Winter and early spring CO₂ losses in a montane peatland are amplified by foehn winds

Abstract

Rapid mid-winter and early spring warming events are emerging as a key but under-recognized driver of carbon loss from cold-region ecosystems. In mountain peatlands, their influence remains largely unknown. Here, we quantify the impact of chinook wind events—warm, dry downslope winds that can raise air temperatures by over 20 °C within hours—on ecosystem respiration in a montane peatland on the eastern slopes of the Canadian Rockies. Using eddy covariance carbon dioxide (CO₂) flux measurements and meteorological data, 13 chinook events were identified between February and May 2021 and segmented each into pre-, during-, and post-event phases. A generalized additive mixed model accounting for temporal autocorrelation showed that CO₂ emission rates increased significantly during chinook events and remained elevated afterward. CO₂ emissions during snow-covered events in February-April were most likely driven by physical degassing from melting snow and thawing surface peat, whereas snow-free events in April and May likely reflected enhanced microbial activity in thawing peat. Our findings demonstrate that regularly occurring cold-season warming events can trigger substantial but short-lived CO₂ releases from mountain peatlands, revealing a climate-sensitive carbon loss pathway likely to intensify as snowpack duration shortens and freeze-thaw regimes shift in mountain regions world-wide.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-50725-6.

Introduction

Peatlands are among Earth’s most important terrestrial carbon sinks, yet they can emit carbon (C) in winter, when biological activity is often presumed near dormant. Covering only about 3% of the land surface but storing approximately one third of global soil carbon ^1,2, peatlands play a vital role in regulating atmospheric carbon dioxide (CO₂). Their C sequestration potential stems from slow decomposition under saturated, anaerobic conditions, which leads to lowered respiratory losses relative to primary production³. However, the cold-season C dynamics of peatlands, particularly in mountain regions, remain poorly understood and are underrepresented in global C models.

Mountain peatlands experience short growing seasons and strong seasonal contrasts, with prolonged snow cover and below-freezing winter temperatures, occasionally interrupted by mid-winter warming events on the eastern slopes of the Rocky Mountains⁴. During winter primary production ceases in peatlands dominated by herbaceous vegetation, and CO₂ fluxes are governed by ecosystem respiration (ER)⁵, including both microbial and root respiration. Despite frozen soils and snowpacks, respiration persists at low rates throughout the non-growing season⁶, contributing a non-trivial amount to the annual C budget. Cold-season CO₂ emissions from boreal peatlands can account for 8–23% of annual emissions^7–9, and up to 12% in the Colorado Rocky Mountains¹⁰. In some cases, these losses are large enough to shift peatlands from annual C sinks to sources¹¹. During early spring, peatlands transition from net CO₂ emission to net CO₂ uptake due to spring phenology, directing the annual balance towards uptake in some cases¹².

The transition from winter to spring is especially dynamic. Microbial respiration persists under snow cover due to cold-adapted communities^13,14, and can increase rapidly as temperatures rise¹⁵. However, temperature is not the sole control on winter ER. Liquid water availability becomes limiting in frozen soils, creating drought-like conditions for microbes¹⁶. Oxygen diffusion is constrained by the water table position during the freeze-up period¹⁷, and aboveground meteorological drivers (e.g., net radiation and wind speed), influencing subsurface CO₂ production¹⁸. During snowmelt, sharp CO₂ pulses may occur^11,19,20 due to both physical degassing^21,22 and biogeochemical activation, including oxygen-rich melt water entering previously anoxic zones²³ and shifts in microbial community composition^24,25. The relevance of these mechanisms in montane systems remains unclear.

In the Rocky Mountains from Alberta to New Mexico, and other mountain regions globally, foehn winds—warm, dry downslope winds known locally as chinooks²⁶—add an additional layer of complexity. Chinooks occur throughout the year but are most noticeable in winter as they can raise air temperatures from well below freezing to over 20 °C within hours^26–29. Chinooks typically last one to several days and drive rapid snowmelt, soil surface exposure, and thawing of near-surface soils. It is not uncommon for chinook winds to deplete over 50% of seasonal snow mass^27,28. On the lee slopes of the Canadian Rockies, nearly a quarter of the coldest winter days are affected by chinooks that raise air temperatures above 5 °C³⁰. These warm events can increase the frequency of freeze-thaw cycles (FTCs)^28,31,32, which can temporarily boost ER due to microbial lysis and substrate release^14,33,34. Yet repeated FTCs may suppress respiration by damaging microbial communities or reducing photosynthetic capacity^34,35, and develop increased FTC tolerance amongst surviving bacteria as observed in chinook-impacted mineral soils³⁰. Despite their frequency and intensity, chinook effects on C fluxes in mountain peatlands have not been studied.

This gap is increasingly important considering climate change. In the Canadian Rockies, warming temperatures are shortening snow duration and diminishing snowpack depth, even as subzero air temperatures persist^36,37. These conditions are likely to enhance the frequency and strength of chinook events, with uncertain consequences for peatland C cycling. For example, the rapid warming over recent decades on the Antarctic Peninsula has been associated with an increase in the frequency of foehn winds³⁸, with similar trends reported in the Alps and Japan^39,40. With potential increases in air temperatures, chinook events may become more frequent (a relationship not yet examined in the Rockies), potentially altering snow cover, soil surface exposure and, consequently, winter thermal and C dynamics. As a result, understanding how these warming pulses influence winter CO₂ fluxes is essential for predicting future ecosystem-climate feedbacks in mountain environments.

Here we investigate the influence of chinook events on ER in a montane peatland in the Canadian Rockies. We hypothesize that chinook-driven warming and snowmelt leads to increased CO₂ emissions during and after chinook events. By characterizing these fluxes across the late winter and early spring transition, and inferring the likely processes responsible, this study aims to clarify the role of this type of synoptic warming in shaping cold-season C dynamics in mountain peatlands.

Results

Chinooks

Thirteen chinook events were identified between February 15 and May 31, 2021, with an average duration of 40.6 h (Fig. 1). The longest event began on April 26 and lasted 109.5 h, while the shortest began on February 16 and lasted 9.5 h. Altogether, chinooks accounted for 25 days, representing 24% of the 106-day study period. Most chinook events (85%) occurred when snow covered the ground and peat remained frozen (Figs. 2 and 3). Event 11 coincided with a transient snow-free period, while event 13 occurred after complete snowmelt, when peat was thawed at all depths (Figs. 2 and 3).

Fig. 1.

Environmental variables during the study period at Sibbald Fen. (A) Air temperature and relative humidity. (B) Wind speed and wind direction. (C) Snow depth and precipitation. Chinook events are highlighted in pink and numbered. All variables are presented as daily mean values.

Fig. 2.

Temporal sequence of snow cover at the study peatland site for our study window. Eddy covariance tower footprint processed for January to October 2021 in red. Imagery: Contains modified Copernicus Sentinel-2 data, retrieved from and processed in Google Earth Engine^41,42.

Fig. 3.

Daily means of ecosystem respiration (ER) (g C m^-2 day^-1) and soil temperatures at 5, 10, 15, and 20 cm depth during the study period. Soil freeze-thaw cycles are represented by red dots. Chinook events are highlighted by season and numbered. No Snow = winter conditions and frozen soil without snow cover. Snow = winter conditions and frozen soils with snow cover. Spring = thawed soil and no snow cover. Positive ER values indicate a release of CO₂ to the atmosphere.

During chinooks, mean wind speed was 2.22 m s^–1 compared to 1.53 m s^–1 during non-chinook periods (Welch’s t-test, p < 0.001), approximately 1.5 times faster (Supplementary Figure S1). During non-chinook periods, winds were primarily down-valley (south-southeast) reflecting the channeling influence of the surrounding topography (Fig. 1, Supplementary Figure S1). Across all chinook events, the air temperature increased by an average of 24.1 °C (SD: 6.83), with the largest change of 36.31 °C occurring during Event 2 on February 18, followed by Event 9 with a change of 35.87 °C on March 30 (Fig. 4). In contrast, the soil temperature at 5 cm depth increased by an average of only 0.4 °C (SD: 0.83), with the greatest rise occurring during event 13 (3.1 °C), which began on May 27. This buffered soil response, despite extreme air warming, reflects a decoupling between air and soil temperatures during most chinook events when peat remained frozen. Once peat thawed in late events (12 and 13), soil temperatures became much more responsive to environmental fluctuations.

Fig. 4.

Plots of (A) air and (B) soil temperature ranges during each chinook event. The temperature range (°C) is indicated above each bar, median temperature is denoted by horizontal black line, and event number is indicated below each bar. Bar width reflects chinook event length.

Freeze-thaw cycles

Frost was observed in the first 20 cm of depth during the study period. Nine FTCs were identified over the study period, with a maximum frost depth of 20 cm. FTCs varied by depth: four at 5 cm, three at 10 cm, and two at 20 cm depth, highlighting more stability at deeper layers. The first FTC occurred during Event 12 on April 29 at 5 cm, following an extended period of persistent frost (Fig. 3). The first FTC at 20 cm depth occurred 13 days later, on May 12. Figure 3 shows that soil temperatures remained below 0 °C at all depths until the end of Event 12 on April 30, after which temperatures at 5 cm rose above 0 °C for 10 days before briefly refreezing. All depths exceeded 0 °C after May 10, with a final temporary refreeze observed on May 22.

Ecosystem respiration response

Daily ER fluxes averaged 2.85 g C m^-2 day^-1 over the study period (Fig. 3). ER was variable yet sustained through late winter with an average of 0.46 g C m^-2 day^-1 (SD = 0.37). Following Event 12, which coincided with soil thaw at 5 cm depth, the average fluxes increased to an average of 4.02 g C m^-2 day^-1 (SD = 3.98).

Table 1 summarizes the generalized additive mixed models used to assess the effect of chinook events on ER. Model 1 was selected based on the lowest AIC, absence of autocorrelation in residuals, and homoscedasticity. This model was considered as the most parsimonious, as increases in complexity did not substantially increase model performance. This model revealed a significant non-linear relationship between soil temperature and ER (edf = 7.292, F = 184.4, p < 0.001; Fig. 5), explaining 64.8% of the variance in fluxes (adjusted R² = 0.648). After accounting for soil temperature, respiration rates were significantly higher during chinook events compared to before (β = 0.00522, SE = 0.00176, t = 2.976, p = 0.00294). Respiration remained slightly elevated after the events as well (β = 0.00581, SE = 0.00197, t = 2.955, p = 0.00315). Back-transformation of the square root ER model indicated that ecosystem respiration increased by ~ 0.056 g C m^− 2 day^− 1 during chinook events, which translates to an approximate 10% rise relative to pre-event conditions. All tested models revealed significant effects of chinook phase and covariates (i.e. net radiation, day/night) on CO₂ emissions.

Table 1.

Models ran for relating chinook phase (PDP) with ER considering the smoothed term for soil temperature at 5 cm depth s(T5), net radiation (NR), and whether it was night or day (ND).

#	Fixed	Random	AR(1)	L	AIC	Ki	Δi	Adj R2
1	s(T5) + PDP	TB	Time | TB	9040.709	-18065.4	8	0	0.648
2	s(T5) + PDP+ND	TB	Time | TB	9200.88	-18078.1	9	12.7	0.655
3	s(T5) + PDP+NR	TB	Time | TB	9048.061	-18383.8	9	318.38	0.686
4	s(T5) + PDP+ND + NR	TB	Time | TB	9206.189	-18392.4	10	326.96	0.690

Time Block (TB) was considered a random effect. Correlation structure AR(1) nests Time within TB. Values include log likelihood scores (L) and number of parameters (K). We assessed models by ranking AIC values (Δ_i) and assessing residuals.

Fig. 5.

(A) Boxplots of ecosystem respiration (ER) during pre, during, and post chinook events across three seasonal periods (winter with snow, winter without snow and spring). ER units represent g C m^− 2 per 30 min. (B) Predicted ER (square root of g C m^− 2 day^− 1) vs. soil temperature at 5 cm by chinook phase; 95% confidence intervals are shown in shaded colours for each phase.

Temperature sensitivity of ecosystem respiration

The calculated Q₁₀ value for soil temperature above − 0.1 °C at 5 cm depth was 7.61 (R² = 0.21, p < 0.0001), indicating a strong temperature sensitivity of ER under thawed conditions (equation of the line: ln(ER) = -3.34 + 0.20T).

Discussion

Our study reveals that chinook wind events, characterized by rapid warming and drying, substantially increase winter and early spring CO₂ emissions from a montane peatland. Using high-resolution eddy covariance and environmental data, we demonstrate that these synoptic events amplify C losses not only during warming but also in the days that follow, even under partially frozen soil conditions. A rough calculation showed that if the entire 0.74 km² peatland responded like the ~ 3050 m² eddy covariance footprint and was dominated by sedges, chinook-triggered ER releases would correspond to an additional ~ 41 kg C day^− 1. Because chinook events represented 24% of the study period, even a 10% increase in ER per event would result into a substantial increase in seasonal C emission during the non-growing season. This finding highlights a climate-sensitive mechanism of non-growing season C release. As climate change is projected to alter snowpack dynamics and increase the frequency of mid-winter thaw events, understanding how chinooks modulate cold-season CO₂ fluxes is critical for predicting future C balances in mountain peatlands globally.

To investigate the drivers of these emissions, ER across 13 chinook events was analyzed and a consistent pattern was found: CO₂ emissions increased significantly both during and after chinook events, relative to pre-event conditions. While temperature increases alone can stimulate microbial activity¹⁵, the sustained emissions observed suggest that additional mechanisms are at play. For the first five events, when peat soils remained deeply frozen (to 20 cm depth) and covered by snow, a possible explanation for CO₂ emission pulses present in the surface is physical degassing: the rapid release of CO₂ that had accumulated beneath the snowpack or within unfrozen peat pores over the winter^21,22. This finding is consistent with patterns observed in other cold-region ecosystems. For example, Elberling and Brandt (2003)⁴² estimated that as much as 14% of CO₂ produced over winter in an arctic tundra site was physically released in a spring pulse. Although microbial respiration can occur under frozen conditions due to unfrozen water films and cold-adapted microbes (e.g.⁴³), the magnitude and timing of early CO₂ emissions suggests that physical release was a major contributor.

As the season progressed and peat began to thaw at depth, a transition toward biological mediated emissions may have occurred, potentially enhanced by percolation of oxygen-rich melt water²³. This was particularly evident after chinook event 12, when soil at 5 cm depth surpassed 0 °C and CO₂ emissions increased noticeably. The calculated Q₁₀ of 7.61 under thawed conditions indicates a sharp increase in respiration with warming and a highly sensitive microbial activity to temperature during the early spring transition. This value exceeds previously reported Q₁₀ values reported for a temperate bog (2.24–4.18;⁴⁴) and permafrost mountain peatlands (1.84–2.51;⁴⁵), suggesting that as thaw progresses, seasonally frozen mountain peatlands may be particularly responsive to warming pulses.

Chinook event 13, the only event to occur under completely thawed conditions across all measured depths, offers a unique perspective. Initially during event 13, CO₂ emissions dropped despite increasing air and soil temperatures in the thawed peat, a counterintuitive response¹⁵. It is hypothesized that this reflects a stress-subsidy dynamic. The preceding freeze event was intense and fast, with < 0 °C occurring down to 20 cm depth lasting 2 days. The freeze event likely caused microbial mortality through cellular damage, reducing respiration (stress phase), but lysis of microbial cells released labile compounds that surviving heterotrophs rapidly consumed once soil thaw resumed (subsidy phase)^33–35. This is supported by the short-lived CO₂ pulse that followed the initial decline. However, the ephemerality of the pulse and the absence of sustained CO₂ increases, despite continued warming, suggest that substrate limitation may have quickly developed after this flush. Alternatively, the shift in CO₂ emissions may reflect a transition in dominant controls, from temperature to water table position or oxygen availability, as thaw progresses⁴⁶. A shift in microbial community composition, perhaps favoring methanogens, may also help explain the flux pattern observed⁴⁷. This site shows a start of the growing season by mid-June based on the green chromatic coordinate⁴⁹. Although phenological timing varies between years, this suggests that autotrophic respiration during event 13 was likely present but minimal, with CO₂ emissions predominantly derived from heterotrophic sources.

CO₂ emissions during winter chinook events are likely shaped by antecedent environmental conditions. Frost depth is influenced by the autumn air temperatures, peat moisture content at the time of freeze-up, and snowpack development⁵⁰. At this site, peat often freezes before snow accumulates, allowing frost to penetrate down to 50 cm depth⁵¹. Snowpacks, when present, insulate the soil and slow heat loss, but thin or melting snow cover driven by chinook events exposes the peat surface to solar radiation and alters freeze-thaw dynamics^27,50. While chinooks are not always sufficient to thaw peat during winter, especially if the peat has a high proportion of water, their cumulative effects on frost depth and snow cover can set the stage for later CO₂ emissions. For example, chinook event 12 was sufficiently warm and long enough to fully thaw peat at 5 cm depth, contributing to elevated emissions in its aftermath.

Although this study is limited to a single site and one winter-spring season, it provides valuable baseline data on how regularly occurring chinook events influence CO₂ emissions in montane peatlands. Results underscore the importance of considering foehn winds as key modulators of C cycling during the non-growing season. These events occur across mountain regions globally, such as the Alps, the Andes, and Southern California⁴, and their frequency and intensity are likely to increase under climate change as it has been predicted in other regions of the world (e.g.^39,40). As such, the patterns documented here may be generalizable across similar cold-region peatland systems.

Future work should include multi-year monitoring to capture interannual variability of chinook occurrence, snowpack, soil thermal regimes, and their influence on peatland CO₂ emissions and budget. Furthermore, comparing CO₂ flux dynamics of chinook events to non-chinook warming episodes would clarify the relative importance of chinooks in peatland C budgets. In this study, we inferred whether CO₂ emissions were biotic or abiotic based on environmental context. However, direct measurements—such as gas flux partitioning, porewater chemistry, and microbial functional profiling—would help quantify source contributions. Additionally, while the focus was on CO₂ emissions, chinook events may also influence methane fluxes. Methanogens often become active during thaw transitions, and methane pulses have been observed in similar contexts⁴⁷.

Finally, it is important to note that the eddy covariance footprint of this study is limited in spatial extent and is dominated by sedge vegetation, which does not reflect the diversity of microhabitats present across the full peatland. Assessing variability across hummocks, hollows, open water, and shrub zones would improve our understanding of landscape-scale responses to chinooks. As climate change reduces snowpack duration and increases the frequency and intensity of chinooks, ecosystem-specific insights will be essential for predicting the future C balance of cold-region wetlands.

Conclusions

This study reveals that chinook wind events are a climate-sensitive trigger for short-lived but significant CO₂ releases from mountain peatlands, increasing emissions during and after warming episodes. The possible drivers may differ with seasonal context, ranging from physical degassing during snowmelt to enhanced microbial activity in thawing upper peat layers. These findings provide the first quantitative evidence of chinook-driven CO₂ loss from a mountain peatland and highlight a mechanism for winter-spring carbon release that is absent from most carbon budgets and models. As snowpack depth reduces and duration shortens, and freeze-thaw regimes intensify in mountain regions worldwide, incorporating the effects of episodic warming events into process-based models will be critical for accurate predictions of the future C balance in cold-region peatlands.

Methods

Study site

Sibbald fen (51°03’29.38” N, 114°52’11.66” W) is a 0.74 km² peatland situated at 1480 m in the montane zone of the eastern slopes of the Canadian Rocky Mountains, Alberta⁵² (Fig. 6). The fen is oriented along a northwest-southeast axis within a glacially carved valley. The peatland is south-sloping (0.7% gradient), with a 60 m elevation difference with the surrounding mountains peaks. It was awarded status as a UNESCO Ecohydrology Demonstration Site in 2024. Characterized by varied microtopography (i.e., hummocks, lawns, and hollows), the peatland receives groundwater inputs from the surficial aquifer and surface water from five inflowing streams. Peat depths range from less than 1 m to 6.5 m and the basal age is ~ 13,000 cal years BP⁵³. Vegetation consists of Salix spp. shrubs, Carex spp. sedges, brown mosses, and limited stunted black spruce (Picea mariana) and Sphagnum spp. moss⁵⁴. The start of the growing season at this site was determined as mid-June based on the timing of increase in the green chromatic coordinate values⁴⁹. A meteorological station equipped with an eddy covariance tower is in the southwest portion of the peatland (Fig. 6). The meteorological station is part of the Global Water Futures Observatory (https://gwfo.ca/).

Fig. 6.

Map of Sibbald Fen in the Rocky Mountains of Alberta, Canada, with a detailed view of the eddy covariance tower footprint processed for January to October 2021. Maps were produced in R⁵⁵ using the ggplot2⁵⁶, sf⁵⁷, ggspatial⁵⁸, and rnaturalearth⁵⁹ packages. Elevation data were retrieved via the elevatr package⁶⁰. Vegetation cover was derived from a supervised classification of UAV-acquired RGB imagery, acquired on 29 August 2021.

CO₂ flux measurements and data processing

Ecosystem-atmosphere greenhouse gas exchange was measured from a 3.42 m tall tower in the southwest part of the peatland using the eddy covariance technique (Fig. 6) between February 15 and May 31, 2021. The distance from the tower to the forested hillslopes was > 100 m in all directions. The tower was equipped with a LI-7500 open path CO₂ and H₂O analyzer (LI-COR, Nebraska, USA), C-SAT 3-D sonic anemometer (Campbell Scientific, Utah, USA), HMP155 air temperature and relative humidity probe (Vaisala, Colorado, USA), HOBO Pendant MX2201 soil temperature loggers installed at 5, 10, 15, and 20–70 at 10 cm increments depth (Onset, Massachusetts, USA), NR Lite2 net radiometer (Kipp and Zonen, Delft, Netherlands), a LI-200R pyranometer (LI-COR, Nebraska, USA) for solar radiation measurements, and a HFP01 soil heat flux plate (HukseFlux: Delft, Netherlands) for ground heat flux.

CO₂ fluxes were sampled at a frequency of 20 Hz, averaged half-hourly, and saved on a CR1000 datalogger (Campbell Scientific, Utah, USA). Fluxes were subsequently corrected for density, sensor separation, self-heating, time lag, spectral attenuation, and coordinate rotation (double coordinate rotation was used) following Fluxnet protocols^61–68. High frequency attenuation was corrected following Moore (1986)⁶⁸. The open-path gas analyzer was mounted within < 10 cm of the sonic anemometer with no intake tubing, minimizing path-averaging and sensor-separation effects. For this configuration at 3.42 m measurement height, theoretical transfer functions indicated high-frequency flux losses of < 5% prior to correction, with residual losses of < 1–2% post-correction. The resulting half-hour fluxes were then processed in a custom R script⁶⁹, which filtered each half-hour to ensure that it had at least 80% of the high-frequency records and that there was no potential for dew formation on the IRGA lenses by either precipitation or by comparing the dew point temperature to air temperature. Additional filtering removed values that were greater than ± 3 standard deviations of a moving average, which consisted of 10 half-hour neighboring values; further filtering was completed to check for physically improbable values.

The Kljun et al. (2015)⁷⁰ footprint analysis was used to constrain the measured fluxes to be within 80% of the desired site boundaries, using their FFP R-functions (Fig. 6). Thereafter, nighttime fluxes with a corresponding friction velocity (u*) below the U50 inflection point (0.11 m s^− 1), derived from a change-point analysis of the nighttime CO₂ flux-u* relationship, were removed from the dataset⁷¹. Following this, the filtered net ecosystem exchange (NEE) data was gap-filled (49% of total) and then partitioned into respiration (ER) and gross primary productions (GPP) using the relationship between night-time respiration and air temperature, following the methods and R-packages from Wutzler et al. (2018)⁷¹. The filtered energy fluxes were forced closed, with the additional energy partitioned based on the Bowen ratio (ß) of sensible heat (Q_h) to latent heat (Q_e)⁷². For each available half-hour, Q_e was gap-filled by scaling potential evapotranspiration, PET⁷³ to ET. The scaler (α) was calculated from the ratio between ET and PET. For periods where no α was available, the ratio was gap-filled using the MDS method outlined in Reichstein et al. (2005)⁷⁴ where Q*, VPD, and u* were used as the gap-filling conditions. When this was not possible, the seasonal mean α was used. Q_h was then gap-filled by solving the energy balance equation using the measured and gap-filled Q*, Q_e and ground heat flux (Q_g). The vegetation across the footprint of the eddy covariance tower is dominated by sedges (Carex spp.), with minor contributions of willow (Salix spp.) and a < 50 m reach of Bateman Creek (Fig. 6).

Chinook identification

Chinook wind events were identified using adapted criteria from Macdonald et al. (2018)²⁷ and Nkemdirim (1991, 1997)^26,29. First, 6-h moving averages of air temperature, relative humidity, wind speed, and wind direction were calculated. A custom R script then flagged potential chinook events based on four conditions: (1) a sustained rise in air temperature; (2) a concurrent drop in relative humidity; (3) increasing wind speeds exceeding 2 m s^− 1; and (4) dominant westerly winds (between 225 and 315°). All flagged events were manually reviewed to ensure they met these criteria and to exclude events occurring during confounding weather patterns such as radiation-driven warming, frontal systems, or katabatic winds.

Snow presence / absence

Due to SR-50 sensor failure in certain dates snow presence was confirmed using high-resolution satellite imagery available via Google Earth Engine⁴¹. Specifically, we used Sentinel–2 multispectral imaging acquired during the study period⁴². A total of 17 images with ≤ 50% cloud cover were independently assessed to verify the presence or absence of snow in the area comprising the peatland and to identify inconsistencies or artifacts (e.g. sensor frost accumulation) in the sensor data. Snow cover classifications derived from the Sentinel-2 imagery were incorporated into our dataset to improve confidence in snow presence/absence determinations during the identified chinook wind events and in the days preceding events.

Environmental conditions prior to the study period

In the autumn preceding the study period, the water table was 78 cm below the surface on October 12, 2020. Prior to freeze-up on November 19, volumetric water content was 52% at 15 cm and 80% at 30 cm depth. These high volumetric water content values suggest that the upper peat profile was nearly saturated with water heading into winter. This indicates that a substantial proportion of the near-surface peat was likely ice-saturated following freeze-up, with ice potentially occupying up to 80% of the total soil volume.

Data analysis

To assess the effect of chinook events on CO₂ emissions, each event was segmented into three phases: pre, during, and post. The pre and post phases were defined as two-day periods preceding and following each event. This window was selected as air temperature changes during chinooks at the site typically precede soil temperature changes by approximately two days. If the post phase from one event overlapped with the pre phase of the next, the intervening period was evenly divided between the two events.

ER data were analyzed using a generalized additive mixed model, implemented in the mgcv R package⁷⁵. Due to non-normality, ER was square root transformed to improve residual distribution. Each 30-minute ER measurement was treated as a replicate. To account for temporal autocorrelation, we included a random effect called Time Block, which grouped observations into two-hour intervals. In addition, an AR(1) correlation structure was incorporated into model within-block autocorrelation. Multiple candidate models were tested that included soil temperature at 5 cm depth (as smooth term), as well as other potential covariates known to influence ER, including net radiation and nighttime vs. daytime. Model selection was based on the absence of autocorrelation in residuals, homoscedasticity, and the lowest Akaike Information Criterion (AIC).

Freeze-thaw cycles

Freeze-thaw cycles were defined as periods during which the 5 cm soil temperature dropped below 0 °C and subsequently rose above freezing, following Robroek et al. (2013)⁷⁶. A single crossing of the 0 °C isotherm in both directions was counted as one FTC.

Q₁₀ calculation

To quantify the temperature sensitivity of ER, we calculated Q₁₀, the factor by which ER increases for every 10 °C rise in soil temperature. We modeled the natural logarithm of ER as a function of soil temperature at 5 cm of depth:

1

where ER is ecosystem respiration (g C m^− 2 per 30 min), T is soil temperature at 5 cm depth (^oC), a is the intercept, and b is the slope of the regression. The Q₁₀ was calculated as:

2

Preliminary inspection of this relationship revealed a clear inflection point close to 0 °C, consistent with a threshold for microbial activity. Therefore, we restricted the analysis to observations where soil temperatures exceeded − 0.1 °C.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by MES. The first draft of the manuscript was written by MES and all authors edited the manuscript. CJW and RMP provided project funding. All authors read and approved of the final manuscript.

Funding

This work was funded by a Canada First Research Excellence Fund sub-grant (Mountain Water Futures), Canadian Foundation for Innovation grant (13163), Alberta Innovates Water Innovation Program grants (222301219 and 3360-E067), and an NSERC Discovery Grant (RGPIN-2017-05873).

Data availability

Ecosystem respiration and meteorological data analyzed in this study are archived in the Zenodo repository, [https://zenodo.org/records/16884435](https:/zenodo.org/records/16884435)⁷⁶.

Competing interests

The authors declare no competing interests.