Corresponding author: N. Kettridge, School of Geography, Earth and Environmental Sciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK. (email@example.com)
 Wildfire represents the single largest disturbance to the ecohydrological function of northern peatlands. Alterations to peatland thermal behavior as a result of wildfire will modify the carbon balance of these important long-term global carbon stores and regulate post-fire ecosystem recovery. We simulate the 3-D thermal behavior of a peatland that has been disturbed by wildfire to identify how changes in peat temperatures emerge from changes to the surface energy balance and peat thermal properties. Peat temperatures are simulated within two adjacent peatlands, one area having burned 4 years previously, the second which has been wildfire-free for 75 years. We demonstrate that there is only a small alteration to the thermal response inSphagnum fuscum hummocks that are not severely burnt within the wildfire. In contrast, wildfire produces important changes to the energy balance of Sphagnumhollows. A large reduction in the latent heat flux post-fire increases surface temperatures by up to 30°C during daytime summer conditions. However, temperatures through the peat profile are insensitive to these increases in surface temperature. The low surface moisture content of near-surface peat insulates the profile from these higher temperatures and, at depths below 0.015 m, only small differences are identifiable between burned and unburned hollow temperatures. Nevertheless, we argue that these alterations to near-surface temperatures and evaporation rates likely substantially influence the thermal and hydrological conditions post-wildfire, impacting the peatland recovery.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Peatlands account for approximately one third of the global soil carbon pool, storing ∼220–460 Pg C [Turunen et al., 2002]. The resilience [Gunderson, 2000] of this carbon store to changing climatic conditions is uncertain [Moore et al., 1998], and peatlands may represent an important source of atmospheric carbon under a changing climate [Strack and Waddington, 2007]. Soil temperatures are a key control on the carbon balance of peatlands, being exponentially related to CO2 and CH4 production [Moore et al., 2007; Dunfield et al., 1993] and linearly related to CH4 oxidation [Dunfield et al., 1993]. Projecting how peat temperatures will respond within a changing climate is therefore crucial to determining the resilience of this global carbon store [Hilbert et al., 2000]. However, the response of boreal peatlands to changing climatic conditions is superimposed upon a regional-scale fire disturbance [Kasischke and Turetsky, 2006]. Wildfire is an essential component of the boreal landscape, with an estimated 10–15 million hectares burning annually within Siberia, Canada and Alaska [Flannigan et al., 2009], burning 185,000 ha of peatlands per annum in western Canada alone [Turetsky et al., 2004]. This regional disturbance alters the ecohydrological functioning of boreal peatlands [Thompson, 2012]; potentially impacting the thermal behavior of these environments for many decades after the original disturbance [Harden et al., 2006]. The response of peat temperatures to changing climatic conditions, and the subsequent resilience of the carbon store, may therefore be strongly influenced by the time since last fire. With the area burnt per annum within the boreal region projected to increase by 75% by 2100 [Wotton et al., 2010], it is essential to determine how peat temperatures respond post-fire, and to project this response under a changing climate.
 Projecting the response of peat temperatures to changing climatic conditions requires a process based understanding of the array of inter connected feedback mechanisms that regulate soil temperatures. Only when feedback mechanisms are quantified adequately, both under current and future climatic conditions, can reliable forecasts of peat temperature be produced. However, the impact of wildfire on the long-term thermal behavior of boreal peatlands is poorly understood and is biased toward shallow Alaskan ecosystems [e.g.,Harden et al., 2006; Yi et al., 2009; Zhuang et al., 2003] that are unrepresentative of the peatlands that dominate a significant proportion of the boreal biome through Siberia and Canada [National Wetlands Working Group, 1987] or low intensity fires as a result of management practices within the United Kingdom [Ramchunder et al., 2009]. Research has also focused on either statistical comparisons of peat temperatures at different positions along fire chronosequences that do not provide a process based account of the observed temperature differences [Harden et al., 2006] or numerical simulations that exclude key feedback mechanisms that will regulate the response of temperature to a changing climate [Zhuang et al., 2003; Yi et al., 2009].
 Numerical simulations offer the opportunity to represent the complex array of interconnected feedback mechanisms that control the thermal behavior of peatlands. However, current post-fire peat temperature models calculate surface temperatures from simplistic empirical relationships with air temperature [Zhuang et al., 2003; Yi et al., 2009] that are calibrated to individual sites along the fire chronsequence [Yi et al., 2009]. This excludes the complex two-way interactions between available energy, turbulent fluxes and ground heat fluxes that likely provide the essential regulation of peat temperatures and which have been observed to vary in response to wildfire [Thompson, 2012]. However, neither the interconnected nature of such fluxes, nor their impact on subsurface peat temperatures, has previously been simulated within a burnt and unburnt peatlands temperature models. Without this process based understanding such models cannot project the thermal behavior of peatlands along a fire chronosequence under future climatic conditions. They cannot be confidently extrapolated to warmer, potentially dryer, conditions nor can they account for medium-term ecohydrological feedbacks that may result under future climatic conditions, including alterations to the vascular surface vegetation cover [cf.Kettridge and Baird, 2008].
 The low, daily, temporal resolution of previous burnt peat temperature models [Zhuang et al., 2003; Yi et al., 2009] prevents simulated temperatures from being applied to accurately determine the carbon balance of peatlands along a fire chronosequence. Peat decomposition is nonlinearly related to diurnal temperatures variations [Dunfield et al., 1993]. Significant errors may therefore be introduced by simulating peat carbon dynamics from temperatures modeled at a low temporal resolution (average monthly temperatures [Zhuang et al., 2003; Yi et al., 2009]) [cf. Kettridge and Baird, 2008]. In addition, peatlands are characterized by a range of Sphagnum species that are often depicted as a microtopography of hummocks, lawns, hollows, and pools [Rydin and Jeglum, 2006] that are not incorporated within the 1-D structure of current temperature models [Zhuang et al., 2003; Yi et al., 2009]. The different water retention properties of the peat laid down by these different Sphagnum species during the development of the peatland [Hayward and Clymo, 1982] produces small-scale spatial variations in combustion across this microtopography [Shetler et al., 2008; Benscoter et al., 2011]. For example, hummock-forming species such asSphagnum fuscum retain more water at higher tensions compared to species such as Sphagnum angustifolium that characterize Sphagnum hollows. As a result, under dryer conditions the surface moisture content of the S. fuscum peat is higher than that of the surrounding S. angustifolium.This higher water content within the hummock increases the energy required to burn the peat, reducing the burn severity compared to surrounding hollows. We would therefore hypothesize that the thermal behavior of hummocks would respond differently to hollows and that these small-scale differences in peatland function are important in determining the resilience of these global carbon stores to changing climatic conditions.
 We monitor and simulate summer peat temperatures through a hummock and hollow within a continental raised bog at two positions along a fire chronosequence. By developing and evaluating a physically based peatland temperature model, we aim to develop a process based understanding of the regulation of peat temperatures within a burnt patterned peatland and to determine how these processes differ from the more widely researched unburnt ecosystems. This will provide the foundation for a physically based numerical model that can project the response of peatland temperature along a fire chronosequence under changing climatic conditions.
2. Overview of HIP Model
 Peat temperatures are simulated here using adaptations of the HIP-Nmann and 3-D HIP models. Details of these models are presented withinKettridge and Baird  and Kettridge and Baird , respectively. A brief summary of these models are presented here followed by a detailed overview of recent model developments.
 The HIP-Nmann model (from this point forward referred to as the 1-D HIP model) is a 1-dimensional peat temperature model. The model includes energy transfers by conduction and advection of vapor. Conduction is simulated using Fourier's first law in which the heat fluxq (W m−2) is given by:
where T is temperature (K), z is depth (m) and the proportionality constant k is the thermal conductivity (W m−1 K−1). The vapor phase advective heat flux is assumed to follow equation (1) except that the thermal conductivity of the soil is replaced by a vapor transfer parameter kv (W m−1 K−1), given by [de Vries, 1963]:
where Lv is the latent heat of vaporization of water (J kg−1), R the gas constant of water vapor (J kg1 K1), κ the diffusion coefficient of water vapor in air (m2 s−1), A the atmospheric pressure (Pa), and pws the saturation vapor pressure (Pa) [de Vries, 1963]. Liquid advective heat transfer in the vertical direction was shown by Kettridge and Baird to have little effect on measured peat temperatures, and was excluded from the 1-D HIP model.
 The thermal conductivity, including allowance for vapor flux, is calculated from Farouki :
where P is the porosity, θ is the volumetric moisture content, fo is the volumetric fraction of organic matter, and the subscripts a, o and w indicators of air, organic matter and water respectively. Values of ko and kw are assumed to be 0.25 and 0.59 W m1 K1, respectively, and ka combines the conduction and advective vapor component and is given by:
The vertical variation in the volumetric heat capacity C (J m3 K1) is calculated by summing the soil constituents multiplied by their respective heat capacities, with the volumetric moisture content simulated from the preferred capillary rise model presented in Kettridge and Baird .
 The surface boundary condition solves the surface energy balance:
where QG (W m−2) is the ground heat flux, α the surface albedo (−), K↓ (W m−2) the incoming short wave radiation, L↓ (W m−2) the incoming long wave radiation, L↑ (W m−2) the outgoing long wave radiation, QH (W m−2) the sensible heat flux and QE (W m−2) the latent heat flux at the peatland surface. Subscript s defines the radiation intercepted by the peatland surface, and excludes that which is intercepted by the vascular vegetation cover. The short wave and long wave energy balance are calculated in accordance with Brock and Arnold , incorporating a light extinction coefficient resulting from the vascular vegetation cover, simulated from Beer's law [cf.Massman, 1992]. The latent heat flux of evaporation is calculated in accordance with the Penman-Monteith model [Oke, 1987], and the sensible heat flux from Newton's law of cooling.
 HIP 3-D simulates temperatures through the center of a hummock, accounting for the important control of horizontal energy transfers (within the hummock and between the hummock and adjacent hollow) on peat temperatures [Kettridge and Baird, 2008]. These horizontal energy transfers are simulated by approximating the 3-D thermal behavior of the hummock and the surrounding hollow in accordance with the HIP 1-D model. Variations in the surface energy balance across the peat microtopography, resulting from spatial variations in the slope, aspect, shading, and from alterations to the aerodynamic resistance, are all incorporated within the model.
3. Development of HIP Model
 Previously, the HIP 1-D and 3-D were applied to simulate summer temperatures within a blanket peatland within the United Kingdom. For the model to be applied to northern Alberta, Canada (see study site), the following model developments were required.
 The HIP 1-D and 3-D models previously simulated the thermal behavior of unfrozen peatlands. However, ice is present through the initial growing season in peatlands throughout northern Alberta and provides an important control on the thermal behavior of these environments.Equation (3) is modified to incorporate the effect of ice on k:
where ki equals 2.22 W m1 K1 and the subscripts i and w indicators ice and water. In accordance with McKenzie et al. , θi is temperature dependent. When T > 2°C, θi is equal to zero. When T < 0°C, θi is equal to θ, and when 0°C < T < 2°C:
We apply the HIP model to simulate the melting of ice and therefore take no account of hysteresis in the above relationship. The latent heat of fusion is incorporated by increasing the volumetric heat capacity when node temperatures are between 0°C and 2°C to account for the energy required to melt the volume of ice per unit temperature increase defined by equation 7 [McKenzie et al., 2007]. The apparent volumetric heat capacity (Capp) is equal to:
where C is the volumetric heat capacity and Lf is the latent heat of fusion.
3.2. Aerodynamic Resistance
 Alberta peatlands are characterized by a tree cover that increases near-surface atmospheric turbulence, reducing the aerodynamic resistance to turbulent fluxes. As a result, the surface roughness of a treeless Swedish peatland, presented byMölder and Kellner and incorporated within the HIP 1-D and 3-D models [Kettridge and Baird, 2008; Kettridge and Baird, 2010], cannot be applied within the subsequent model simulations. In addition, the aerodynamic resistances measured across the hummock surface and incorporated into the 3-D HIP model will likely overestimate the hummock aerodynamic resistance within the treed system. With hollow micro-features, we therefore apply aerodynamic resistances calculated from vertical wind profiles measured within each study site [Thompson, 2012]. Within hummock micro-features, for simplicity, and in the absence of further evidence, we assume that the reduction in aerodynamic resistance across the studied hummocks is proportional to the reduction observed byKettridge and Baird  within hummocks of a similar size.
3.3. Peat Moisture Content
 Vertical moisture contents within the HIP 1-D and 3-D models were simulated from measured water table depths, applying a development of the empirically based capillary rise model ofGranberg et al. , presented by Kettridge and Baird . The model is developed here to simulate unsaturated water content profiles within the study sites. Surface moisture contents, θ0, are calculated from the average near-surface (0–10 cm) water retention curves presented byThompson , assuming the peat surface is in equilibrium with the water table depth, i.e., water can be supplied adequately to the peat surface to match the evaporative demand:
where zwt is the water table depth (m) and α and nare empirical constants equal to 3.7 and 1.28, respectively. In accordance within the HIP 1-D and 3-D models, the volumetric moisture content at depth z,θ(z), is equal to:
where Pis the average porosity. Unlike previous simulations of 1-D and 3-D HIP model, surface moisture contents are not restricted to a minimum volumetric moisture content of 0.2, but are instead fully dependent on the water table depth and the peat moisture retention.
3.4. Initial Conditions
 Initial temperatures within the 1-D HIP model are linearly interpolated from measured temperature at the beginning of the simulation period. Within the 3-D HIP model, the subsurface is categorized as being beneath either a hummock or a hollow. Initial temperature profiles beneath the hummock are assumed to be equal to the measured hummock temperature profile through the hummock center. Temperatures beneath the hollow are assumed to be equal to the measured hollow temperature profile.
4. Study Site
 Measurements were conducted at a burned and unburned site within a single raised ombrotrophic peatland complex in central Alberta, Canada (55.8°N, 115.1°W; Figure 1) between 15 July and 15 August 2010 (d.o.y. 196–227). The unburned site has not burned since 1935 (hereafter referred to as BC35). It is characterized by S. fuscum hummocks and S. angustifolium hollows, a vascular vegetation cover that includes Ledum groenlandicum, Rubus chamaemorus, Maianthemum trifolia, Vaccinium oxycoccus and V. vitus-idea, and a tree cover of black spruce (Picea mariana) with a basal area of 11 m2 ha−1 and an average height of 2.3 m. Hummocks covered 72% of the peatland by area, and hollows 28% of the peatland by area. The burned site, defined here as BC06, burned as part of a 44 ha wildfire in September 2006. The fire killed all trees within the fire scar and burned the peat to an average depth of 75 mm [Thompson, 2012], calculated using the adventitious root method [Kasischke et al., 2008]. This depth of burn varied substantially between different micro-features.S. angustifoliumhollows were consistently burned, producing bare ‘peat’ surfaces characterized by an approximately 5 mm thick layer of dark-colored char. In comparison, the impact toS. fuscumhummocks was more varied, demonstrating the greater resistance of these micro-features to fire under dry conditions [Shetler et al., 2008]. While many S. fuscumhummocks burned, a significant proportion were only singed by the fire and did not burn to any definable depth. In the 4 years post-fire, widespreadSphagnum recolonization has not occurred at BC06 and Sphagnumis principally a remnant of the pre-fire ecosystem. The true mossPolytrichum strictumhas colonized approximately 10% of the hollow surface within 4 years of fire. The black spruce tree cover killed within the fire also remains as remnant stands with no branches or foliage. However, substantial recolonization by vascular vegetation species has occurred within both the hummock and hollow micro-features. Recolonization of the burned peat surface had been primarily byL. groenlandicum, R. chamaemorus, and V. vitus-idea, with the other vascular species largely absent. Within BC06, hummocks covered 62% of the peatland by area, and hollows 38% of the peatland by area. BC35 and BC06 shared very similar synoptic weather conditions, being less than 700 m apart. However, the denser black spruce canopy of BC35 increased turbulent mixing within the surface boundary compared to BC06. The roughness length of momentum was 0.22 at BC35, but decreased to 0.14 m at BC06 [Thompson, 2012].
 At each site, peat temperatures were measured within the 2010 growing season within one hummock and one adjacent hollow and peat moisture was measured within the second hummock and the second adjacent hollow. Within BC35, the S. fuscum hummock used for temperature measurements was 0.30 m in height and approximately 1.0 m in diameter. Peat temperatures were measured through the center of the hummock at depths of 0.0, 0.015, 0.03, 0.07, 0.13, 0.23, 0.4, 0.6 and 1.0 m using copper constant type T thermocouple wire (error ± 0.5°C). Within the adjacent S. angustifoliumhollow, temperatures were measured at a distance of 1.0 m to the northeast of the hummock center at depths of 0.0, 0.015, 0.03, 0.07, 0.13, 0.23, 0.4, 0.6 m. The chosen hummock and hollow were not substantially influenced by shading from the black spruce canopy. Within the second hummock and hollow, peat moisture contents were measured at a depth of 0.05, 0.15 and 0.30 m using Campbell Scientific CS616 water content probes. Standard meteorological measurements of air temperature, humidity, incoming short wave radiation, net radiation, rainfall, and wind speed at 10, 5 and 2 m were also measured, averaged over 20-min intervals. Water table depth relative to the peat surface was also monitored using an Odyssey water level recorder. Within the 2011 growing season, peat surface temperatures were monitored in three separate hummocks and three separate hollows within BC35 and in three separate hummocks and three separate hollows within BC06.
 The instrumented hummocks and hollows within BC35 were considered representative of how the BC06 hummocks and hollows may have appeared today if they had not burnt in 2006. This was based both on the interpretation of laboratory investigations showing the different resistances of S. fuscum and S. angustifolium to fire [Benscoter et al., 2011], and from a visual comparison between the BC35 and BC06 sites. Both hummocks monitored within BC06 were similar in height and diameter to those instrumented within BC35. Within the hummock center, the capitula of the S. fuscum were singed during the fire, and more distinguishable burning had occurred around the hummock edges. The depths of burn within the S. angustifolium hollows were more substantial and were both characterized by a bare ‘peat’ surface. Peat temperatures, moisture content, water table depth and meteorological conditions were measured at BC06 in accordance with BC35.
5.1. Observed Temperatures
 The surface temperatures of the BC35 and BC06 hummocks fluctuated between 3°C and 36°C during the experiment period (Figure 2; representative three day period shown for clarity). At night there is strong correspondence between hummock surface temperatures. During the day, temperature differences increase, with surface temperatures within BC06 exceeding those of BC35 by up to 5°C (Figure 2). Through the hummocks (depth 0.04–0.13 m), temperature differences between BC35 and BC06 are small and there is a close correspondence in the diurnal temperature fluctuations. Deeper within the peat profile, temperatures are cooler within the BC35 hummock, by up to 2°C at a depth of 0.4 m (Figure 2). These temperature differences reduce with depth and no clear difference in peat temperature is evident at a depth of 1.0 m (data not shown).
 The surface temperature of the BC35 hollow is comparable to the surface temperature of the BC35 and BC06 hummocks, with temperatures ranging between 0°C and 34°C (Figure 3; note difference in scale compared to Figure 2). In comparison, the maximum daytime surface temperatures of the BC06 hollow are substantially higher (Figure 3). Surface temperatures within the BC06 hollow increase rapidly after sunrise and at midday are up to 25°C warmer than the surface of the BC35 hollow. This large diurnal variation in BC06 hollow temperatures reduces substantially within the top 0.015 m of the peat profile and, at a depth of 0.015 m, peat temperature are very similar within the BC35 and BC06 hollows; differences in peat temperatures reduce to a maximum of 2°C. Despite the much higher surface temperatures within the BC06 hollow, deeper within the peat profile, at a depth of 0.4 m, peat temperatures are actually 2°C cooler in the BC06 hollow compared to the BC35 hollow.
 The differences between surface temperatures and air temperatures at 13:00 LT (Ts-a) are compared within a range of hummocks and hollows within BC06 (Figure 4) and BC35 (Figure 5). Temperatures are compared during similarly dry hydrological conditions during the 2010 and 2011 growing season. During these periods, the water table depth within a BC35 hollow ranged between 0.17 and 0.24 m in 2010 and 0.12–0.24 m in 2011 (Table 1). Within a BC06 hollow, the water table depth ranged between 0.14 and 0.29 m in 2010 and 0.14–0.31 m in 2011 (Table 1).
Table 1. Maximum and Minimum Water Table Depths During Comparison Periods for the 2010 and 2011 Growing Seasons
 Within the BC06 hummocks, median Ts-a ranged between 0.3 and 8.3°C (Figure 4). Although Ts-a is lowest within the 2010 hummock (that which provides the focus of this manuscript), its value is comparable to the other hummocks measured during 2011. Median Ts-a is significantly (p < 0.05) higher within the BC06 hollows than the BC06 hummocks and ranges between 8.3°C and 24.8°C. Median Ts-a is highest within the 2010 hollow (hollow 8; Figure 4), although its value is comparable in magnitude to hollow 6 and 7 (Figure 4), while hollow 5 shows a lower median Ts-a of 8.3°C. Failure of the logging instrumentation prevented a full comparison of temperature at the six locations within BC35 (Figure 5). However, Ts-a was similar between the functioning probes within both the hummocks and hollows. Median Ts-a was equal to −1.1°C and 3.6°C in BC35 hummocks, and 6.7°C and 5.4°C within the BC35 hollows.
5.2. Observed Water Contents
 Measured moisture contents varied little during the 2010 study period within both BC35 and BC06 (Table 2). For example at a depth of 0.05 m within the BC35 hummock, moisture contents ranged between 0.12 and 0.125, with only a small spike of 0.14 during a 23.6 mm rainfall event on d.o.y. 214. Moisture contents at a depth of 0.05 m were lower within the BC06 compared with BC35, and lower within the hummocks compared to the hollows (Table 2). Moisture contents average 0.07 within the BC06 hummock, increasing to 0.15 m within the BC06 hollows. In comparison, the moisture content at a depth of 0.05 m average 0.12 in the BC35 hummock compared to 0.24 in the hollow.
Table 2. Maximum and Minimum Water Contents Within the Hummocks and Hollows of BC35 and BC06
5.3. Simulated Peat Temperatures
 The 1-D and 3-D HIP models provide a reasonable representation of measured temperatures within the hummock and adjacent hollow of BC35 (Figure 6). Within the BC35 hollow, average surface temperatures and the diurnal temperature variation are well represented by the 1-D HIP model. However, average peat temperatures are slightly underestimated through the profile. At a depth of 0.07 m, the model provides a good approximation of the maximum daily temperature, but underestimates the diurnal temperature variation by ∼2°C. At a depth of 0.23 m, peat temperatures are well represented between d.o.y. 196 and 201. Measured temperatures then increase relative to the simulated temperatures and the model underestimates peat temperatures by ∼1°C. Deeper within the peat profile, at a depth of 0.4 m, the model provides a good approximation of measured temperatures.
 The 3-D HIP model simulates cooler maximum surface temperature within the BC35 hummock than the BC35 hollow, resulting from a reduction in the aerodynamic resistance across the hummock. However, measured maximum surface temperatures within the BC35 hummock are not cooler than the adjacent hollow and, as a result, the 3-D HIP model underestimates maximum hummock surface temperatures (Figure 6). Further through the peat profile, the 3-D HIP model provides a reasonable representation of peat temperatures. At a depth of 0.07 m, the 3-D HIP model provides a good estimation of the average temperatures and its diurnal variation, although with the simulated diurnal temperature variation lagging slightly behind measured temperatures. At a depth of 0.23 m and 0.4 m, peat temperatures are overestimated by ∼2°C. The importance of ice on deeper peat and its spatial distribution through the hummock hollow microtopography temperatures is clearly evident within the peat profile. By excluding the effect of the latent heat of fusion within the 3D HIP model, simulated temperatures are increased by ∼3°C at a depth of 0.4 m, leading to an overestimation of peat temperature by ∼5°C (Figure 6). Near the peat surface, the impact of ice on peat temperatures is substantially reduced (data not shown).
 Within BC06, the model performance varies greatly through the surface microtopography (Figure 7). Hummock surface temperatures are well represented by the 3-D HIP model with the model providing a good approximation of the average temperatures and the diurnal temperature variation. Within the hummock, for example at a depth of 0.07 m, the model provides a good approximation of the average peat temperatures (Figure 7). However, simulated temperatures substantially underestimate the magnitude and timing of the measured diurnal temperature variation. Deeper through the peat profile the 3-D HIP model continues to provide a reasonable representation of average hummock temperatures, but underestimates the diurnal temperature variations (for example, at a depth of 0.23 m). At a depth of 0.60 m, the 3-D HIP model provides a good representation of the peat temperatures at the beginning and end of the study period, although simulated temperatures demonstrate an apparent step change associated with ice melt rather than the gradual increase in observed temperatures.
 Within the BC06 hollow, the 1-D HIP model does not account for the large diurnal surface temperatures variation. The model provides a reasonable approximation of the minimum nighttime surface temperatures, but underestimates the maximum daytime temperatures by up to 30°C. While there is a large error in the ability of the 1-D HIP model to represent these high near-surface temperatures, at a depth of just 0.015 m, the model provides a good approximation of peat temperatures. With increasing depth, the 1D HIP model even over estimates the measured peat temperatures by ∼1.5°C at a depth of 0.13 m and ∼1.0°C at a depth of 0.23 m.
 Simulations of the thermal behavior of an unburned and burned Alberta peatland identify important questions regarding the ecohydrological function of these ecosystems. Areas where the HIP models fail to provide a reasonable representation of measured temperatures suggests a likely alteration to the thermal behavior of the peatland ecosystems from that which the model was originally developed and evaluated. While differences between modeled and measured temperatures are evident within BC35, errors in the model performance occur primarily within BC06, suggesting that wildfire significantly influences the peatland thermal system and/or the ability of the HIP model to replicate this system. We indicate below the important differences between modeled and measured temperatures and discuss how these errors inform our understanding of the thermal behavior of the peatland pre- and post-fire.
6.1. Underestimation of Surface Temperatures of the BC35 Hummock
 Measured daytime surface temperatures do not vary substantially between the BC35 hummock and hollow. In comparison, model simulations suggest cooler hummock surface temperatures associated with the 42% reduction in aerodynamic resistance across the hummock. The underestimation of hummock surface temperatures could result from (i) higher surface resistances within the hummock that reduces evaporation rates, (ii) lower vascular vegetation cover at the point of installation of the surface thermocouple that maximizes the intercepted short wave radiation, and/or (iii) a similar aerodynamic resistance between the BC35 hummock and adjacent hollow. The addition of trees into the peatland ecosystem significantly increases the surface roughness (roughness length = 0.22 m), reducing the aerodynamic resistance to evaporation (ranging between 26 and 367 s m−1). However, within BC35, this increase in the surface roughness is associated with an increase in the displacement height, from zero within untreed [Mölder and Kellner, 2002] or burned peatlands, to a height of 0.63 m at BC35 [Thompson, 2012]. This displacement height is over twice that of the simulated hummocks. Differences in the aerodynamic resistance observed between hummocks and hollows within untreed peatlands may therefore not be appropriate within the treed peatland ecosystems.
6.2. Larger Diurnal Temperature Variations Within BC06 Hummock
 The 3-D HIP model underestimates the diurnal temperature variation through the BC06 hummock (Figure 7). Because hummock surface temperatures are well represented by the 3-D HIP model, this underestimation of the magnitude of the diurnal temperature variations results either from the erroneous representation of the hummock thermal properties or from the inaccurate representation of horizontal energy transfers within the hummock. The former appears improbable as wildfire would likely cause a drying of the peat surface [Thompson, 2012], insulating the subsurface peat from the diurnal temperature variations (see below). In addition, measured temperatures do not vary substantially between the BC35 and BC06 hummocks, suggesting similarity in the functioning of the thermal system. The 3-D HIP model accounts for spatial variations in the available energy across the hummock associated within variations in slope aspect and shading. It appears likely that model errors result from the inability of the model to adequately account for the horizontal energy transfers through the BC06 hummock, resulting from the complexity of the BC06 surface microtopography.
 Wildfire does not burn hummocks evenly [Shetler et al., 2008; Benscoter et al., 2011]. Variations in the depth of burn are often evident between the top and sides of hummocks associated with variations in the near-surface moisture content associated with differing rates of evaporation [Kettridge and Baird, 2010]. Differences in the depth of burn are further enhanced by variations in the slope and aspect of the surface. A surface angled toward the flaming front receives a greater radiation pulse, increasing the depth of burn compared to a flat surface [Dupuy and Maréchal, 2011]. These factors cause a transformation of hummocks shapes pre- and post-fire, producing hummocks with more vertical sides. The 3-D HIP model incorporates the affect of such changes in the hummock shape on the available energy, associated with changes in slope and aspect. However, because the 3-D HIP model was developed to simulate smooth, unburned hummocks, it assumes that this spatial distribution in solar heating is applied to the surface of a hummock that is discretized into 0.05 m deep nodes. A hummock with vertical sides is thus simulated as a stack of nodes with energy applied to the top of the hummock that propagates vertically through to its interior. In reality, solar radiation would directly heat the sides of such a hummock and that energy propagates directly into the hummock center, increasing the diurnal temperature variations. Therefore, the difference between modeled and measured temperatures within the burned hummock likely results from a failure of the 3-D HIP model to simulate the complex thermal dynamics within the hummock, rather than an alteration to the thermal dynamics of a hummock as a result of fire.
6.3. Insensitivity of BC06 Peat Profile to Surface Temperatures
 The HIP model substantially underestimates the diurnal variation in BC06 hollow surface temperatures. Such large surface temperature variations would, if applied to the HIP model, propagate through the peat profile producing large diurnal temperature variations with depth. However, within BC06, such diurnal temperature variations are substantially reduced at a depth of 0.015 m and are equivalent to the diurnal temperature variations observed within the BC35 hollow.
 To determine why the deeper peat profile is so resistant to the large surface temperature variations, we adjusted the surface boundary of the 1-D HIP model. Instead of determining the ground heat flux from the surface energy balance, a dirichlet boundary condition [Kettridge and Baird, 2008] was applied where surface temperatures were driven from measured temperatures (equivalent to the basal boundary condition). This model is comparable to the HIP-Dlet model [cf.Kettridge and Baird, 2008]. If the top 0.015 m is removed from the simulated peat profile, and the model driven from measured temperatures at a depth of 0.015 m, a reasonable representation of peat temperatures is produced (data not shown). This demonstrates that the large diurnal variation in peat temperatures observed at the peat surface is dampened within the top 0.015 m of the profile and the functioning of the thermal system below does not differ substantially from an unburned hollow.
 This dampening results from the low near-surface moisture content of the peat. The HIP model assumes that the near-surface moisture content is in equilibrium with the water table depth. While moisture contents measured at a depth of 0.05 m (ranging between 0.19 and 0.13) do not differ substantially from the equilibrium moisture content, the near surface of the BC06 hollow was substantially water stressed. The thermal properties of the drier surface peat will, therefore, insulate the profile below. We incorporated such a low surface moisture content within the HIP-Dlet model, reducing the peat moisture content in the top 0.02 m of the profile to zero (C = 12500 J m−3 K−1, k = 0.036 W m−2 K−1). This dry near-surface layer reduces the diurnal temperature variations (Figure 8). The average temperature and the magnitude of the diurnal temperature variation are well represented by this model at a depth of 0.07 m. At a depth of 0.23, simulated temperatures are overestimated by approximately 1°C.
6.4. Cause of Warmer BC06 Hollow Surface Temperatures
 While the dry near-surface peat dampens the large diurnal variations in surface temperature within the BC06 hollow, it does not substantially influence the magnitude of the surface temperatures. Therefore, we incrementally reparameterized the 1-D HIP model, initially parameterized for the BC35 hollow (but with a dry near surface), in an attempt to replicate observed surface temperatures within the BC06 hollow.Figure 9 compares average surface temperatures at 13:00 LT between the different model parameterization. Surface temperatures of the initial BC35 parameterization averaged ∼24°C (Figure 9.1). Modifying the initial temperatures and the basal boundary condition of the model to the BC06 site (Figure 9.2), and subsequently changing the meteorological inputs from BC35 to BC06 (Figure 9.3) had little impact on simulated surface temperatures. Increasing the transmissivity from 0.64 to 0.78 to account for the more sparse vascular vegetation cover at BC06 (Figure 9.4) and then reducing the surface albedo from 0.13 to 0.06 to account for difference in albedo between the Sphagnumsurface of BC35 and the dark-colored char of BC06 (Figure 9.5) increased the average surface temperature by a further 0.7°C. Subsequently altering the surface roughness to account for the impact of the reduced tree canopy cover increased the surface temperatures by 0.5°C (Figure 9.6). Overall, these alterations to the model parameterization increased the average temperatures by 2.1°C, only 15% of the 15°C increase necessary to reach that observed.
 Surface temperature can be increased by altering two further model parameters; surface resistance and transmissivity (vascular vegetation cover). The surface resistance is parameterized for an unburned peatland in which the water is readily available to the peat surface [Kettridge and Baird, 2008], maximizing rates of evaporation. The surface resistance provides a key regulation of surface temperatures, particularly within Albertan peatlands where, at BC35 and BC06, the humidity regularly dropped below 40% during daytime conditions. We suggest that within burned peatlands, water evaporates from the peat surface during the day at a rate faster than it can be replenished. As the peat dries, the peat near-surface hydraulic conductivity is substantially reduced [Price et al., 2008] and the surface becomes hydrologically disconnected from the subsurface water supply. It has been shown in bare soil environments that under such conditions the evaporation surface, originally at the soil surface, propagates down through the soil profile [Choudhury and Monteith, 1988]. This substantially increases the surface resistance because moisture must diffuse through the dry near-surface zone before it is transported via turbulent fluxes in the air above. Increases in surface resistance of 1500 s m−1 have been observed previously within mineral soils with a lowering of the evaporation surface of only 8 mm [Aluwihare and Watanabe, 2003]. Within the burnt hollow, this would induce an approximately 16-fold increase in surface resistance. We incorporated this feedback within the 1-D HIP model by reducing evaporation from the peat surface to zero. This produces a substantial increase in the average surface temperature of 6°C and cumulatively accounts for 54% of the observed increase in average daily surface temperatures (Figure 9.7). No other alteration to the peat surface boundary condition can account for such a large increase in measured surface temperatures.
 The surface transmissivity of the 1-D HIP model (equal to 0.77) is parameterized for the average vascular vegetation cover of the BC06 hollow. However, the vascular vegetation cover is not distributed uniformly within the micro-feature. Within the BC06 hollow, the vascular vegetation is patchy with areas with no vascular vegetation (Figure 1). While the small-scale spatial variations in near-surface temperatures will average with depth, such spatial variations will impact model performance when simulations are compared with point-scale measurements of surface temperature. Surface temperatures within the BC06 lawn were measured within a small exposed area of the BC06 hollow with no vascular vegetation cover. The 1-D HIP model was therefore additionally parameterised without a vascular vegetation cover. This caused a further 2.7°C increase in peat temperatures (Figure 9.8) and in total accounts for 70% of the observed temperature increase. As such a small-scale variations in the vascular vegetation cover will likely be average with depth, this additional paramterisation is not incorporated within the modified 1-D HIP model.
 By incorporating the dry near-surface peat and the reduction in the evaporation rate, the 1-D HIP model provides a reasonable representation of peat temperatures at the peat surface and through the peat profile (Figure 10). The model approximates the large surface diurnal temperature variations and the dampening of these variations within the top 0.015 m. The model also provides a good representation of the magnitude and diurnal variation in peat temperatures through the profile (Figure 10).
6.5. Response of Peat Temperatures to Rainfall
 Because the near-surface moisture content and surface resistance provide important controls on the thermal behavior of the burned hollow, one might expect an alteration to this behavior associated with rainfall events - wetting the surface peat and increasing evaporation. During the study period, there was one substantial rainfall event of 23.6 mm on d.o.y. 214 (Figure 10). However, the impact of this rainfall on the thermal behavior of the peatland was relatively small. At a depth of 0.4 m, the modified 1-D HIP model provides a good approximation of peat temperatures up to the rainfall event, after which the model underestimates peat temperature by ∼0.7°C. Such a transition would be expected with an increase in the near-surface moisture content, increasing energy transfer into the peat profile. The impact of the rainfall event on surface temperatures is less clear (Figure 11). During the rainfall event, surface temperature are substantially dampened associated with the reduced solar radiation. The day after the rainfall event, the peat temperatures more closely approximate the simulated values. The difference between modeled and measured temperatures then increases slightly in the subsequent days (Figure 11). However, this impact is unclear.
 So why is the impact of the rainfall event so small? Within the BC06 hollow, changes in the measured near-surface moisture content associated with the 23.6 mm rainfall event are relatively small; 0.7% at a depth of 0.05 m. This therefore suggests that most rainfall is propagated directly down through the peat profile producing the observed 0.14 m increase in the water table position. If the near-surface peat is unable to respond rapidly to such wetting, this poses important questions to the response of the near-surface to extreme drought early within the growing season and even the potential hydrophobicity of this near-surface layer.
 We have highlighted a previously undocumented negative feedback to disturbance within a northern peatland whereby hydrological disconnection of the near surface restricts water loss, maintaining saturated conditions through much of the peat profile. While this substantially increases surface peat temperatures, subsurface temperatures remain cool due to the insulating effect of the dry near surface. This negative feedback therefore maintains the cool, saturated conditions necessary for carbon storage. Although surface temperature measurements are limited in their spatial extent, the observation of increased surface temperatures within multiple burnt hollows strongly suggest that this disconnection is widespread within the burnt peatland. However, within hummocks, the thermal functioning of these micro-features is not substantially altered from its unburned form and the peat surface continues to evaporate freely.
 There is a clear need to improve our understanding of the hydrology of the surface peat to account for this essential hydrological feedback within both burned and unburned peatland ecosystems. Specifically, we suggest there is an important need to: (i) quantify the magnitude of the hydrological stress required to potentially transfer the ecosystems to a new hydrological state, (ii) determine whether this hydrological stress differs between disturbed and undisturbed peatlands, and (iii) identify the ability of a peatland to recover to standard hydrological functioning post-wildfire. While it is clear that this understanding of near-surface unsaturated moisture dynamics is necessary to accurately simulate the thermal response of these peatland ecosystems to fire, the HIP models do provide a significant advance over current modeling simulations. They are able to incorporate the impact of the surface microtopography on burnt peat temperatures and approximate diurnal temperature variations. The latter of which is essential to account for the strong non linearity between peatland carbon dynamics and temperatures [Dunfield et al., 1993]. Moreover, the model simulations also suggest that peat temperatures provide an exciting tool in understanding and monitoring such small-scale hydrological dynamics, which can be difficult to measure directly using traditional measurement methods.
 This research was funded by a Natural Science and Engineering Council (NSERC) of Canada Discovery Grant and Discovery Accelerator Supplement to JMW. We thank Steve Baisley and James Sherwood for assistance. We also thank Drs. Paul Morris, Merritt Turetsky, Brian Benscoter and Mike Wotton for helpful discussions with this research and the Editor, Associate Editor and two anonymous reviewers for their comments on a previous version of this manuscript.