Effects of fire disturbance on soil respiration in the non-growing season in a Larix gmelinii forest in the Daxing'an Mountains, China

In boreal forests, fire is an important part of the ecosystem that greatly influences soil respiration, which in turn affects the carbon balance. Wildfire can have a significant effect on soil respiration and it depends on the fire severity and environmental factors (soil temperature and snow water equivalent) after fire disturbance. In this study, we quantified post-fire soil respiration during the non-growing season (from November to April) in a Larix gmelinii forest in Daxing'an Mountains of China. Soil respiration was measured in the snow-covered and snow-free conditions with varying degrees of natural burn severity forests. We found that soil respiration decreases as burn severity increases. The estimated annual C efflux also decreased with increased burn severity. Soil respiration during the non-growing season approximately accounted for 4%–5% of the annual C efflux in all site types. Soil temperature (at 5 cm depth) was the predominant determinant of non-growing season soil respiration change in this area. Soil temperature and snow water equivalent could explain 73%–79% of the soil respiration variability in winter snow-covering period (November to March). Mean spring freeze–thaw cycle (FTC) period (April) soil respiration contributed 63% of the non-growing season C efflux. Our finding is key for understanding and predicting the potential change in the response of boreal forest ecosystems to fire disturbance under future climate change.


Introduction
Soil respiration (Rs) is the second-largest carbon flux in most terrestrial ecosystems-the amount of CO 2 released by soil respiration is more than ten times that released by global fossil fuel combustion [1,2]. It contributes 20%-40% of the CO 2 input to the atmosphere [3]. Soil respiration is estimated to be 80-98 Pg CÁyr -1 [4]. Therefore, slight changes in soil carbon may influence the global carbon balance. Many studies focus on forest soil respiration during the growing season [5][6][7][8][9], and estimated the annual soil respiration by assuming the respiration flux near zero during the non-growing season [10]. However, recent studies have shown that a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 elevation in this area ranges from approximate 500 m to 800 m. The climate is the cold temperate continental monsoon. The average annual temperature is -3˚C, with an extreme minimum temperature of -48˚C. There are approximately 2500 annual sunshine hours, and the frost-free period is approximately 90 to 100 days. The annual total precipitation varies from 350 mm to 500 mm. Snow is composited of 10% to 20% of the annual total precipitation. The snowfall is mainly in December to March of the next year.
In April 2006, four forest fires were caused by lightning in Songling forest farm within the Nanweng River Forest Ecological Station. We selected three sites in each of the unburned (control), low burn severity, or high burn severity. Fire severity in the study area was classified according to the classification proposed by Keeley [57]. In the low severity area, fires burned 25% of the understory shrubs, bark char height was 1.8-2.4 m, and 20% tree mortality occurred. The high severity area experienced the complete consumption of understory shrubs and litter and duff layers, 2.5-5.5 m bark char height, and 85% tree mortality. As the high heterogeneity of soil respiration, we established three 20 m × 20 m replicate sites within each of the burn severity areas. To accomplish all measurements of soil respiration on the same day, and thus avoid the influence of day-to-day variation, replicated sites were 200 meters within each other. The vegetation compositions in the study sites were shown in Table 1. All sites were located at the southwest aspect with the 10-15˚(17-27%) of slope of and 463 m elevation. Within each severity area, nine permanent automatic data measurement systems were installed to record the soil temperature at five depth every 30 minutes during the period 2011-2012.

Ethics statement
The research complies with all laws of the country (China) in which it was approved by the National Natural Science Foundation of China (permit number: 31070544, 31470657). The authority responsible for a national park or other protected area of land or sea, the relevant regulatory body concerned with protection of wildlife. We state clearly that no specific permissions were required for these locations, because these locations are uncultivated land. We confirm that the field studies did not involve endangered or protected species.

Soil respiration measurement
A portable LI-8100-1032 and LI-8100 Automatic Measuring Systems (Li-Cor Inc., Lincoln, NE, USA) were used to measure soil respiration flux. In early May 2010, five PVC (polyvinylchloride) soil rings (inner diameter 19 cm, height 7 cm) were systematically installed in each 20 m × 20 m site (Fig 1) (denoted snow-covering Rs), a total of 15 soil respiration rings in each burn severity type. The top of the PVC ring was 2-3 cm above the litter surface, and the PVC ring remained in the same position throughout the measurement period. At the beginning of November 2011, three plastic sheds (Fig 2) were randomly set up in each site to simulate a snow-free condition. Nine snow-free soil rings (3 sheds × 3 replicates; Fig 1, denoted snow-free Rs) measured soil respiration in each burn severity under snow-free conditions. The sheds consisted of a transparent plastic membrane, is open to the environment at both ends to allow light transmittance and air flow. We maintained all sheds to ensure no snow underneath them throughout the measurement period. The sheds were 1.5 m high and the horizontal size of 2 m × 2 m. A PVC soil respiration ring was placed in the middle of each shed. One end of the soil respiration ring was sharpened and pressed into the soil. The soil respiration rings were remained in the same position during the entire study period. Eight soil respiration rings were used to measure soil respiration in each site, with five of them for snow-covering condition (Fig 1, snow-covering Rs) and three of them for snow-free sheds (Fig 1, snow-free Rs). Soil respiration was measured one time per month during the non-growing season from November 2011 to April 2012. Once snow started to thaw in early April, soil respiration rates were measured every 10 days for three measurements: early April (Apr-E), middle of April (Apr-M), and late April (Apr-L). The soil respiration measurement time was approximately 2 minutes for each soil respiration ring. Each measurement was completed between 9:00 AM and 11:00 AM. The snow-covering Rs and snow-free Rs were measured on consecutive days. Snow thickness and mass and soil temperature (at 5 cm depth) were measured near each soil respiration ring at the same time as the Rs measurement. Snow mass was determined using a PVC cylinder with an inner diameter of 19 cm. The PVC cylinder was inserted vertically into snow until it reached soil. Then a small shovel was inserted into the bottom edge of the cylinder. The snow in the cylinder was then emptied into a plastic bag and the snow mass was measured using an electronic scale (resolution 0.01 kg, measuring range 0-45 kg). Snow thickness was the vertical depth from snow surface to soil measured with a plastic ruler [58]. Soil temperature was measured using a JM624 portable digital thermometer (resolution 0.1˚C, measuring range -50 to 199˚C; JinMing Instrument Co, LTD, China). Fire disturbance on soil respiration in the non-growing season The year was divided into periods based on air temperature and soil temperature. Winter was defined as the snow-covering period, and the daily mean air temperature below 0˚C lasted at least 5 consecutive days. The freeze-thaw cycle (FTC) period in spring was defined as the period from the daily maximum air temperature above 0˚C (i.e., the start of snowmelt), to the daily minimum soil temperature (at 5 cm depth) above 0˚C (i.e., the end of soil freezing at the 5cm depth) [20]. The growing season was defined as the period from the end of the spring FTC period to the beginning of winter. The non-growing season was defined as winter days and the FTC period, the rest days is the growing season period ( Table 2).

Data analysis
The data were processed and analysed using SPSS19.0 (SPSS Institute, Inc., Chicago, IL, USA). Differences in variables between burnt and control sites were tested using ANOVA and  comparisons between means were performed with the least significant differences (LSD) test. All statistical analyses were performed with a significance level of 0.05. Snow water equivalent is a measurement of the amount of water contained in snow pack. It is considered as the depth of water that would theoretically result if the whole snow pack instantaneously melted [59]. The snow water equivalent is calculated by the following equations [60]: where M s is snow mass (g); V is volume (cm 3 ); ρ s is the snow density (gÁcm -3 ); d is the snow thickness (cm); and SWE is the snow water equivalent (mm). Because the snowfall mainly occurred in December of the previous year to March of the current year, and started to melt at the beginning of April. The snow water equivalent was only calculated during the main snowcovering period (Nov 2011-Mar 2012). Model fitting for soil respiration rate and soil temperature during the non-growing season was fitted by an exponential models. The goodness-of-fit of the models were determined by the coefficient of determination (R 2 ) and residual analyses. The regression model between soil respiration and soil temperature was shown as Eq 3 [61-63]: where Rs is the soil respiration (μmol CO 2 Ám -2 Ás -1 ), T is the soil temperature at 5cm (˚C), and α and β are regression coefficients. After consideration of the common model forms [3,64,65], exponential models were used for analyses of fire severity, soil temperature, snow water equivalent, and the interaction effects of soil temperature and snow water equivalent during the snow-covering period (Nov 2011-Mar 2012). Logarithmic transformation of Rs was required to achieve linearity and homoscedasticity. The regression model is shown as Eqs 4 and 5: where Ln(Rs) is logarithmic transformation of Rs that was applied to achieve linearity and homoscedasticity; T is soil temperature (˚C); SWE is snow water equivalent (mm), T × SWE is the interaction effect of T and SWE, and α, β, ε, ω are regression coefficients. A stepwise regression procedure was performed to remove insignificant terms (P = 0.05). The accumulated C efflux (g C m -2 yr -1 ) was estimated based on parameters shown in Eq 6 [66] and calculated for both the growing and non-growing seasons.
where 12 is the molecular weight of carbon and 1800 is a constant value (unit: second) based on the automatic data acquisition systems recording soil temperature every 30 minutes for one year. Q 10 is the temperature-sensitive coefficient representing the increase in a process as result of temperature increase at each 10˚C. We used Eqs 3 and 7 to calculate Q 10 [67]: where β is the regression coefficient calculated from Eq 3 and e is the exponential base.

Effects of fire disturbance on soil respiration and environmental factors
Rs increased from May to July, then decreased until March, and increased until the end of April in all three kinds of snow-covering sites (Fig 3a). Soil temperature was relatively high in July and August and relatively low in February (Fig 3b), which was consistent with the changes in Rs. The mean Rs in the growing season were 4.06, 3.22, and 2.92 μmol CO 2 Ám -2 Ás -1 for control, low, and high burn severity sites, respectively. The mean non-growing season Rs were 0.29, 0.23, and 0.13 μmol CO 2 Ám -2 Ás -1 for control, low, and high burn severity sites, respectively ( Table 3). The variations of soil respiration during the non-growing season (Nov to Apr-L) were similar for all kind of sites (Fig 3a). Rs in the control sites increased from 0.19 to 1.11 μmol CO 2 Ám -2 Ás -1 from early April to late April, a six-fold increase. In the low burn severity sites, Rs increased from 0.13 to 1.03 μmol CO 2 Ám -2 Ás -1 , an eight-fold increase, from early April to late April. In the high burn severity sites, Rs increased from 0.04 to 0.95 μmol CO 2 Ám -2 Ás -1 , an increase of 24 times from early April to late April. The average non-growing season Rs at the low and high burn severity sites were 79% and 45% of that in the control sites Rs respectively. Compared with control sites, the average non-growing season Rs for high burn severity sites was significantly lower than that in the control site (P<0.05), while that of the low burn severity sites was not significantly different from control sites (P>0.05). The mean non-growing season soil temperatures in the control, low, and high burn severity sites were -5.76, -5.71, and -8.53˚C for control, low, and high burn severity sites, respectively (Table 3). Temperatures tended to decline from August to February, rise until July, and then stabilize (Fig 3b). The minimum soil temperature occurred in February, when the soil temperatures for control, low, and high burn severity sites were -11.24, -10.54, and -15.4˚C, respectively. The maximum non-growing season soil temperatures occurred in the late April, and were 0.34, 1.16, and 1.98˚C in the control, low, and high burn severity sites, respectively.
The snow water equivalent (SWE) increased with time during winter, with averages across treatments 5.1-18.9 mm (Fig 4). The SWE of control and low burn severity sites was significantly higher than that of high burn severity sites during November and December (P<0.05), but there was no measurable difference in SWE among different kinds of sites during January to March (P>0.05). The SWE accumulations of the whole non-growing season were 72.72, 69.56, and 62.68 mm for control, low, and high burn severity sites, respectively. The SWE accumulation of control and low burn severity sites were 14% and 10% higher than that in the high burn severity sites, respectively.

Soil respiration in snow-free conditions
The mean non-growing season Rs for the snow-free treatment were 0.16, 0.14, and 0.16 μmol CO 2 Ám -2 Ás -1 for control, low, and high burn severity, respectively ( Table 3). The overall trends in soil respiration in the snow-free condition were similar to those in the snow-covering condition (Fig 5). Rs in different burn severity sites was significantly different (P = 0.05). The SWE significantly affected the Rs in the snow-covering and snow free treatment (P<0.05). The interactive effects of the burn severity and SWE also significantly affected Rs (P<0.05). The mean snow-free soil temperatures were -7.82, -7.91, and -9.76˚C for control, low, and high burn severity sites, respectively ( Table 3). The minimum snow-free soil temperatures all occurred in January or February, at -14.13, -15.07, and -18.37˚C for control, low, and high burn severity sites, respectively. The maximum snow-free soil temperatures of all sites occurred in late April, when the soil temperatures for control, low, and high burn severity sites were 1.07, 1.93, and 2.53˚C, respectively. Soil temperature in the snow-covering treatment was significantly higher than that in the snow-free treatment from November to March (P<0.05). The mean in the snow-covering Rs was twice that in the snow-free Rs during the period from November to March (Fig 6).   Fire disturbance on soil respiration in the non-growing season

Relationships between soil respiration and environmental factors
The exponential regressions with T as a single controlling factor of Rs were significant (P < 0.01) for all kinds of sites (control, low burn severity, and high burn severity sites), which could explain 66%-78% of the variation of the Rs in the snow-covering condition ( Table 4). The exponential regressions with SWE as a single explanatory variable of Rs was also significant in all sites (P<0.01). Regression models with SWE as a single controlling factor could explain 29%-58% of the variation of the Rs in the snow-covering condition ( Table 4). The models fitted with T and SWE explained 73%-79% of the variation of the Rs in different fire burn severity sites (Table 4). Temperature and snow water equivalent together improved the coefficients of regression models in control and low burn severity sites, while the exponential Fire disturbance on soil respiration in the non-growing season regressions with T as a single controlling factor could better explain the variation of the Rs in high burn severity sites (Table 4).

Temperature-sensitive coefficient (Q 10 ) and C efflux
The Q 10 in the growing and non-growing season snow-covering and non-growing season snow-free treatments were shown in Table 5. The non-growing season Q 10 in each treatment was significantly higher than that in the growing season (P<0.05). Burn severity in all treatment significantly affected the Q 10 of the Rs (P<0.05). In the growing season, the Q 10 of the Rs significantly increased and decreased after the low and high burn severity fire disturbances (P<0.05). Whether in snow-covering or snow-free condition, the non-growing season Q 10 shown the similar trend that the Q 10 significantly decreased with burn severity (P<0.05). Fire disturbance on soil respiration in the non-growing season Fire disturbance on soil respiration in the non-growing season

Fig 8. Regression of soil respiration (Rs) and soil temperature (T) fitted models of (a) control, (b) low burn severity, and (c) high burn severity in the non-growing season.
The soil C efflux during the growing season, non-growing season, and spring FTC period (Fig 9) was calculated based on the exponential relationship between the soil respiration and soil temperature shown in Figs 7 and 8, and Table 4. We estimated that non-growing season C efflux in the control sites was 37.23 g CÁm -2 Áyr -1 , 31.3 g CÁm -2 Áyr -1 in the low burn severity sites, and 28.24 g CÁm -2 Áyr -1 in the high burn severity sites, contributing 4%, 4%, and 5% of annual C efflux in the control (933.11 g CÁm -2 Áyr -1 ), the low burn severity (771.71 g CÁm -2 Áyr -1 ), and the high burn severity (562.70 g CÁm -2 Áyr -1 ) sites, respectively. Compared with control sites, the annual C efflux of the low burn severity sites was 16% lower in the non-growing season and 17% lower in the annual C efflux budget. Compared with the control sites, the annual C efflux of the high burn severity sites was 24% lower in the non-growing season and 40% lower in the annual C efflux budget (Fig 9). Meanwhile, spring FTC period C efflux was 23.41 g CÁm -2 Áyr -1 for control sites, 16.07 g CÁm -2 Áyr -1 for the low burn severity sites, and 11.03 g CÁm -2 Áyr -1 for the high burn severity sites, contributing 3%, 2%, and 2% of annual C efflux of the control sites, the low burn severity sites, and the high burn severity sites, respectively. ; and α, β, ε, and ω  Fire disturbance on soil respiration in the non-growing season

Discussion
Effect of fire disturbance on soil respiration during the non-growing season The mean non-growing season Rs in the Larix gmelinii forest was 0.29 ± 0.06 μmol CO 2 Ám -2 Ás -1 (Table 3). This result was consistent with Wang et al. [20], who found that the average winter Rs rate of seven forest ecosystems in northern China was 0.28 μmol CO 2 Ám -2 Ás -1 . Many studies have found that the Rs of forest ecosystems during the non-growing season ranged from 0.15 to 0.67 μmol CO 2 Ám -2 Ás -1 [12,13,16,68]. The non-growing season Rs varies with the duration of the snow-covering period, and is also affected by climate change [69]. We should be cautious when comparing non-growing season Rs among different ecosystems because the many definitions of the non-growing season will lead to great discrepancy among different studies [70]. Fire reduces Rs to a degree dependent on the fire severity and duration [71][72][73]. In the nongrowing season, high burn severity resulted in significantly decreased Rs, approximately 55% of that of the control sites. This result may be mainly due to the decline of autotrophic Rs in high burn severity sites. Richter et al. [74] found that Rs after fire disturbance was half that of unburnt areas in boreal forest across similar latitudes, largely because of the decline in autotrophic Rs. Severe fires have more significant impacts on autotrophic Rs than smaller fires because they cause more serious damage to plant roots. Fire restrains autotrophic Rs due to mortality of fine roots; this effect can be shrouded by the increase in heterotrophic Rs as a result of the fire in the growing season [75,76]. Several studies have shown that heterotrophic Rs accounts for approximately 50%-70% of the total Rs during the growing season [77][78][79]. Heterotrophic Rs is largely caused by soil microbial activity that is partly suppressed during the non-growing season, when the majority of total Rs was provided by autotrophic Rs [24, 80, Fire disturbance on soil respiration in the non-growing season 81]. Mikan et al. [82] reported that temperature is the predominant contributor to microbial respiration in arctic tundra soil, and low temperatures will suppress the heterotrophic Rs.
The soil C efflux during the non-growing season in our forest represented approximately 4%-5% of the annual C efflux, depending on burn severity (Fig 9). Our result was consistent with the previous studies in other forest ecosystems, which accounted for 2%-37% of annual total C efflux during the non-growing season [17][18][19][20]. The annual C efflux was lower in burned treatment areas than in the control, particularly when burning was high severity. This result indicates that the influence of fire disturbances on ecosystem annual soil C efflux in boreal forests should not be ignored.

Effect of temperature on soil respiration during non-growing season
During the growing season, the interaction between T and soil moisture content is the primary factor reported to restrict Rs [71,83,84]. However, the water in the soil is frozen and thus virtually ineffective for Rs during the non-growing season. Some studies have found that T is the only environmental factor that impacts Rs and have shown that Rs models using both T and soil moisture may overestimate the Rs for the non-growing season and annually [85,86].
Regression models show that T is the most important abiotic factor determining Rs in the non-growing season [29,87]. In this study, Rs in the non-growing season had a much closer relationship with T than it did in the growing season (Figs 7 and 8). Higher Rs is generally triggered by the higher temperature. Although T seems to be a better predictor than SWE, while SWE still is a significant factor affecting Rs. SWE could explain 29%-58% of the variation of Rs during the snow-covering period, which revealed the interaction of T and SWE. Fig 10  shown that with the increase of T, the SWE decreased. Fig 11 shown that the exponential model with interactions of T and SWE and the exponential model with T as a single controlling factor both could better estimate the actual measured Rs than the exponential model with SWE as a single controlling factor. Table 4 shown that the exponential model of the interaction of T and SWE were the better fit to explain the Rs in control and low severity burn sites, while the exponential model with T as a single controlling factor of Rs was the better fit to explain the Rs in high burn severity sites. There results suggested that the main controlling factor in the control and low burn severity sites was the interaction of T and SWE in snow-covering condition and the main controlling factor of high burn severity sites was still T.
The global annual Q 10 of soil respiration is 2.4 [3] and many studies have indicated that the Q 10 of Larix gmelinii forest in the northern China during the growing season ranged from 1.5 to 5.7 [88,89]. However, the Q 10 of soil respiration could be as high as 63-207 under the cold conditions [82]. Our value of the Q 10 during the non-growing season was 76.71 and decreased with and the burn severity. Soil respiration during the non-growing season is more sensitive to temperature than that during the growing season (i.e. the high Q 10 ). This may result from the high carbon availability in thawing soil from the snow-covered to snowmelt period [87,90]. According to recent studies, the Q 10 not only reflects the soil respiration sensitivity to temperature but also expresses the combined response to fluctuations in temperature, root biomass, moisture conditions, and substrate quality [83]. The variation of the Q 10 after fire disturbance may result from the effects of fires on root material because the high burn severity fires will destroy root structures, cause the loss of the labile fraction of soil organic carbon (SOC) into the atmosphere, and decrease the Q 10 [91,92].

Soil respiration change during spring FTC period
Mean Rs during the spring FTC period in this study accounted for 63% of the C efflux during the non-growing season and 3% of annual total C efflux. Wang et al. [20] performed a similar Fire disturbance on soil respiration in the non-growing season study at the Maoershan Forest Ecosystem Research Station (127˚40´E, 45˚24´N, 400 m a.s.l.), also in northeastern China. The much colder temperatures of our research area may be the main reason behind the lower C efflux in winter and the larger proportion of total non-growing season C efflux being produced during the spring FTC period than Wang's findings (39%). The changes of Rs during the spring FTC period are much more complex. Snow starts to melt in early April, the Rs in the snow-free condition is higher than that in the snow-covering condition. The rapid increase in air temperature led to an increase in soil surface T and moisture [13,93]. The increase in T greatly affects soil microbial activities, fine root growth, and the biochemical processes of soil [94]. In the early part of the FTC period (here the first 10 days of April), snow has not melt completely yet. A brief thawing results in several centimetre thick layer of ice and greatly increase the density of remaining snow. This permits greater loss of heat, and results in lower T and deeper freezing than succeeding snow cover [85].
Our findings suggest the change patterns of Rs during the spring FTC period and determine whether climate fluctuations several years after fire disturbance extend the spring FTC period and shortens the non-growing season of Daxing'an Mountains should be further explored. In addition, the mechanisms of fire disturbance on Rs and how fires change the biological (e.g., Fire disturbance on soil respiration in the non-growing season microbe, soil fauna, and plant root activity) and non-biological (e.g., soil temperature and moisture content) factors that relate to Rs in the Daxing'an Mountains also need to be studied. Our results provide a basic data for accurate understanding the effects of fire disturbance on boreal forest ecosystems under future climate change.
Supporting information S1 Data. Soil respiration data. (XLSX) 29. Monson  Fire disturbance on soil respiration in the non-growing season