The FCA presents an estimate of the annual change of organic C in forest ecosystems (ΔC), and is considered here in terms of a pool-based approach:
where ΔPh, ΔD, and ΔSOC are annual C changes in phytomass, dead vegetational organics and soils, respectively, and by using a flux-based approach, describing the C fluxes between ecosystems and atmosphere, lithosphere, and hydrosphere as:
where NBP and NPP are net biome and net primary production, HR is ecosystem heterotrophic respiration, D is fluxes generated by disturbances, and L is lateral fluxes. Major C pools and fluxes, which are considered in this study, are presented in Fig. 1.
Two major information sources covering the entire country were used in this study: forest inventory data and an integrated land information system (ILIS), either independently or in different combinations (Fig. 1). Forest inventory data (called the forest inventory approach) were presented in aggregated form by the State Forest Account (SFA), which contains the comprehensive characteristics of forests (dominant species, age structure, levels of productivity, etc.) by forest enterprises (about 1800), administrative units (89), a number of large subdivisions (e.g., European and Asian Russia) and the whole country. The SFA data are available for 1961, 1966, 1973, 1978, 1983, 1988, 1993 and 1998 (SNKh SSSR, 1962; Gosleshoz SSSR, 1968, 1976, 1982, 1986; Goscomles SSSR, 1990, 1991; FSFMR, 1995, 1999; data for 1 January of each year). Details of the Russian forest inventory system and the specifics and reliability of the information are discussed in Shvidenko and Nilsson (2002). The ILIS consists of a multi-layer Geographical Information System (called GIS approach) for Russian land, developed by IIASA in 1993–2000 (Nilsson et al., 2000). The GIS components include digitized maps (forests, soils, landscapes, vegetation, land-use, on-ground organic layer, etc.), at the scale of 1:1–1:4 million, accompanied by attributive databases, numerous long-term statistical data, including wood harvest, disturbances, etc.; auxiliary modeling systems (e.g., for assessing phytomass, net and gross growth, and mortality); data on typified soil horizons; measurement results of CWD and dead roots; and relevant semi-empirical aggregations and scientific results. The classification of forests (FA) used 125 classes of forest vegetation applied to about 13 000 GIS polygons. Background data used for the GIS development were basically collected between the late 1970s to mid 1990s, and the major maps used were published about the 1990s. The forest map has been connected to the State Forest Account data for 1993. Thus we approximately reference the major GIS components for the 1990s.
2.1. The FCA for 1990
The FA phytomass pool was assessed by two independent methods: forest inventory data and GIS tools (Nilsson et al., 2000). A special modeling system was developed for assessing forest phytomass by fractions based on forest inventory data. About 2250 sample plots and more than 250 regional studies were used for generating non-linear multidimensional regression equations of major forest-forming species for aggregated eco-regions of Russia. As forest inventory data is the only reliable source of forest phytomass inventories at the national scale, the regressions were presented by the ratio (conversion factor) between the mass of phytomass fractions and growing stock volume. The statistical accuracy of the equations, which included age, site index and relative stocking, was rather high. The multiple non-linear correlation coefficients and the regression coefficients of variables were, as a rule, statistically significant with a probability of >0.95; the adequacy was checked by analyzing the distribution of residuals and their correlations with the variables included in the equations. The calculation of phytomass (divided into stem wood, branches, bark, roots, foliage, understory and green forest floor) was provided by combining the inventory data for forest enterprises into data for the eco-regions and application of the regression equations to these aggregated data. The second estimate of phytomass was made by GIS tools. In this approach, we used the average densities of phytomass fractions for forest vegetation classes [calculated independently basically using the original database by Bazilevich (1993)] and areas from the forest map. The calculations were provided for initial polygons of the map and aggregated to eco-regions and seven bio-climatic zones. The regression equations mentioned above are available from Shvidenko et al. (2001a) and the aggregated conversion factors and carbon densities by major forest forming species are presented in Table 10 in the Appendix.
Table 10. Carbon density and conversion factors by major forest forming species
|Species and groups||Density by age groups, kg C m−2||Conversion factor, Mg C m−3|
| Hard deciduous||3.41||7.49||8.46||9.79||0.6326||0.5201||0.4884||0.5043|
| Soft deciduous||1.30||5.18||6.86||6.77||0.4917||0.4288||0.3878||0.3688|
| Other species||3.33||5.60||10.59||16.33||0.7515||0.5444||0.5416||0.5772|
| Hard deciduous||2.63||6.23||6.42||6.25||0.6164||0.5275||0.5190||0.5449|
| Soft deciduous||1.05||4.13||5.90||6.64||0.5058||0.4395||0.4125||0.4113|
| Other species||1.98||1.37||5.97||15.28||0.7014||0.5044||0.5312||0.5764|
In order to estimate CWD we used forest inventory data aggregated for individual forest enterprises, data of CWD measurements on sample plots available in publications and archives, as well as our own measurements. The GIS was used for up-scaling the “point” measurements. The mass of dead roots was assessed in a similar way, based on IIASA's forest map and average estimates of dead roots by species and eco-regions (Shvidenko et al., 2000).
The GIS approach was used to assess the three C pools of soils (indicated in Fig. 1) by overlapping the soil map at the scale of 1:2.5 million (digitized by the Dokuchaev Soil Institute, Moscow) with the IIASA forest map that was connected to typified soil profiles and to an on-ground organic layer database (Nilsson et al., 2000). The reference data were collected for undisturbed soils. In order to take into account the impact of disturbances, we introduce an index of the severity of disturbance regimes (ISDR) as the ratio ISDR = 100 (BA + ASF + GG)/FL, where BA, ASF and GG are area of burnt and dead forests, anthropogenic sparse forests, and grassy glades and barrens, respectively. Dependence between correction coefficients and the ratio (separately for on-ground organic layer and 1 m top soils) was parametrized based on fragmented regional data (Table 11 in the Appendix). The corrections for 1990 were −10.9% for C of on-ground organic layer and −1.5% for the 1 m soil.
Table 11. Auxiliary analytical expressions used in calculations
|A1||L11 (t) = 174.82 − 1.8003t+ 0.14865 t2||CWD1, 0 ≤t≤ 30 for 1960–1990, M11(1961) = 3239 Tg C|
|A2||L12 (t) = 254.1 + 1.3621t− 08161t2||CWD1, 0 ≤t≤ 8 for 1990–1998, M12(1998) = 5001 Tg C|
|A3||L21 (t) = 20.68 − 0.2226 t+ 0.0183t2||CWD2, 0 ≤t≤ 30 for 1960–1990, M21 (1961) = 304 Tg C|
|A4||L22 (t) = 31.40 + 0.1684t− 0.1009t2||CWD2, 0 ≤t≤ 8 for 1990–1998, M22 (1998) = 381 Tg C|
|A5||L31 (t) = 338 + 1.9t||RL, 0 ≤t≤ 30 for 1960–1990, M31 (1961) = 3778 Tg C|
|A6||L32 (t) = 395 − 3.25t||RL, 0 ≤t≤ 8 for 1990–1998, M32 (1998) = 4113 Tg C|
|A7||L41 (t) = 740 + 5t||GPL1, 0 ≤t≤ 30 for 1960–1990, M41(1961) = 1222 Tg C|
|A8||L42 (t) = 890 + 0.6t||GPL2, 0 ≤t≤ 8 for 1990–1998, M42 (1998 = 1490 Tg C|
|A9||H1= 150 +t||Input to soil C compartment, 0 ≤t≤ 30 for 1960–1990|
|A10||H2= 180 +t||Input to soil C compartment, 0 ≤t≤ 8 for 1990–1998|
|A11||ΔOGL= 10.03 − 0.18x+ 0.05667x2||Correction for soil C (1 m layer) transformation, x= ISDR|
|A12||Δ1MTL= 0.836 + 0.120x||Correction for on-ground organic C transformation, x = ISDR|
|A13||KNSF= 1.029 − 0.0052x||A13–A16 give corrections for C of 1 m top layer of UFA; see notes below|
|A14||KASF= 1.02 − 0.00333x|| |
|A15||KBA= 1.03 − 0.0048x|| |
|A16||KGG= 1.025 − 0.0046x|| |
|A17||KONSF= 1.135 − 0.016x||A17–A20 give corrections for C of on-ground organic layer of UFA; see notes below|
|A18||KOASF= 1.044 − 0.0074|| |
|A19||KOBA= 1.2 − 0.036x|| |
|A20||KOGG= 1.23 − 0.028x|| |
The major C fluxes (NPP, HR and L) were also estimated based on the GIS approach. Because the fluxes should be assessed for definite areas and a given period of time, we applied statistical methods by using georeferenced polygons and the field measurement averages of the calculation indicators. The database, which was used for the NPP assessment, includes the results of measuring NPP on about 1600 sample plots by three aggregated fractions (total green parts, above-ground wood, and underground parts). We used the original database, developed by Bazilevich (1993), which was supplemented by measurements made during the last decade (e.g., Karelin et al., 1995; Gower et al., 1995, 2000; Schulze et al., 1999).
It is evident that NPP quantified in this way describes a certain “quasi-stable” state, as both GIS data and weather conditions used in the calculations are averaged over a certain period, data on forest disturbances are incomplete and have time lags, etc. To improve the accuracy of the results, the most important natural and anthropogenic factors that affected the forest ecosystems during the assessment period (1988–1992) were analyzed. In order to estimate NPP, we considered the change in forest productivity after disturbances (particularly after a fire on permafrost areas), impact of wetland amelioration, and loss of the actual NPP in areas affected by major types of disturbances during the assessment period; these corrections increased the total NPP on FL for about 7% (Shvidenko et al., 2001b). For many years, NPP located in wood has played and continues to play a crucial role in C sequestration and is an important component in assessing the NEP and NBP. For this reason, as well as for validating the results obtained by using GIS technologies, we estimated the gross and net increment on FA based on forest inventory data (Shvidenko et al., 1995a, 1997).
Heterotrophic respiration (HR) was assessed based on the soil map polygons and average annual accumulated fluxes calculated by soil types and aggregated land cover classes. The assessment was provided in two independent ways using different databases and methodological approaches. The first approach did not include winter fluxes, i.e., the period with an air temperature of < 0 °C and using a linear decrease in CO2 soil evolution at the temperature interval of 5 to 0 °C (Nilsson et al., 2000; Stolbovoi, 2002b). The second approach accounted for total yearly estimates of soil C fluxes and attempted to include the specifics of different vegetation types (Kurganova, 2002).
UFA included burnt areas, dead stands, cutovers, glades and barrens, natural and anthropogenic sparse forests, and unstocked planted forests. The average densities of the main indicators (phytomass, NPP, CWD, SOC, etc.) were estimated by the UFA categories and bio-climatic zones and applied to these aggregated units.
A unified approach was used to assess the fluxes caused by major disturbances (fire, insect and disease outbreaks, site effects of harvest, and abiotic factors). The total carbon flux TCFρ,t1 during year t1 generated by a disturbance ρ (for annual time steps) was calculated as:
where DFρ,t1 is the direct flux during year t1, and PDFρ,t<t1 is the post-disturbance, as a rule biogenic, flux generated by disturbance ρ that occurred during previous years t < t1. The values of DFρ,t1 and PDFρ,t<t1, as well as the explicit form of eq. (4), depend on the type, strength and extent of ρ, the conditions under which ρ occurs, the type and specifics of the ecosystem, and on the approach and structure of the model used. As an example, the direct flux due to forest fires (for the year of fire) is defined as (Shvidenko and Nilsson, 2000):
where Cilkq are the coefficients for the consumed forest combustibles during a fire, Silkq is the estimate of burned vegetation areas, (FC)ilkq is the storage of forest combustibles (t/ha, dry matter), and γ is the coefficient for recalculating dry organic matter to C units [we used 0.5 for forest combustibles and 0.45 for the remaining vegetation (Vonsky, 1957; Filippov, 1968; Telizin, 1973)]. The indexes are: i= territorial units for which the calculations are made; l= aggregated land-use classes; k= types of forest fire; and q= types of forest combustibles.
Post-fire fluxes are caused by the decomposition of both incombustible (dead) residuals and post-fire die-back (mortality), as well as by changes in the structure and content of soil organic matter. Let Oij (t) be a function that describes the amount of dead organic matter entering a decomposition pool j in year t, and Oij (t*) be the value of this function in year t*. Using a simple exponential model (the more advanced approaches available (Melillo et al., 1989; Aber et al., 1990) were not used due to the lack of data for the diversity of soil–vegetation groups of Russian forest lands), the process of decomposition of organic matter of pool j is described as:
where Gij (t*,τ) is the mass of organic matter that did not decompose during period τ, αij is the constant of decomposition, and τ is the number of years between the year of the fire and the year of the PDF estimation, e.g., τ*=t*−t1. Evidently, for eq. (6), the time for decomposition of 95% of the decomposition C pool T0.95 depends only on αij, T0.95= ln20/αij. Thus, the post-fire biogenic flux to the atmosphere during year t1 caused by fires during previous years can be estimated by:
where χ, 0 < χ < 1, is the share of C from decomposed organic matter that is taken up by the atmosphere, ϕ= int [T0.95] (the integer part of T0.95), and δSOC is the post-fire change of heterotrophic soil respiration during year t1. There is not sufficient data for regional estimates of χ, so we used the average value of 0.88 [based on measurements by Chagina (1970) of 0.92 for old growth Siberian cedar (Pinus sibirica) forests, Vedrova (1995) of 0.75–0.92 and 0.77–0.88 for 25-yr-old coniferous and deciduous plantations, respectively, and Kurz et al. (1992) of 0.82 for Canadian forests]. To estimate the actual post-disturbance fluxes, a retrospective period of 200 yr is needed for the taiga and forest tundra zones. In the framework of the pool-based account, the changes of soil organic C were calculated as (ΔSOC)ijt1= 1.05(1 −χ)(PDF −δSOC)ijt1+ Cch,t, where the first component provides the change of soil C caused by the decomposition of post-fire die-back, and the second provides the input of charcoal during the year of the fire. Background data used in the calculation are presented in Tables 12 and 14 in the Appendix. More details on the topic can be found in Shvidenko et al. (1995b), Shvidenko and Nilsson (2000) and Nilsson et al. (2000).
Table 12. Coefficients of organic matter decomposition
| ||Coefficients aij||T0.95, yr|
|Bioclimatic zones||Green parts||CWD2 (medium)||CWD1 (slow)||Roots||Green parts||CWD-medium||CWD-slow||Roots|
|Forest tundra, sparse and northern taiga||0.145||0.043||0.024||0.066||21||70||125||45|
|Semi-desert and desert||4.5|| ||0.175||0.33||0.7|| ||17||7|
Table 14. Extent of major types of disturbances in Russian forests (×106 ha, average for 1988–1992)
| || ||Biotic factors|| |
| ||Wild fire by types|| ||Including|| |
|Bioclimatic zones||Crown fire||OGF||Peat fire||Total||Total||Insects||Diseases||Harvest|
|Forest tundra, sparse and northern taiga||0.06||0.39||0.15||0.60||1.12||0.05||0.02||0.05|
|Semi-desert and desert||—||—||—||—||0.01||—|| ||—|
Two independent attempts were made to assess the fluxes caused by the decay of forest products. The first attempt used the slightly modified approach described above for disturbances (Shvidenko, 1997), and the second used a specially developed model (Obersteiner, 1999). Both approaches are based on as comprehensive as possible assessment of harvested wood (including domestic consumption), wood products and wastes, which are separated in three decay pools of different decomposition rates (fast, medium and slow).
The lateral fluxes were estimated in an aggregated form taking into account that the results of measurements are poor for big regions. Dissolved and particulate organic carbon (DOC and POC, for which the difference is defined by a boundary size of 0.45– 0.5 μm) is transported with surface and below-ground run-off to the hydrosphere (rivers and inner water reservoirs), lithosphere deposits on geochemical barriers, and to deep (outside soil profiles) below-ground water. Based on lysimetric measurements, the content of DOC and POC in soil water is rather high in forest soils of the boreal zone, and varies on average from 50 to 100 mg L−1, sometimes significantly more (e.g., Ponomareva and Plotnikova, 1972; Djakonova, 1972; Glazovskaya, 1996). Concentrations of DOC and POC in rivers are significantly lower: from 10 to 30 mg L−1 (Vinogradov et al., 1998; Kassens et al., 1999; Romankevich and Vetrov, 2001). This means that part of DOC and POC is absorbed by mellow deposits, where the content of organic matter is often high (0.5–1.5%); the results of direct estimates of this C flux to the lithosphere are fragmentary (Glazovskiy, 1983; Glazovskaya, 1996; Rapalee et al., 1998). The assessment of lateral fluxes were provided by aggregating the eco-region data. A part, which is supposedly transported to the hydrosphere by FL, was assessed based on areas of catchments and the corresponding share of FL. The C reaching ecosystems in dry and wet deposition (DOC + POC) was assessed by using published data (Meyback, 1982; Lychagin, 1983; Saet and Smirnova, 1983; Nilsson et al., 1998; Labutina and Lychagin, 1999; Russian official reports on the state of the environment for the last decade). We excluded from the consideration some processes due to their supposedly small impact on the forest FCB and/or contradictive opinions about the sign of fluxes generated by these processes (e.g., soil erosion, cf. Schlesinger, 1995; Smith et al., 2001).
2.2. The FCA for 1961–1998
The State Forest Account data for 1961–1998 were used as the basis for assessing the dynamics of vegetational C in phytomass, CWD, and dead roots. Because of the growing stock volume presented by the Russian forest inventory has a bias, which changes over time (Shvidenko and Nilsson, 2002), the calculations were carried out in two ways (for official forest inventory data and for “restored dynamics”) in an endeavor to eliminate this bias. It should be noted that this correction was only done for growing stock on FA; the remaining forest inventory data do not have any significant biases in this respect. In order to estimate phytomass dynamics on FA, the above-mentioned modeling system on forest phytomass was applied to the forest inventory data for 1961–1998 by aggregated eco-regions. Phytomass on UFA was defined by major UFA land categories using average zonal values previously estimated for 1990 (Tables 8 and 9 in the Appendix). To check the consistency of our restored dynamics and to quantify the CWD input to decomposition pools, we calculated a wood balance based on forest inventory data, growth indicators, wood consumption data and the impact of disturbances (for details, see Shvidenko and Nilsson, 2002).
Fluxes due to litter dynamics were modeled based on a linear feedback theory (Olson, 1963) as:
where Mj(t) is the mass of the litter, Lj(t) is the litter input during year t, α presents the zonal decomposition coefficients by four decomposition pools j (foliage and green forest floor, two pools of CWD: medium-fast and slow pools, with a top diameter of wood residuals at 1 ≤d≤ 8 cm and d > 8 cm, respectively, and roots). In order to calculate the integral of eq. (8), Lj(t) were approximated by the polynomials at the intervals [0 ≤t≤ 30] and [0 ≤t≤ 8] for 1960–1990 and 1990–1998, respectively. Coefficients of analytical expressions for Lj(t) and average zonal values of αj are given in Table 12 in the Appendix.
The dynamics of soil organic C of FL were calculated based on land cover change, densities of soil C and the severity of disturbance regimes, expressed by ISDR. Supporting information is presented in Table 11 in the Appendix. For comparison, we examined a simple one-compartment model of soil C dynamics, which was presented as:
where Cs(t) is the mass of C in soil organic matter at time t, 1–χ is the share of C entering the soil (which is partially humified and partially transported out of forest ecosystems), and β is the decomposition rate (mineralization) of the SOC compartment.
The basic idea of this simple approach has been used in a number of studies, e.g., for Canadian forests in the CFS-CBM (Kurz et al., 1992) and for arable crops in central Sweden in the ICBM (Andren and Katterer, 1997). In spite of its simplicity, the approach has evident advantages: outflows from the pools follow first-order kinetics, and eqs. (8) and (9) can be integrated analytically. However, while empirical regional data for α are rather numerous (e.g., Grishina, 1986; Kobak, 1988; Orlov, 1990), available estimates of β are not sufficient and we considered the behavior of eq. (9) for β varying from 0.006 to 0.002 (e.g., for T0.95 from 500 to 1500 yr, cf. Andren and Katterer, 1997).