the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 May 2026
Tales from the past: remapping dynamic tree- and forest lines in response to changing climate and current land use
Abstract. Average temperatures are rising more rapidly in high-latitude and alpine regions than elsewhere, leading to a gradual compression of the alpine bioclimatic zone due to the upward shift of the tree- and forest lines (TFLs). While most studies on TFL dynamics indicate advance, the rate at which regional and local treelines respond to climate warming remains uncertain. Furthermore, not every empirical TFL is determined by climate alone; edaphic conditions, species traits, and, in particular, land use, affect tree growth and distribution. In many regions, domestic grazing and other forms of traditional mountain summer farming have historically depressed forest lines. Previous research has been limited by the comparison of data sampled with mixed methods and poor temporal data coverage. Additionally, there is still a lack of studies accounting for time-lags, thus including data spanning a long time.
In this study, we used consistently remapped in situ measurements of mountain birch (Betula pubescens ssp. czerepanovii) in Norway, dating from 50 to 130 years back, to: (1) document the rate of TFL change; (2) understand the impact of land use and climate change on regional TFL dynamics; and (3) discuss regional aspects and quality components of the data.
We find that Norwegian TFLs are advancing at rates exceeding 0.5 m yr−1, primarily driven by climate change for TLs and land use for FLs. Still, the rates of TFL change vary considerably between regions, likely due to stochastic disturbances (e.g. snow avalanches, insect outbreaks, pests, landslides, rockfalls).
We highlight the need for better quantification of time lags in treeline responses and for consistent definitions and methodologies when assessing long-term TFL dynamics in boreal–alpine ecotones.
- Preprint
(1840 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2717', Anonymous Referee #1, 15 Aug 2026
-
AC1: 'Reply on RC1', Ingrid Vesterdal Tjessem, 24 Sep 2026
The manuscript by Tjessem et al. reports changes in treelines (TLs) and forest lines (FLs) and investigates the roles of climate change and land use using long-term in situ observations of mountain birch in Norway. The authors find that these tree- and forest lines (TFLs) have been advancing at rates comparable to those reported in previous studies. Based on regression analyses, they further conclude that changes in TLs are primarily driven by climate change, whereas changes in FLs are primarily driven by land use.
While this study leverages a unique long-term in situ dataset to investigate changes in TFLs, I found that some of the main conclusions are not sufficiently supported by the analyses presented. In addition, several aspects of the methodology and datasets are not described clearly enough to allow the reader to fully understand or evaluate the analyses. Please see my detailed comments below.
Response: We appreciate the positive and constructive comments by this reviewer. From the "inside", it is often difficult to see which explanations will need clarification and which are clear. As will become apparent below, we have responded to each of the reviewer's concrete suggestions. In addition, we have gone through the manuscript ourselves and made a few changes to improve clarity.
Major comments:- Key datasets (TFLs and land use) are not described with sufficient clarity.
For the TFL data, how frequently were these sites surveyed, and how are the observations distributed temporally? The description in L150 (“the first, second, and third TFL record”) and the three registrations shown in Fig. 2 are somewhat confusing. L204 mentions the change is calculated with only the first and last record, were other records used in the analysis? Please clarify the temporal sampling scheme and explain how these different records/registrations are defined and used in the analysis.
L150: ”The climate data were processed to account for time-lags in the response of TFLs to climate (Alexander et al., 2018) by first calculating the mean value for each climate variable over a 20-year period prior to the first, second, and third TFL record.”
L204: “To explore the impact of land use and climate change on TFL dynamics, we analyzed the full rate of change (i.e. the change in m from first to last record) using linear regression.”
Response: We will improve the description of how the sites were surveyed, and also improve the text explaining key datasets. We will also clarify the text explaining the temporal sampling “scheme”, and how they were used in the analyses.
For the land-use data, are observations available only for the years listed in L163 and L166? If so, how are the 20-year averages prior to each TFL record calculated if land-use information is only available for selected years? A clearer description of the temporal coverage, interpolation/aggregation procedure, and how these data are matched with the TFL observations is needed.
L163: “The first dataset contains the number of livestock grazers on rangelands in each municipality in 2003, 2004, and 2019, according to applications for agricultural subsidies (data downloaded from the Norwegian Agriculture Agency (https://www.landbruksdirektoratet.no, accessed May 2020).”
L166: “The second dataset contains the number of livestock grazers registered in each municipality in 1891, 1907, 1917, 1929, 1939, 1949, 1959, and 1969.”
Response: We see that our text is somewhat unclear with respect to how the land-use variables have been prepared, so we appreciate this comment. We will therefore improve and clarify how the land-use data has been generated based on the available sources.
- The explanatory power of the statistical analyses is very low, making the attribution of TFL changes to specific drivers difficult to support.
For example, the correlation in Figure 4 is less than 0.25. The linear regression models also just explain very limited variability in TFL changes (R2 up to 0.2). Although the authors also mention other factors such as disturbance, the poor model performance makes it difficult to draw any convincing conclusions regarding the drivers of TFL changes.
Response: We agree that the relationships revealed by our analyses are not very strong, and that this suggests that factors other than those we have represented by explanatory variables may be in operation. We have tried our best to make clear that the results provide clear evidence for an upward movement of TFLs and that temperature changes explain some of that variation. However, we also acknowledge that the reasons for the variation in the magnitude of TFL movement remain complex. During the revision, we will make sure that the uncertainty with respect to reasons for variation of TFL change is clearly expressed.
Minor comments:L24: TL and FL should be spelled out at first appearance. Likewise in the main text.
Response: We will correct this in the abstract and main text.
L53: This sentence is ambiguous. It is unclear which region the factors described after “due to” refer to. Please clarify.
L53: “The locally highest-situated TFL locations occur at lower elevations in the northern and western parts of the country rather than the south-central mountain massif, due to either lower mean summer temperatures, ‘summit syndrome’ (Körner, 2012: i.e. that TFLs close to a summit tend to be regulated by non-thermal abiotic disturbance (for empirical support, see Wistrand, 1962)), proximity to the coast, or a combination of multiple factors (Odland, 2015).”
Response: We will rephrase the sentence to remove ambiguity.
Table 1: The selection of growing season length variables (Monthly temperature > 5, 5.5, 6, … °C) appears somewhat arbitrary. Please provide a justification for the choice and range of these temperature thresholds.
Response: As the first basis for this selection, we chose values for the parameter T (the temperature at which a month would be considered part of the growing season) based on Körner and Paulsen (2004; https://doi.org/10.1111/j.1365-2699.2003.01043.x), which documents a range of mean air temperatures at the treeline of 5.5–7.5 °C. We expected the relationships between the parameter Ts and Kendall's rank correlation coefficients to be unimodal. Based on that expectation, we generated climate data with parameter values distributed around T = 6.0 °C. The ranges and the parameter values within the ranges were based on our intention to capture the unimodal distributions. As can be observed in Figure 4, these relationships were unimodal or close to unimodal, which suggests that our chosen parameter values captured the ranges successfully. We will include a brief justification in the manuscript.
Table 1: “Annual temperature deviation” is confusing by its name, for example, it could mean the standard deviation of temperature within each year or across years, although Table A1 provides a more detailed definition. Since this variable represents the temperature difference relative to the 1981–2010 reference period, a term such as “temperature change” may be clearer.
Response: We agree that the term "Annual temperature deviation" could be misinterpreted as standard deviation across or within years. However, since the variable represents the difference relative to the climatological standard normal for 1981–2010, we consider "Temperature anomaly" to be a more precise scientific term than "Temperature change", which could inadvertently imply a temporal trend. To ensure complete clarity, we have renamed "deviation" to "anomaly" for all relevant temperature and precipitation variables in Table 1, Table A1, and throughout the manuscript.
L159: Could you briefly explain why NNE 22.5° is least favourable?
Response: Energy supply through radiation is one of the major factors influencing growth, species abundances and, hence, the species composition of terrestrial ecosystems. In vegetation ecology, several studies find that the optimal (i.e. the "warmest") aspect is about SSW (202.5°). Examples of such studies are Dargie (1984) and Heikkinen (1991). The corollary of this is that around NNE, or 22.5°, is the least favourable aspect. We will incorporate this justification in the text if desired.
Dargie, T.C.D. 1984. On the integrated interpretation of indirect site ordinations: a case study using semi-arid vegetation in south-eastern Spain. ‒ Vegetatio 55: 37-55.
Heikkinen, R.K. 1991. Multivariate analysis of esker vegetation in southern Häme, S Finland. ‒ Annls bot. fenn. 28: 201-224.
Citation: https://doi.org/10.5194/egusphere-2026-2717-AC1
-
AC1: 'Reply on RC1', Ingrid Vesterdal Tjessem, 24 Sep 2026
-
RC2: 'Comment on egusphere-2026-2717', Prajwal Khanal, 03 Sep 2026
Please find the review in the attached pdf.
-
AC2: 'Reply on RC2', Ingrid Vesterdal Tjessem, 24 Sep 2026
The manuscript "Tales from the past: remapping dynamic tree- and forest lines in response to changing climate and current land use" provides a valuable analysis of ecotone shifts. The authors effectively demonstrate an advancement of approximately 0.5 m/yr, linked to an extended growing season (specifically, mean temperatures exceeding 6.5 °C).
What I really enjoyed about the study is the thorough statistical analyses, that the authors perform to verify their claims, complemented by a transparent discussion section that meticulously addresses methodological rationale and study limitations.
Response: We very much appreciate Dr Prajwal's positive evaluation of our work. His two questions are addressed below.
When examining Figure 5, the central graph in the manuscript, it seems the coefficient of determination (R2=0.20) for the growing season length (mean temperature >6.5 degree C) is relatively low compared to what is typically expected in hydrological studies. However, when weighed against the statistical significance demonstrated in Table 2, these findings still provide a credible level of confidence in the underlying temperature-treeline relationship.
Furthermore, given the extreme scarcity of long-term in-situ historical records, creatively analyzing this data to yield insights into temperature-driven treeline shifts significantly adds to our understanding of ecotone shifts.
Only point for me requiring further clarification pertains to the confounding effect of observation duration and spatial positioning, given that the highest rates of change occur in the longest and northernmost record (JMN, 130 years) while shorter records (such as BAA, 50 years) are situated in the southeast.
This raises two questions for me:
- To confirm that the higher rates of change in longer records are not an artifact of observation window length, could the authors perform additional analyses using a standardized record length (e.g., truncating all records to a uniform 50-year period, where possible)?
Response: Thanks, in fact we have only mentioned this briefly in line 311‒312, where we state:
“However, TFL dynamics cannot be explained by time alone if the drivers do not change during the study period.”
We will include a short discussion about this and include the test results. The short story is that if we remove the oldest dataset (JMN), there is no correlation between time window and TFL change. It is most likely not the time available, but the temperature or land use change that explain the TFL dynamics. However, there is a weak correlation between the time window and change in the oldest dataset (JMN). But to remove collinearity, and because the time window is not explained by the other datasets, we excluded time as a variable. As stated above, we will include a short explanation/discussion in the revised manuscript.
A uniform 50-year analysis would be optimal, but the challenge is that we lack TFL data from periods other than those sampled, making a uniform 50-year period analysis difficult. We do not know exactly when, between the sampled periods, the TFLs actually advanced, retreated, or were stable. We can therefore only use the existing time windows.
- Currently, only aspect favorability carries spatial information; could the authors include latitude as a covariate in the model to ensure that macro-geographic gradients aren't confounding the local aspect signal?
Response: We appreciate the suggestion. However, latitude primarily acts as a proxy for macro-climatic conditions (such as temperature and growing season length). Because our models already include high-resolution (1x1 km), locally adjusted meteorological data that explicitly capture these thermal gradients, the macro-geographic variation is already accounted for physically. Introducing latitude as an additional covariate would likely introduce collinearity without providing additional mechanistic explanatory power. We will add a brief sentence in the methods section to clarify this rationale.
Citation: https://doi.org/10.5194/egusphere-2026-2717-AC2
-
AC2: 'Reply on RC2', Ingrid Vesterdal Tjessem, 24 Sep 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 198 | 119 | 41 | 358 | 30 | 30 |
- HTML: 198
- PDF: 119
- XML: 41
- Total: 358
- BibTeX: 30
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript by Tjessem et al. reports changes in treelines (TLs) and forest lines (FLs) and investigates the roles of climate change and land use using long-term in situ observations of mountain birch in Norway. The authors find that these tree- and forest lines (TFLs) have been advancing at rates comparable to those reported in previous studies. Based on regression analyses, they further conclude that changes in TLs are primarily driven by climate change, whereas changes in FLs are primarily driven by land use.
While this study leverages a unique long-term in situ dataset to investigate changes in TFLs, I found that some of the main conclusions are not sufficiently supported by the analyses presented. In addition, several aspects of the methodology and datasets are not described clearly enough to allow the reader to fully understand or evaluate the analyses. Please see my detailed comments below.
Major comments:
1. Key datasets (TFLs and land use) are not described with sufficient clarity.
For the TFL data, how frequently were these sites surveyed, and how are the observations distributed temporally? The description in L150 (“the first, second, and third TFL record”) and the three registrations shown in Fig. 2 are somewhat confusing. L204 mentions the change is calculated with only the first and last record, were other records used in the analysis? Please clarify the temporal sampling scheme and explain how these different records/registrations are defined and used in the analysis.
For the land-use data, are observations available only for the years listed in L163 and L166? If so, how are the 20-year averages prior to each TFL record calculated if land-use information is only available for selected years? A clearer description of the temporal coverage, interpolation/aggregation procedure, and how these data are matched with the TFL observations is needed.
2. The explanatory power of the statistical analyses is very low, making the attribution of TFL changes to specific drivers difficult to support.
For example, the correlation in Figure 4 is less than 0.25. The linear regression models also just explain very limited variability in TFL changes (R2 up to 0.2). Although the authors also mention other factors such as disturbance, the poor model performance makes it difficult to draw any convincing conclusions regarding the drivers of TFL changes.
Minor comments:
L24: TL and FL should be spelled out at first appearance. Likewise in the main text.
L53: This sentence is ambiguous. It is unclear which region the factors described after “due to” refer to. Please clarify.
Table 1: The selection of growing season length variables (Monthly temperature > 5, 5.5, 6, … °C) appears somewhat arbitrary. Please provide a justification for the choice and range of these temperature thresholds.
Table 1: “Annual temperature deviation” is confusing by its name, for example, it could mean the standard deviation of temperature within each year or across years, although Table A1 provides a more detailed definition. Since this variable represents the temperature difference relative to the 1981–2010 reference period, a term such as “temperature change” may be clearer.
L159: Could you briefly explain why NNE 22.5° is least favourable?