Prioritization of Climate-Smart Agricultural Technologies at Local Scale
Methodology and Assessment
CCAFS 4th December 2013
Session Outline 1. Introductions 2. Biophysical model data requirements
– Q&A on data aspects
3. Demonstration tool overview – Mathematical Programming Toolkit – Model Overview
4. Tool exploration exercise 5. Comment: Upscaling these approaches 6. Discussion
– Are these tools relevant? – Challenges to uptake and implementation – Capacity building
• Simulation + what-if? analysis
– What would the farmers select?
– Select best from constrained option set
– If the farmers selected, what would be the outcome?
• Optimisation + do-what? Analysis
– What should the farmers be using?
– Search for and select best portfolio from large (potentially infinite) option set
– Manual OR Automated procedure (e.g. LP)
Optimisation
Why Mathematical Programming?
Simulation
Model Fundamentals (Classical) Classical toolkit of agricultural sector LP modelling tools
dating back over 60 years
• Activity selection for land-use planning + Technical coefficient generator
• Linearized market-price effects • Discounting and net-present value • Risk measures (e.g. TARGET-MOTAD) • Returns to capital investment • Interactive Multiple Goal Linear Programming
• Key Text: Hazel & Norton (1986) [IFPRI]
𝒙𝒄𝒓𝒐𝒑𝑙,𝑡,𝑐,𝑎,𝑓𝑠,𝑝,𝑘
Model Fundamentals (Extensions) Innovative modelling approaches
• Spatially-explicit crop-models, climate-forecasts and greenhouse-gas emissions calculators
• Dynamic optimization with technological investment, land-use change and technology uptake
• Stochastic-dynamic modelling to support planning with uncertainty in future climate – Minimax / Maximin / Low-Regret – Real Options analysis: Value the wait and see
• Multi-objective optimization and identification of the efficient frontier + gradient
• CPU++ Computational tractability ++ resolution
Land-Unit Constraints Land Availability Crop Suitability
Technological Suitability Farm-Size Technology Access Production Area Protection
State-Level Constraints Domestic Market Demand
Export Limits Rate of Land-Use Change
Development Targets
Spatial-Dynamic Land-Use Model (3) Multi-Scale Constraints
District Level Constraints Water Availability Labour Availability
Land-Units broken down further by rainfed/irrigated area and
farm-size categories
CSA Prioritization Toolkit Model Structure
Spatially-Explicit Bio-physical
Database
Model Input Database
Farm Size Breakdown
Target Demand Forecasts
Crop Nutrition
Prices and Elasticities
Labour Forecasts
Minimum data boundary
Constrained Production
Land available by area and type
Crop-water demand + irrigation available
Crop labour demand + population supply
Crop yields and
emission factors
Modular Model Code
Investment Cost Module
Markets Module + Growth Targets
Risk-Objectives (TARGET-MOTAD)
Spatial Allocation Constraints
Calibration Multi-Objective
Analysis Model Engine
COIN-OR CBC LP Solver
Post-Solve Output Analysis
Demo Tool Setup Note: Only tested on Excel 2010+ versions
1. Place CSA Priotization Demo_v1.0 in desired model folder
2. Unload contents of folder OpenSolver21 into same model folder
3. Open blank Excel workbook
4. Double click OR drag OpenSolver.xlam add-in file into open workbook – This should load the Opensolver menu under Data tab
5. Activate the default Excel solver add-in – Goto File-Options-Add-Ins – Select Manage “Excel Add-Ins” and click Go – Activate the Solver Add-In
6. Open CSA Prioritization Demo_v1.0
See: http://opensolver.org/
Efficient Frontier Cannot improve in one objective
without sacrificing another
Optimal Space
Run 2: Min Emission
= Min SSR Run 1: Max SSR = Max Emissions
Tradeoff Analysis: Overview Priority Means AND Ends
Running Tradeoff Analysis
1. Run model to optimize primary objective – Suggested: Maximize production or margin
2. In sheet <Variables> record the current objective levels (Cells E17:E21)
3. Select tradeoff objective and specify a desired bound level <Variable> (Cells H17:H21) – Example: Record production max level of CO2,eq and
set bound at 80% of that level
4. Re-run the model for primary objective - now under additional constraint
Upscaling Tool to Project
Resources required: • Minimum data specification
• Algebraic Programming Language – Algebraic Modelling Systems, Modeling and Solving Real World Optimization Problems,
Josef Kallrath (Ed.) (2012)
• Computational tools (NEOS, Kestrel, CPLEX Studio, Solver Studio etc.,) – http://solverstudio.org/
– http://www.neos-server.org/neos/
• Modelling programme management – Quality Assurance (QA)
• Analytically literate policy audience – Structured policy engagement + facilitation
Discussion Points
1. Do people see promise in this approach to support prioritization of climate-smart investment?
2. What do people envisage as the challenges to implementing these approaches more widely?
3. If needed what do people and institutions need to take this approach forward? (Tools? Programming skills? Data?)