Assessment of Multimodel Ensemble Seasonal Hindcasts for Satellite-Based Rice Yield Prediction

• Published in: Journal of Agricultural Meteorology Vol. 72; no. 3-4; pp. 107 - 115
• Main Authors: CHUN, Jong Ahn, KIM, Sung, LEE, Woo-Seop, OH, Sang Myeong, LEE, Hyojin
• Format: Journal Article
• Language: English
• Published: The Society of Agricultural Meteorology of Japan 2016
• Subjects: MME (Multimodel Ensemble); MODIS NDVI; Rice Yield; Statistical Downscaling
• ISSN: 0021-8588; 1881-0136
• DOI: 10.2480/agrmet.D-15-00019

Several pre-harvest rice yield estimation methods have often failed to accurately estimate rice yields due to weather variability. We attempted to assess the APEC Climate Center Multimodel Ensemble (APCC MME) seasonal hindcasts to a satellite-based rice yield prediction model to timely provide estimates of rice yields for efficacious intervention plans. The developed model by a multiple regression analysis is Yield = 5.635NDVI - 0.0012P9 + 0.91 (where yield is the white rice yield in t ha-1 and P9 is the observed monthly precipitation in September in mm month-1). The goodness-of-fit measures were 0.66, -0.14%, 0.13 t ha-1, and 2.25%, for adjusted R2 (coefficient of determination), Percent bias (PBIAS), Root Mean Square Error (RMSE), and Mean Absolute percentage Error (MAPE), respectively. A statistical downscaling method using Empirical Orthogonal Function Analysis (EOFA) and Singular Value Decomposition Analysis (SDVA) was used to predict monthly precipitation hindcasts in September required for the developed model. Even though the estimates of rice yield using the predicted monthly precipitation for whole study period were not as good as the estimates using the 9.15 sampling method, the estimates for the two years of 2008 and 2009, when the 9.15 sampling method largely underestimated, were better than those using the 9.15 sampling method. It is concluded that the proposed approach can be used to timely provide rice yield estimates that reflect the meteorological conditions for more effective intervention plans in the rice market.