Extended Abstract (564K)P3.6
Forecasting cotton yields over the southeastern US using NCEP Climate Forecast System
Guillermo A. Baigorria, University of Florida, Gainesville, FL; and M. Chelliah, C. C. Romero, K. Mo, J. W. Jones, J. J. O'Brien, and R. W. Higgins
Making monthly and seasonal climate forecasts potentially useful to farmers and agriculturists remains a major challenge. Beyond their experience, intuition and climatological information, farmers can take advantage of critical seasonal climate forecasts and information that may actually help them with their crop management (i.e. planting, control of pest and diseases). The partnership between the Climate Prediction Center and the Southeast Climate Consortium took the challenge made by the Climate Test Bed to use the improved NOAA climate forecast products to enhance decision-making of farmers and agriculturists. Previous research indicated some potential of predicting cotton yields over the southeastern United States based on regional atmospheric circulation in the upper troposphere and surface temperatures. In this study, de-trended cotton yields from 57 counties in Alabama and Georgia (1987-2006) were hindcasted using June 0.5-3.5 (July-August-September) retrospective forecasts of 2m temperatures from the NOAA/NCEP/Climate Forecast System (CFS). NOAA/NCAR/Reanalysis (Reanalysis) from 1976 to 2006 (July-August-September) were used to assess the potential predictability of cotton yields using 2m temperatures and also as training periods during the retroactive validation. The first principal component of Reanalysis of 2m temperatures in the geographical domain of the southeastern USA, corrected by observed total rainfall amounts from 62 weather stations during the cotton's growing phenological phase (April-May-June), significantly predicted cotton yields in 31 counties. The same principal component but replacing Reanalysis by CFS, significantly hindcasted cotton yields in 48% of those counties predicted using Reanalysis. Using a retroactive validation, Goodness-of-fit indices were 0.5274 and 0.5247 respectively for the significantly predicted and hindcasted counties. Root mean square errors of the significantly hindcasted counties ranged from 13.6% to 28.2%. It is concluded that cotton yields in some areas of the southeastern USA can be forecasted three months before harvesting using the first principal component of the CFS' forecasted 2m temperatures corrected by observed rainfall during the first 3-month cotton growing phenological phase.
Supplementary URL: http://plaza.ufl.edu/gbaigorr/GB/GBPublications.htm
Poster Session 3, Climate Model and Prediction Poster Session
Wednesday, 23 January 2008, 2:30 PM-4:00 PM, Exhibit Hall B
Previous paper Next paper