the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Jun 2025
Northward shift of boreal tree cover confirmed by satellite record
Abstract. The boreal forest has experienced the fastest warming of any forested biome in recent decades. While vegetation–climate models predict a northward migration of boreal tree cover, the long-term studies required to test the hypothesis have been confined to regional analyses, general indices of vegetation productivity, and data calibrated to other ecoregions. Here we report a comprehensive test of the magnitude, direction, and significance of changes in the distribution of the boreal forest based on the longest and highest-resolution time-series of calibrated satellite maps of tree cover to date. From 1985 to 2020, boreal tree cover expanded by 0.844 million km2, a 12 % relative increase since 1985, and shifted northward by 0.29° mean and 0.43° median latitude. Gains were concentrated between 64°– 68° N and exceeded losses at southern margins, despite stable disturbance rates across most latitudes. Forest age distributions reveal that young stands (≤36 years) now comprise 15.4 % of forest area and hold 1.1–5.9 Pg of aboveground biomass carbon, with the potential to sequester an additional 2.3–3.8 Pg C if allowed to mature. These findings confirm the global advance of the boreal forest and implicate the future importance of the region’s greening to the global carbon budget.
- Preprint
(2765 KB) - Metadata XML
-
Supplement
(3523 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2025-2268', Richard Fernandes, 09 Jun 2025
The results presented in Figure 1 and 2 rest on the temporal precision and temporal stability of the tree cover change estimates (see https://library.wmo.int/idurl/4/58111 for definition of these quantities). The Supplementary information Figure S8 panel b compares calibrated and reference tree cover. This raises three questions that should be addressed:
1. Figure S8b only shows comparisons between model predictions and LIDAR reference data. Comparisons should also be presented and summarized for the high resolution reference imagery,
2. Spatial drift - what is the change in the agreement between predicted and reference tree cover for hold out areas at the tree lines. By hold out I mean completely removing a LVIS track for an ecozone and completely removing data from a high resolution sample tile.
3. Stability - stability is defined as the change in bias over time (e.g. %/year). This is CRICTIAL to understand the uncertainty in trends shown in Figure 2. This requires having time series of reference measurements at sites rather than a spatial sampling only. This is important since the approach presented does not take steps to standardize differences between Landsat imager spectral response, phenological impacts due to changes in intra annual sampling dates, and persistent haze due to forest fires that are often not accounted for well with atmospheric correction.
Citation: https://doi.org/10.5194/egusphere-2025-2268-CC1 -
CC2: 'Comment on egusphere-2025-2268', Richard Fernandes, 09 Jun 2025
The results assess trends in tree cover (%/year) using linear regression and aggregated by latitude. The authors should explain why
1. Mann-Kendall regressions were not used to account for measurement errors as done for many other climate studies.
2. Why report trends by Latitude given that the boreal zone is defined not just by latitude but by growing season length, growing degree days, in the North and often disturbances in the south. It may be more useful to report trends at the northern and southern boundaries relative to a baseline year if one wants to test theories related to the northward shift of the boreal forest.
Citation: https://doi.org/10.5194/egusphere-2025-2268-CC2 -
CC3: 'Comment on egusphere-2025-2268', Richard Fernandes, 09 Jun 2025
The results related to forest age (Figure 4) have a RMSE of 17.46 years and bias of -3.27 years. This is substantial and should be reported in Results S3.3. Given that Fig 4. indicates Age ranges from 2-36 years it is not clear if the RMSE of the estimator of age is sufficiently precise to represent the Map and area of forest age without error bars / identifying areas where the age uncertainty is too large to be useful. Even aggregate ages (Fig 4 bottom) would need to first quantify bias in forest age by region and age level.
I feel this result is based on input estimates that are currently both too uncertainty and not well characterized to include them in this study.
Citation: https://doi.org/10.5194/egusphere-2025-2268-CC3 -
CC4: 'Comment on egusphere-2025-2268', Richard Fernandes, 09 Jun 2025
Publisher’s note: this comment is a copy of CC3 and its content was therefore removed on 16 June 2025.
Citation: https://doi.org/10.5194/egusphere-2025-2268-CC4 - RC1: 'Comment on egusphere-2025-2268', Anonymous Referee #1, 27 Jun 2025
-
RC2: 'Comment on egusphere-2025-2268', Anonymous Referee #2, 06 Sep 2025
This is an important study investigating tree cover change, northward shifts, and post-disturbance recovery in the boreal biome between 1985 and 2020 using the Landsat archive. The results confirm a large-scale northward advance of boreal forests and highlight their growing importance for the global carbon budget. A particularly interesting finding is that “the distribution of disturbance—while varying strongly among years—remained broadly stationary over time” (line 171), whereas several studies based on post-2000 time series report increasing disturbance regimes in the boreal biome. The longer time series used here may provide a broader perspective on the impacts of disturbance.
Tundra ecoregions adjacent to the Arctic Ocean are excluded from the analysis. It would be interesting to know whether tree cover advances can be observed there, particularly along rivers and south-facing slopes.
Specific comments:
- Line 45: Consider citing Pan et al., 2024 (https://doi.org/10.1038/s41586-024-07602-x).
- Line 49: Please reference the more recent IPCC Sixth Assessment Report (IPCC, 2023).
- Line 113: Forest age is defined as the “year of the most recent significant forest gain,” but trees already have age >0 when first detected as forest in remote sensing. This approach may underestimate stand age. Could you clarify, or indicate whether you tested which minimum tree age is recognized as tree cover?
- Line 114: The criterion (“no forest cover or loss within a 150-m radius earlier in the time series”) may misclassify forests regenerating from large-scale pre-1984 disturbances, potentially overestimating the area of “new” tree cover.
- Lines 123–124: The three scenarios of stand age (“100 years yielding 35.8–80.5 Pg C”) would benefit from references or a clearer rationale.
- Line 139 vs. 142: You report boreal tree cover as 7.997 million km² in 2020 under a 30% threshold, but then 9.41–13.26 million km² for the 10–30% threshold. Why does the first figure not fall within the second range?
- Line 151: In Fig. 1, much of the tundra biome is missing compared to Dinerstein et al. (2017). Tundra extends to the Arctic Ocean coast.
- Lines 181–191: Consider citing more recent studies showing changing trends since 2010, e.g. https://www.nature.com/articles/s41561-022-01087-x.
- Supplementary Fig. S1: According to Dinerstein et al. (2017), most tundra ecoregions are omitted (e.g., Scandinavian Montane Birch forest and grasslands). Only ~7 of 48 tundra ecoregions are shown. It may help to clarify that the analysis focuses on the boreal biome and includes only selected tundra ecoregions where tree cover change was observed.
Citation: https://doi.org/10.5194/egusphere-2025-2268-RC2
Data sets
Boreal tree cover dataset Min Feng https://doi.org/10.3334/ORNLDAAC/2012
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 538 | 77 | 21 | 636 | 29 | 17 | 27 |
- HTML: 538
- PDF: 77
- XML: 21
- Total: 636
- Supplement: 29
- BibTeX: 17
- EndNote: 27
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1