Detecting treeline dynamics in response to climate warming using forest stand maps and Landsat data in a temperate forest

Do forest treelines respond to global warming? The vegetation of treelines and alpine ecotones is among the most vulnerable and sensitive to the impact of global warming on earth (Masek 2001, Bharti et al. 2012, Weisberg et al. 2013, Gaire et al. 2017, Mishra and Mainali 2017). The movement of forest treelines towards higher altitudes is mainly related to thermal variations and other factors (Zinnert et al. 2011, Du et al. 2017, Gaire et al. 2017, Holtmeier and Broll 2017, Skowno et al. 2017), such as enhanced atmospheric CO₂ and nitrogen deposition (Masek 2001, Grace et al. 2002, Salzer et al. 2009, Capers et al. 2013, Skowno et al. 2017). Simultaneously, increases in these factors usually result in the migration of alpine plants to upper altitudes. Moreover, downward shifts in treelines and decreases in treeline stability have been reported due to harsh environmental conditions (Zhang et al. 2009, Du et al. 2017, Gaire et al. 2017) and human disturbances (Jump and Penuelas 2005, Li et al. 2016b). Indeed, these shifts probably have a remarkable influence on plants that are a carbon sink major source (Hanberry and Hansen 2015, Li et al. 2016a). In fact, the reliable assessment of forest treeline dynamics is essential to conserve the ecological condition of alpine ecotone biodiversity for sustainable management and forest conservation strategies (Grace et al. 2002, Holtmeier and Broll 2010, Mishra and Mainali 2017, Skowno et al. 2017).

Various methods have developed and used in various parts of the world to determine woody plant encroachment in response to global warming, including methods using dendroecological techniques (Danby 2007, Salzer et al. 2009, Du et al. 2017, Gaire et al. 2017, Wang et al. 2017b), aerial photograph assessments (Lutz et al. 2013, Solár 2013, Ropars et al. 2015), field plot resurveys (Grabherr et al. 1994, Holtmeier and Broll 2005, Danby and Hik, 2007a, Elmendorf et al. 2012), satellite images (Masek 2001, Holtmeier and Broll 2005, Fraser et al. 2011, Zinnert et al. 2011, Patrick Shawn Sawyer 2014, Mishra and Mainali 2017, Skowno et al. 2017), repeated photographic evidence (Kullman 1993), palynological data investigations (Grace et al. 2002), seedling recruitment studies (Miller et al. 2017), and model scenarios (Masek 2001, Trivedi et al. 2008). Furthermore, the number of successful investigations that have integrated remote sensing images and historical maps is increasing (Wallentin et al. 2008, Zinnert et al. 2011, Weisberg et al. 2013). Simultaneously, remote sensing data provide both fine- and coarse-resolution data (Fraser et al. 2011, 2012). Moreover, Landsat satellite data are the most effective data sets and cover large areas (Arekhi et al. 2017, Chen et al. 2017). Landsat images, which have been recorded since 1972 and are freely available for a period of more than 45 years with high spatial-temporal resolution and a 16-day revisit time (Masek 2001, Wang et al. 2017a), represent one of the most valuable data sources for determining woody plant encroachment. Likewise, they provide a great resource for depicting timberline dynamics and treeline movements with various methods and vegetation indices (Holtmeier and Broll 2005, 2017, Zinnert et al. 2011, Bharti et al. 2012, Elmendorf et al. 2012, Gaire et al. 2017, Wang et al. 2017a). The NDVI is the most common vegetation index, and it is well known as a good representative of plant biomass. Additionally, the NDVI is directly indicative of plant photosynthetic activity (Zinnert et al. 2011, Du et al. 2016, Li et al. 2016a, Mishra and Mainali 2017). Therefore, monitoring of changes in the NDVI values across the forest timberline border provides a quantitative estimate of the tree biomass and density over time (Singh et al. 2011, Bharti et al. 2012, Sawyer and Stephen 2012, Iverson and McKenzie 2013), which demonstrates trends in historical shifts that can be detected by remote sensing techniques (Singh et al. 2011, Raynolds et al. 2013, Ropars et al. 2015, Morley et al. 2017).

Combining forest stand maps with long-term Landsat data provides a good opportunity for detecting and monitoring historical shifts in the forest structure, timberline border, and treeline movement (Zinnert et al. 2011, Solár 2013). Additionally, another aspect of remote sensing data accessibility is that its data include LST, which provides an unprecedented opportunity to investigate and compare forest dynamics (van Leeuwen et al., 2011) among certain interval times. The importance of LST in climate change analyses has been proven (Cristóbal et al. 2008, Brabyn et al. 2014)(Shuman and Comiso 2002, Parastatidis et al. 2017). However, LST anomalies are poorly investigated, and the spatial and temporal monitoring of LST can be used in treeline monitoring studies. The Mediterranean region is recognized as an area impacted by global warming in Turkey, and it is facing increasing summer temperatures and winter precipitation (Solomon et al. 2007, Talu et al. 2010, Parolly 2015). Moreover, evidence of the upward movement of treelines has been reported in the eastern Karadeniz region (Black Sea), and treelines may move approximately 200–400-m according to climate change predictions (Terzioglu Salih 2015). Some climate change scenarios have indicated that the mountain flora in the western and central Taurus mountain regions (Parolly 2015) and the eastern Mediterranean region are threatened (Ozturk et al. 2015). Conventionally, a treeline vertical movement of approximately 640 m is expected to occur with a 4.5 °C warming rate per century and 33 m of movement per 1 °C increase. Holtmeier and Broll 2017 stated, “82% of coarse scale (regional) variation in treeline elevation is associated with thermal conditions”. However, there is evidence that accelerated glacier melt in sensitive, snow-covered mountain areas due to increasing temperature impacts treeline structure and composition (MacDonald et al. 2008, Capers et al. 2013, Wilmers et al. 2013, Holtmeier and Broll 2017). However, there are a few studies investigating the case of forest treelines in the Mediterranean region, and there is a gap in knowledge about forest treeline dynamics and structural shifts.

The present study objectives were 1) to quantify treeline movement rates from 1970 to 2013 by the combination of NDVI maps with forest stand maps; 2) to explore whether the increasing trends in LST and NDVI coincided with treeline advancement from 1984 to 2017; 3) to determine promising tree species in the timberline border in four times (1970, 1992, 2002, 2013) by overlaying forest stand maps on NDVI maps.

This paper reports the feasibility of detecting treeline dynamics and structural shifts through the integration of Landsat data with forest stand maps. Additionally, the present study sets aimed to address the importance of measured forest stand maps at certain time intervals in forest treeline dynamics investigations, which has been less considered by prior studies. The current study found that the treeline expanded from 1970 to 2013 according to the performed difference analysis between the time periods. The resultant maps showed the geographical expansion of the uppermost treeline over 43 years, with an expansion of approximately 20 to 40 m each decade; the treeline altitude increased to 2442.5 m from 2335 m (Table 2). However, changes in some regions were not uniform (Virtanen et al. 2010, Elmendorf et al. 2012, Hanberry and Hansen 2015, Bruggeman et al. 2016). The findings of the current study are consistent with those of previously conducted investigations that have either used remote sensing data or field survey data, such as tree species dendrological measurements (Danby and Hik, 2007a, Du et al. 2017), alpine ecotone biology observations (Virtanen et al. 2010), repeated photo evidence (Grace et al. 2002, Fraser et al. 2011), and remote sensing method (Masek 2001, Grace et al. 2002, Fraser et al. 2012, Li et al. 2016b, Holtmeier and Broll 2017). Moreover, the movement of the treeline to higher elevations has been reported as a response to global warming, mainly the increase in air temperature. The trend analysis of monthly average temperature (T_MIN, T_MAX, T_MEAN) displayed a statistically positive trend over the study period, especially in the monthly average mean July temperature (tau = 0.62, ρ< 0.00) (Table 1). As discussed, our research supports that climate warming is accompanied by the advancement of treelines towards higher elevations (Danby 2007, Virtanen et al. 2010, Singh et al. 2011, Monleon and Lintz 2015, Du et al. 2017). Our study findings are consistent with those of other scientific studies that have recorded the upslope movement of trees, particularly those in the northern hemisphere (MacDonald et al. 2008, Solár 2013). Similar results achieved by other investigations have emphasized the potential of recorded historical Landsat data to provide an extraordinary opportunity for observing and determining forest timberline and treeline shifts (Masek 2001, Weisberg et al. 2013). Previous studies have suggested and documented an altitudinal advance of the treeline worldwide between 140 and 700 m over the past century (Grace et al. 2002), with an average change of 3 m per year (Singh et al. 2011). Moreover, for some cases, a mean altitudinal change for tree species of nearly 29 m has been reported. A 1 °C increase in temperature leads to 11–40 m of treeline expansion (Grabherr et al. 1994, Lenoir et al. 2008). Similarly, our finding depicted a 4 °C increase over 40 years, which confirmed the treeline increased between 20 to 40 m per decade, if we assume that temperature increased 1 °C per decade. In other findings, the expansion rate of the lowest treeline altitude was between 20 m to 106 m, which needs to be investigated in further studies (Table 2). In addition, numerous investigations have reported that the upward trend in treeline movement is likely to result in the natural succession of trees and reach an optimum in the Early Holocene, which must be considered in further investigations (Grace et al. 2002, MacDonald et al. 2008, Holtmeier and Broll 2010). These shifts would have a significant impact on the landscape, affecting the plants, microorganisms, and animals that live in the tangible mountain ecosystem. Additionally, if the upward expansion of trees continues, they will reduce high-latitude albedo and increase global warming (MacDonald et al. 2008). Therefore, they are possibly contributing to further climate warming (Liess et al. 2012).

The selected plots showed an increasing trend in the mean NDVI and LST values over the study period. However, the results indicate that the increase in the mean NDVI was consistent with the increase in the mean LST (Fig. 5a). A moderately positive correlation between the mean NDVI and mean LST values at most sample plots in all study periods for each year (1984, 1992, 2013, 2017) was observed, except for in 2002. There was a statistically significant positive correlation between the mean NDVI and the mean LST (r = 0.634; ρ< 0.05) in 1984 (Table 3). The increasing trends in mean LST and NDVI had R² values of R² = 0.94 and R² = 0.96, respectively (Fig. 6). The increase in the mean NDVI was positively correlated with the increase in the mean LST using per pixel correlation analysis in the nine time periods between 1984 and 2017 (r = 0.75, ρ< 0.05) (1984–1988–1992-1997-2002-2007-2010-2013-2017). The correlation coefficient was positive and significant (r = 0.901–0.777, ρ< 0.05) between the forest stand density and summer temperature in (Danby and HIK, 2007b) the study, with values of r = 0.537, ρ< 0.05 (Li et al. 2016b). Based on the field investigations, a significant and positive trend and meaningful correlation was observed between the temperature and tree growth in July (Körner 1998, Li et al. 2016b), which is in agreement with the results obtained for the 20 sample plots and July LST temperature over the study period using per pixel trend analysis (r = 0.75, ρ< 0.05). Therefore, according to per pixel trend analysis, the increasing LST played a considerable role in the increasing NDVI trend during the study period in the selected plots (Fig. 6). The warmer the pixel temperature, the higher the tree species growth and seedlings (MacDonald et al. 2008, Monleon and Lintz 2015, Miller et al. 2017). This finding suggests that the tree expansion in the study area can be attributed to climate change factors, especially temperature (Fraser et al. 2011). The monthly average temperature in July was the variable that experienced the largest significant shift among the climate factors in our study (tau = 0.62, ρ< 0.00) (Table 1). The control of the treeline situation by the summer temperature (Harsch et al. 2009), mostly the mean July temperature, has been proven in Northern Eurasia (MacDonald et al. 2008). The monitoring of treelines has delineated that altitudinal forest expansion is evident in all of the study sites, with a consistent warming trend (Liess et al. 2012, Morton et al. 2012, Hanberry and Hansen 2015) in the LST values from 1984 to 2017. Concurrently, it seems that the temperature increases possibly met the growth needs of tree species (Grace et al. 2002). However, year-to-year changes in climate factors, mainly temperature, can be anticipated to have a significant influence on vegetation shifts (Virtanen et al. 2010). In previous studies, it has been noted that rising temperature is the main reason for shifts in the timberline ecotone. It is likely that the expansion of temperate forests reflects increases in temperature as an indicator of global warming (Salzer et al. 2009, Singh et al. 2011, Hanberry and Hansen 2015).

There is obviously mounting evidence of shifting landscape structures at the sample plots (Fig. 8). In the research area, during the last three decades, Cedrus libani and Juniper communis spp. have moved to higher elevations in all directions more quickly than other coniferous species, according to the comparison of the available forest stand maps for each decade (Fig. 8). These shifts can be considered increasing greenness in the resultant NDVI maps (Fig. 4). However, in some stands, shifts in the observed tree species cannot be distinguished by the NDVI values. An examination of 20 sample plots showed a consistent increase in the NDVI greening trend and vascular plants, especially Juniper communis spp. and Cedrus libani. Moreover, the expansion of Juniper spp. into grassland was reported by multitemporal analysis of vegetation indices extracted from Landsat data (Wang et al. 2017a). In addition, the expansion of forest into treeless regions was recorded (Wallentin et al. 2008). Similar to the increasing NDVI trends in the sample plots, findings showed that the forest stand tree species covers expanded, especially in sample plots with already enduring tree species stands (MacDonald et al. 2008, Fraser et al. 2011, Solár 2013, Ropars et al. 2015), mostly based on the NDVI obtained using Landsat data (Bharti et al. 2012, Morton et al. 2012, Fraser et al. 2014, Patrick Shawn Sawyer 2014). Interestingly, according to previously conducted investigations, Cedrus libani is a promising tree species that is able to sustain even the severe cold of winter and extended summer droughts. In addition, Cedrus libani depicted a high compatibility with climate change conditions, and it is likely to provide permanent and productive forest stands (Hajar et al. 2010, Güney et al. 2015, Messinger et al. 2015). Furthermore, the expansion of Cedrus libani to higher elevations was recorded by the model scenario investigations, and its adaptation is discussed (Hajar et al. 2010). Various tree species responded differently, which needs to be considered in further studies. Another important finding is the expansion of Cedrus libani on the rocks more than Juniperus communis spp. according to forest stand maps. Three tree species (Quercus sp., Pinus brutia and Pinus nigra) were replaced with Juniper communis spp. and Cedrus libani (Fig. 8). The expansion of woody species has been reported at a global scale (Zinnert et al. 2011). The present research proved that it is possible to detect and estimate the compositional and structural shifts in tree species types under climatic warming using forest stand maps, and our findings are consistent with those of previous scientific studies conducted at the treeline ecotone (Virtanen et al. 2010, Singh et al. 2011). However, it cannot be denied that the response of each tree species to warming varies, which is likely related to their particular ecological characteristics and basic requirements and the site states. However, high mountains in the eastern Mediterranean are seriously affected by climate warming, so the effects of anthropogenic impacts and forest fires (Ozturk et al. 2015) need to be investigated in further studies. Both data sets complement each other and add a reality and validity to the obtained results. This significant aspect is discussed in other studies (Fraser et al. 2011, Iverson and McKenzie 2013). It should be noted that the available forest stand maps which is generate with high accuracy (Cakir 2006, Ozkan and Yesil 2016) or repeated field survey measurements validate the NDVI greenness obtained from remote sensing data for vegetation cover (Fraser et al. 2011) and high aerial photos, such as Bing Aerial maps. Integrating other ecological factors, dendrological studies and tree species seedling recruitment and putting experimental plots in the study area are suggested. In further investigations, treeline sensitivity to shifts in environmental parameters should be considered in various forest types, such as the tropical and boreal forest. Furthermore, the importance of treeline history needs to be considered in treeline dynamics studies. It has been emphasized, considering the post-glacial history of the research area, that impacts on treeline position shift the existing ecotone structure (Fraser et al. 2014, Holtmeier and Broll 2017). However, the research area has experienced a rapid warming, and its glaciers disappeared in 1980 (Yavasl횤 et al. 2015). If the upward movement of the treeline continues, it is likely to threaten the biodiversity of the plant belts in the higher altitudes. It should be noted that the present study has its limitations. One of its weaknesses is the lack of enough meteorological stations in the study area. According to the obtained results, one alternative to meteorological stations is the use of LST maps with Landsat thermal bands, which provide temperature information for each 30 m × 30 m pixel. However, many studies have proved the feasibility of using LST instead of air temperature in remote regions (Shuman and Comiso 2002, Cristóbal et al. 2008, Brabyn et al. 2014). Moreover, the LST potential in climate change investigations has been reported in forest monitoring studies (Shuman and Comiso 2002, Parastatidis et al. 2017).

The present research illustrates the feasibility of monitoring treeline expansion rates and structural shifts using forest stand maps, long-term Landsat data (NDVI, LST), and temperature data for more than 43 years. The results were verified by available high-resolution Bing Aerial images in the Qgis environment and existing forest stand maps. The uncertainties associated with the use of only Landsat data or forest stand maps are likely reduced by integrating two datasets with Bing Aerial maps based on the applied method. In the present study, three significant sciences, namely, GIS techniques, remote sensing data, and field-based measurements, were integrated. Prior studies have been conducted based on just remote sensing and greenness or just field-based methods (dendrology and biology studies). This study showed that LST, an indicator of temperature, and NDVI, a surrogate of the green biomass of the treeline ecotone, demonstrated an overall increasing trend from 1983 to 2017 at the plot scale. The research results show that treeline temperature anomalies can be detected successfully in the ecotone zones with LST by using per pixel trend analysis and linear model. Additionally, shifts in the heterogeneity of the forest structure over space and time can be detected by integrating two data sources. Moreover, the present study depicted an apparent change and increase in both Cedrus libani and Juniperus communis spp. and determined two pioneer tree species by both NDVI maps and forest stand maps. In the current study, Cedrus libani became a pioneer tree with Juniperus communis spp. the results showed that warmer temperatures enhanced the NDVI values, which supports the hypothesis that climate warming leads to the upward movements of treelines. Therefore, determining how the timberline border vegetation responds to increasing temperatures helps establish some developmental adaptation strategies for forest managers in the treeline ecotone. Future investigations should focus on the comparison of NDVI and other vegetation indices to examine their performance in detecting timberline and treeline movements. Further studies should investigate the expansion and reduction in various tree species, such as Cedrus libani, and Juniperus communis spp., and their competition and compromises with other tree species. Additionally, investigating other alpine vegetation shifts, such as shifts in shrubs and herbs, is recommended. These findings suggest that, in general, these outputs can be used effectively in niche modeling studies, although tree species modeling scenario investigations are based on stimulation rather than experimental data. In addition, future research should examine the contributions of other factors on timberline and treeline movement dynamics. Moreover, the use of medium-resolution data, namely, ESA’s Sentinel 2A and radar remote sensing data, that provide more reliable information coinciding with optical remote sensing data is recommended.