelsa dataset has a resolution of～1 km(Karger et al.,2017,2018).This dataset has the advantage of capturing the climate trend from the treeline.Moreover,the precipitation algorithm includes wind fields,valley exposition,and boundary layer height (Karger et al.,2017).Local weighted regression (LOESS) lines were used to represent temperature and precipitation trends over the years.The De Martonne aridity index(IDM)was calculated based on the annual temperature and precipitation(1979-2013)from the treeline using the following formula:

We used this index to determine if climatic conditions at the treeline have changed,as high values indicate more humid conditions and low values indicate drier conditions.

2.3.Aerial images

The availability of the aerial imagery for three reference years(1945,2008,and 2015) from the archive of the National Cadastre &Mapping Agency S.A(Ktimatologio,2016)allows the use of greyscale(8 bit)and color (24 bit) orthorectified and co-registered imagery with Root Mean Square Error (RMSE) of RMSEx ≤1.00 m,RMSEy ≤1.00 m RMSExy≤1.41 m and an RMSE of ≤2.44 m,for CE 95%.The images from 1945 have a resolution of 2 m,while those from 2008 and 2015 have a spatial resolution of 1 m and 25 cm.The historical aerial images (1945) have been further co-registered with the imagery of 2008 and 2015 to eliminate slight shifts in flight lines.Characteristic landscape features,e.g.,rock formations,were used,and,for each image,5 to 6 such reference ground control points were selected,resulting in an RMSE<4 m for(CE 95%).In each region,the tree individuals found at the highest elevation and having a size of 2 m,were selected manually,using visual interpretation of the ortho-rectified aerial images by expert forest scientists,using QGIS Desktop Software v.2.18.20.We thus measure treeline elevation by the elevation of individual trees that form the treeline.These treeline measurements were grouped according to their study periods(1945,2008,and 2015).We mapped over 550 individual trees(Table 1).We considered only those trees with a crown diameter greater than 1.5 m or those with typical crown shading.

2.4.Topographical characteristics and data analysis

A series of variables and indices were calculated based on the Digital Elevation Model (DEM) at 5-m pixel size (Ktimatologio,2016),such as elevation,slope,aspect,wind exposition index,topographic wetness index,and the topographic profile using ArcMap 10.7.1 and SAGA GIS 6.3.0 software.Wind direction and velocity next to the ground are influenced by the land surface (B?hner and Antoni?,2009).Therefore,we used the wind exposition index as a proxy.Wind exposition index values >1 indicate areas exposed to wind and <1 indicate wind shadowed areas.The topographic wetness index is used as a proxy for soil moisture(Kopecky et al.,2021).The distance to the coast was calculated based on the location of each tree and the coastline using the tool“Near”,method geodesic in ArcMap 10.7.1.

A Shapiro-Wilk test was employed to check the normality of the data and a Levene test to check the homogeneity of variance,both requirements for a valid analysis of variance (ANOVA) test.In case these assumptions were not met,a non-parametric Kruskal-Wallis test was conducted.ANOVA and Kruskal-Wallis tests were applied to compare treeline elevation values across years,aspect,slope,topographic wetness index,distance to the coast across study sites,and wind exposition index across the aspect.If the Kruskal-Wallis test was significant,the difference between groups was tested using Dunn’s Kruskal-Wallis multiple comparisons test.A linear regression model with the square root transformation of the explanatory variable was applied to assess the relationship between the treeline elevation and wind exposition index,based on the four study sites(N,S,E,and W).All statistical analyses were conducted with the software R 3.6.3 (R Core Team,2020) and the additional packages pgirmess,FSA,ggpubr,and ggplot2 v3.2.2.

3.Results

The graphical representation of the treeline for the eastern(E)study area is shown in Fig.2a.Whereas,in Fig.2b and c,a section from the E treeline is shown on the aerial images from 1945 and 2015 representing the tree location.We found no significant treeline shifts between the study years (1945,2008,and 2015) (Fig.2d,p=0.05) and within the study sites (Fig.2e,p>0.05).However,the treeline elevation was significantly different between N,E,W,and S study sites (Fig.3a,p<0.001).The northern study site is placed at 15 km from the sea,whereas the southern study site is only at a 4-km distance from the coastline.The slope was significantly lower on the E and higher on N,S,and W study sites (p<0.05),whereas both S and N study sites,did not present significant differences in terms of slope (p>0.05) (data not shown here).The topographic wetness index was not statistically significant for none of the study sites (p>0.05) (data not shown here).The distance to the coast decreased from N to S and every study group was significantly different (Fig.3b,p<0.001).The topographic profile represents the relief of Crete in a cross-section(30 km)from the south to the north coast of the island.On this profile,the treeline elevation from the S and N study site is indicated (Fig.3c).Moreover,the treeline has a diffuse form in each study site,e.g.tree density decreases with increasing elevation.

The mean annual temperature in the investigated area increased by 1°C (Fig.4a) and the average precipitation varied over the years with a mean of 630 mm between 1979 and 2020 for the Souda meteorological station (Fig.4b).This covers the period of a substantial repercussion of global climate to emissions anthropogenic greenhouse gases.However,there was a relatively stable annual mean temperature between 1979 and 1995,but it increased steadily from 1995 to 2020.Such a delayed response in warming could be related to the marine environment of the island.In addition,temperature and precipitation at the treeline followed the same pattern as near the coast(Fig.4c and d).However,the average precipitation decreased slightly over time.

Temperature increased in each study area from 1979 to 2013,while average precipitation decreased.The mean temperature decreases from N (10.08°C) to S (8.10°C).Thus,there is a 2°C difference in mean temperature between these sites (Fig.5a).The increase in temperature was significant for all sites (R2=0.82,p<0.001).The temperature increase between 1979 and 2013 was 0.81°C for the E site,0.75°C for the N site,0.76°C for the S site,and 0.7°C for the W study site.Precipitation decreased across all sites,with an average of 170 mm between 1979 and 2013.The decrease in precipitation was significant at the N and E sites,while the W and S sites showed the same pattern but was not significant(Fig.5b).Moreover,treeline elevation correlates strongly with the wind exposition index (Fig.5c,R2=0.74,p<0.001).In Fig.5d,the De Martone Drought Index shows a shift at the treeline from very humid to humid climate over the years.All study sites were statistically significant(p<0.001).Therefore,the trees grow on steep slopes(with a mean angle of E=39°,N=48°,S=52°,W=60°) with high topographic wind exposure.The W,followed by the S and N study areas were the steepest(p<0.001,an angle betwen 30°-70°,data not shown).On SW facing slope the wind exposition index is significantly higher than on N,NE,S,and SE facing slopes(Fig.6a,p<0.001).Individual trees are located at a higher elevation on the S,SE,and SW facing slopes(Fig.6b,p<0.001).In addition,slopes facing S,SW,W,N,NE,and NW have higher slopes than the one facing E and SE (p<0.001,data not shown).

Fig.2.Treeline elevation across years and study sites. (a) Graphical representation of the treeline in the eastern (E) study site.The graphical imagery was obtained from Google Earth Pro?version 7.3.3.7786(https://www.google.com/earth/).(b)Treeline mapped on the aerial image for 1945 and(c)2015 that corresponds to the highlighted square(white)from(a).Boxplot comparing treeline elevation across(d)study years,1945(n=559),2008(n=569),and 2015(n=571)and(e)study sites.Boxplot components:medians (black lines),interquartile range (whiskers),and outliers (black dots) are shown.

4.Discussion

Our results show that the treeline has remained stable despite a significant increase in temperature over the past 70 years.These results could have several explanations,which are discussed further.

Temperature increased by 0.81°C,while precipitation decreased by 170 mm at treeline sites,indicating a warming and drying trend that exacerbates moisture stress by accelerating evapotranspiration.Although temperature and precipitation can accelerate treeline shifts,an increase in temperature and a decrease in precipitation at the treeline showed to be a bottleneck factor.Tree establishment is limited by precipitation,especially in semi-arid areas with low water storage capacity (e.g.,limestone),as it increases evapotranspiration,which exacerbates moisture stress(Pe～nuelas and Sardans,2021).Recently,several studies have shown that despite an increase in temperature,precipitation and thus soil moisture limit establishment at the treeline (Moyes et al.,2015;Rees et al.,2020;Sigdel et al.,2021).The climate in the Samaria National Park,Crete,is predicted to continue to change in the future (Hoffmann et al.,2019;Hoffmann and Beierkuhnlein,2020).As a result,the area will continue to warm and become drier,which will affect tree establishment at the treeline.Already in the summer of 2021,heat waves with temperatures above 46°C and forest fires were recorded in the Mediterranean region,especially in Greece (Copernicus Land Monitoring Service,2021).

Fig.3.Treeline elevation (m a.s.l.) across(a)study sites and(b)distance to coast(km)for the four study sites (N=north,E=east,W=west,S=south).N=120,E=177,W=202,S=72,n=571.Boxplot components as defined in Fig.3.The letters above boxplots indicate significant differences between boxplots as calculated by Dunn's Kruskal-Wallis multiple comparisons test.(c) Topographic profile from the south to the north coast of the island of Crete,crossing through the S and N study sites.The treeline elevation in the S and N study site is indicated by the green symbols as a representation of the tree species (Pinus brutia and Cupressus sempervirens).The topographic profile is based on the Digital Elevation Model(DEM)at 5 m resolution.(For interpretation of the references to color in this figure legend,the reader is referred to the Web version of this article.)

Islands are generally less affected by global warming because the marine environment buffers temperature increases,i.e.,the surface temperature difference between land and the nearby ocean results in constant ocean breeze and precipitation that can buffer the effects of climate change on coastal regions (Sutton et al.,2007).However,continental islands such as the island of Crete are likely more affected by warming than oceanic islands due to shorter distances to the mainland(Harter et al.,2015).We also found distance to the coast to be a potential factor for treeline elevation on the island of Crete.The distance to the coast affects local temperatures,i.e.cooling in summer and warming in winter.The distance from the coast is higher on the N study site(15 km)and lower on the S study site (4 km),with N >E >W >S.This might explain why in the S study site,the treeline elevation is at a higher elevation than in the N study site,which is further away from the sea and thus exposed to more extreme and growth-limiting temperatures.

Establishment of trees at higher elevations may be limited by microclimatic factors,such as lack of shelter from neighboring trees that can increase wind exposure (McIntire et al.,2016).At the study sites,trees grow mainly in areas with high topographic wind exposure(Fig.5c),therefore this might already limit their further expansion.The wind exposition index showed that the treeline from the S,E,and W study sites is more exposed to wind(values>1,Fig.5c)and thus cool and humid air masses,which could explain a higher treeline elevation.The island of Crete is also exposed to a series of winds throughout the year,such as Khamsin (south winds),Etesian (north winds),and F?hn.Khamsin winds are coming from Libya and are associated with the extreme dust episodes from the Sahara.This contributes to the formation of F?hn winds (hot and dry) on the leeward side (north Crete) (Nastos et al.,2017)which leads to low humidity in northern Crete(Prezerakos,1994).The north winds,Etesian,come from the Aegean Sea and are dry and cool in summer.Therefore,the mountain peaks are characterized by very arid,hot,and dry conditions in summer,whereas in winter,they are covered by snow.That conforms to the aridity index indicating that the western part of the island experienced a humid climate from 1951 to 1990,while the east side had a sub-humid dry climate (Nastos et al.,2013).

Topographic,geomorphologic and pedological factors,such as steep slopes with shallow soils,limit germination,establishment,and survival at the treeline (Holtmeier and Broll,2012;Cudlín et al.,2017).The limestone slopes on which the trees grow have an angle of mainly>50°,which prevents the formation of thick soil layers.Treelines tend to be lower on oceanic volcanic islands than on continental islands and the mainland because oceanic islands are more remote and isolated,have lower mountain mass effect,and more severe drought conditions in their alpine zone due to trade winds(Leuschner,1996;Irl et al.,2016).This is even more pronounced in the study area because the geologic substrate is represented by highly karstified limestone,which has limited ability to store water.This constraint could be responsible for the temporal lag in treeline change recorded in this study area.Therefore,topographical effects might considerably shape the Crete treeline ecotones and drive their dynamics.

The treelines of the four study sites have different elevations,with N

Fig.4.(a)Mean annual temperature(°C)and(b)average precipitation(mm)trend between 1979 and 2020 for the Souda meteorological station,Crete,Greece.(c)Mean annual temperature (°C) and (d) average precipitation (mm) trend for the treeline based on the CHELSA timeseries (1979-2013).The climatic trend is represented using local weighted regression (LOESS) lines.

Fig.5.(a)Annual temperature and(b)average precipitation(mm)for each study site from 1979 to 2013.(c)Wind exposition index versus treeline elevation for the four study sites(N=north,E=east,W=west,S=south).Wind exposition index values>1 indicate areas exposed to wind and<1 indicate wind shadowed areas.(d)De Martonne aridity index(IDM).IDM values between 10 and 20 indicate semi-arid,20-24 mediterranean,24-28 semi-humid,28-35 humid,35-55 very humid,and >55 extremely humid climate.

Traits of treeline-forming species could also be key to understand treeline dynamics.Pinus brutiaandCupressus sempervirensare remnants of the natural population and both grow up to high elevation on Crete,being part of the treeline.P.brutiacones are serotinous and seeds production decreases with increasing elevation (Boydak,2004).Seed production and dispersal appeared to be a limiting factor also in other mountain systems,as they decreases with elevation (Neuschulz et al.,2018;Anadon-Rosell et al.,2020).Compared toP.brutia,C.sempervirensis less flammable and can have both serotinous and non-serotinous cones(Lev-Yadun,1995).Both species are thus well adapted to the temperate Mediterranean climate with hot and dry summers and fire (Boydak,2004;Baldi et al.,2011).We could,however,not analyze the role of drought and fire in forming this treeline as there is no appropriate data on changes in the drought and fire regime on Crete over such a long period.Forest fires in Crete mainly occur in scrub and pasture areas with little vegetation,mainly in phrygana and maquis.Based on this and local ecological knowledge,forest fires do not have a major impact on the shape of the treeline.

Fig.6.(a)Wind exposition index across aspect.Wind exposition index values>1 indicate areas exposed to wind and<1 indicate wind shadowed areas.(b)Treeline elevation across aspect.Boxplot components as defined in Fig.3.Letters above boxplots are defined as in Fig.3.

Human land use is another well-known driver of treeline dynamics(Gehrig-Fasel et al.,2007).Grazing limits the regeneration of bothP.brutiaand C.sempervirens(Brofas et al.,2006).Although overgrazing with goats was a common practice in ancient times on Crete,it decreased over time and nowadays,local pastoral activities are concentrated in mid-elevation areas with sufficient vegetation for food.In the Samaria Nationl Park grazing is restricted both in the core zone (25% grassland area) and in the peripheral area (51% grassland area) (Spanos et al.,2008).Furthermore,some parts have been abandoned after the declaration of the National Park in 1962.This difference in land use has resulted in significant changes in forest area and density ofP.brutiain Crete.For instance,forest area and density ofP.brutiaincreased in the protected area of Samaria National Park as a result of land abandonment,but decreased in the unprotected area of Mount Ida(2,456 m)due to the increase in the number of sheep and goats over the past 50 years(Papanastasis,2004).Thus,the effect of soil trampling from grazing activities is also reduced in our study areas.This offered the tree species the opportunity to establish through natural regeneration.

Compared to unprotected areas,highly protected areas experience reduced human pressure and less forest loss which promotes natural regeneration (Beloiu and Beierkuhnlein,2019;Leberger et al.,2020).However,despite the protection status since 1962,past and present grazing activities might still limit seedling establishment and development at the treeline.In another protected area of Greece,the Olympus Mountains,both an upward shift and a retreat of the treeline were observed,which could not be explained by climate alone(Zindros et al.,2020).However,a major difference between our study region and Olympus Mountain is the topography and the sufficient soils that can promote sapling establishment.While the study areas from the Lefka Orimountains are dominated by exposed limestone and steep slopes instead of a developed layer of soil(Fig.1).Steep slopes and absence of suitable substrate can be major limiting factors in treeline advance(Macias-Fauria and Johnson,2013;Cudlín et al.,2017).

The treeline on the continental island of Crete has a diffuse form and tree density decreases with increasing elevation.Diffuse treelines are more sensitive to changes in temperature and are more likely to exhibit earlier upward shifts than krummholz,island treelines,and abrupt treelines (Harsch and Bader,2011).Consequently,our results were unexpected and thus stimulate the fundamental debate on treeline dynamics as a result of recent climatic changes highlighted in the introduction.In the global context,the Mediterranean treeline generally occurs at exceptionally high temperatures and very low elevations,e.g.on Mount Olympus,Greece,at 2,320 m a.s.l.and 8°C;on Mount Helmos,Greece,at 2,100 m and 11.1°C;on the Maiella massif,Apennines,Italy,at 1,820 m a.s.l.and 10.5°C (K?rner,2012).However,the treeline on Crete is formed at a significantly lower elevation(between 1,270 and 1,1884 m a.s.l.),with a mean elevation of 1,536 m a.s.l.in Samaria National Park.The elevation of the Crete treeline shows signifciant stability between 1945 and 2015 even though it occurs at much lower elevation than most other Mediterranean treelines.Thus,the observed treeline elevations on Crete is below the globally modelled potential treeline.Its temporal stability thereby suggests a temporal lag in treeline response due to relatively stable climate conditions until 1990 and topographical drivers that keep the treeline elevation low.

In the Northern Hemisphere,66%of the treeline showed an increase,34% showed no shift (Hansson et al.,2021),and other studies even showed a retreat of about 1%(Harsch et al.,2009;Lu et al.,2021).Hence,the general trend is that the treeline is advancing at 0.35 m?year-1,but even that is too slow to keep pace with climate warming (Rees et al.,2020;Lu et al.,2021).We showed that treeline elevation remained stable regardless of temperature increase and precipitation variation.Therefore,tree species growing at the treeline in Crete do not track contemporary climate change.Similar to these results,many studies from Central and Northern Europe found a stable or even declining trend in treeline elevation(Harsch et al.,2009;Hansson et al.,2021).Moreover,several ecological models and field observations showed no climate effects on ecosystem composition and species shift in the Mediterranean basin (Camarero and Gutiérrez,2004;Gritti et al.,2006) and at high latitudes(Zhu et al.,2012).Nevertheless,we also found that the treeline ecotone was enriched with a few tree individuals over the years.Consequently,temperature increases alone cannot promote treeline advancement if precipitation decreases(Moyes et al.,2015;Sigdel et al.,2018;Rees et al.,2020;Lu et al.,2021).

5.Conclusions

Our results have revealed that,based on high-resolution aerial images,there was no shift in the treeline elevation between 1945 and 2015 on the continental island of Crete.Therefore,tree species growing at the treeline are unable to keep up with rising temperatures.The absence of treeline shift is mainly due to a combination of climatic and topographic factors,such as decreasing precipitation with increasing temperature,and the lack of a shelters effect due to high topographic wind exposure.The temporal lag in treeline shift could impose risks as the optimum elevation changes.Aerial imageries proved particularly suitable on inaccessible steep terrain where in-situ monitoring is limited or impossible,a distinct advantage in mountainous areas.Our findings demonstrate the benefits of using historical high-resolution remote sensing images in stimulating the controversial debate about treeline dynamics under climate change.The continental island of Crete is home to particularly many endemic species and thus very important for biodiversity conservation.

Data availability

Datasets analyzed in the current study are available online on https://zenodo.org/record/4404269#.X-zbEtgzaUk.

Authors’contributions

D.P.and C.B.conceived the idea,D.P.and M.B.led the writing process.M.B.processed the climate data,mapped the treeline on the aerial images,did the statistical analysis,and prepared the figures.P.L.,A.B.,and A.T.analyzed the aerial images.S.H.contributed to the writing.C.B.,S.H.,D.K.,and N.C.provided feedback on the manuscript.N.C.and C.B.supervised the research activity.All authors revised and approved the final version of the manuscript.

Competing interests

The authors declare no competing interests.

Acknowledgments

We acknowledge support from the ECOPOTENTIAL project-EU Horizon 2020 research and innovation program,grant agreement no.641762.