Measuring and Monitoring Soil Carbon - Soil Carbon: Science, Management and Policy for Multiple Benefits - page 194

Following up from the review of Euro-
pean SMNs (Morvan et al ., 2007), the Joint
Research Centre of the European Commis-
sion has launched an initiative to sample
the topsoil at 22,000 points of the Land Use/
Cover Area Survey (LUCAS project, see
Montanarella et al ., 2011). LUCAS is based
on the visual assessment of land-use and
land-cover parameters that are deemed rele-
vant for agricultural policy. The soil sam-
pling at the LUCAS points carried out in
2009 will produce the first coherent pan-
European physical and chemical topsoil
database. This topsoil survey resulted in a
consistent spatial database of the soil cover
across Europe, based on standard sampling
and analytical procedures. A stratified sam-
pling design was implemented to produce
representative soil samples for major land-
forms and types of land cover of the partici-
pating countries.
Further initiatives for national SMNs
were retrieved from a questionnaire that
was sent to participants of a special session
at the SOM 2009 conference in Colorado
Springs, USA ( Table 16.2) . Some SMNs are
designed to estimate country-specific land-
use or management effects on SOC stocks,
while others collect soil carbon and ancil-
lary data to provide a nationally consistent
assessment of SOC conditions across the
major land-use types. The SMNs in Brazil
and Canada use a paired-site approach in
order to detect the SOC response to specific
land management (no-till in Canada and con-
version from forest to agriculture in Brazil).
These are stratified by ecoregion or typical
farming system. Three out of eight national
inventories have a grid design, and the re-
mainder are stratified according to land use,
soil type and climate regions.
Spencer et al . (2011) give a comprehen-
sive overview of the statistical consider-
ations to be taken into account for an SMN.
The three possibilities are simple random
sampling, stratified sampling or grid-based
sampling. Although random sampling is
conceptually the simplest option, it can be
difficult to implement and carries the risk
of leaving aside some regions. Grid-based
sampling is a practical and efficient tech-
nique and generally results in a better esti-

even distribution across the whole domain.
A stratified approach allows for allocation of
a greater number of samples in strata with a
higher variability in SOC stocks. Generally,
samples are allocated randomly within strata.
Strata can be defined according to major cli-
mate, land-use and soil-type combinations.
Such an approach has the advantage that
SOC stock changes can be linked directly to
the categories used for reporting by the UN-
FCCC (see Ravindranath and Ostwald, 2008).
Using the results of Century model runs,
Spencer et al . (2011) discuss the statistical
power of different attribution approaches for
sampling points to the strata.
It has been shown that marking indi-
vidual sampling sites with either a physical
marker (e.g. ball marker 3M, Austin, Texas)
or precise positioning using a Differential
Global Positioning System (DGPS) is the most
efficient in order to decrease the MDD for
eventual re-sampling in the future (Fig. 16.1 ).
Generally, a composite sample, which in-
volves taking subsamples and bulking them,
is taken according to a fixed spatial pattern
( Table 16.2 ). Studies of subsampling error of
monitoring sites are crucial for interpretation
of results and changes (Arrouays et al ., 2012).

Error propagation

Error propagation methods have been used
to estimate the contribution of the different
variables required to calculate SOC stocks
(C concentration, bulk density, stoniness
and soil depth) (Goidts et al ., 2009). Overall,
the spatial variability of topsoil SOC stocks
is larger in grasslands than in croplands
(Schrumpf et al ., 2011). Although the main
source of uncertainty in the topsoil SOC stock
varied according to scale, the variability of
SOC concentration and of the stone content
were the largest. When assessing SOC stock at
the landscape scale, one should focus on the
precision of SOC analyses from the labora-
tory, reducing the spatial variation of SOC,
and use equivalent masses for SOC stock
comparison (Goidts et al ., 2009).

Sampling depth

Organic layers at the soil's surface need to