Carbon dioxide evolution from snow-covered agricultural ecosystems in Finland

Hiilidioksidin vapautumista lumipeitteisesta ja paljaasta maasta mitattiin suorasti ja epasuorasti Suomen maaperasta. Mittaukset tehtiin turve-, hiekka- ja savimaasta. Suorissa mittauksissa kaytettiin suljetun kammion menetelmaa ja epasuorissa Fickin diffuusiomallia.Kaikkien em. maalajien talvilampotila vaihteli 2 cm:n syvyydessa Oja +1 °C valilla. Hiilidioksidipitoisuus lumikerroksen sisalla oli sita korkeampi mita paksumpi lumikerros oli. Lumen hiilidioksidipitoisuudesta mitattu keskimaarainen hiilidioksidivirtaus vaihteli 10-27 mg CO2 m-2 h-1 diffuusiomallilla ja 18-27 mg CO2m-2 h-1 suljetun kammion menetelmalla mitattuna. Taman tutkimuksen mukaan lumi on hyva maan lampoeriste, ja CO2:n tuotanto jatkuu maassa lampo-tilojen pysyessa lumipeitteen alla jaatymispisteen ylapuolella. Maaperasta vapautuu ilmakehaan hiilidioksidia lumipeitteen lapi. Hiilidioksidin muodostumista talvella ja kesalla kuvaavan suhdeluvun arvioitiin olevan noin 4 %. Taten merkittava osa vuotuisista hiilidioksidipaastoista syntyy talvella.

man 1990). Thus the contribution of CO, from alpine or arctic regions in winter has not been considered important in calculations of global carbon balances. Some reports have, nevertheless, shown that C0 2 concentrations are high at the base of snowpacks in arctic and temperate regions (Kelley et al. 1968, Solomon andCerling 1987). In addition, there is some evidence that microorganisms in soils beneath the snow continue to respire at temperatures close to O°C (Taylor andParkinson 1988, Sommerfeld et al. 1991). It is currently thought that the soils under alpine and sub-alpine snowpacks evolve CO, to the atmosphere throughout the snow-covered period (Sommerfeld et al. 1993, Mariko et al. 1994).
Snow can cover between 44% and 53% of the land area of the northern hemisphere, and in alpine and sub-alpine regions it may be several meters deep for more than half the year (Barry 1992). Sommerfeld et al. (1993) reported that C0 2 fluxes through snowpacks in these regions ranged from 31 to 84 mg C0 2 nv 2 h ', as calculated from Pick's law. If snow-covered soils show some C0 2 evolution in winter, the C0 2 released can strongly influence the annual global carbon budget. This flux needs to be measured urgently for a more complete understanding of the global carbon balance. However, not only C0 2 fluxes from the snow surface but also C0 2 concentrations in the snowpack have rarely been measured (Coyne andKelley 1974, Solomon andCerling 1987). In particular, no researcher has measured the fluxes directly with chamber methods.
The aim of the present study is to measure directly or indirectly the CO, released from the snow surface in winter and the soil surface in summer in three different soil types in agricultural ecosystems in Finland. The closed chamber (CC) method (Bekku et al. 1995 was used for the direct measurement. For methodological comparison, the C0 2 fluxes were measured with Pick's diffusion model (DM), which so far has been used as an indirect method (Roiston 1986).

Study site
The investigations were carried out in experimental fields of the Agricultural Research Centre of Finland (ARCF) in July 1992 and March 1994. The fields were located at Jokioinen (60°9'N, 23°0'E, 104 m asl) in southern Finland. Meteorological data at the ARCF site showed that annual precipitation averaged 581 mm, the mean annual temperature was 3.9°C and the average snow depth in March was 39 cm during the period from 1961 to 1990. The monthly mean temperature in July 1992 was 16.0°C and in March 1994 -3.2°C. The warmth index at this site was 38.1 degree-months, indicating that the site is in the subarctic climate range. Three barley fields consisting of three different soil types were investigated. Classified on the Finnish and FAO system, the soils were peat (histosol), sand (cambisol) and heavy clay (cambisol). Table 1 lists the cultivation methods for each soil type in the agricultural ecosystems. Spring barley was cultivated in one field from the middle of May to the end of August, after which the field was fallowed from September to the following April.
After the measurements of CO, flux in winter, snow pits were dug beneath five flux-measurement points in each soil type to assess snow depth and porosity (Sommerfeld et al. 1993). At the same time, air temperatures were measured 50 cm above the snow or at the soil surface, snow temperatures at 10-cm depth and/or soil temperatures at 2-cm depth in winter and summer. The temperatures were measured with a digital-reading, spike stem, dual metallic thermometer.

Measurement of C0
2 flux using the CC method The evolution of CO, from the soil surface in summer and from the snow surface in winter was 422 AGRICULTURAL AND FOOD SCIENCE IN FINLAND Vol. 5 (1996) The flux was measured at five points in each soil type during a diurnal period lasting from 10:00 to 14:00. The C0 2 flux was determined directly in situ with the CC method, because this method can be used at sites where electricity or special equipment is not available. In this method, a closed chamber is placed over the soil or snow surface and the increase in the concentration of C0 2 within the chamber is measured as a function of time. The CO, flux is calculated from Eq. (1).
where Fis the C0 2 flux (mg C0 2 nr 2 h 1), V is the volume of air within the chamber (m 3 ), A is the area of the soil or snow within the chamber (m 2 ), and AC/At is the rate of change in the C0 2 concentration in the air within the chamber (mg CO, m 3 h ')• A polyvinyl-chloride cylinder (15-cm high, 21-cm internal diameter) was placed on the soil or snow surface. About an hour later, the cylinder was closed with a polyvinyl-chloride lid, and a rubber-capped (blood collection) needle was fitted onto an air sample port on the top of the lid. Air in the chamber was aspirated through the needle into an evacuated vial (5 ml) three times at equal intervals (2 or 5 minutes), and the C0 2 concentration in the vial air was measured as described in the previous paper (Mariko et al. 1994). The concentration of C0 2 was plotted against the time (Fig. 1).   Profile of C0 2 concentration in snowpack The concentration gradient of C0 2 in the snowpack and atmosphere was determined on the same day as the direct determination of the C0 2 flux. Air was sampled, using the gas collector described in a previous paper (Mariko et al. 1994), at various depths below or above the closed chamber. The air sampling ports were installed at different snow depths. Snow air was drawn through the needle with an evacuated sampling vial (5 ml). Some air (about 2 ml) in the gas collector was previously removed with a gastight syringe connected to the needle at the end of the gas collector. The air samples were taken twice at two lateral locations at the same snow depth to obtain a more representative concentration at a particular depth.
Calculation of C0 2 evolution from snow surface using Fick's DM The evolution of C0 2 from the snowpack in winter was calculated assuming simple diffusion, average C0 2 gradients, average snow depth and average porosities (Sommerfeld et al. 1993, Mariko et al. 1994. The diffusion of gases in the snowpack can be described by Fick's first law, which states that the steady state transport of gas by diffusion through a unit area is proportional to the concentration gradient measured along the line normal to the area: where Fis the C0 2 flux (mg C0 2 nr 2 h '), a is a constant for unit conversion (36 as a value), Dp is the diffusion coefficient (cm 2 s ') and dC/ dz is the vertical concentration gradient of C0 2 (mg C0 2 m 3 cm 1 ). The term dC/dz can be determined directly by drawing a straight line through the average concentrations from the snow surface to 20-cm depth (C0 2 profile) and by calculating the slope.
The diffusion coefficient Dp is influenced by the atmospheric pressure and temperature in situ. Dp for C0 2 in snow can be estimated from Eq. (3): where D 0 is the diffusion coefficient of C0 2 in air (cm 2 s '), 0 is the porosity of the snow measured (cm 3 cm 3 ), and Ais the empirical tortuosity factor (cm cm ') (van Bavel 1951). D 0 was corrected for the temperature ( T,°K) and the atmospheric pressure (P, hPa) at each study area from Eq. (4): T and P were estimated from the meteorological data of the Meteorological Observatory of the Agricultural Research Centre of Finland. Dg(S) is the diffusion coefficient at the normal state (the empirical value 0.135 cm 2 s' 1 for C0 2 ) and an exponent n is 1.71 for C0 2 (Osozawa 1987). The tortuosity factor for snow was not measured here. However, the constant value of 0.69 estimated from the data of Sommerfeld et al. (1993) was adopted. This value is appropriate since the porosities influencing tortuosity were almost constant over the snow depths measured, and similar to those (c. 0.58) of Sommerfeld et al. (1993) and Solomon and Cerling (1987). The average porosity of snow was between 0.53 and 0.59 at all measurement points.

Measurement of C0 2 concentration of air stored in the sampling vial
The concentration of C0 2 in the air stored in the sampling vial was determined with an infrared gas analyser, IRGA (Model ZRC, Fuji Elec. Tokyo, Japan). A 1or 2-ml aliquot of air in the vial was removed with a gas-tight syringe (2-ml volume). During the withdrawal, the air removed was replaced with distilled water, adjusted at pH4.O with phosphoric acid (H,PO 4 ). The withdrawn air was injected into a gas line purged with 424 AGRICULTURAL AND FOOD SCIENCE IN FINLAND Vol. 5 (1996): 421^30. pure N 2 gas at a flow rate of 0.5 ml min 1 . This gas line allowed the injected air to reach the IRGA in several seconds. The electric output from the IRGA was recorded as a steep peak, like a spike on a pen recorder (Servocorder 5R6312, Graphtec, Tokyo, Japan). The spike heights (X) were used for quantitative analyses ofCO, concentration. A standard curve was preliminarily prepared by injecting the same volume of C0 2 gas (1 or 2 ml) in various known concentrations (T), which gave a linear response curve to the spike heights (T=ll.OBX + 20.15, r=0.99, pcO.001).

Carbon and nitrogen contents in soil
After cultivation of the fields, soil samples from each field were collected at five flux-measurement points from seven depths, 2.5, 10, 20, 30, 40, 50 and 60 cm. Samples of dried soil (105°C) were ground to determine the carbon and nitrogen contents. Each dried sample was sieved through a 0.2-mm mesh sieve, care being taken to sieve the entire sample. This portion of the sample was analysed for total carbon and nitrogen using an automatic carbon and nitrogen analyser (C-N Corder MT-600, Vanako, Kyoto, Japan) Figure 2 shows the vertical distribution of the carbon and nitrogen contents in three soil types. The soil carbon and nitrogen contents were constant to a depth of 20 cm, but below that decreased with depth in each soil. The contents differed, however, between the soil types. The values of carbon to 20-cm depth were approximately 20% for peat, 1.7% for sand and 3.5% for clay soil, and those of nitrogen were approximately 1.3% for peat, 0.13% for sand and 0.25% for clay soil. Storage ofcarbon within the upper 60-cm layers of soil amounted to 54.4 kgC m 2 in the peat, 16.9 kgC m 2 in the clay and 8.4 kgC nr 2 in the sandy soil.  The average snow depth on the measurement days was approximately 20 cm for the peat, 15 cm for the sand, and 25 cm for the clay soil. Table 2 shows the temperatures of air, snow and soil on the measurement days in each soil type. The air temperatures in winter were between -1 and +2°C in the three soil types, and those in the summer between 17 and 19°C. The summer air temperatures were slightly higher than the average maximum air temperature of the warmest month (July, 15.8°C) as is normal for these soil types. The winter soil temperatures at 2-cm depth were between 0 and I°C in each soil type, and the summer soil temperatures between 18 and 19°C. The snow temperatures below 10-cmdepth were almost O°C in all soil type fields ( Table 2).

Results
Concentrations of CO, in the snowpack (average of five measurement points) increased with snow depth and its gradients differed from one soil to another (Fig. 3). The concentration decreased linearly between the snow surface and 15-cm or 20-cm depth. The gradient was smallest in the sand soil, but largest in the peat soil; it was intermediate in the clay soil.
The winter fluxes of C0 2 from the snowpack (average of five measurement points) differed depending on the soil types, ranging from 18 to 27 mg CO, nr 2 h 1 for the CC method and from 10 to 27 mg CO, m 2 h 1 for the DM method (Table 3). Both methods showed a similar differential pattern between soil types. The snow-covered fields in the peat and sand soils produced the largest and smallest amounts ofC0 2 , respectively. The clay soil had intermediate winter fluxes. The variance in the winter fluxes within the same soil types was relatively larger in the DM than in the CC method (Table 3). The results obtained with the CC method indicated that the average summer C0 2 evolution was 14-24 times higher than the winter evolution.

Discussion
Carbon dioxide in soils derives primarily from root respiration and microbial oxidation of organic matter (Witkamp and Frank 1969,Wildung et al. 1975, Kowalenko et al. 1978, Heinemeyer et al. 1989, Koizumi et al. 1993. Therefore, low temperatures during the snow-covered period retard C0 2 production in soils. Based on some empirical or theoretical studies (Steudler et al. 1989, Bouwman 1990), most C0 2 budgets have been calculated by assuming that C0 2 exchange stops when the soil is covered with snow or soil temperatures drop to around O°C. The present study, however, suggests that the assumption for the winter CO, budget is incorrect. In sub-alpine and cool-temperate regions, soils at 5to 10or 20-cm depth, or the active layer of soil respiration (Crill 1991), never freeze during the winter because of the thermal insulation effect of the snow cover (Sommerfeld et al. 1993). This allows soil microorganisms to produce enough CO, to form gradients within the snowpack (Fig. 3). It is suggested that each soil type serves as a source of CO, in winter.
Table 3 also indicates that the CO, evolution rates in winter were higher in the peat than in the sand or clay soil fields. The C0 2 evolution rates in the peat soil were about twice as high as those in the sand soil. These results suggest that the higher availability of organic matter in the peat soil results in higher rates of C0 2 production in that soil (cf. Fig. 2).
Using the DM method it was estimated that the snow-covered fields in southernFinland can evolve 10-27 mg of CO, per m 2 per hour on some winter days (Table 3). The winter fluxes show somewhat lower values than those calculated by Sommerfeld et al. (1993) from the CO, profiles in the snowpacks of sub-alpine meadow, southeasternWyoming. Nakane (1978) estimated C0 2 evolution rates from Japanese cooltemperate beech/fir forest soils beneath the snowpack to be between 50 and 70 mg per m 2 per hour. Mariko et al. (1994) also demonstrated that the C0 2 evolution rates from snow-covered soils in four Japanese cool-temperate deciduous broadleaved and evergreen needle forests ranged from 20 to 75 mg per m 2 per hour.
Here, direct measurement with the CC method also demonstrated remarkable fluxes of 18-27 mg per m 2 per hour (Table 3), indicating the validity of the results obtained with indirect measurements or estimation. Thus the winter fluxes represent an important part of the annual Table 3. Average and standard deviation of C0 2 evolution rate (mgC0 2 m 2 h') from snow surface (winter) and soil surface (summer) in three soil types. The flux was measured directly using the closed chamber method (CC method) and calculated using Pick's law from C0 2 profiles over a 0-20 cm layer within the snowpack (DM method   (Coyne and Kelley 1974, Nakane 1978, Mariko et al. 1994 Still, it has not been established whether C0 2 evolution through the snow surface changes during the winter period. Sommerfeld et al. (1993) stated that, as soil microorganisms remain active without change during the winter because there is no fluctuation in soil temperature (Solomon and Cerling 1987), the results obtained during limited study periods are representative of the entire winter. The actual winter fluxes, however, seem to change seasonally depending on snowfall. This is suggested by the findings of the previous study (Mariko et al. 1994) showing that the C0 2 gradient within a thick snowpack is small compared with that within a thin snowpack, even in areas with almost the same vegetation (Kelley et al. 1968, Solomon andCerling 1987). To understand the contribution of winter to the annual carbon budget, the flux must be determined throughout an entire winter period.
Seasonal comparison of C0 2 evolution from each soil type can be made using data obtained with the CC method in the subarctic climate range. On the basis of these data, winter C0 2 evolution in subarctic climate ranges is estimated to be between 4.2% and 7.2% of evolution in summer, when it is almost totally due to soil respiration. These percentages represent the extreme minimum fluxes on winter days in relation to the maximum fluxes on summer days with the highest annual temperature. Therefore, the winter/summer ratio of the C0 2 evolution rate calculated from seasonally-integrated fluxes is estimated to exceed 4%.
The snowpack influences winter CO, evolution in both positive and negative ways. First, the snow insulates the soil thermally, allowing CO, production to continue throughout the winter at temperatures above freezing. Second, the snow supplies water that participates in weath-ering reactions in the soil, allowing CO, production to increase (Solomon and Cerling 1987). Third, the snow causes the CO, concentration in snow and soil to rise by partially capping the snow-soil column and by lowering the soil diffusion coefficient as the soil pores are filled with water from the melting snowpack. Moreover, there is the interesting suggestion of Sommerfeld et al. (1993), who pointed out that the snow feeds fungi and bacteria, making them able to respire there.
Direct measurement (CC method) showed the same trend in the difference in winter fluxes between sites as the indirect measurement (DM method). The CC method tended to show larger absolute average fluxes than the DM method even when the measurements were made at the same points and on the same day. The variance in the fluxes within each soil type tended to be slightly larger with the DM than with the CC method. These differences may be due to the qualitative difference in their measurement principles, which result in different kinds of measurement errors (Roiston 1986). Of great practical interest is to know which of the methods is more reliable in situ. This will be determined in a further study to be conducted by the authors. The present study does at least suggest, however, that the CC method is acceptable for the practical measurement of C0 2 flux through the snow surface.