Spatial variability of soil organic carbon in grasslands: implications for detecting change at different scales

https://doi.org/10.1016/S0269-7491(01)00265-2Get rights and content

Abstract

Extensive data used to quantify broad soil C changes (without information about causation), coupled with intensive data used for attribution of changes to specific management practices, could form the basis of an efficient national grassland soil C monitoring network. Based on variability of extensive (USDA/NRCS pedon database) and intensive field-level soil C data, we evaluated the efficacy of future sample collection to detect changes in soil C in grasslands. Potential soil C changes at a range of spatial scales related to changes in grassland management can be verified (α=0.1) after 5 years with collection of 34, 224, 501 samples at the county, state, or national scales, respectively. Farm-level analysis indicates that equivalent numbers of cores and distinct groups of cores (microplots) results in lowest soil C coefficients of variation for a variety of ecosystems. Our results suggest that grassland soil C changes can be precisely quantified using current technology at scales ranging from farms to the entire nation.

Introduction

Soil organic C is the largest C reservoir in many terrestrial ecosystems including grasslands, savannas, boreal forests, tundra, some temperate forests, and cultivated systems, comprising as much as 98% of ecosystem C stocks in some systems (Schlesinger, 1977). Globally, the amount of C stored in soil is equal to the amount stored in vegetation and in the atmosphere combined (Schimel, 1995). A substantial portion of C fixed by vegetation is transferred to the soil annually (Raich and Nadelhoffer, 1989), a portion of which is refractory material with long turnover times (Falloon and Smith, 2000, Paul et al., 1997); the rest decomposes relatively rapidly and is returned to the atmosphere as CO2. Thus soil C is a large, relatively dynamic component of terrestrial C stocks.

Historically, intensive cultivation in the US has resulted in the transfer of about 1 Pg of soil organic matter to the atmosphere in the form of CO2 (Kern, 1994). Soil organic matter (SOM) losses due to conversion of native grasslands to cultivated agriculture are both extensive and well documented (Davidson and Ackerman, 1993, Haas et al., 1957, Kern and Johnson, 1993, Schlesinger, 1986) and losses due to overgrazing and poor pasture management have also been observed (Abril and Bucher, 1999, Conant and Paustian, 2001, Fearnside and Barbosa, 1998). However, historical SOM losses can potentially be reversed, and atmospheric C sequestered, with improved agricultural management. In the United States, agricultural conservation practices such as reduced tillage, improved fertilizer management, elimination of bare fallowing, use of perennials in rotations, and use of cover crops can potentially sequester large amounts of atmospheric C (Paustian et al., 1997). Similarly, cultivated areas converted to well-managed permanent grassland, as pastures or rangelands, constitute potential C sinks. Within established pastures, soil C can be increased by eliminating soil disturbances and increasing primary production through improved grazing management, fertilization, sowing improved forage species and legumes, or irrigation (Conant et al., 2001).

While improved agricultural or pasture management can sequester C in soils (Sampson et al., 2000), there are several challenges associated with verifying changes in soil C. Many factors influence soil C, including temperature, precipitation, NPP, and soil physical characteristics (Parton et al., 1987), all of which are spatially variable. The result is substantial variability in soil C, with coefficients of variation as high as 20% even in a visually uniform cultivated field (Robertson et al., 1997). As variability increases, the minimum number of samples needed to detect a given level of change increases. Furthermore, short-term changes in soil C are usually small relative to the amount of C in soil. In grasslands, for example, soil C may be sequestered at a rate of around 0.46 Mg C ha−1 year−1 in surface horizons (Conant et al., 2001), against a background of 30–80 Mg C ha−1 in most temperate grassland soils (Conant et al., 2001).

Measurement and verification needs vary between broad-scale applications, such as national-level greenhouse gas inventories and design of government policy, versus project-based sequestration efforts that might be undertaken by private landowners or cooperatives at a more local scale. In the former instance, the main emphasis is likely to be on accurate quantification of carbon changes nationally, with less need for attribution of changes to specific localities and/or practices. In the latter, project-level, case, attribution to specific landowners and practices is likely to be important. In both instances, however, there are tradeoffs between the ability to detect change (precision) and the number of samples required (which is directly related to cost).

In this paper we evaluate some limitations of sampling to quantify soil C changes in managed grasslands for broad-scale inventories/assessments and smaller-scale project-level quantification. First, we used a large compilation of soil data for the entire US, the USDA/NRCS (US Department of Agriculture/Natural Resource Conservation Service) pedon database (NSSC, 1997), to examine broad-scale soil C variability and the implications for detecting soil C changes at national, state, and county levels. Second, farm-level analysis, based on analysis of soil C variability within plots sampled to assess soil C levels in managed pastures, was used to evaluate the efficacy of different sampling schemes designed to maximize sensitivity to field-level changes in soil C following changes in land use or management. These spatially intensive sampling schemes could be used for benchmark site sampling, model validation, or to assess impacts of land management changes as part of a mitigation project. Results presented here are part of a larger effort to integrate quantitative measurements of changes in soil C with various tools to extrapolate and interpolate those measurements.

Section snippets

NRCS PEDON database

The USDA/NRCS pedon database is a compilation of soil information collected during soil mapping or detailed soil investigations and assembled by the USDA/NRCS (NSSC, 1997). Each record, collected to ‘represent the central concept of a soil series, the central concept of a map unit but not of a series, or to bracket a range of properties within a series or landscape’ (NSSC, 1997), contains information about horizonation, geographic location, and many soil physical and chemical characteristics.

NRCS PEDON database

General characteristics of soil C from the USDA/NRCS pedon database are shown in Table 2. More than 2700 samples from uncultivated grassland-derived soils in the USDA/NRCS pedon database contained soil C data (Table 2). Though the number of samples decreased dramatically for smaller sample areas, sampling density for Dundy County (0.49 samples per 1000 ha) was slightly greater than that at the state level (0.42 samples per 1000 ha) and much greater than at the national (0.012 samples per 1000

Discussion

Our results suggest that changes in soil C brought about by changes in pasture management can be quantified precisely using current technology at scales ranging from individual farms to the entire nation. As with any sampling design, a tradeoff exists between the number of samples required and the desired precision, and the number of samples collected and analyzed is directly related to costs associated with verification. This work demonstrates that small changes in soil C at a very fine scale,

Conclusions

This research evaluated two methods of detecting changes in soil C driven by changes in agricultural management: (1) regional (county-, state-, or national-level) analysis relying on previously collected data and (2) field-level sampling. Statistically derived expectations about our ability to detect changes with future resampling were assessed for each method. Our results show that detection of changes in soil C likely to occur over a 5–10 year period can be accomplished at broad scales

Acknowledgements

Discussions with N.A. Scott, K.R. Tate, G.R. Smith, and M. Sperow aided significantly in the development of this project. Thanks to participants of the Canadian Prairie Soil Carbon Balance Project (B.G. McConkey, in particular) for assisting in implementation of the intensive farm-level sampling method. The comments of N.A. Scott, B.G. McConkey, and one anonymous reviewer on an earlier version of this manuscript improved the manuscript. J. Zumbrunnen's statistical assistance was instrumental in

References (31)

  • M.D. Eve et al.

    A national inventory of changes in soil carbon from national resources inventory data

  • M.D. Eve et al.

    An inventory of carbon emissions and sequestration in US cropland soils

  • P.D. Falloon et al.

    Modeling refractory soil organic matter

    Biology and Fertility of Soils

    (2000)
  • R.F. Follett et al.

    The potential of US grazing lands to sequester soil carbon

  • H.J. Haas et al.

    Nitrogen and Carbon Changes in Great Plains Soils as Influenced by Cropping and Soil Treatments

    (1957)
  • Cited by (144)

    • Grazing management for soil carbon in Australia: A review

      2023, Journal of Environmental Management
    View all citing articles on Scopus
    View full text