final_program_abstracts[1]

11 IMSC Session Program
Expressing uncertainty in climate analysis – A pilot of climate
impacts on Panamanian Maize
Wednesday - Plenary Session 5
L. DeWayne Cecil 1 , Alex C. Ruane 2 , Radley M. Horton 2 , Cynthia Rosenzweig 2 ,
Peter A. Parker 3 , Brian D. Killough 3 , Ray McCollum 4 and Doug Brown 4
1 U.S. Geological Survey, Idaho Falls, USA
2 NASA Goddard Institute for Space Studies, New York, USA
3 NASA Langley Research Center, Hampton, USA
4 Booz, Allen, and Hamilton, Hampton, USA
Decision makers often sift through multiple analytical model results in an effort to
determine the best course of action. In the area of climate analysis, the problem is
complicated with uncertainty in the data measurements and input provided from an
array of Global Climate Models (GCMs) that each show different results. The
uncertainty from model to model is often a key consideration for decision makers and
provides a gauge of reliability in the simulated projections.
The objective of this presentation is to create a standard process to analyze and
express uncertainty in climate impacts analysis such that decision makers can better
understand the role it plays in decision support.
The project analyzed uncertainty involved in the analysis of 5 GCMs used to produce
scenarios for agricultural decision support on a climate change time scale (the 2030s).
To ensure that the results were relevant, the model was based on a location in Panama
with an extensive climatological record. Data were analyzed and displayed to provide
policy makers with the means to mitigate and adapt to projected climate changes in
this challenging region.
To date, the GISS team has constructed climate change model projections for the
uncertainty analysis. Probability functions were derived from the climate model data.
Cumulative density functions were derived for Maize yield and conditional
probability functions were derived for hot and cold climate conditions. The
conditional functions helped explain the impact of temperatures rising above, or
falling below, specified levels. All answers were specified in terms relevant to a
decision maker so that one clear picture of the future could be presented to the
decision maker based on the inputs from multiple models.
The outcome of the project has demonstrated that variability from model to model
was insignificant compared to the variability in the predicted climate patterns at the
Panamanian location. Model-to-model variability was actually much smaller and all
models give similar results when compared to overall prediction uncertainties.
The next step in the process is to classify and express uncertainty throughout the
climate prediction process. Preliminary analysis of the data has shown that CO2 level
variation plays a greater role in the sensitivity of the existing maize yield model than
anticipated. This revelation and other study findings point to the importance of
understanding and classifying uncertainty throughout the process of climate
prediction.
Abstracts 165