In recent years, climate change phenomena caused by global warming have been emerging in various forms. Due to the impacts of climate change, abnormal weather phenomena, such as droughts, floods, typhoons, and heavy snow, are becoming stronger. The frequent occurrence of these abnormal weather phenomena will act as a major threat. For example, it could cause a decrease in major food crop productivity and quality in agricultural crop producing areas. Reduction of crop production caused by abnormal weather in major crop producing areas is acting as a stressor on the increase of international grain prices. Other factors in the increasing grain price include increase of food consumption; high transportation costs; usage for bioenergy; and intervention of speculative funds into grain markets. The sudden change in grain price tends to cause a food crisis and can sometimes ruin governments of grain importing countries that heavily rely on grain imports. In order to ensure a stable and adequate grain supply from major export countries in a precarious international grain market, the domestic feed and food industries in Korea are demanding reliable and frequent information on the trends in the international grain market from the government and research institutes. The Korea Rural Economic Institute (KREI) has built an international grain observation system and publishes a monthly observation report on the international grain trend to help ensure stable grain supplies to livestock farmers and the food industry. However, the reports do not consider meteorological phenomena such as droughts, floods, and cold weather between yield prediction time and crop harvest time, that directly affect the crop yield.
The APEC Climate Center (APCC) produces and offers multi-model ensemble (MME) seasonal forecasts by collecting long-term climate prediction data produced by 16 research institutions in 10 APEC member countries. However, the application of this long-term climate prediction data in the agriculture and water resources sectors is still very low. In order to improve the reliability of the prediction results, APCC performs verification of the forecast data every month. 3-month and 6-month lead time forecast data provided by APCC are generated in 2.5°×2.5° spatial units and monthly time units. In order to take advantage of the long-term grain yield forecast, it is necessary to develop the temporal and spatial refinement (downscaling) technique. Different climate models produce different predictions due to the systematic error of each model in the long-term climate prediction data. In order to reduce the difference in predictions, we developed the Simple Bias Correction (SBC) method for the long-term climate prediction data. We applied SBC to the Hindcast (1983-2006) and Forecast (2013-2015) prediction periods.
Daily weather data is necessary in order to simulate crop yields in to the crop model. Since APCC currently offers long-term climate prediction data in a monthly mean data format, a temporal downscaling process is needed to offer data in a daily mean data format. First, we selected the top 5 years from NCEP/NCAR observed monthly temperature data (1951-2014) that are the closest to the forecast monthly temperature data. Next, we selected top year from NCEP/NCAR observed monthly precipitation data (top 5 years) that is the closest to the forecast monthly precipitation data. Forecast daily data were resampled from the selected year’s daily data. We performed Temporal Correlation Coefficient (TCC) and Root Mean Square Error (RMSE) analysis to estimate correlation and accuracy of the observed and predicted values of the climate prediction model. As a result, predictability for the temperature was higher in Mexico, a region in a low-latitude area, than in the United States and China. The Modified Global Agro-ecological Zones (M-GAEZ) model is an estimation model for potential global crop yields. The potential crop yields are calculated based on conditions, such as the climate, soil, and input levels. The M-GAEZ model can estimate the potential yield for 19 varieties of maize. Some varieties of maize differ in growth periods depending on the cultivated area. Daily climate data is used in the M-GAEZ model at a spatial resolution of 1º×1º latitude/longitude. In order to evaluate the reliability of the maize yield prediction, we predicted the yield according to the country, using Hindcast climate prediction data (1991-2000) from six climate prediction models (MSC_CANCM3, MSC_CANCM4, NASA, NCEP, PSNU, POAMA) as inputs to the M-GAEZ model. In the United States and China, the TCC values of the climate prediction models were distributed mainly around 0.4; in Mexico, the TCC values were distributed between 0.4-0.8; and in Brazil, the TCC values were distributed between 0.0-0.6. Maize yield predictability in Mexico was the best in four countries. We also carried out the maize yield prediction using the forecast climate prediction data (2013-2015), according to country.
