연구보고서
- 저자
- 신용희 박사
- 작성일
- 2016.01.23
- 조회
- 243
- 요약
- 목차
Increased occurrence of abnormal weather caused by climate change has been acting as a big menace especially in the agricultural sector. Crop yields mainly under the influence of agroclimatic factors (e.g., temperature, precipitation and insolation) and technological factors (e.g., improvement of breed, agrichemicals, fertilizer and irrigation system). Reduction of damage scale and frequency of production in the major crops production area due to the recent abnormal weather has emerged as a major issue in the international community. The serious reduction damage of crop production to occur in the major crops production area will lead to regional food supply imbalance led to soaring international grain price. South Korea leans on imports for most of the consuming crop and grain self-sufficiency rate has recorded a record low of 23.1% in 2013. In order to provide a stable supply of grain from abroad in pure grain importing countries, such as South Korea, high accuracy crop monitoring and crop yield forecasting system for major grain exporting countries are required. Recently, Korea Rural Economic Institute (KREI) has built an international grain observation system to help ensure stable grain to livestock farmers and the food industry. KREI began to offer a supply and demand trends and forecast information for major crops. However, since there is no consideration of the meteorological phenomena such as drought, flood, cold weather from prediction time to crop harvest time that directly affect the crop yield, there are considerable difficulties exist in policy decisions regarding the grain supply.
APEC Climate Center (APCC) is producing and offering a multi-model ensemble (MME) seasonal forecasts that is evaluated the predictive performance of world-class, but the utilization of climate prediction information for applications such as agriculture and water resources is still very low. To assess the reliability of hindcast prediction data, comparison of 6 months hindcast data and B.C. NCEP reanalysis data were made during 24 years between 1983 and 2006 for average temperature of July September in main rice growing periods. As a result of having analyzed the temperature prediction by prediction model, the uncertainty range in accordance with the prediction model appeared quite large. Uncertainty range is reached in the case of the USA about 6.3°C. In order to take advantage of seasonal climate forecast directly in agricultural applications, bias correction process is essential. Simple bias correction method has been used for about 6 months Hindcast bias correction to take advantage of the seasonal forecasts in agriculture production research. The reliability of the prediction was evaluated through the analysis of RMSE and TCC for biascorrected temperature forecasts by country. In the case of China, reliability of predictions about the average temperature of July to September is relatively higher in nasa and ncep model. In the case of India, reliability of predictions is relatively higher in can3 model. For crop yield simulation, temperature, precipitation, diurnal temperature, solar radiation and wind speed data that are required by the crop model were reproduced to daily unit and reliability of prediction was evaluated. Results of reliability evaluation of reproduced daily prediction data, reliability for the average temperature in September were relatively high in China. RMSE and TCC analysis results are about 0.4 °C and 0.7 respectively in can3, nasa, ncep model. However, since the reliabilities of other climate were considerably lower than temperature this problem must be solved before applying to application study.
The M-GAEZ model was modified by National Institute for Environmental Study (NIES) based on the Global Agro-ecological Zones (GAEZ) model that is an estimation model for potential crop yields on a global scale developed by the Food and Agriculture Organization (FAO) and the International Institute for Applied Systems Analysis (IIASA). The potential crop yields are calculated based on conditions, such as the climate, soil, and input level. Specific climate conditions, such as the daily mean temperature [°C], daily precipitation [mm/ day], daily mean irradiation [W/m²], and daily mean windspeed [m/s], are used as the input information. The M-GAEZ model can estimate the potential yield for 8 varieties of rice. Some varieties of rice differ in growth periods depending on the cultivated area; 4 varieties (growing periods of 105, 120, 135 and 150 days) for Japonica, 4 varieties (105, 120, 135 and 150 days) for Indica are treated distinctively. M-GAEZ model requires the daily climate scenarios at a spatial resolution of 1º×1º latitude/longitude for each climate conditions as the input data. We created every mesh of a daily climate data using APCC seasonal climate predictions.
Rice yield was calculated by M-GAEZ model inputting the reproduced daily prediction data using bias corrected seasonal forecast data and NCEP reanalysis data and reproduced daily prediction data through anomaly inflation. In the cast of China with the most rice amount of production, results of rice yield in the 1990s calculated from B.C. NCEP reanalysis data substantially similar to FAOSTAT agricultural statistics. However, yield prediction results showed significant uncertainty depending on the prediction model. As a result of having carried out RMSE and TCC analysis for the reliability of rice yield, RMSE is generally large to 200-400 kg/ha and TCC appeared between 0.6-0.8 highly. In the case of India, RMSE is less than China to 80-140 kg/ha and TCC showed generally low correlation between 0.1 to 0.6. In the case of South Korea and the USA, RMSE are significantly large to 400-550 kg/ha, 400-700 kg/ha respectively and TCC showed significantly low correlation to –0.2-0.2 and 0.0-0.3 respectively.