the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Jul 2026
SwissPhenoCam: A country-scale dataset of tree-level phenocam greenness captures species-specific phenological variation along elevation gradients in Switzerland
Abstract. Vegetation phenology, the seasonal timing of recurring plant life-cycle events, is a key regulator of ecosystem functioning, carbon and water cycling, and species interactions, while also serving as a sensitive indicator of climate variability and climate change. However, existing phenological observations often face trade-offs between species-level detail, temporal frequency, and spatial coverage. Networks of near-surface digital repeat cameras (phenocams) deliver robust, high-frequency phenological signals, but their spatial coverage remains a limitation. In addition to expanding such phenocam networks, this limitation can be tackled by repurposing the outdoor cameras already deployed for other purposes. Here, we present SwissPhenoCam, a country-scale dataset of tree-level phenological greenness observations derived from a network of outdoor digital cameras distributed across Switzerland. By repurposing cameras originally installed for meteorological monitoring, tourism, and other non-ecological applications, SwissPhenoCam demonstrates the potential of existing imaging infrastructure for large-scale phenological observation.
The dataset comprises imagery from 34 sites spanning broad elevational, climatic, and biogeographic gradients, with archives extending up to 15 years. Individual trees were delineated and identified to species level where possible, enabling the extraction of species-specific greenness time series and phenological transition dates. The curated dataset contains 5,855 tree-years of observations representing more than 20 tree species and captures substantial environmental variability across Switzerland.
Validation against independent observations from the Swiss Phenology Network demonstrates strong agreement between camera-derived and ground-based phenological transition dates, confirming that non-specialized webcams can provide reliable phenological information. Analyses of the dataset reveal pronounced species-specific responses to environmental gradients. Spring green-up is consistently delayed with increasing elevation across deciduous species, whereas autumn senescence shows comparatively weak elevational trends. As a result, growing-season length decreases primarily because of later spring onset rather than earlier autumn decline.
The SwissPhenoCam dataset is openly available through Zenodo (https://doi.org/10.5281/zenodo.20451225) (Garnot et al., 2026) and is intended as a living resource that will expand as observations accumulate. By providing species-resolved, individual-tree phenological observations across a topographically complex landscape, the dataset offers a valuable benchmark for evaluating remote-sensing products, improving phenological models, and assessing climate-change impacts on forest ecosystems. Future work can use this resource to investigate long-term phenological trends, species-specific climate sensitivities, and ecosystem responses to ongoing environmental change.
- Preprint
(32716 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2026-435', Anonymous Referee #1, 02 Aug 2026
-
RC2: 'Comment on essd-2026-435', Anonymous Referee #2, 16 Aug 2026
The manuscript by Sainte Fare Garnote and coauthors describes the Swiss PhenoCam network, and presents a dataset derived from that camera imagery. The network is notable for capturing phenological variation along strong environmental gradients occurring across relatively short distances, and for the focus on extracting individual-tree data from canopy-level imagery. The ability to resolve “how co-occurring species differ in their seasonal timing, greenness trajectories, and growing-season length along Switzerland’s steep elevation gradients” is an important strength of this dataset. The manuscript is well-written, clear, and easy to read. I appreciated that the authors made their code for the entire image processing pipeline available on GitHub. The description of the dataset and its organization appear to be adequate for relatively easy re-use. The comparison against SPN (Swiss Phenology Network) data will be useful for many readers. The manuscript is appropriate for ESSD (the recent PhenoCam V3 dataset by Young et al was published here in 2025).
My main concern with this paper is its claims of novelty, and specifically the question “can reliable phenological time series be extracted from outdoor webcams not originally deployed for ecological monitoring?” As noted below (see my comment re L77), papers on this question were first published 15+ y ago. The conclusion by Sainte Fare Garnote and coauthors that “outdoor webcams not originally deployed for ecological monitoring can also deliver phenological transition dates in strong agreement with independent ground observations.” But in my opinion, this is not always guaranteed to be the case. The quality of phenological data that can be obtained from different brands and models of digital cameras is highly variable, and in some cases although the imagery may look beautiful, there is so much internal processing going on that the phenological data are compromised (this is based on my own experience with several models of cameras by Axis (Sweden), and also some of the “bullet” cameras sold by Campbell Scientific, USA)). I'd like to see the authors either develop quantitative or even qualitiative metrics that could be used to decide when data from a given camera "is good enough", as I think this could make a novel contribution. Alternatively, I would recommend just deleting this research question from the manuscript.
In my comments below, I try to point to some specific papers that the authors might find useful.
L60. Maybe this is pedantic, but I disagree with the statement that phenocam-type monitoring has emerged “over the last decade”. The idea itself, I think, can be traced to Dennis Baldocchi, who presciently wrote in a 2005 paper (https://link.springer.com/article/10.1007/s00484-005-0256-4) “we 10.1016/j.agrformet.2011.09.009each day.” This was followed by a paper by Graham et al 2006 (https://www.journals.uchicago.edu/doi/abs/10.1086/503786) who used vegetation indices derived from repeat digital photography to track the seasonal variation of photosynthetic activity of some moss on a rock. Early “phenocam”-type applications include work in Switzerland (e.g., Ahrends, Eugster, et al 2008 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2007JG000650 and also later papers by Ahrends and Eugster), Japan (Ide and Oguma 2010 https://www.sciencedirect.com/science/article/abs/pii/S1574954110000762) and the US (Richardson et al 2007 https://link.springer.com/article/10.1007/s00442-006-0657-z).
L65. Many, cameras are capable of a fourth (IR) channel. Petach et al. (https://www.sciencedirect.com/science/article/abs/pii/S0168192314001257) is just one of many examples.
L77. Regarding leveraging existing cameras for monitoring traffic etc.– these data have been used in multiple publications. please see Graham et al 2009 (https://doi.org/10.1111/j.1365-2486.2010.02164.x), Morris et al 2013 (https://www.mdpi.com/2072-4292/5/5/2200), as well as work by Robert Pless’s group e.g. Jacobs et al 2009 (https://doi.org/10.1145/1653771.1653789). Furthermore, the “open question” that the authors refer to is not really an open question—it is clear that some, but not all, cameras deployed for other purposes can provide phenological information, but the quality of data varies across cameras (Sonnentag et al. 2012, 10.1016/j.agrformet.2011.09.009), and automated/internal processing of imagery (as the authors of this paper note on L165). The Jacobs paper referred to above (see also Seyednasrollah et al 2018 https://www.nature.com/articles/s41597-019-0229-9) has excellent examples of how bad auto white balancing algorithms can obscure the phenological signal. Looking at the time series plots in the Appendix to the paper by Sainte Fare Garnote and coauthors, it appears to me that both the image-to-image, and day-to-day variability in Gcc and Rcc is higher for some cameras than others. A quantitative analysis of these patterns might be useful. Beyond images per day, can Sainte Fare Garnote and coauthors provide conctrete recommendations on “how good is good enough” based on that image-to-image and day-to-day variability?
L425. There is something going on with the formatting of the reported rates of change (e.g. 4.21,days,hm−1).
L518. As much as I like the decibel concept here, I think it would be easier for most readers to interpret if the SNR was reported just as a ratio.
L551. I think the paper’s contribution would be more accurately summarized by rearranging this sentence, as follows; “The SwissPhenoCam dataset represents a significant step forward in observing vegetation phenology in Switzerland. This is achieved combining high-frequency imagery with tree-level data across elevation gradients and diverse biogeographic zones.”
L576. I don’t understand the point about accuracy vs. precision, or how the proposed approach achieves this. I also would have thought that both accuracy AND precision are important for a national network.
Throughout the manuscript, it feels to me like there could be a few more citations – either to acknowledge the source of ideas that are not the authors’ own, or to point to relevant literature in the context of the point being made.
Citation: https://doi.org/10.5194/essd-2026-435-RC2
Data sets
SwissPhenoCamDataset Vivien Sainte Fare Garnot et al. https://doi.org/10.5281/zenodo.20451225
Model code and software
SwissPhenoCam-greenness-processing: initial release Authors/Creators Vsainteuf https://doi.org/10.5281/zenodo.20435338
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 173 | 72 | 12 | 257 | 14 | 8 |
- HTML: 173
- PDF: 72
- XML: 12
- Total: 257
- BibTeX: 14
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
First of all, I sincerely appreciate your hard work and fantastic study. The authors collected long-term phenology images at multiple sites in Switzerland and then analyzed the characteristics of spring and autumn phenology along the altitude gradient.
While this study is not novel, its uniqueness lies in combining various types of camera networks in Switzerland, including cameras not intended for phenological research. They also sincerely validated the data quality and uncertainty for analyses of RGB time-series data.
From the viewpoint of transparency and reproducibility, their study is unquestionable. I believe this study will support our deeper understanding of the interaction among ecosystem functions and services, and biodiversity under rapid climate changes. However, I have some comments to highlight the significance of this study. First, the volume of this manuscript is too large. I guess readers will struggle to find the significance of this study. Therefore, I strongly recommend arranging the structure of the main text. In other words, moving some parts to the appendix would be beneficial. Second, you should add more suitable references to strengthen your logic. In addition, some of your knowledge is outdated. Please see in my minor comments. Finally, you show many figures. Regarding my first concern, I recommend moving some figures to the appendix. Despite my three concerns, the authors will quickly improve their manuscript.
1: line 34 How about the ecosystem services (cultural) and biodiversity?
2: lines 53-58 Satellite remote sensing has already observed individual tree-level phenology. Please see the following papers.
Chen, B., Y., Jin, and P., Brown. 2019. “An enhanced bloom index for quantifying floral phenology using multi-scale remote sensing observations.” ISPRS Journal of Photogrammetry and Remote Sensing 156: 108–120. https://doi.org/10.1016/j.isprsjprs.2019.08.006.
Miura, T., S., Nagai, M., Takeuchi, K., Ichii, and H., Yoshioka. 2019. “Improved characterisation of vegetation and land surface seasonal dynamics in Central Japan with Himawari-8 hypertemporal data.” Scientific Reports 9: 15692. https://doi.org/10.1038/s41598-019-52076-x.
Pan, B., X., Xiao, S., Luo, et al. 2025. “Identify and track white flower and leaf phenology of deciduous broadleaf trees in spring with time series PlanetScope images.” ISPRS Journal of Photogrammetry and Remote Sensing 226: 127–145. https://doi.org/10.1016/j.isprsjprs.2025.05.013.
Wang, J., Y., Li, M. M., Rahman, et al. 2024. “Unraveling the drivers and impacts of leaf phenological diversity in a subtropical forest: A fine-scale analysis using PlanetScope CubeSats.” New Phytologist 243: 607–619. https://doi.org/10.1111/nph.19850.
Zhao, Y., C. K. F., Lee, Z., Wang, et al. 2022. “Evaluating fine-scale phenology from PlanetScope satellites with ground observations across temperate forests in eastern North America.” Remote Sensing of Environment 283: 113310. https://doi.org/10.1016/j.rse.2022.113310.
3: line 75 We also have a phenology network in Brazil.
https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2023.1223219/full
4: line 78 We have this kind of camera network in Japan.
https://www.sizenken.biodic.go.jp/index_en.php
https://www.sciencedirect.com/science/article/abs/pii/S1574954110000762?via%3Dihub
5: Figure 1 It's quite informative. It's not so easy to understand quickly.
6: Figure 2 I recommend moving this to the appendix.
7: Equations. From a mathematical viewpoint, your descriptions are perfect. However, I guess many readers won't understand them. Please add some explanations. For example, "arg max(g)"
8: Line 223 Please add suitable references.
9: Figure 8 I recommend moving this to the appendix.
10: Table 3 R2 >- R2
11: Figure 12 Please write the scientific name in italics.
12: Table 4 Please write the scientific name in italics.
13: Figure 13 Please write the scientific name in italics.
14: Line 516 Please briefly explain "SNR".
15: Table 6 Please write the scientific name in italics.
16: Figure 15 I recommend moving this to the appendix.
17: Line 557 Please add suitable references.
18: Line 580 How about the ecosystem services (cultural) and biodiversity?
19: Line 587 Regarding my second comment, please revise it. I guess you may find the latest study in Switzerland using Sentinel-2 satellites. Please survey it.
20: Line 589 Please add suitable references.
21: Line 597 Please add suitable references.
22: Line 600 Please add suitable references.
23: Line 602 Please add suitable references.
24: Line 623 Please add suitable references.
25: Conclusion The following papers may support your conclusions.
https://onlinelibrary.wiley.com/doi/10.1111/j.1466-8238.2008.00398.x
https://www.sciencedirect.com/science/article/abs/pii/S1470160X19304674
https://www.nature.com/articles/s41598-025-14547-2
https://link.springer.com/article/10.1007/s00484-020-01953-6
https://www.frontiersin.org/journals/forests-and-global-change/articles/10.3389/ffgc.2021.659910/full
26: Fig A2, A3, A8, A11, A15, and A17. You lost the colour bar in the figures. In addition, please explain the numbers of x and y axes.
27: Line 724 iif -> if
28: Table B1 Please explain such as BEB, BEA, and BLB.
Best wishes,
Shin