The number of soil pits or cores that must be sampled varies according to the topographic variation and spatial heterogeneity in the distribution of soils. At the large (regional/national) scale, a stratified random sample scheme is statistically efficient for soil data, provided that the basis for stratification accounts for a significant fraction of the variation in soil carbon contents. It is essential that the sample is accurately placed within the stratum, which requires the use of Geographic Information Systems (GIS) and Geographical Positioning Systems (GPS) or accurate field survey. For validation purposes, all sample locations should be geo-referenced as accurately as possible. The actual plot location should be marked to allow relocation of the plot reference point, for resampling in close proximity to prior sampling. High precision is requiring to avoid confounding by the high level of fine-scale variation that is common in soil carbon. Where an a priori randomly determined location is found to be inaccessible, it must be replaced with a sample having similar characteristics. The number of samples and the choice of location depend on the land-use and soil management systems.
At the project level, sampling strategies could be based on representative toposequences or catena for major land uses within the ecoregion of interest. The sampling strategy should be determined when the spatial distribution of the overlying plant community is known. Material should be collected at the end of each growing season (at the maximum stage of decomposition and with the maximum litter input), but at a minimum at the beginning and end of each commitment period. Uncertainty associated with differences in stock between two times can be substantially reduced by using paired samples; in other words, samples taken at time 2 should be from very close to the samples taken at time 1 and the difference calculated per pair, rather than on the accumulated averages (Lal et al., 2000).
Except in unusual circumstances and in peats, the amount of soil organic matter declines exponentially with depth (Nakane, 1976). Globally, only about one-third of the organic carbon in the surface 2 m is found at a depth of 1-2 m (Batjes, 1996). Losses or accumulations of soil carbon are greatest in the upper soil profile (0-15 cm), which should be sampled most intensively (Richter et al., 1999). In the case of management changes on agricultural land (e.g., Smith et al., 1997b, 1998), samples must be taken from lower in the profile because accumulations of carbon in the surface horizons may be balanced by losses of carbon at depth (Powlson and Jenkinson, 1981; Ismail et al., 1994; McCarty et al., 1998). Defining, a priori, a global soil depth to which carbon should be analyzed is not practical, however. The sampling depth should be below the depth that significant change in carbon is expected to occur.
Soil carbon changes in some situations may occur at substantial depths (e.g., 2-5 m in deep, tropical soils), whereas in cropped soils undergoing management change, almost all of the change may occur in the top 30 cm. The important issue is that the depth used in one inventory should match the depth used for that location or land-use type in the next inventory. The depth definition should be left open, subject to consistency between inventory times. Determining the effective sampling depth is not easy when the surface level is subsiding or aggrading (e.g., in highly organic bog soils, soils under arable agriculture, or soils that are subject to rapid erosion). Simultaneous measurements of bulk density are essential in all cases (Lal et al., 2000); such measurements can be used to calculate an effective soil depth in mineral soils. Measuring soil depth as a function of mass, rather than distance from the surface, may be desirable because management activities may significantly change soil bulk density between measurements, substantially altering the amount of carbon found in different increments of distance from the soil surface. Failure to account for bulk density changes could lead to large artifacts in apparent carbon storage (Ellert and Bettany, 1995). In organic soils, an absolute reference level is required, which must be surveyed from a stable datum.
Other reports in this collection