Soil Infrared SpectroscopyApplications in Africa
International Workshop
Soil Spectroscopy: the present and future of Soil Monitoring
FAO HQ, Rome, Italy, 4-6 December 2013
Soil spectroscopy to monitor the state of soil resources in the
present and in the future
Keith D Shepherd
Surveillance Science• Measure frequency of problems and associated risk factors in populations
using statistical sampling designs & standardized measurement protocolsUNEP. 2012. Land Health Surveillance: An Evidence-Based Approach to Land Ecosystem Management. Illustrated
with a Case Study in the West Africa Sahel. United Nations Environment Programme, Nairobi.
http://www.unep.org/dewa/Portals/67/pdf/LHS_Report_lowres.pdf
Identify problem
Develop case
defintition
Develop
screening test(s)
Measure prevalence
(no. cases/area)
Measure incidence
(no. cases/area/time)
Confirm risk factors
Measure
environmental
correlates
Differentiate risk
factors
Infrared spectroscopy
Shepherd KD and Walsh MG (2007) Infrared
spectroscopy—enabling an evidence-based diagnostic
surveillance approach to agricultural and environmental
management in developing countries. Journal of Near
Infrared Spectroscopy 15: 1-19.
• Increase sample density
• Measure soil functional
properties at landscape
scales
• Direct prediction of soil-
plant responses to
management
Spectral shape relates to basic soil properties
• Mineral composition
• Iron oxides
• Organic matter
• Water (hydration,
hygroscopic, free)
• Carbonates
• Soluble salts
• Particle size distribution
Functional properties
Infrared spectroscopy
Dispersive VNIR FT-NIR FT-MIR Robotic FT-MIR Portable
Handheld MIR ?Mobile phone cameras
?
Shepherd KD and Walsh MG. (2002) Development
of reflectance spectral libraries for characterization of
soil properties. Soil Science Society of America
Journal 66:988-998.
Brown D, Shepherd KD, Walsh MG (2006). Global
soil characterization using a VNIR diffuse reflectance
library and boosted regression trees. Geoderma
132:273–290.
Terhoeven-Urselmans T, Vagen T-G, Spaargaren O,
Shepherd KD. 2010. Prediction of soil fertility
properties from a globally distributed soil mid-
infrared spectral library. Soil Sci. Soc. Am. J.
74:1792–1799
CalibrationSoil organic carbon
Spectral pretreatments
• Derivatives, smoothing
Data mining algorithms:
• PLS +
• Support Vector Machines
• Neural networks
• Multivariate Adaptive
Regression Splines
• Boosted Regression
Trees
• Random Forests
• Bayesian Additive
Regression Trees
Training Out-of-bag
validation
Soil pH
R package soil.spec
Soil spectral file
conversion, data
exploration and
regression functions
✓60 primary sentinel sites➡ 9,600 sampling plots
➡ 19,200 “standard” soil samples
➡ ~ 38,000 soil spectra
➡ 3,000 infiltration tests
➡ ~ 1,000 Landsat scenes
➡ ~ 16 TB of remote sensing data to date
AfSIS
Spectral libraries
Spectral prediction performance
Spectral Lab
Network
•IAMM, Mozambique
•AfSIS, Sotuba, Mali
•AfSIS, Salien, Tanzania
•AfSIS, Chitedze, Malawi
•CNLS, Nairobi, Kenya
•ICRAF, Nairobi, Kenya
•CNRA, Abidjan, Cote D’Ivoire
•KARI, Nairobi, Kenya
•ICRAF, Yaounde, Cameroon
•Obafemi Awolowo University,
Ibadan, Nigeria
•IAR, Zaria, Nigeria
•ATA, Addis Ababa, Ethiopia (+ 5
on order)
•IITA, Ibadan, Nigeria
•IITA, Yaounde, Cameroon
•ICRAF, Nairobi, Kenya
Planned
•Eggerton University,
Kenya
•MoA, Liberia
•IER, Arusha, Tanzania
•FMARD, Nigeria
•NIFOR, Nigeria
•CNLS, Nairobi
•BLGG, Kenya (mobile
labs)
• Submit batch of spectra
online
• Uncertainties estimated for
each sample
• Samples with large error
submitted for reference
analysis
• Calibration models improve
as more samples submitted
Soil-Plant Spectral Diagnostics Lab
• 500 visitors/yr
• 338 instruction
• 13 PhD, 4 MSc training
Land Health Surveillance
Consistent field
protocol
Soil spectroscopyCoupling with remote
sensingPrevalence, Risk factors, Digital mapping
Sentinel sites
Randomized sampling schemes
Markus Walsh
Probability topsoil pH < 5.5 ... very acid soils
prob(pH < 5.5)Africa Soil
Information Servicewww.africasoils.net
Markus Walsh
Calibrating plant response to IR
http://afsis-dt.ciat.cgiar.org
IR applications Vital signs
Cocoa - CDIParklands Malawi
National surveillance
systems
Regional Information Systems
Project baselines
Ethiopia, Nigeria
Rangelands E/W AfricaSLM Cameroon MICCA EAfrica
Global-Continental Monitoring Systems
CGIAR pan-tropical sites
AfSIS
Private sector soil testing
Critical success factors• Consistent field sampling protocol
• Soil-Plant sample labeling, drying,
preparation, sub-sampling, shipping, back-up
storage
• Data management, linking
• Judicious selection of samples for reference
analysis
• Consistency of reference analyses
• Stable spectrometer technology and
protocols
• Training in all steps and follow-up support
http://worldagroforestry.org/research/land-health/spectral-diagnostics-laboratory