the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Geographically divergent trends in snowmelt timing and fire ignitions across boreal North America
Brendan M. Rogers
Rebecca C. Scholten
Stefano Potter
Thomas A. J. Janssen
Sander Veraverbeke
Abstract. The snow cover extent across the Northern Hemisphere has diminished while fire extent and severity has increased over the last five decades with accelerated warming. However, the effects of earlier snowmelt on fire is largely unknown. Here, we assessed the influence of snowmelt timing on fire ignitions across 16 ecoregions of boreal North America. We found spatially divergent trends in earlier (later) snowmelt led to an increasing (decreasing) number of ignitions for the northwestern (southeastern) ecoregions between 1980 and 2019. Similar northwest-southeast divergent trends were observed in the changing length of the snow-free season and correspondingly the fire season length. We observed increases (decreases) over Northwest (Southeast) boreal North America which coincided with a continental dipole in air temperature changes between 2001 and 2019. Earlier snowmelt induced earlier ignitions of between 0.22 and 1.43 days earlier per day of earlier snowmelt in all ecoregions between 2001 to 2019. Early-season ignitions (defined by the 20 % earliest fires per year) developed into significantly larger fires in 8 out of 16 ecoregions and 77 % larger across the whole domain. Using a piecewise structural equation model, we found that earlier snowmelt is a proxy for earlier ignitions but may also result in a cascade of effects from earlier desiccation of fuels and favorable weather conditions that led to earlier ignitions. This indicates that snowmelt timing is an important trigger of land-atmosphere dynamics. Future warming and consequent changes in snowmelt timing may contribute to further increases in western boreal fires while the number of fires in eastern boreal North America may increase too with climate change.
- Preprint
(3502 KB) - Metadata XML
-
Supplement
(3298 KB) - BibTeX
- EndNote
Thomas D. Hessilt et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-1741', Anonymous Referee #1, 17 Sep 2023
Major comments
The study by Zeng et al., titled “Geographically divergent trends in snowmelt 1 timing and fire ignitions across boreal North America,” reported the influence of snowmelt timing on fire ignitions across the ecoregions of boreal North America. They found spatially divergent trends in early (late) snowmelt that led to an increasing (decreasing) number of ignitions in the northwestern (southeastern) ecoregions between 1980 and 2019. Early snowmelt is a proxy for early ignition but may also result in a cascade of effects from early desiccation of fuels and favorable weather conditions that lead to earlier ignition. This indicates that snowmelt timing is an important trigger for land–atmosphere dynamics.
Overall, this paper is logical and worthy of publication. However, minor revisions are required prior to publication.
Please note the following points.
- The major limitation of this paper was the incorrect use of the term “snowmelt.” While snowmelt is also associated with snow cover, the authors must understand that the majority of snowmelt occurs at 100% snow cover. The term “snowmelt” used in this paper is incorrect and should be corrected throughout the paper, as it is more accurate than the terms “snow disappearance date” or “snow disappearance timing.” Otherwise, the readers may fail to understand the authors’ analyses. I believe that the title also needs to be revised.
- The authors used surface data from ERA5 as the climate drivers of snowmelt and ignition timing, but ERA5 is a model estimate, and surface data are known to have bias. The bias is particularly large at high latitudes because of the lack of land-based observations and thus differs from ground-based data. It is necessary to demonstrate the validity of using ERA5 in this study with a reference.
- (4) Surface relative humidity was used to model snowmelt timing, but I do not consider it as a good indicator of atmospheric dryness because surface relative humidity varies significantly depending on temperature. Instead, I recommend using the surface saturation deficit. The model in Figure 6 should also be recalculated using the saturation deficit because temperature and relative humidity vary almost identically, which is not desirable as a variable in a hypothesized model.
- Figure 4d shows that there was more variation than a positive correlation, but we cannot say that there was a positive correlation. I would like to have this figure removed or a rationale added to say that the correlation was strong for those with large variations.
- I could not understand what is written in 3.4 the Ignition timing and fire size, or what is expressed in Figure 5. It is necessary to explain the view shown in Fig. 5 in more detail in the text.
Comments on specific points are provided below.
Line 35-36: After "... are heterogeneous across the Northern Hemisphere" in this sentence, we suggest citing Suzuki et al. (2020, doi:10.1002/hyp.13844).
Line 73: We recommend adding Bartsch et al. (2009, doi:10.1088/1748-9326/4/4/045021) to this citation.
Line 108-140: The entire section needs to be revised because snow disappearance timing, and not snowmelt timing, is mentioned here.
Line 220-223: To demonstrate the validity of using ERA5 surface data, please cite the application papers on ERA5 surface data used in such analyses.
Line 260: “andthe…” should be revised to “and the..”
Line 399-401: P may represent a narrow temporal window. Isn't it?
Line 435-437: It does not appear consistent with Figure 4d.
Line 447: "Boreal Interior, Taiga Plain, and Western Taiga Shield..." should be preferably revised to "Boreal Interior (Fig. 5A), Taiga Plain (Fig. 5E), and Western Taiga Shield (Fig. 5N)...".
Citation: https://doi.org/10.5194/egusphere-2023-1741-RC1 -
RC2: 'Comment on egusphere-2023-1741', Quinn Barber, 29 Sep 2023
Geographically divergent trends in snowmelt timing and fire ignitions across boreal North America
General comments: This study by Hessilt et al., titled “Geographically divergent trends in snowmelt timing and fire ignitions across boreal North America”, investigated the temporal trend in snowmelt timing and fire season length across 16 ecoregions of North America, including Canada and Alaska. Using an ambitious combination of climatological data, fire polygons, MODIS snow-cover products, and MODIS active-fire detections, the authors discuss spatially-variable trends in snowmelt timing and the fire season. Using piecewise structural equation modelling, they show that the strongest factor driving snowmelt date is air temperature, which is in turn the strongest factor driving ignition timing, but that snowmelt timing also drives a cascade of effects including fuel desiccation and favourable fire weather, thereby further influencing ignition timing.
Overall, this manuscript is well-written and deserves publication in EGUsphere with minor changes. The authors have done an excellent job of presenting their research with good-quality writing and excellent figures. One general criticism I can offer is that one of the findings, a spatially-divergent trend in ignition timing, has been reported on an Ecozone-level in Hanes et al. (2019)(see Table 3), which has not been adequately cited and discussed here. There are also several methodological limitations which will need to be addressed before publication. In spite of these limitations, this manuscript advances understanding of drivers and trends in ignition and snowmelt seasonality, and is well worthy of publication in EGUsphere.
Specific comments:
- (Lines 115-117) The use of a 15% threshold for 14 consecutive days is problematic at high latitudes. As MODIS angle of observation increases, the degree of noise present in the signal also increases. I’ve attached a snapshot of Terra NDSI from 2018 at a random point in the Northwest Territories, 65 degree latitude. It is apparent that the signal for defining snowmelt jumps above 15% NDSI even deep into August, which would confound the 14-consecutive day algorithm. One would expect this would cause erroneously-late fire snowmelt dates in the northern parts of the study area. However, since this error is applied equally across the study period 2001-2019, it is not likely to significantly bias the findings. I recognize that recalculating NDSI snowmelt timings is outside the scope of this paper, since the authors are using the Verbyla (2017) product. Therefore, I would recommend that the authors discuss the limitations of using MODIS-derived snowmelt products at high latitude in the discussion, possibly in a new “Limitations” section or integrated in the discussion.
- The use of the Canadian Large Fire Database (CLFD) is surprising, and very well may be a typo. The CLFD is a deprecated product which only includes fires from 1959 to 1999. The authors may be referencing the Canadian National Fire Database (CNFDB) or the National Burned Area Composite (NBAC), both available at https://cwfis.cfs.nrcan.gc.ca/datamart.
- (Lines 169-171) Eliminating fires which occurred prior to snowmelt is a potentially confounding factor, and I would suggest this is the biggest weakness of the manuscript. Although fires occurring in deep winter are rightfully excluded, noise in the snowmelt data (see (1)) and long-lasting snow pockets can result in MODIS-detected snowmelt being delayed far after true fire season start (Pickell et al. 2017 provide discussion of this). That is to say, snowmelt detections have a high false negative rate, and should not be used to truncate ignitions except for those ignitions which are obviously impossible. The authors’ calculation of correlation between fire season start and snowmelt is dependent on this assumption, since the authors use 1% cumulative fire ignitions to indicate fire season start. Impacts on trends in fire season length, and fire season start are also possible. I suggest that the authors rerun their calculations including only ignitions which occur at most 7, 14, or 21 days prior to snowmelt, selecting a threshold appropriately.
- (Lines 169-179) Ignition causes are not in line with established proportions, as listed in the NBAC or NFDB. Not counting Alaska, approximately 17% of the fires from 1980-2019 over 200 ha in final area are classified as anthropogenic in the NFDB, although the authors report only 4% have an anthropogenic cause. It’s possible that the authors are identifying some fires in ABoVE which are not represented in the NFDB, but it is very unlikely that the NFDB missed thousands of fires over 200 ha in final area. This is difficult to disentangle without knowing what the true source of Canadian fire polygon data is, see (2).
- (Lines 535 - 537) The idea that earlier snowmelt could lead to a persistent ridge is interesting, but is not supported by your results. This speculation should be cut, or discussion should be changed to indicate that it is speculation, with language such as “Although our results do not clearly indicate…”
- Although this manuscript is on “boreal North America”, it includes the Western Cordillera which is not boreal for most of its area. The authors must change the text throughout to reflect that this study does not encompass only the boreal forest, such as using the word “super-boreal”. Alternatively, a sentence in section 2.1 could be written to define the study area, such as ‘hereafter, “boreal North America” for readability’.
- Some further discussion of the study limitations is warranted, possible in a new ‘Limitations’ subsection but this is up to the authors discretion.
- The lack of significance in many Ecoregions is glossed over, and it should be mentioned that conclusions are being drawn from these insignificant trends (Fig 2). However it must be acknowledged that these trends are logical and are limited by the time series length, so are still worth reporting.
- The authors used pSEM well, but they provide insufficient evidence to conclude that early snowmelt is driving meteorological processes given the limited relationship strength for these variables and the interlinked nature of these variables. It is likely that this relationship varies between ecoregions, which are pooled in the pSEM. Some discussion of this limitation is needed.
- See item (1), the use of a 15% threshold for defining snowmelt.
Technical comments:
- (Lines 10-11) There is no consensus that fire severity has increased (Guindon et al. 2021), rephrase.
- (Lines 18-19) “Earlier snowmelt induced earlier ignitions…” This topline result may be unduly influenced by the decision to cut all ignitions prior to snowmelt. Given the importance of this result it is necessary to recalculate including ignitions prior to MODIS snowmelt.
- (Line 26) “The number of fires in eastern boreal North America” I’m not sure your results support this (Fig 2), it may be better to leave it out.
- (Line 260) Missing space between “and” and “the”.
- (Lines 352-357) These sentences need to be proofread, in particular the “promoted to a distinct” segment is very difficult to read and understand.
- (Lines 459-461) You haven’t shown that the subset pSEMs (Tables S11, S12) are weaker than the full pSEM (Table S10). Is statement based on AIC? AIC is sensitive to sample size, and so cannot be compared between the three models. It would be best to compare model fit using a comparative fit index such as one described in Hu & Bentler (1999). You could alternately cut this comparison entirely, as the intercomparison of the models is not a key finding.
- (Lines 476-478) This sentence is unclear. “DC influenced ignition timing positively” would suggest that higher DC delayed ignitions. “earlier ignitions generally occurred under wetter DC conditions” is also unclear, in the context of the pSEM. Please clarify.
- (Lines 481-482) You can abbreviate CAPE here.
- (Lines 535-537) Fix this grammatical error, there are two sentences here that need to be split. Also, these processes describe the extensive eastern Canadian fires in 2023, and the authors may want to mention that, although it’s not strictly necessary.
- (Lines 617-622) These discussions neglect to mention that the southern and eastern ecoregions have a much higher deciduous proportion than the northwestern ecoregions. Although this is outside of the scope of the study, it’s necessary to discuss that these are a factor influencing the drivers of early-season flammability, particularly in the Boreal Plains.
- (Figure 1) In panel B, you have used the Ecoregions without a split in the softwood shield, this should be changed for consistency.
- (Figure 3) The caption says that the ignitions on these plots correspond to the annual 20th percentile of ignitions, but some of the ecoRegions have ignitions extending past the 300th julian day. How is this possible? Is the caption wrong, or have you accidentally included all the ignitions in some of the plots? Please correct.
- (Figure 4) You are missing an “M” label in panel B. Panel D is not readable, please increase the size of the points significantly or change the points to their corresponding letter label (e.g. “A”, “B”).
- (Figure 5) The grey dashed line is too light and difficult to read. All dashed lines should be darkened and plotted on top of the data. Also, I’m assuming the coloured fire data represents early-season ignitions, while the grey data represents all other ignitions. If this is true, why do some coloured lines appear mixed into the late-season data (e.g. in panel L)? Please fix or clarify.
- (Figure 6) Why are you reporting M-R2 and C-R2 separately for fire weather and for weather? Were these two models run independently? If so, please indicate it somewhere.
- (Works cited) Most of your academic citations do not have a publication year, please add it.
Works cited:
Guindon, L., Gauthier, S., Manka, F., Parisien, M.A., Whitman, E., Bernier, P., Beaudoin, A., Villemaire, P. and Skakun, R., 2021. Trends in wildfire burn severity across Canada, 1985 to 2015. Canadian Journal of Forest Research, 51(9), pp.1230-1244.
Hanes, C.C., Wang, X., Jain, P., Parisien, M.A., Little, J.M. and Flannigan, M.D., 2019. Fire-regime changes in Canada over the last half century. Canadian Journal of Forest Research, 49(3), pp.256-269.
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural equation modeling: a multidisciplinary journal, 6(1), 1-55.
Pickell, P.D., Coops, N.C., Ferster, C.J., Bater, C.W., Blouin, K.D., Flannigan, M.D. and Zhang, J., 2017. An early warning system to forecast the close of the spring burning window from satellite-observed greenness. Scientific Reports, 7(1), p.14190.
Verbyla, D., 2017. ABoVE: Last Day of Spring Snow, Alaska, USA, and Yukon Territory, Canada, 2000-2016. ORNL DAAC.
Thomas D. Hessilt et al.
Thomas D. Hessilt et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 250 | 71 | 12 | 333 | 16 | 3 | 3 |
- HTML: 250
- PDF: 71
- XML: 12
- Total: 333
- Supplement: 16
- BibTeX: 3
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1