Exploring Uncertainties Associated with Scaling Crop Systems Modeling Results from Point to Region KEMIC TEAM Job Kihara (Soil scientist) Jawoo Koo (Crop modeling) Jonathan Hickman (Crop modeling) Charles Vanya (Climatologist) Dilys MacCarthy (Soil scientist) Julius Mangisoni (Economist) Edward Yeboah (Soil scientist) P2R Spontaneous Inception Workshop - H A L F B A K E D -
16
Embed
Exploring Uncertainties Associated with Scaling Crop Systems Modeling Results from Point to Region KEMIC TEAM Job Kihara (Soil scientist) Jawoo Koo (Crop.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Exploring Uncertainties Associated with Scaling Crop Systems Modeling Results from
Point to Region
KEMIC TEAMJob Kihara (Soil scientist)
Jawoo Koo (Crop modeling)
Jonathan Hickman (Crop modeling)
Charles Vanya (Climatologist)
Dilys MacCarthy (Soil scientist)
Julius Mangisoni (Economist)
Edward Yeboah (Soil scientist)
P2RSpontaneousInception Workshop
- H A L F B A K E D -
AFRICA IS LARGER THAN U.S., EUROPE, CHINA, INDIA… COMBINED
Background Africa is big; points are small. And, we do not
have, and won’t have, the complete picture. Yield estimates being made at sentinel sites
(points) need to be aggregated to provide regional/global-scale input data to the economic models.
Scaling-up options are available (or being developed). Many choices and assumptions need to be made; their uncertainties and consequences are not well known.
Research Questions
1. How much uncertainties are we introducing to the point-to-region aggregates, depending on:– Where we simulate (sentinel sites vs. grids, or both)– Choices of soil, climate, and management– How we simulate crop productivity– How we aggregate
2. What are the best options for the reasonable representation of mean and variance in aggregates of point-based estimates?
Objectives1. To create an independent false “Truth” maize productivity
data on 10 km grids for 5-year period
2. To create various point-to-region aggregates generated from using:A. Selected points or uniform gridsB. Choices of aggregation methodsC. Choices of model input dataD. Assumptions of management practices
3. To compare their uncertainties by comparing with the aggregated false “Truth” data.
4. Develop a grid-based crop modeling framework that can be used to test/develop adaptation scenarios for future climate.
Methodology
Assumptions
• 10 km grids adequately represent local variability of soil, climate, and management
Study Area
• Malawi
Generation of grid-level “Truth” of maize productivity on 10 km grids
CERES-Maize + CENTURY District-level production statistics for 2000-2005 Spatial Production Allocation Model
(area/production/yield; disaggregated production statistics on 10 km grids; four levels of input systems)
Gridded soil profile database from HarvestChoice (HWSD + WISE; 10 km grids)
Global fertilizer rate database (60 km grids) Irrigation extent Random noise (to take into account model errors)
1
Calibration of locally used maize varieties for four sentinel sites
AfSIS Diagnostic Trials Millennium Village Project Trials SIMLESA Project
2
Grid-based characterization of maize production systems in the region
For each grid cell, for each of four input systems :Variety choiceSeasonality + Rainfed/IrrigatedUse of fertilizer:inorganic and organicSoil fertility(SOM fractions)…
3
Grid-level DSSAT-based maize yield estimates from various cases, such as…
For 2000-2005:
1.Source of soil data
2.Source of climate data3.Assumption of soil fertilityTSBF + HarvestChoice + MVP
4.Seasonality Rainfed-only Plus, (hidden) irrigated 2nd season
5.Fertilizer application rate
6.…
4
Aggregate the point-level data (gridded outputs) various ways
1. Four sentinel sites DistrictAgMIP protocol; bias correction and match to statistics
2. All sites District
3. …
5
Notes1. This study will only focus on crop modeling (no TOA) with one
crop model (no APSIM) on current climate (no CC).2. Outstanding needs
Full understanding of AgMIP Aggregation Protocol Looking at reality; we’re ambitious (yes) We do not have budget for this; will need to explore sources to bring
members together.
3. Plan Sep: Straight-out workplan Oct: Present at the Rome meeting, seeking feedback and possible
contribution to the global-scale aggregation team Nov: Get all the data ready Feb: First round of results ready for review Apr: Finalize the study Oct: Publication (TBD)