Measuring and Monitoring Soil Carbon - Soil Carbon: Science, Management and Policy for Multiple Benefits - page 197

Upscaling and Modelling

soil-monitoring programmes. Visible and
near infrared (Vis-NIR) and mid-infrared
(MIR) reflectance spectroscopy have pro-
duced good results for the prediction of SOC
content (McBratney et al ., 2006; Shepherd
and Walsh, 2007; Viscarra et al ., 2010). Air-
borne imaging spectroscopy has been used
for mapping topsoil properties (Ben-Dor
et al ., 2008). This technique provides good
results for mapping the SOC content in the
plough layer of bare soils (i.e. in seedbed
condition).
The prediction of topsoil SOC content
from remotely acquired spectral data is
generally based on an empirical approach.
Reference soil analyses of samples col-
lected in the field are related to the spectral
information through a multivariate calibra-
tion model used to predict the SOC values
at locations for which there are no meas-
ured SOC data (Stevens et al ., 2012). Sev-
eral multivariate calibration models were
developed to predict the SOC content in
the plough layer of bare cropland fields in
an airborne imaging spectroscopy scene of
420 km 2 in Luxembourg. For such large
areas, the model performance depends
strongly on the validation technique. The
root mean square error of the most stringent
validation procedure (excluding the fields
used in the calibration) was equal to
4.7 g C kg − 1 . Although this uncertainty is
probably not good enough for the estima-
tion of SOC stocks in individual fields, it
can be used for regional mapping of SOC
content and provides a unique insight into
the spatial pattern of SOC content within
fields (Fig. 16.2; Stevens et al ., 2010). In all
cases, however, conventionally measured
SOC (dry combustion) in reference labora-
tories is necessary to calibrate the new
techniques, and to build spectral libraries
needed for the extension of spectral meas-
urements in unsampled areas (Shepherd
and Walsh, 2007; Bartholomeus et al ., 2008;
Terhoeven-Urselmans et al ., 2010). Further,
as techniques and standards for soil ana-
lyses are evolving continuously, it is good
practice to preserve soil samples from
SMNs so that they may be re-analysed in
the future (McKenzie et al ., 2002; Shepherd
and Walsh, 2002; Arrouays et al ., 2012).

Typically, soil-monitoring activities encom-
pass several decades of measurements (which
implies long-term commitment from fund-
ing agencies and researchers). Appropriate
data management tools are required to store
the various data, check for errors and re-
trieve selected data for sharing and analysis
(Cools et al ., 2006; Lacarce et al ., 2009; Batjes
et al ., 2013). The range of soil and ancillary
data collated through SMNs and similar field
sampling programmes should be stored in a
(freely accessible) information system to
support geostatistical analyses and model-
ling (see Chapter 17, this volume). At present,
however, external access to SMN data is often
restricted to the metadata (Morvan et al ., 2008;
Panagos et al ., 2013), thereby greatly reducing
their value to the scientific community and
society.
In particular, there is a need for glo-
bally consistent protocols and tools to
measure, monitor and model SOC changes
and GHG emissions so that funding agen-
cies and other organizations can assess
uniformly the possible impacts of land-use
interventions and climate change, as well
as the associated uncertainties, across the
range of world climates, soils and land
uses. An example of such an integrated fa-
cility is the online Carbon Benefits Project
tool developed for the Global Environmen-
tal Facility (GEF). It includes both empir-
ical as well as process-based modelling
approaches, which can be chosen based
on the user requirements and available
data through a user guidance module (Milne
et al ., 2010, 2012). Whether monitoring
or modelling SOC dynamics/processes, the
key issue is how to address complex issues
of spatial and/or temporal variability at the
scale of interest (Cerri et al ., 2004; Maia
et al ., 2010; Smith et al ., 2012). Oppor-
tunities include the use of ancillary data,
scale-speciic methods, development of
spectral libraries, digital mapping of soil
carbon and better integration of remote-
sensing technologies into empirical and
simulation SOC models (Grunwald et al .,
2011; Croft et al ., 2012; Minasny et al .,
2013).