| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Division of Ecosystem Sciences and Center for Assessment and Monitoring of Forest and Environmental Resources (CAMFER), 137 Mulford Hall, College of Natural Resources, Univ. of California, Berkeley, CA 94720-3114
* Corresponding author (earthy{at}nature.berkeley.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: GIS, geographical information system • LRR, land resource region • MRLA, major land resource areas • NRCS, Natural Resources Conservation Service • OM, organic matter • SIC, soil inorganic carbon • SOC, soil organic carbon • STATSGO, state soil geographical database
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Soil inorganic C is also a large C pool. However, studies on SIC storage and content have only focused on local or regional assessments (Schlesinger, 1982; Grossman et al., 1995; Monger and Matrinez-Rios, 2000). Estimates of the SIC pools at a national or global scale have been more tentative than estimates of SOC pools (Lal et al., 1998b). Nonetheless, most of the SIC, which exists as carbonates, is believed to occur in soils of arid and semiarid regions (Grossman et al., 1995; Schlesinger, 1997; Lal et al., 1998b). Monger and Matrinez-Rios (2000) estimated the amount of soil carbonate in grazing lands of the USA by focusing on the woodlands, shrublands, and grasslands that occur within aridic, ustic, and xeric moisture regimes using random sampling for at least 25 sites per ecoregion. An estimation of SIC storage for the entire USA has not been made.
The STATSGO database is not only amenable for exploring the national distribution of soil properties, but also for examining soil properties within LRR and among the taxa within Soil Taxonomy categories. At this time, no systematic studies of SOC and SIC partitioning by LRRs or soil orders at a national scale have been performed.
The STATSGO database used in this study is a geographic information system (GIS) based relational database compiled by the National Resources Conservation Service (NRCS), which was made by generalizing detailed soil survey data. The level of detail in STATSGO is based on its intended use for planning and management covering state, multi-state, and regional areas. Most importantly, it is the only soil database currently available for evaluating national soil resources (SCS, 1992; Reybold and Gale, 1989). The mapping scale for the STATSGO data is 1:250 000 (with the exception of Alaska) with a minimum mapping unit area of 6.25 km2, equivalent to a square cell of 2.5 by 2.5 km. The basic structure of STATSGO is the map unit and its components. Components are the finest horizontal entities (units) for data recording. A map unit may contain 1 to 21 components. In the conterminous USA (excluding water, urban land, bare rock, and other non-soil bodies), there are 10 441 STATSGO map units (74 590 polygons) and 111 247 components (regions within the map units). For each component, its area percentage (%) within the map unit, its soil classification (Soil Survey Staff, 1999), and its properties for each soil layer ("O" horizon excluded) are reported in a relational database format by experienced local soil scientists based on soil survey results.
The purpose of this study is to calculate total SOC and SIC inventories, as well as the contents (e.g., concentrations, kg m–2), within three depth intervals (0–0.2, 0–1.0, and 0–2.0 m) for the conterminous USA using the STATSGO database, and to examine the partitioning of SOC and SIC pools by natural land resource region and by soil orders in Soil Taxonomy. Our analysis of SOC expands previous STATSGO SOC analyses, and our analyses of SIC for the nation is, to our knowledge, an entirely new contribution to the soil carbon inventory literature.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Determination of Soil Carbon Storage and Variation
To calculate the soil C (SOC or SIC) storage, the OM and CaCO3 data (reported on a <2.0-mm fraction) was normalized for gravel content. Soil and rock fragment fractions for each soil layer of a component are reported as inch10 (>250 mm), inch3 (75–250 mm), and ‘no10’ (<2.0-mm fraction from that which has passed through a 75-mm sieve). The high (H) and low (L) gravel estimates (%) for each fraction (USDA-NRCS, 2003) are recorded in STATSGO.
The low and high fractions of the soil in the <2-mm diameter fraction for a given layer of the soil components within STATSGO were calculated as follows:
 | [1] |
 |
The midpoint fraction of soil <2 mm in diameter for each layer of the component Qijkp(M) as well as the rock fragment conversion factor (fijkp) used to adjust for the volume of rocks in a given layer were calculated in the way used by Bliss et al. (1995).
Soil organic C and SIC storage for each layer of a component was calculated:
 | [2] |
The SOC and SIC storage in the 0- to 0.2-, 0- to 1-, and 0- to 2-m depths of each soil component was calculated by summing the SOC and SIC of the corresponding soil layers, weighted by depth.
The SOC (or SIC) content to 0.2, 1, and 2 m of each soil component (kg m–2) was calculated as follows:
 | [3] |
Total SOC (or SIC) storage SCTij(ZD) and content SCDij(ZD) in each map unit were estimated as follows:
 | [4] |
Soil organic C (or SIC) storage SCTi(ZD) and content SCDi(ZD) in the ith state were estimated as follows:
 | [5] |
Soil organic C (or SIC) storage SCT(ZD) and content SCD(ZD) for the conterminous USA were calculated in a way similar to Eq. [5].
The variance SCS2 (D)i, and the coefficient of variation CV(D)i, of SOC (or SIC) among soil components in the ith state using the midpoint approach was calculated as follows (Gnedenko and Khinchin, 1962):
 | [6] |
The area weighted variance SCS2(D), and the coefficient of variation CV(D), of SOC (or SIC) among soil components in the conterminous USA determined by the midpoint approach were calculated in a way similar to Eq. [6].
Determination of Total Soil Carbon within a Land Resource Region
Land resource regions are geographically associated land resource units defined by USDA–NRCS (USDA-SCS, 1981). Land resource regions are designated by capital letters and identified by a descriptive name. Land resource regions, A through U, with the exception of Q, are found in the conterminous 48 states. Each LRR is further divided into Major Land Resources Areas (MLRAs). We calculated the SOC (or SIC) for each MLRA. Then, the SOC (or SIC) in each LRR was calculated from the MLRAs within the LRR. The area weighted variance and coefficient of variation for SOC (or SIC) of the soil components within each LRR, and within LRRs, using the midpoint approach were determined in the way similar to the Eq. [6].
Determination of Soil Carbon Storage within Each Soil Order
The six taxonomic categories (in order of increasing detail: order, suborder, great group, subgroup, family, and series) of each soil component are given in STATSGO. We calculated the SOC and SIC storage and content for each soil order. The variance and coefficient of variation of SOC (or SIC) among soil components in each taxon of the Soil Taxonomy were estimated in the way similar to Eq. [6].
Treatment of Missing Data
Each empty record (blank or zero value) of 12 fields (OML, OMH, BDL, BDH, no10L, no10H, inch3L, inch3H, inch10L, inch10H, CaCO3L, and CaCO3H) was checked to determine the completeness of the dataset.
The following assumptions were used to determine if an empty record in OML and OMH fields is a missing datum: (1) OML and OMH should be zero for the following textures: WB (weathered bedrock), UWB (unweathered bedrock), CEM (cemented), and IND (indurated); (2) a zero value is acceptable for OMH if the texture is ICE (ice or frozen soil), FRAG (fragmental material), G (gravel), and CIND (cinders); (3) a zero value for OML is acceptable in mineral or inorganic, but not for organic or organic-modified textures. All other empty records were considered to be missing data if found in OML and OMH fields. If a missing datum occurs in the middle layer of a soil profile, the average OM (OML or OMH) values of its next (upper and lower) layers were used to fill in the missing data. For the remaining missing records determined in the OMH and OML fields, the method of Amichev and Galbraith (2004) was used to estimate values for the missing data.
An empty bulk density record in the BDL and BDH fields is considered to be missing data if the other soil properties such as OM (OML, OMH), no10 (no10L, no10H), etc. in the same layer have non-zero values. The missing bulk density values were first estimated according to Brejda et al. (2001). The method of Amichev and Galbraith (2004) was then used to estimate values for missing data that were unable to be determined by the method of Brejda et al.
An empty soil fraction record (<2 mm in size) in no10L and no10H fields is considered as a missing data point if the records of the other soil properties such as OM (OML, OMH) and bulk density (BDL, BDH) in the same layer have a non-zero value. If a missing data point occurs in the middle layer of a soil profile, the average no10 value (no10L or no10H) of the adjacent (upper and lower) layers was used to fill in the missing data. For the remainder of the missing data in the no10L and no10H fields, the method of Amichev and Galbraith (2004) was used to calculate the missing data.
An empty rock fragment fraction record (>250 mm is size) in the inch10L and inch10H fields is considered to be missing data if the TEXTUREs-left (rock fragment modifier) code is ST (stony), STV (very stony), STX (extremely stony), BY (bouldery), BYV (very bouldery), and BYX (extremely bouldery), indicating that the layer should contain 
15% volume of stones. Missing data was assumed if there is an empty record in the smaller sized rock fragment fraction (75–250 mm in size) of the inch3L and inch3H fields when the TEXTUREs-left code is CB (cobbly), CBA (angular cobbly), CBV (very cobbly), CBX (extremely cobbly), CN (channery), CNV (very channery), CNX (extremely channery), FL (flaggy), FLV (very flaggy), and FLX (extremely flaggy). It was also assumed that a soil layer with stones should also contain smaller size rock fragments. The missing data of the rock fragment fraction (inch10L, inch10H, inch3L, and inch3H) were estimated according to Amichev and Galbraith (2004).
If all data layers of a component are empty in CaCO3L and CaCO3H fields when the soil has the formative element "Calc" at great group or subgroup taxonomy levels, or has the "carbonatic" element at family level of Soil Taxonomy, the component is considered to have at least one missing datum in the CaCO3L and CaCO3H fields. This missing component was then estimated using the average CaCO3L (or CaCO3H) values of the same layer in other soil components that have the same taxon (filling in priority: the series, family, subgroup, and great group in Soil Taxonomy) within the same map unit, or within in nearby map units of the same MLRA, and or the map units of the same land resource region.
The 368 942 layers of data for 111 247 components (excluding water, urban land, bare rock, and other non-soil bodies) in STATSGO were checked, and the missing values were filled in based on the assumptions and the filling methods described above.
The original projection of STATSGO was retained except the datum was changed from NAD27 to NAD83 using ARC/INFO software (Environmental Systems Research Institute, 1998). All calculations were processed using programs written by the senior author using the visual basic language in Microsoft Access (Microsoft Corporation, 2000) and Avenue language in ArcView (Environmental Systems Research Institute, 1999).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
Use of multiple data sources or methods to estimate the U.S. SOC pool should improve the confidence in these estimates, though STATSGO is the only national soil database presently available. In previous work, Bliss et al. (1995) used midpoint values from STATSGO to determine the SOC storage (total in the soil profile) in 40 states. Lacelle et al. (2001) generated a map of SOC in the upper 1 m of North America (the U.S. portion was again based on STATSGO midpoint values). While the midpoint value may yield reasonable stock estimates, low and high limits provide conservative bounds to the U.S. soil C stocks.
Most estimates of SOC in the USA are limited to the upper 1 m. Here, we calculated that the SOC in the upper 1.0 m of the conterminous USA ranges from 254 to 1131 x108 Mg, with a midpoint value of 639 x 108 Mg. Using laboratory data from 3700 pedons, Kern (1994) estimated that the SOC in the upper 1 m in the USA is between 621 and 845 x108 Mg, a range obtained by scaling up the pedon data using three approaches: ecosystem, great-group taxonomic unit, and soil map of world-based methods. In general, Kern's result is similar to our estimated midpoint value suggesting that the STATSGO data and approach are reliable for C inventory analyses.
The spatial distribution of SOC and SIC in the upper 2.0 m is presented in Fig. 1 and Fig. 2 . The spatial SOC distribution is similar in most details to the map generated by Kern (1994) to 1.0 m and to the SOC map of North America by Lacelle et al. (2001). The eastern Great Plains and Midwest have the highest SOC content, though some high SOC regions in the East Coast, Gulf Coast, and Pacific Northwest also occur. For SIC, the highest SIC storage is in Texas, though the Midwest also has a high SIC storage in the Upper 2.0 m. There are some obvious SIC changes across state boundaries—for example between Iowa and South Dakota. This indicates that some SIC data might be still missing in these states or the soil survey results should be correlated between states since current results were based on surveys conducted separately in each state.
|
|
|
It is widely recognized that SIC occurs in soils of arid and semiarid regions (Grossman et al., 1995; Schlesinger, 1997; Lal et al., 1998b; Monger and Matrinez-Rios, 2000), a pattern also observed here for the upper 1.0 m. However, when SIC to 2.0 m in soil depth is considered, our results show that there is a large SIC pool in the Midwest, where mean annual precipitation (MAP) is about 700 to 1000 mm. While the SIC in the upper 1.0 m is generally leached out in these climates (Jenny and Leonard, 1934), the deeper depth increments still retain a mixture of both primary and secondary carbonates. In the Midwest, the SIC to 2.0 m strongly correlates spatially with the extent of the last glaciation (Schruben et al., 1998), suggesting these recently rejuvenated areas retain carbonate derived from calcareous sediments of various types. The SIC pattern in the south central plains (particularly Texas) matches the pattern of bedrock (Schruben et al., 1998). There is little SIC in the East and Southeast to 2.0-m depths because of the high mean annual precipitation (MAP). In contrast, there is high SIC in the West due to the arid and semiarid climate and to the bedrock and aerosol sources of carbonate (Monger and Matrinez-Rios, 2000).
Quantity and Spatial Variability of Soil Carbon in the Land Resource Regions
The C storage and content in each LRR are presented in Table 3. About 12 to 20% of total U.S. SOC is in LRR M (Central Feed Grains and Livestock Region) and 9 to 10% is in both the LRR T (Atlantic and Gulf Coast Lowland Forest and Crop Region) and the LRR K (Northern Lake States Forest and Forage Region) regions. The highest SOC content (2.0 m, midpoint method) is LRR U (Florida Subtropical Fruit, Truck Crop, and Range Region) with 39.6 kg m–2. Other regions with remarkable SOC contents are: LRR T (Atlantic and Gulf Coast Lowland Forest and Crop Region) with 35.3 kg m–2, LRR K (Northern Lake States Forest and Forage Region) with 25.2 kg m–2, and LRR L (Lake States Fruit, Truck, and Dairy Region) with 20.9 kg m–2. In terms of SIC, 21 to 23% of the total is in LRR D (Western Range and Irrigated Region), 17 to 20% is in LRR H (Central Great Plains Winter Wheat and Range Region), and 8 to 13% is in LRR M (Central Feed Grains and Livestock Region). The highest SIC contents (to 2.0 m by midpoint) are in the LRR I (Southwest Plateaus and Plains Range and Cotton Region) with 34.8 kg m–2, LRR J (Southwestern Prairies Cotton and Forage Region) with 29.6 kg m–2, and LRR H (Central Great Plains Winter Wheat and Range Region) with 16.6 kg m–2.
|
|
|
|
The patterns of SOC and SIC storage vs. soil depth vary with the soil orders. Inceptisols and Alfisols have 35 and 39% (midpoint value) of their SOC in the upper 0.2 m, while only 16% of SOC is in the upper 0.2 m of Histosols. Unlike SOC, most of the SIC storage is in the deeper layers. However, there are some exceptions: Andisols and Entisols have 19 and 15% (midpoint value) of their total SIC in the upper 0.2 m.
There is a large spatial variability of SOC and SIC in each order and at all depths (Table 6). Standard deviation (Std) describes the absolute variability of SOC and SIC within each order, and the coefficient of variation indicates the relative variability of SOC and SIC, which can be used to compare the differences in the variation of SOC and SIC among the orders, since the means of SOC (or SIC) in each order are different. Entisols and Inceptisols have the largest CV (or relative variability) among the orders. In terms of SIC, Andisols and Spodosols have the smallest standard deviation due to their very low SIC content. Relative variability of SIC is much larger than that of SOC within any order.
|
The area-weighted variability of SOC in taxa at each taxonomic categorical level of each order is presented in Fig. 3 . The variability of SOC in taxa decreases as taxonomic category decreases in all soil orders, which is especially obvious moving from the family to the series categories. In Entisols and Inceptisols, variability in taxa, at any taxonomic level, is larger than that in the other soil orders.
|
The results obtained in this study (based on the analysis of 111 247 soil components) suggest that predicting the SOC pool using the LRR-based method will require a larger sample size than the taxonomy-based method to arrive at similar levels of accuracy. The coefficient of variation for SOC in LRRs is 103, 156, and 178% for the 0.2-, 1-, and 2-m depths, respectively (Table 4). In contrast, the coefficient of variation for SOC in the orders is 82, 107, and 125% for the same depths (Table 7), a modest improvement over the LRR approach. The coefficient of variation for SOC substantially decreases when estimates are based on soil order approach. This is undoubtedly due to the fact that taxonomic designations are successful at grouping soils of similar characteristics, whereas resource regions may indeed have one dominant state factor, while many others (which may have affects on soil C pools) vary considerably. Therefore, a higher accuracy estimate of SOC can be expected when taxa of lower taxonomic categories are used to estimate the SOC pool over a geographical area.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
To estimate soil C for a large area, we have observed that LRR (land cover or ecosystems)-based methods will need a larger sample size than the taxonomy-based method to achieve the same level of accuracy since the variation of SOC in a LRR population is larger than that in the Soil Taxonomy population. Variation within soil taxa becomes smaller as taxonomic category becomes more detailed, especially from the family to the series categories. Due to high variability, there will be especially large inaccuracies in SIC estimations based on taxa at higher taxonomic categories. An unanticipated finding was that a substantial SIC pool exists in the central USA between depths of 1 to 2 m. When SIC in the 2.0-m soil is considered, a large SIC pool was found in the Midwest where the mean annual precipitation (MAP) is about 700 to 1000 mm. While the SIC in the upper 1.0 m is generally leached out in these climates, the deeper depth increment still retains some combination of primary and secondary carbonates.
We conclude by noting that the patterns of SOC and SIC across the landscape are determined by the widely varying combinations of vegetation, climate (precipitation and temperature), topography, soil parent materials, and landform age (Jenny, 1994) that occur across the country. In this paper we have first focused only on soil C storage and its partitioning among LRRs and soil orders, with little discussion as to why the trends are present. In a companion paper, we examine the factors controlling soil C distribution and discuss the implications with respect to global change and land use activities.
|
ACKNOWLEDGMENTS |
|---|
Received for publication May 16, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. L. Jespersen and L. J. Osher Carbon Storage in the Soils of a Mesotidal Gulf of Maine Estuary Soil Sci. Soc. Am. J., March 12, 2007; 71(2): 372 - 379. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
