By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 We measured soil CO2 concentration at half-hour intervals with infrared gas analyzers buried in soil at four depths throughout the snow cover season extending from early December to early April in a deciduous temperate forest. We evaluated soil CO2 efflux or total soil respiration, topsoil (the A-horizon) respiration, and subsoil (the C-horizon) respiration using a modified flux gradient method. Thereby, we investigated seasonal and diurnal variations in these soil respirations under snowpack. Soil CO2 concentration and soil respiration changed dynamically under the condition of constant soil temperature. Topsoil respiration decreased rapidly in late autumn and relatively constant until mid-winter, whereas it increased in late winter when snowmelt progressed. On the other hand, subsoil respiration decreased gradually until mid-winter and increased slightly in late winter. Both topsoil and subsoil respirations showed similar diurnal variations with a peak in early or mid-afternoon, respectively, independently of soil temperature. These seasonal and diurnal variations in soil respiration were inferred to result from the supply of labile carbon compounds, which were respiratory substrates for microorganisms, into soil from litter with meltwater. The seasonal sum of topsoil, subsoil, and total respirations for the snow cover period of 4 months were 21.4, 48.0, and 69.4 gC m−2, respectively; these accounted for 3.0, 27.3, and 7.7% of the annual sum, respectively. The ratio of topsoil respiration to total soil respiration was 0.31 on average during the snow period, which was considerably lower than that in the summer season.
 Soil temperature remains relatively high under snowpack even in midwinter because the snowpack serves as an effective heat insulator for soil from the cold atmosphere. Although soil surface temperature is almost constant around the freezing point throughout the snow cover season, soil CO2 efflux showed seasonal variation with the minimum in early winter and the maximum at the onset of snowmelt in alpine and subalpine areas [Sommerfeld et al., 1996; Mast et al., 1998], but no seasonal and diurnal variations in a temperate forest [Mariko et al., 2000]. Abrupt increase in soil CO2 efflux following the onset of snowmelt was also measured in a Siberian forest [Shibistova et al., 2002]. However, these temporal variations in soil CO2 efflux resulted from intermittent measurements at long intervals, such as once a month, or short-term measurements. Similarly, wintertime soil CO2 effluxes have been inferred based on few measurements. Many studies have applied the flux gradient method using diffusion coefficients and CO2 concentration gradients of snowpack to measure soil CO2 efflux [Sommerfeld et al., 1993, 1996; Brooks et al., 1997; Mast et al., 1998; Fahnestock et al., 1999; van Bochove et al., 2000; Wickland et al., 2001]. However, gaseous diffusion coefficients of snowpack, which are necessary for the flux gradient method using diffusion theory, are difficult to determine [van Bochove et al., 2000]. The effects of turbulence-driven pressure pumping on CO2 flux through snowpack may not be negligible [Massman et al., 1997]. Alternatively, the chamber method has been applied to snow surfaces on the presumption that CO2 storage change, production, and consumption are negligible in snowpack [Oechel et al., 1997; Mariko et al., 2000; Kurganova et al., 2003]. Nevertheless, this method is extremely sensitive to chamber installation on snow surfaces because of the effect of horizontal convective airflow in the upper snowpack [Mast et al., 1998]. Therefore knowledge about soil CO2 efflux or soil respiration remains incomplete for the snow cover season because of qualitative and quantitative insufficiency of field measurements.
 For better understanding soil CO2 efflux and soil carbon cycles under snowpack, we measured soil CO2 concentration at half-hour intervals with infrared gas analyzers buried at four depths in soil throughout the snow season between December 2000 and April 2001 in a deciduous temperate forest. We evaluated soil CO2 efflux half-hourly from the soil CO2 profile using a modified flux gradient method [Hirano et al., 2003]. This method is independent of the physical properties of snowpack, which are usually unsteady and difficult to estimate. Moreover, we partitioned soil CO2 efflux or total soil respiration into topsoil respiration and subsoil respiration. Using a large number of half-hourly field measurements, we investigated diurnal and seasonal variations in these respirations, and the dependence of these respirations on soil temperature and moisture. The soil respiration shown in this paper, which was measured at one point, is insufficient as the representative value of the forest because of large spatial variation in soil respiration [Hanson et al., 1993; Rayment and Jarvis, 2000; Law et al., 2001]. Multipoint measurement is necessary to provide a statistical sampling. However, detailed information on temporal variation in soil respiration under snowpack, which was derived from the field measurement with high time resolution conducted throughout the snow season, has academic contributions.
2. Study Site
 The study was conducted in a temperate deciduous forest in the Tomakomai Experimental Forest, Hokkaido University. The forest is located at 42.4°N, 141.4°E near Tomakomai City in Hokkaido, northern Japan. It is about 8 km from the Pacific Ocean; its elevation is about 70 m above sea level. Dominant tree species are oak (Quercus mongolica), magnolia (Magnolia obovata), Japanese elm (Ulmus davidiana), maple (Acer mono), and heartleaf hornbeam (Carpinus cordata). Young oak and maple trees and ferns (Dryopteris crassirhizoma) grow sparsely, forming the understory. The annual mean air temperature is 6.5°C; the monthly means are below 0°C from December through March. Annual precipitation is about 1200 mm; 70% falls in the summer season. Soil is classified as a volcanogenous regosol with high water permeability and poor nutrients. Under 0.01- to 0.02-m thick litter, the A-horizon and C-horizon are stratified in order; the B-horizon is lacking. The A-horizon thickness is 0.15–0.20 m. The field experiment was conducted at a flat place with an average stand density. Snow depth in the experimental period was 0.4, 0.7, 0.35, and 0.35 m on 25 December, 16 February, 22 March, and 30 March, respectively. Soil was not frozen on those days.
 CO2 concentration of soil air had been measured since May 2000 directly with a small infrared gas analyzer (IRGA) (GMD20; Vaisala, Finland) buried at depths of 0, −0.02, −0.13, and −0.17 m in the A-horizon, respectively, at one point (Figure 1) [Hirano et al., 2003]. Depths of 0 m and −0.17 m denote the boundaries between the A-horizon and the litter layer, and the A- and C-horizons, respectively. The IRGA comprises a small plastic box containing electric circuitry and a plastic tube projecting from the center of the box. The plastic tube contains an infrared source, optical filter, and detector; it was 155 mm long and 15 mm in diameter. A 50-mm-long and 4-mm-wide slit is cut on the tube and covered with a membrane; CO2 exchange between the inside and outside of the tube occurs through this membrane by diffusion.
 The measurement was conducted at one point on the forest floor with no replicates. The place where the sensors were installed was chosen for reasons of flat ground, the midpoint of surrounding trees, and the mean thickness of litter and the A horizon. A pit was carefully dug down about 0.35 m. On a soil profile shaped in the pit, horizontal holes were made at depths of −0.02, −0.13, and −0.17 m. A 170-mm-long PVC tube, which was plugged at one end, was inserted into the horizontal hole to prevent the membrane from contacting soil and to protect the horizontal hole. The PVC tube had a slit of equal size to that of the IRGA. The IRGA tube, or sample cell, was inserted into the PVC tube set in advance. Both slits were directed downward together for protection against rainwater penetration. After installation, the pit was filled back with extreme care to restore the soil layer structure. An IRGA was also placed on the soil surface at a depth of 0 m. The pit was dug again 3 times until the end of May 2001 for replacement of a broken IRGA. The pit digging and filling necessarily cut roots and disturb soil structure to some extent in spite of careful work. This disturbance influences soil respiration and soil CO2 distribution. However, the slit of IRGA for gas exchange was located 0.08–0.09 m from the pit wall, and the soil around the slit was undisturbed except for the insertion of the PVC tube. Moreover, the pit was usually filled up. Thus the effect of pit digging on soil CO2 measurement was probably small, although soil CO2 concentration was decreased by atmospheric mixture for half a day after the pit digging and filling [Hirano et al., 2003].
 IRGA measuring range was 0–2000 ppmv (= μmol mol−1) for 0 m, 0–5000 ppmv for −0.02 m, and 0–10000 ppmv for −0.13 and −0.17 m. IRGA resolution was 0.5% of the range, which corresponds to 10 ppmv for 0 m, 25 ppmv for −0.02 m, and 50 ppmv for −0.13 and −0.17 m. Although the IRGA measures volumetric CO2 density, IRGA output is volume concentration of CO2 expressed in ppmv. Conversion between measurements and output is made by a calibration curve obtained in a factory under a condition of approximately 1013 hPa and 25°C. Thus the output was corrected using the following equation with measured soil temperature (Ts, °C):
where Cs is corrected soil CO2 concentration (ppmv), Cu is IRGA output (ppmv), Bc is atmospheric pressure during factory calibration (hPa), Bs is atmospheric pressure in soil (hPa), and Tc is air temperature during factory calibration (°C). Bc and Tc were fixed at 1013 hPa and 25°C, respectively. Also, Bs was fixed at 1013 hPa.
 The IRGA was operated intermittently to prevent soil temperature rising by heat emitted from the IRGA. Because output signals from the IRGA became stable within 5 min after being switched on, the IRGA was operated for 7 min every 30 min. The output signal was measured every 5 s for the last 90 s of each period in this periodic 7-min operation; the mean was recorded with a datalogger (CR10X; Campbell Scientific, Inc.). The IRGAs were calibrated before and after the snow season by the method described by Hirano et al. . Calibration revealed that shifts in the span and zero of the IRGAs were typically within 1 and 2% of the range, respectively. In addition, soil temperature (Ts) and volumetric soil moisture (θ) were measured every 15 min with the same datalogger at depths of −0.02, −0.07, and −0.15 m with thermocouple thermometers and TDR sensors (CS615; Campbell Scientific, Inc.); Ts was also measured at 0 and −0.1 m. The TDR sensors were calibrated by the gravimetric method. In addition, hourly data of Ts at −0.5 and −1 m were used. Those data were measured at a meteorological station in the forest. For Ts and θ, subscripts denote depth.
 Assuming horizontal homogeneity, CO2 flux (F: μmol m−2 s−1) caused by diffusion is expressed as Fick's first law,
where D is the gaseous CO2 diffusion coefficient (m2 s−1), c is CO2 density (μmol m−3), and z is depth (m). In the same way as Hirano et al. , we calculated soil CO2 efflux from the soil surface (Fs) and CO2 flux from the C-horizon to A-horizon (FC) every 30 min from the profiles of soil CO2 concentration (Cs) measured at one point (Figure 1). We applied equation (2) to the upper surface and bottom of the A-horizon with a thickness of 0.17 m and considered CO2 storage change of the A-horizon for the calculation. Cs was converted to c using Ts on the supposition that Bs is constant at 1013 hPa. D was calculated from the following equation [Campbell, 1985].
Therein, D0 is the CO2 diffusion coefficient (m2 s−1) in the atmosphere at 1013 hPa and 273.2 K. Bs was fixed at 1013 hPa. Relative gaseous diffusion coefficients (D/D0) were determined from air-filled porosity (e) of soil using their linear relationship [Hirano et al., 2003], which was obtained with undisturbed soil cores by the diffusion chamber method [Currie, 1960; Osozawa, 1987]. The e was determined as residual of volume fraction of solid and water (θ). Soil was inferred to be unfrozen because soil temperature was higher than 0°C, even at −0.02 m, throughout the snow season (see Figure 3 in section 4.2). Therefore the relationship derived from unfrozen soil cores is available. Considering CO2 storage change, we determined the CO2 production rate of the A horizon (PA) as the difference between Fs and FC [Hirano et al., 2003]. Topsoil respiration, subsoil respiration, and total soil respiration are represented as PA, FC, and Fs, respectively. This method for flux evaluation is fundamentally independent of the physical properties and depth of snowpack. Moreover, the availability of this method was shown by Liang et al. , based on comparisons with other methods.
4.1. Dynamic Variations
 Soil CO2 concentration and soil respiration dynamically changed even under snowpack. Figure 2 shows half-hourly variations in Cs, Fs, FC, and PA for four successive days in January (left panels), February (middle), and March (right) 2001 along with that of soil moisture. A developed depression passed between 9 and 10 January. During those 2 days, Cs first decreased by a strong wind and then increased because of an unseasonable rainfall of 15 mm, which decreased the air-filled porosity of the soil and snowpack. Fine weather lasted between 14 and 17 February. Snow depth was about 0.7 m. In accordance with the diurnal pattern of wind speed, Cs decreased from the late morning until the late afternoon at all depths. Although Ts and θ were almost constant, PA and Fs showed a diurnal pattern with a peak early in the afternoon. However, no diurnal pattern was observed in FC. In late March, θ increased in the afternoon as a result of water supply from snowmelt. Fs and PA showed similar diurnal patterns with those in February. In addition, FC increased slightly in the afternoon.
4.2. Seasonal Variations
Figure 3 shows seasonal variations in daily means of Cs, Fs, FC, PA, and the ratio of PA to Fs (PA/Fs) between 1 November 2000 and 30 April 2001 along with those in Ts,θ, and the gaseous CO2 diffusion coefficient of soil. From Ts and precipitation data, we inferred that snow cover began on 5 December and lasted until 6 April; snow cover continued for four months (123 days).
Cs always increased concomitant with soil depth. It gradually decreased from November to early December with a decrease in Ts. On the contrary, Cs began to increase immediately after the onset of snow cover. Although Ts decreased slightly or was almost constant, Cs continued to increase up to 1409, 1488, 2303, and 2498 ppmv in mid-March at 0.00, −0.02, −0.13, and −0.17 m, respectively, on a daily-mean basis. These peak concentrations at −0.02, −0.13, and −0.17 m were about half the summer values [Hirano et al., 2003]. During late March and early April, Cs decreased concomitant with snow depth because of snowmelt progress. No rapid decrease was observed in Cs after the end of snowmelt because topsoil temperature increased rapidly.
PA decreased rapidly between early November and mid-December. It remained less than 0.2 μmol m−2 s−1 on a daily-mean basis until late February, whereupon it increased gradually up to 0.5 μmol m−2 s−1 during March and early April. PA increased rapidly with increasing Ts immediately after the end of snowmelt. FC decreased from 0.7 to 0.4 μmol m−2 s−1 in November. Its slight decrease continued during December and January. It was almost constant in February and increased gradually until the end of snowmelt. The sum of PA and FC, which is Fs, decreased drastically from early November to early December. It was almost constant around 0.5 μmol m−2 s−1 under snowpack between mid-December and late February. Fs began to increase in early March and showed a rapid rise after the end of snowmelt in the same manner as PA. PA, Fs, and FC in January and February were less than 4, 8, and 40%, respectively, in comparison with their peak values in summer [Hirano et al., 2003]. The magnitudes of seasonal variations in PA and FC are different. Therefore, PA/Fs decreased from 0.7 to 0.2 between early November and January. It increased gradually during February and March, and exceeded 0.7 after the end of snowmelt.
Ts never dropped below 0°C during the snow cover season. Temperature decrease through the snow season was 0.0, 0.6, 0.9, 1.2, and 5.5°C at −0.02, −0.07, −0.10, −0.15, and −1 m, respectively, for 4 months. This decrease indicates that the temperature of the A-horizon was almost constant under snowpack. On the other hand, θ increased on three occasions between late December and early January because of unseasonable rain. The θ decreased gradually between mid-January and late February; it increased again in March because of water supply from snowmelt through the litter layer. We infer that soil remained unfrozen during the winter because no sudden drop in θ was observed in early winter; also, Ts was higher than 0°C.
4.3. Diurnal Variations
Figure 4 shows monthly-mean diurnal variations in Cs, Fs, PA, FC, Ts, air temperature, and wind speed in February when climate and snowpack conditions were relatively steady. Diurnal variations in Cs, PA, and Fs became clear, as in Figure 2. Cs decreased gradually from the early morning and reached minima in the late afternoon or in the evening. Minimum Cs values occurred, slightly lagged, with depth below −0.02 m. In addition, the diurnal range of Cs decreased with depth, which was 75, 56, 50, and 43 ppmv at 0, −0.02, −0.13, and −0.17 m, respectively. This diurnal pattern was caused by the change of convective airflow through snowpack [Massman et al., 1997]. In the daytime, wind speed increased on an average, and relatively strong wind produced convective airflow and induced more fresh air penetration into snowpack, which decreased the CO2 concentration in snowpack. This process engendered Cs decrease after increasing wind speed. Temperature dropped below 0°C at the soil surface in the morning because fresh air around −10°C penetrated into the snowpack. However, no temperature drop was observed in topsoil even at −0.02 m. This fact indicates that the dominant process of mass and energy transfers in soil was not convective airflow, but diffusion and conduction under snowpack. Therefore the direct effect of wind on the transfer process was probably negligible in soil. Because the daytime depletion in Cs decreased and lagged with depth, the vertical gradient of Cs peaked in the afternoon. This diurnal variation in the Cs gradient mainly caused the diurnal variations in Fs and FC with a peak in the afternoon. In addition, PA showed a similar diurnal variation; PA was calculated as the sum of CO2 storage change in the A horizon and the difference between Fs and FC. The diurnal patterns of Fs and PA are similar to those observed in the summer season, which followed Ts [Hirano et al., 2003].
 Mean diurnal variations in PA, Fs, and FC were calculated for each month in the snow cover season (Figure 5). PA had a peak early in the afternoon. Its diurnal pattern was more obvious in February and March than in December and January. The diurnal range extended from 0.09 μmol m−2 s−1 in December and January to 0.16 μmol m−2 s−1 in February and March. FC also showed diurnal patterns within the range between 0.35 and 0.41 μmol m−2 s−1 in December, January, and March. FC reached a peak in the mid- or late afternoon, 2 or 3 hours after PA. Because the diurnal range was much larger in PA than FC, Fs showed a very similar pattern with that in PA, which had a peak early in the afternoon. Its diurnal range increased from 0.10 to 0.20 μmol m−2 s−1 between December and March.
4.4. Dependence on Soil Temperature and Moisture
Figure 6 shows relationships between PA and Ts at the middle of the A-horizon (−0.07 m), and FC and Ts at −1 m on a daily-mean basis. PA decreased exponentially as Ts,−0.07 decreased with Q10 of 16.9 (r2 = 0.94) between early November and early December before the snow cover season. However, under snowpack, PA fluctuated between 0.01 and 0.5 μmol m−2 s−1 in spite of constant Ts. Moreover, PA increased up to 0.8 μmol m−2 s−1 immediately after the end of snowmelt, before Ts increased. During mid- and late April, PA increased exponentially as Ts,−0.07 increased with Q10 of 1.9 (r2 = 0.80). Thereby the temperature sensitivity of PA was quite different before and after the snow season: The two curves crossed around 9°C. Although a similar relationship was obtained even using Ts at −0.02, −0.10, and −0.15 m, use of Ts,−0.07 showed the highest level of significance. On the other hand, FC decreased exponentially as Ts,−1 decreased with Q10 of 2.0 (r2 = 0.81) until the end of February regardless of snow cover conditions. However, a negative relationship was obtained after March, when Ts,−1 was lower than 3°C.
PA is plotted against θ−0.02 during the snow cover season (Figure 7). Although the range of θ−0.02 was small between 0.39 and 0.45 m3 m−3, a significant positive linearity was found (r2 = 0.48). The r2 of the linear regression was larger for θ−0.02 than for either θ0.07 or θ−0.15. This relationship indicates that θ−0.02 accounted for 48% PA fluctuation under snowpack.
5.1. Seasonal Variations
 Topsoil respiration (PA) decreased rapidly in late autumn until snow covered the soil surface. Under snowpack, PA was almost constant until late February, whereas it increased gradually in March when snowmelt progressed (Figure 3). After the soil surface was exposed again to the atmosphere, PA increase was accelerated by rapidly increasing soil temperature (Ts) in April. Although PA changed exponentially with Ts both before and after the snow cover period, its magnitude was larger after the snow period when Ts at −0.07 m was lower than 9°C (Figure 6). On the other hand, subsoil respiration (FC) decreased gradually from early November to late February with decrease in subsoil temperature (Ts,−1) and showed no increase in April (Figure 3). Total soil respiration or soil CO2 efflux (Fs), which is the sum of PA and FC, changed concomitant with PA during the snow season because the seasonal range of FC was considerably smaller than that of PA. The means and standard deviations for the snow season of 123 days were 0.17 ± 0.12, 0.38 ± 0.04, and 0.55 ± 0.14 μmol m−2 s−1 on a daily-mean basis for PA, FC, and Fs, respectively. The mean soil CO2 efflux of 0.55 μmol m−2 s−1, which was obtained from single-point measurement, is compatible with measurements during the snow cover season in other ecosystems (Table 1). They were 0.1–0.3 μmol m−2 s−1 for a boreal forest in Canada with completely frozen soil [Winston et al., 1997], 0.16–0.85 μmol m−2 s−1 for alpine and subalpine ecosystems in Wyoming [Sommerfeld et al., 1996], 0.17–0.49 μmol m−2 s−1 for a subalpine wetland in Colorado [Mast et al., 1998], 0.17–0.31 μmol m−2 s−1 for a temperate deciduous forest in Japan [Mariko et al., 2000], 0.15–0.25 μmol m−2 s−1 for a conifer forest in central Siberia [Shibistova et al., 2002], 1.1 and 0.89 μmol m−2 s−1 for an agricultural field and a deciduous forest in southeastern Canada, respectively [van Bochove et al., 2000], and 0.2–0.3 μmol m−2 s−1 for some agricultural fields and forests in southeastern Canada [Risk et al., 2002]. Except for the Canadian boreal forest, soil temperature was around 0°C, as in this study. In this study, FC was more than twice as large as PA on average. Therefore, PA only accounted for less than a third of Fs. This ratio of PA to Fs contrasts with that during the summer, which was around 0.9 [Hirano et al., 2003]. Risk et al.  also reported that subsoil respiration became more important during winter than summer, but soil respiration of very shallow topsoil dominated soil CO2 efflux even under snowpack in Canadian agricultural fields and forests. The seasonal pattern of the relative importance of FC is explained partly by the fact that the vertical profile of Ts; Ts was higher during summer and lower during winter in topsoil than subsoil.
Table 1. Comparison of Soil Respiration Under Snowpack Among Seasonally Snow-Covered Ecosystems
 During the snow cover season, PA changed between 0.01 and 0.5 μmol m−2 s−1 on a daily-mean basis. The PA variation was not attributed to Ts variation because topsoil temperature was constant (Figure 6). In contrast, a significant positive relationship between PA and volumetric soil moisture (θ) was found, and θ at −0.02 m accounted for 48% of the PA variation under snowpack (Figure 7). The relationship shows that the change of θ−0.02 from 0.39 to 0.45 m3 m−3 increased PA from 0.06 to 0.35 μmol m−2 s−1. However, during the snow season, Ts was low (between 0 and 2°C) and soil was relatively wet in the A-horizon. The response of soil respiration to soil moisture was weak at temperatures below 5°C [Schlentner and van Cleve, 1985]. For that reason, it is difficult to directly relate PA to θ; instead, we discuss other explanations for PA variation. The value of θ increased by the supply of liquid water from snowmelt through the litter layer located between the A-horizon and snowpack because soil was unfrozen. The water carries labile carbon compounds into the A-horizon from the litter layer decomposed during winter through the cycles of freezing and thawing events [Moore, 1983; Taylor and Jones, 1989; Brooks et al., 1997]. During the snow period of 123 days, 61 days had the daily maximum Ts above 0.5°C and the daily minimum Ts below −0.5°C at 0 m. This fact suggests that the freezing and thawing frequently occurred at least in a part of the litter layer. The labile carbon compounds were fed to microorganisms. Consequently, they stimulated heterotrophic respiration in the A-horizon [Brooks et al., 1997], engendering PA increase following θ increase. Increase in soil CO2 efflux during snowmelt periods in late winter or early spring was reported at several sites [Sommerfeld et al., 1996; Winston et al., 1997; Mast et al., 1998; Fahnestock et al., 1999]; it was suggested that the increase in soil CO2 efflux was attributable to increasing soil moisture [Mast et al., 1998]. Moreover, Shibistova et al.  measured large increase in soil CO2 efflux immediately after snowmelt and related it to increased heterotrophic respiration because of enhanced substrate availability as a consequence of repeated freezing and thawing. Similarly, topsoil respiration (PA) increased immediately after snowmelt in this study, even before Ts largely increased (Figures 3 and 6). Daily mean Ts at −0.07 and −0.02 m increased only by 0.3°–0.4°C during 2 days after the end of snowmelt, whereas daily maximum Ts increased by 1°–2°C. The increase in daily maximum Ts possibly engendered the sudden increase in topsoil respiration. In addition, topsoil respiration was larger in early spring than late autumn under the same condition of Ts (Figure 6). Difference in labile carbon availability probably caused this discrepancy in topsoil respiration before and after the snow period. Labile carbon compounds frequently entered topsoil with meltwater from litter, which continued to decompose under snowpack [Moore, 1983; Taylor and Jones, 1989]. The labile carbon would be stored in topsoil throughout the snow season, especially during the snowmelt period in late winter, because soil respiration was low under snowpack. This fact suggests that labile carbon availability in the A-horizon for heterotrophic respiration was higher after the snow season than before. However, the stored substrate was insufficient to support relatively high PA over a long period; therefore, the discrepancy in PA vanished until the end of April. On the other hand, subsoil respiration (FC) increased in March, but Ts decreased at −1 m (Figure 6). The FC increase accelerated in response to Ts drop at mid-March (Figure 3), which shows that cold snow water reached to −1 m with labile carbon compounds. Therefore the increase in subsoil respiration was probably caused by the supply of respiratory substrates. Moreover, lack of response of FC to increasing Ts,−1 after the snowmelt in April may result from the shortage of substrates for microbiological respiration.
5.2. Diurnal Variations
 Topsoil respiration (PA) showed a clear diurnal variation with a peak early in the afternoon under snowpack (Figures 2, 4, and 5). Its diurnal range was more than 60% of its daily mean for each month (Figure 5). Similarly, Ts at −0.02 m showed a slight diurnal variation having a peak in the afternoon with a diurnal range of 0.1–0.15°C on a monthly-mean basis (Figure 4). However, no diurnal variation in Ts was observed at −0.07, −0.10, or −0.15 m. In addition, θ had a tendency to increase in the afternoon (Figure 2). Its diurnal range was 0.01 to 0.03 m3 m−3 at −0.02 m on a monthly-mean basis. θ gradually increased from January through March. The diurnal pattern of θ suggests that respiratory substrates were supplied into topsoil with the inflow of meltwater through the litter layer in the afternoon. Therefore the daytime increase in θ probably increased PA through enhanced heterotrophic respiration by the substrate supply because the amplitude of diurnal Ts variation was very small. Moreover, the diurnal range of PA extended from January to March (Figure 5), whereas that of Ts,−0.02 had no change during the period. Regarding root respiration, its diurnal pattern under snowpack is unknown. However, root respiration may have some contribution to the diurnal pattern of PA because diurnal variations in PA and stem temperature were similar each other; stem temperature may have stimulated root respiration. Subsoil respiration (FC) also showed diurnal variation with a peak in the mid-afternoon under snowpack (Figure 5). Its diurnal range was less than 20% of the diurnal range of PA. This diurnal pattern of FC was likely caused by infiltration of meltwater into subsoil, similarly to that of PA. The time lag of peak occurrences between diurnal patterns of PA and FC indicates that labile carbon compounds traveled downward from topsoil to subsoil. Finally, total soil respiration or soil CO2 efflux (Fs) had a diurnal range of 20–30% of its daily mean (Figure 5). Therefore soil respiration was not steady under snowpack even on an intraday basis. However, such a diurnal pattern of soil respiration was not measured in previous studies [e.g., Mariko et al., 2000]; also, the diurnal variation has been ignored. It is notable that the diurnal variation was not always negligible in the snow cover season, especially in the snowmelt season.
5.3. Cumulative Soil Respiration for the Snow Cover Season
Figure 8 shows cumulative changes in PA, FC, and Fs from 1 November 2000 to 30 April 2001. During that period, cumulative FC increased linearly, whereas the cumulative curve of PA had two inflection points in early November and early April when topsoil temperature changed rapidly (Figure 3). The sums of PA, FC, and Fs for the snow cover period between 5 December and 6 April of 123 days were 21.4, 48.0, and 69.4 gC m−2, respectively. Soil respiration for the snow season accounted for 3.0, 27.3, and 7.7% of the annual sum for PA, FC, and Fs, respectively; the annual sums were 722, 176, and 898 gC m−2 between 28 March 2000 and 27 March 2001 for PA, FC, and Fs, respectively [Hirano et al., 2003]. This fact indicates that the importance of subsoil respiration is greatly enhanced in the snow season to investigate the carbon balance of this deciduous forest. Although the annual Fs of 898 gC m−2 is compatible with results from EUROFLUX sites [Janssens et al., 2001], it is approximately twice as large as an annual sum for this forest [Tanaka et al., 2001], which was estimated using an exponential relationship between soil temperature and soil respiration measured by the static chamber method in the growing season. The difference between two annual Fs values is probably due to single-point measurement in this study and underestimation by the static chamber method [Nay et al., 1994; Rayment and Jarvis, 2000]. The seasonal sum of soil CO2 efflux (Fs) under snowpack is almost within the range of reported values from other ecosystems. Winter soil CO2 efflux was 70 gC m−2 in Alaskan moist tundra for 7 months [Oechel et al., 1997] and 44–55 gC m−2 in a Canadian boreal forest for 6 months [Winston et al., 1997]. It was between 35 and 174 gC m−2 in alpine areas for 235 days [Sommerfeld et al., 1993]. For deciduous forests, it was 66–103 gC m−2 in southeastern Canada for 4–5 months [van Bochove et al., 2000] and less than 100 gC m−2, which was less than 15% of the annual sum, in Japan [Mariko et al., 2000].
 We thank Yumiko Tanaka, Honghyun Kim, and the staff of Tomakomai Experimental Forest, Hokkaido University, for assisting in field experiments, and Nobuyoshi Ishikawa for his continuing support. This study was supported by Grants-in-Aid for Scientific Research (11490001 and 12878089) from the Japanese Ministry of Education, Culture, Sports, Science and Technology.