7.8 Value based Ensemble Analysis and Potential Individual versus Simultaneous Climate Change Effects on Agriculture

Thursday, 23 June 2005: 9:45 AM
North & Center Ballroom (Hilton DeSoto)
Robert Mera, NCSU, Raleigh, NC; and D. Niyogi and F. Semazzi

In this study we evaluate and apply the NCAR Community Atmosphere Model, using an the economic value (V) of the probabilistic NCAR GCM seasonal 'forecasts' over a tropical belt and then a midlatitudinal region. For the tropical case, we use CMAP for 18 equal-area tropical regions spanning the equatorial belt. The assessment was based on the ability of the NCAR GCM in predicting the observed 'above normal' rainfall terciles. For the midlatitudes, we have selected use of climate information for agriculture as a potential end-user with specific focus on soybean and maize crops in North Carolina with a DSSAT: Decision Support System for Agrotechnology Transfer CROPGRO (soy) and CERES-Maize (maize) models. We varied the climate input to the DSSAT by: (i) systematic variations in the radiation, temperature and precipitation amounts; and (ii) using ensemble simulation output from the GCM for the three variables. For the tropics, we find that results are relatively insensitive to reasonable uncertainties in defining the end points of a tercile, but are sensitivity to the geographical location of the geographical. We derived a mathematical condition for the lower and upper limits of (C/L) that ensure positive end-user useable skill of the model-based prediction system. A unanticipated result is the dramatic variation in the performance among the ensemble members of the GCM. The analysis of the sensitivity on geographical region in the tropics shows that the Western-Central Pacific region has the greatest model prediction skill. The other regions with high skill are Central Indian Ocean, Atlantic Ocean, and South America. The three regions with lowest predictability are the maritime continental region centered over Papua New Guinea, the eastern region of the Indian Ocean, and the Amazonian. For the agricultural application, we found that, crop yield is most sensitive to precipitation changes, radiation impact is non-linear, and temperature has limited impact and the response is non-linear. The response varies by the crop (soybean or maize i.e. photosynthesis pathway), and the response studied from previous one-at-a time sensitivity study may be quite different than that seen due to simultaneous changes in the climate forcing. A ensemble input based DSSAT analysis is currently underway and results will be discussed. We have also proposed a modified and more stringent condition based on the ROC for evaluating the skill of a prediction system. In addition to the false alarm and hit rate information it also incorporates end-user economic parameters.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner