User Guide: Orion™ Weather Station (Columbia Weather Systems)
Modelling pig and poultry production systems: computational and conceptual challenges
1. Modelling pig and poultry production
systems: computational and conceptual
challenges
M. Gilbert (& T. Van Boeckel)
Université Libre de Bruxelles
http://lubies.ulb.ac.be/Spatepi.html
T. Robinson
International Livestock Research
Institute
2. Livestock Human population
Spatial epidemiology &
invasion ecology
Catherine Linard
Yann Forget Jean
Artois
Clément Tisseuil
Gaëlle Nicolas
Weerapong
Thanapongtharm
Post
docs
PhD
http://lubies.ulb.ac.be/Spatepi.html
3. Intensified livestock production systems and
the emergence of Highly Pathogenic Avian
Influenza
Favour infections
High
density &
contacts
Genetic
similarity
Living &
health
condition
HPAI emergence mostly documented in intensive poultry
production systems
4. Intensified livestock production systems and
agricultural antimicrobial use
Favour infections
High
density &
contacts
Genetic
similarity
Living &
health
condition
Marginal gains due higher off-take rates do
make a difference over large volume
(but see Graham et al. 2007)
Feed
conversion
rate
matters
Fast prod.
cycles
High
inputs /
high
outputs
Higher use of antimicrobials in intensive systems
(preventive, curative, feed additive)
7. Outline
Context
• Intensification has taken place rapidly in the past
• Strong projected changes in demand will lead to further
intensification
• Changes are structured geographically
Objectives
• Better document the geographic distribution of intensive
livestock production
• Develop tools for making projections
Methods
• Mapping the global distribution of livestock
• Disaggregating in production systems
8. Livestock distribution:
Gridded Livestock of the World (GLW 1.0)
• General principle
• Collection of sub-national livestock
census data
• Many variables correlated to
livestock farming are mapped at high
resolution (e.g. land cover).
• Statistical models are based on high
resolution GIS predictors and
applied to downscale census values
by pixel (stratified multiple linear
regressions)
• Previous developments
• GLW 1.0 published by FAO in 2007,
mostly based on census data < 2005
(Wint & Robinson 2007)
• Global extent, 5 km resolution
9. Livestock distribution:
Gridded Livestock of the World (GLW 2.0)
• Recent developments
• More recent & higher resolution census data
• Spatial modelling @ 1km resolution
• Automation of the methodology in R
• Disseminated through the Livestock GeoWiki
• http://www.livestock.geo-wiki.org/
• New species division
• Cattle
• Pig
• Chicken
• Duck
• Sheep
• Goat
Robinson, T., W. Wint, T, G. Conchedda, T. P. Van
Boeckel, V. Ercoli, E. Palamara, G. Cinardi, L. D’Aietti,
& M. Gilbert (2014) Mapping the Global Distribution of
Livestock. PLoS ONE 9(5): e96084.
doi:10.1371/journal.pone.0096084
10. Livestock distribution:
Gridded Livestock of the World (GLW 3.0)
• In progress…
• New machine learning algoritm (Random Forest)
• Systematic evaluation (years of CPU time in 4 months)
• 180 models for Asia chicken and Africa cattle
• Processing on ILRI cluster (parrallelized)
• Full integration of metadata
• Spatial modelling & dissemination @ 1 km & 10 km resolution
• Toward global runs instead of continental tiles
• Revision of predictor variable to include more anthropogenic
factors
11. Livestock distribution:
Gridded Livestock of the World (GLW 3.0)
• In progress…
• New machine learning algoritm (Random Forest)
• Systematic evaluation (years of CPU time in 4 months)
• 180 models for Asia chicken and Africa cattle
• Processing on ILRI cluster (parrallelized)
• Full integration of metadata
• Spatial modelling & dissemination @ 1 km & 10 km resolution
• Toward global runs instead of continental tiles
• Revision of predictor variable to include more anthropogenic
factors
14. Outline
Context
• Intensification has taken place rapidly in the past
• Strong projected changes in demand will lead to further
intensification
• Changes are structured geographically
Objectives
• Better document the geographic distribution of intensive
livestock production
• Develop tools for making projections
Methods
• Mapping the global distribution of livestock
• Disaggregating in production systems
17. Conceptual framework (3)
The % ext. chicken is predicted at
national level by the GDP model
Ext. raised chicken are distributed
equally across rural population
Intensively raised poultry is estimated
by the difference with the total
21. Validation: chicken extensive
Van Boeckel, T., W. Thanapongtharm, T. Robinson, L. D’Aietti & M. Gilbert (2012). Predicting the distribution of intensive
poultry farming in Thailand. Agriculture, Ecosystem and Environment. Doi: 10.1016/j.agee.2011.12.019
22. Validation: chicken intensive
Van Boeckel, T., W. Thanapongtharm, T. Robinson, L. D’Aietti & M. Gilbert (2012). Predicting the distribution of intensive
poultry farming in Thailand. Agriculture, Ecosystem and Environment. Doi: 10.1016/j.agee.2011.12.019
28. Disaggregating between extensive and
intensive production systems
• Limitations
• Uncertainty in the GDP model (& other important variables ?)
• Ignore sub-national GDP variations
• Assumption of equal number of Ext. Chicken / rural population
29. Validation: chicken extensive
Van Boeckel, T., W. Thanapongtharm, T. Robinson, L. D’Aietti & M. Gilbert (2012). Predicting the distribution of intensive
poultry farming in Thailand. Agriculture, Ecosystem and Environment. Doi: 10.1016/j.agee.2011.12.019
30. Discussion (1)
People
• Number
• Weatlh
• Diet
Livestock
• Number
• Production
systems
Impact
• Amonia pollution
• GHG emissions
• EIDs
• Antimicrobial
resistance
31. Drivers of change in
spatial distribution
Drivers of change in
number
Demand
Discussion (2)
People
Livestock
Demography Wealth
# Consumers
Dietary
preferences
Urbanization of
consumers
Change in
stock
Change in
productivity
Urbanization
Vertical integration and
distribution of inputs
and demand
32. Future work (1)
Livestock
products
Change in
stock
Change in
productivity
Vertical integration and
concentration of
demand
•Methodological improvements
•Using agricultural population
•Using sub-national GDP where appropriate (e.g.
China, India)
•Forward and backward predictions
33. Future work (2)
2000
log GDP per capita c. $ 2.9
% extensive c. 83 %
2000
2030
2030
log GDP per capita c. $ 3.8
% extensive c. 18 %
Chicken production in
China
34. Future work (3)
Livestock
products
Change in
stock
Change in
productivity
Vertical integration and
concentration of
demand
•Methodological improvements
•Using agricultural population
•Using sub-national GDP where appropriate (e.g.
China, India)
•Forward and backward predictions
•GDP data & projections (national / sub-national)
•Spatial concentrations (peri-urban, access to
port, founder effect)
35. Future work (4)
People
• Number
• Weatlh
• Diet
Livestock
• Number
• Production
systems
Impact
• Amonia pollution
• GHG emissions
• EIDs
• Antimicrobial
resistance