The main purpose of this study is to build a crop yield forecasting model for rice, soybean, and the summer Chinese cabbage cultivated in highlands with meteorological elements taken into account. The difference compared with the previous studies and the main outcome of this study are as follows: first, this study used a stochastic method to overcome the drawback of point estimation by using meteorological scenarios; second, model specification was changed from a linear function to a quadratic function form (such as the concave function to the zero point) to find optimal points on each element; third, a panel data analysis was used to enhance the degree of freedom.
The panel analysis was used with a two-way fixed effect model considering cross-section and time period to find unbiased and consistent estimates of meteorological elements.
Based on the analysis results, this study developed a stochastical crop yield forecasting model with a variety of meteorological scenarios. In addition, the possibility of introducing an EPIC(Erosion Productivity Impact Calculator) model is reviewed in order to overcome the limitations of our model.
An estimation of rice yield was made using a regional panel data of over the recent 10 years and with mean temperature, effective cumulative temperature, sunshine hours, and daily temperature range as independent variables. The results using a fixed effects model confirm that the average temperature is a quadratic form and that the rice yield is highly affected by effective cumulative temperature before the grain filling stage and by the mean temperature during the grain filling stage.
We set up a total of 1,296 scenarios at the end of September in 2011 to forecast rice yield based on weather data provided by Korea Meteorological Administration. Reducing the scenarios and replicating the estimation as the forecast time goes close to the target time, which was the end of September in 2011, we finally obtained the result that the target rice yield would be 496kg based on our scenario in the middle of September in 2011. Our forecast is not different from the real rice yield of the target time, 497kg, announced by the National Statistical Office through an actual inspection.
An estimation of yield for the summer Chinese cabbage was made using a main producing district's panel data of over the recent 10 years and with high temperature and rainfall as independent variables. The results using a fixed effects model confirm that both high temperature and rainfall take a quadratic form. The result of the yield forecasting model of summer Chinese cabbage shows that the yield is higher than 2010's but less than average year.
A soybean yield estimation was made using a regional panel data from 1980 to 2010, and except for Jeju, and with average temperature, sunshine duration, and rainfall as independent variables. The results using a fixed effects model confirm that both average temperature and rainfall take a quadratic form. This model used provincial weather data because we could not access the yield data of cities. So we can not determine the relationship between weather component and soybean yield.
The yield forecasting model of rice, soybean, and summer Chinese cabbage cultivated in highlands has an implication that we can know their yield before the National Statistical Office announces its researched yield results. Therefore, we can use yield forecasts as basis data to prepare measures for any imbalance in supply and demand.
The limitation of this study is that it did not consider the damage caused by diseases and insects and climate change because we don't have experiment data on yields of rice, soybean, and the summer Chinese cabbage cultivated in highlands. Therefore, a study for obtaining the experimental data has to be conducted by the Rural Development Administration beforehand. In the future study we need to merge EPIC models and the statistic yield forecasting model of this study to enhance the accuracy of crop yield forecasting.


Researchers: Sukho Han, Byounghoon Lee, Misung Park, Junho Seung, Hyunseok Yang, Sungchul Shin
Research period: 2011. 7. - 2011. 12.
E-mail address: shohan@krei.re.kr