Scaling Agronomy for Smallholder Recommendation Systems using Mid-Infrared and Total X-Ray Fluorescence Spectroscopy for Rapid, Low Cost Soil-Plant Analysis
Scaling Agronomy for Smallholder Recommendation Systems
1. 86: Symposium – Smallholders Managing Soil Health for Climate Resilience
2018 ASA, CSSA, and CSA Annual Meeting (Nov. 4-7) in Baltimore, MD, USA
Scaling Agronomy for Smallholder Recommendation
Systems using Mid-Infrared and Total X-Ray
Fluorescence Spectroscopy for Rapid, Low Cost Soil-
Plant Analysis
Keith Shepherd, Erick Towett, Andrew Sila
Africa Soil Information Service
2. Improving relevance of soils information for users
Limitations
• Inference space of recommendations not known
• Uncertainty not represented or communicated
• Soil science knowledge not integrated into
economic decision making
Shepherd KD. How soil scientists can do a better job of making their research useful. The Conversation
(Science & Technology) 14 August 2018.
3. Africa Soil Information Service
Statistically sound
sampling schemes
Sample diversity
Unbiased prevalence
data
Shepherd et al. (2015). Land health surveillance and response: A framework
for evidence-informed land management. Agricultural Systems 132: 93–106
5. Hengl T, Leenaars JGB, Shepherd KD, Walsh MG, Heuvelink GBM, Mamo T, Tilahun H, Berkhout E,
Cooper M, Fegraus E, Wheeler I, Kwabena NA. 2017. Soil nutrient maps of Sub-Saharan Africa:
assessment of soil nutrient content at 250 m spatial resolution using machine learning. Nutrient Cycling in
Agroecosystems 109:77–102.
Digital soil mapping of soil nutrients
https://soilgrids.org/
8. Spectral Shape Relates to
Basic Soil Properties:
• Mineral composition
• Iron oxides
• Organic matter
• Carbonates
• Soluble salts
• Particle size distribution
These properties are the determinants of most functions!
MIR spectral fingerprints
10. MIR soil spectral profiling
0.0
0.5
1.0
6 7 8
pH
density
Clu
0.00
0.05
0.10
0.15
20 40 60 80
Clay (%)
density
Cluster
A
B
0.00
0.05
0.10
0.15
0 25 50 75 100
CEC (ECD)
density
Cluster
A
B
0.0
0.5
1.0
1.5
2.0
0.5 1.0 1.5
K (mg/kg)
density
Clu
Machakos County, Kenya (Technoserve Ltd)
11. NIR Plant N calibration in yam trials
YAMSYS Plant N calibration
13. Foliar pXRF as diagnostic
One Acre Fund trials in Western Kenya: Low P, K, S, Cu, Zn
K P S Mg Ca Cu Zn Fe Mn
14. Application levels for spectral technology
• Digital mapping of soil constraints, crop nutritional deficiencies,
spectral soil types
• National scale
• Refinement at county / district level
• Local scale - UAV hyperspectral calibration / indices
• Cost effective soil-plant testing services for farmers
• National labs
• Rural soil-plant spectral testing labs – walk-in service to farmers
• Low cost sensors for community knowledge workers, private
enterprises
15. Spectral lab network & capacity development
Country Lab
Benin AfricaRice
Cameroon IITA; ICRAF
Cote D’Ivoire CNRA; ICRAF
Ethiopia ATA/NSTC (5); Mekelle Uni;
Ghana CSRIO-SRI
Kenya KARLO; One Acre Fund; CNLS, ICRAF
Madagascar Antananarivo Uni (collaborative).
Malawi CARS/ DARTS
Mali IER
Morocco Mohammed Vi Polytechnic /OCP (in progress)
Mozambique IAMM
Nigeria Obafemi Awolowo Un; IITA; IAR; FDMA&RD (2)
South Africa KwaZulu-Natal Dept A
Tanzania SARI; Min Ag (4); Sokoine Uni
Outside Africa Australia (CSIRO); China (YPC); India (CIMMYT; ISSS-ICAR);
Peru (IIAP); UK (Rothamsted)
Soil archiving system
Training courses; lab audits
17. Represent & communicate uncertainty
• Use distributions not averages
• Communicate uncertainty to users
• Maintain links to original data
• Validate recommendations
• Focus further measurement on areas of
uncertainty that matter
18. Principles for taking agronomy to scale
•Define the decision dilemma
•Define the region of interest
•Sample it to provide a sound basis for inference
•Measure using rapid, low cost, reproducible methods
•Represent & communicate the uncertainty in results
•Validate recommendations using independent samples
•Maintain the link to the original data
•Focus further sampling to reduce uncertainty that matters
http://worldagroforestry.org/landhealth
19. Decision-focused agricultural research
• Identify the decision goals & alternatives
• Risk-Return analysis of intervention options
• Holistic - all relevant factors considered
• Quantifies uncertainties and risks; combines expert
knowledge with data
• Quantifies trade-offs - $
• Value-of-information analysis
• Where to measure & how much to spend on it
• Guides adaptive monitoring
Tools developed
• Monte Carlo simulation R package
• Bayesian Networks with value-of-information analysis
Using uncertainty and value-of-information analysis to define data needs
Luedeling E and Shepherd KD. 2016. Decision-Focused
Agricultural Research. The Solutions Journal 7: 46-54
20. Examples
•Shepherd K, Hubbard D, Fenton N, Claxton K, Luedeling E, De Leeuw J, 2015. Development goals should enable decision-making.
Nature 523, 152-154.
•Luedeling E and Shepherd KD. 2016. Decision-Focused Agricultural Research. The Solutions Journal 7: 46-54.
•Yet, B., Constantinou, A., Fenton, N., Neil, M., Luedeling, E. and Shepherd, K. 2016. A Bayesian Network Framework for Project Cost,
Benefit and Risk Analysis with an Agricultural Development Case Study. Expert Systems With Applications 60: 141–155.
•Rosenstock,T.S., Mpanda, M., Rioux J., Aynekulua, E., Kimaro, A.A., Neufeldt, H., Shepherd. K.D., Luedeling. E. 2014. Targeting
conservation agriculture in the context of livelihoods and landscapes. Agriculture, Ecosystems and Environment 187: 47–51
•Luedeling, E., Oord, A., Kiteme, B., Ogalleh, S., Malesu, M., Shepherd, K. D., De Leeuw, J. (2015). Fresh groundwater for Wajir – ex-ante
assessment of uncertain benefits for multiple stakeholders in a water supply project in Northern Kenya. Frontiers in Environmental
Science 3: 16.
•Favretto, N., Luedeling, E., Stringer, L. C., & Dougill, A. J. (2017). Valuing ecosystem services in semi-arid rangelands through stochastic
simulation. Land Degradation and Development 28, 65–73.
•Tamba Y, Muchiri C, Shepherd K, Muinga G, Luedeling E. 2017. Increasing DryDev’s Effectiveness and Efficiency through Probabilistic
Decision Modelling. ICRAF Working Paper No 260. Nairobi, World Agroforestry Centre.
•Tamba Y, Muchiri C, Luedeling E, Shepherd K. 2018. Probabilistic decision modelling to determine impacts on natural resource
management and livelihood resilience in Marsabit County, Kenya. ICRAF Working Paper No 281. Nairobi, World Agroforestry Centre
•Wafula J, Karimjee Y, Tamba Y, Malava G, Muchiri C, Koech G, De Leeuw J, Nyongesa J, Shepherd K and Luedeling E. (2018) Probabilistic
assessment of investment options in honey value chains in Lamu County, Kenya. Frontiers in Applied Mathematics and Statistics 4: 6-
11
•Whitney CW, Lanzanova D, Muchiri C, Shepherd KD, Rosenstock TS, Krawinkel M, Tabuti JRS, & Luedeling E. (2018).Probabilistic
decision tools for determining impacts of agricultural development policy on household nutrition. Earth’s Future 6: 359–372.
Notas del editor
One of the key innovations that has underpinned the Africa Soil Information Service is soil-plant spectral diagnostics, developed by ICRAF’s Soil-Plant Spectral Diagnostics Lab.
Spectral technology …
Applications . . .
Digital mapping for liming recommendations in Tanzania crop lands
One of the key innovations that has underpinned the Africa Soil Information Service is soil-plant spectral diagnostics, developed by ICRAF’s Soil-Plant Spectral Diagnostics Lab.
Spectral technology …
Applications . . .
Digital mapping for liming recommendations in Tanzania crop lands
One of the key innovations that has underpinned the Africa Soil Information Service is soil-plant spectral diagnostics, developed by ICRAF’s Soil-Plant Spectral Diagnostics Lab.
Spectral technology …
Applications . . .
Digital mapping for liming recommendations in Tanzania crop lands
One of the key innovations that has underpinned the Africa Soil Information Service is soil-plant spectral diagnostics, developed by ICRAF’s Soil-Plant Spectral Diagnostics Lab.
Spectral technology …
Applications . . .
Digital mapping for liming recommendations in Tanzania crop lands