Modeling AgricultureEnvironment Interactions - PowerPoint PPT Presentation

Title: Modeling AgricultureEnvironment Interactions


1
Modeling Agriculture-Environment
Interactions John M. Antle Department of Ag Econ
Econ Montana State University
Presented as the Distinguished Fellows Lecture at
the annual meetings of the Australian
Agricultural and Resource Economics Society,
February 2005
2
This presentation is based on work supported in
part by NSF, USAID, USDA, USEPA and the Montana
AES. Also, thanks to my colleagues who
contributed to various pieces of the work
presented here Susan Capalbo, Montana State
U. Charles Crissman, Int. Potato Center Bocar
Diagana, Montana State U. Kara Gray, Montana
State U. Sian Mooney, U. Wyoming Keith Paustian,
Colorado State U. Jetse Stoorvogel, Wageningen
U. Roberto Valdivia, Montana State U.
3
In 1990, Richard Just and I proposed a conceptual
framework for modeling agriculture-environment
interactions. This framework integrates
physical and economic models at a disaggregate
level necessary to capture the heterogeneity of
the physical environment and the economic
behavior of farmers. (Just and Antle, AEA
Proceedings, 1990, p. 197). My lecture is about
issues that arise in modeling agriculture-environm
ent interactions in a way that is useful to
support policy decision making.
4

• The paper contains 3 sections
• Designing Models to Support Informed Policy
  Decision Making
• A Spatially-Explicit Model of the Supply of
  Environmental Services
• Incorporating Important Features of Agricultural
  Production Systems into Econometric-Process
  Simulation Models
• Apologies to the many colleagues who have done
  similar work

5
Point of reference loosely-coupled agricultural
production systems models
6

• Designing Models to Support Informed Policy
  Decision Making
• How do you do it? Coordinated Disciplinary
  Research.
• A team concept define boundaries, units of
  measurement.
• Black-box approach counter productive.
• The TOA approach a process that links decision
  makers with scientific teams to define
  indicators, tradeoffs, scenarios.
• TOA software provides a modular approach to
  modeling.

7
Indicators, Tradeoffs and Scenarios
Soil Carbon
Is non-market valuation necessary or useful? Does
BCA or MCDA inform or obscure?
Conventional
Econ Returns
TOA on-line course at www.tradeoffs.nl
8
Tradeoff Analysis software
The Tradeoff Analysis software is a tool to model
agricultural production systems by integrating
spatial data and disciplinary simulation models.
It helps scientific teams to quantify and
visualize tradeoffs between key indicators under
alternative policy, technology and environmental
scenarios of interest to policy decision makers
and other stakeholders.
www.tradeoffs.nl
9

• Designing Models to Support Informed Policy
  Decision Making
• Timeliness of analysis minimum data methods and
  other strategies (IPCC) Need order-of-magnitude
  accuracy
• Credibility reproducibility and transparency.
• Need an open, modular approachnot large,
  complex, undocumented, proprietary models.
• Modeling representative individuals versus
  populations
• Need information about populations, often tails
  of the distributions are most important
  (vulnerability, poverty, food security, health
  and environmental risks)
• Heterogeneity is a key feature

10
The soil fertility puzzle distribution of
mineral fertilizer use in a sample of farms in
the Peanut Basin of Senegal. (Source Gray, 2005)
11
Mean versus coefficient of variation of net
returns by Montana sub-MLRA, for climate change
(CC) and CO2 fertilization scenarios with (A) and
without (N) adaptation. (Source Antle et al.,
Climatic Change, 2004).
12

• Designing Models to Support Informed Policy
  Decision Making
• Optimization, hypothesis testing and prediction
• Economists focus too much on optimization,
  estimation hypothesis testing as opposed to
  prediction.
• Example the literature on risk attitude
  estimation
• tests the hypothesis of RA using static f.o.c.
• does not show RA models predict better than RN
  models (identification problem)

13
Designing Models to Support Informed Policy
Decision Making Many important policy questions
involve out-of-sample prediction in heterogeneous
populations, in systems with unobserved
non-linearities, thresholds, and multiple
equilibria Example tech adoption literature
- representative, static risk models
typically usedbut models over-predict
adoption. - alternative explanations for
non-adoption include heterogeneity, thresholds
and multiple equilibria. Economic models imply
farmers do not operate near critical thresholds,
hence empirical models fail to capture key
non-linearities needed to predict out of
sample
14
Spatial Heterogeneity and Technology Adoption
Carbon rate (?C)
Net return (v)
Soil scientist should sequester C where ?C
highest Economist farmers would sequester C
where ?v/?C lowest
15
Thresholds and non-linearities
The effect of differences in the thickness of the
fertile A-horizon on the dry matter production of
potatoes as simulated with the DSSAT crop model
in the northern Andean region of Ecuador.
16
The temporal dynamics in carbofuran leaching in
the northern Andean region of Ecuador. (Source
Antle and Stoorvogel, Environment and Development
Economics, in press).
17
Impact of Transaction Costs on the Carbon Supply
Curve (Source Antle, Capalbo, Paustian and Ali,
2005).
18
A Spatially-Explicit Model of the Supply of
Environmental Services

• Simple model of supply of enviro services to
  illustrate logical basis of production models
  with spatial heterogeneity
• Key insight land-use decisions in a
  heterogeneous population are characterized by the
  joint spatial distribution of opportunity cost
  and environmental services

19
Figure1. Derivation of the Supply of
Environmental Services from the Spatial
Distribution of Opportunity Cost
20
Incorporating Important Features of Agricultural
Production Systems into Econometric-Process
Simulation Models

• Goal simulate spatial distribution of opp cost
  ?(?v, ?e ?)
• Approach Use econometric prodn models to build
  spatially-explicit economic simulation models
  that capture key features of the systems and
  farmer decision making
• Discrete land use (extensive margin) choice among
  multiple uses
• Discrete and continuous inputs use (extensive
  margin)

21
The structure of loosely-coupled agricultural
production systems models
EPM
22
Issues in development of econometric-process
simulation models

• Modeling heterogeneity
• Bio-Physical linking crop/livestock sim models
  to economic models to characterize spatial
  variation in productivity
• Decision maker spatial distribution of
  population characteristics
• Economic spatial price distributions

23
Issues in development of econometric-process
simulation models

• Dynamics
• Intra-seasonal, inter-seasonal
• Between sub-systems
• Bio-physical economic
• Crop livestock
• Methods for simulating dynamics
• Modular models
• Loose coupling (crop to econ to env)
• Loose coupling with feedback
• Close coupling (integrated, non-modular)
• Do benefits justify added complexity?
• Complexity Thresholds multiple equilibria
• Explaining technology adoption, non-adoption
  dis-adoption

24
Issues in development of econometric-process
simulation models

• Zero values for inputs and outputs
• Conventional model single, positive output,
  essential inputs
• Most actual systems have
• Discrete land use
• Multiple outputs in varying combinations over
  time, crop failure
• Non-essential, sequential inputs
• Need estimation methods suitable for simulation
• Efficiency vs feasibility
• Primal vs dual
• Minimum data methods to estimate spatial
  distribution of opportunity cost
• Systems of behavioral equations vs moment
  estimation
• Field scale vs aggregated data
• Matching model scale to decision scale

25
Comparison of Montana Soil C Supply Curves based
on EP and MD Models (Rho covariance of returns)
26
Comparison of Carbon Supply Curves for
Field-Scale and Aggregate EP Models (Montana)
27
Carbon Supply Curves Estimated with County-level
C Rates and Regional Mean C Rate (Source Antle,
Capalbo, Paustian and Ali, 2005).
28
Matching Model Scale to Decision Scale
Prediction errors in soil C sequestered with a
mean soil C rate for the corn-soy-feed system in
the Central United States with a carbon price of
50/MgC.
29
Concluding thoughts Policy-relevant science
will lead to more informed policy decisions that
must balance competing interests. Agriculture
is a complex system best understood through
coordinated disciplinary research.
30
This presentation and related information are
available at www.tradeoffs.montana.edu
www.tradeoffs.nl www.climate.montana.edu