The Institute for Agro-Environmental Sciences, NARO (NIAES), in collaboration with the APEC Climate Center (APCC), has developed a new method to forecast the variations in crop yields globally. The yields of maize, soybean, rice, and wheat were predicted using the seasonal forecasts created by five weather agencies. The forecasts were performed three months before harvest in about 1/3 of the global harvesting area thereby reliably capturing the variations in yield relative to the previous year. For about 1/4 of global crop producing countries including Japan's import destination like Australia (wheat), it is now possible to forecast country-level yield variations. A global crop yield forecasting service using this method is being developed for international organizations, and APCC and NARO jointly plan a test experiment for a 2-year period until 2020.
Overview
In recent years, there has been increase in the import volume of grain in many countries. The rise in the international market prices due to the crop failure in exporting countries has become a major risk in securing food availability for importing countries, especially in the developing regions. In response to the deterioration of the food situation caused by abnormal weather, and to promptly deal with the increase of domestic stockpile volume, it is important to forecast the grain yield not only in own country but also in other countries. Hence global crop yield forecast is demanded by food agencies in food importing countries and international organizations.
NARO, in collaboration with the APEC Climate Center (APCC), has developed a new method to forecast the global crop yield variations using seasonal forecast data of temperature and precipitation created by five weather agencies in the Asia-Pacific region. Yield fluctuations from weather conditions is relatively easier to capture than yield in absolute terms. In the new method, the amount of seasonal forecast data used is 5 times larger than the earlier method, and hence improves prediction skill. Hence, in addition to wheat and rice, maize and soybean can also be predicted, and the area where yield fluctuation can be forecasted three months before the harvesting has increased from about 30% to 40% of the world's harvest area. Also, the area of yield prediction for rice has increased by 2.5 times. Here, prediction skill is evaluated by comparing forecast value and actual yield variation over the past 27 years (1984 - 2010).
Unless forecasts for a large spatial domain are reliable, it is difficult to predict the country-level yield variation. But with the new method, it is now possible to predict the yield of all four crops in about a quarter of global crop producing countries, including some major exporters. The food supply-demand forecast released by food agencies such as U.S. Department of Agriculture uses data by country but does not consider the influence of abnormal weather. Hence global yield forecast by this method is expected to complement these existing prospects.
By this result, forecast of yield variation for major crops has become technically possible. NARO is aiming to develop a global crop yield forecasting service for international organizations such as FAO, to achieve the actual operation at APCC, and a test operation is scheduled in 2019-2020.
Publication
Toshichika Iizumi, Yonghee Shin, Wonsik Kim, Moosup Kim and Jaewon Choi (2018) Global crop yield forecasting using seasonal climate information from a multi-model ensemble. Climate Services https://doi.org/10.1016/j.cliser.2018.06.003.
For Inquiries
Contact: http://www.naro.go.jp/english/inquiry/index.html
Reference Information
Fig. 1. Conceptual diagram of yield variation forecast based on seasonal forecast data
APCC obtains seasonal forecast data every month for the following three months from the weather agencies of APEC member economy (blue arrow). Crop yield variation is forecasted for areas and crops where the three month span of temperature and precipitation forecast period coincides with the three month span just before harvesting (within-season prediction). In order to secure the time (lead time) to respond users to the predicted yield variation, within-season prediction is carried on the month before the beginning of the 3 months until harvesting.
Fig. 2. Regions where yield variation could be predicted 3 months before harvesting
The forecast data of temperature and precipitation prepared by APCC three months before harvest were fed into the yield prediction model, the actual yield fluctuation region could be predicted (orange color) and unpredictable region(black color) can be found. Yield fluctuation prediction was done every 120 km mesh. In the evaluation by mesh, crop yield fluctuation can be predicted where predictive skill is significantly higher than the skill for 27 years (1984-2010) which is not based on seasonal forecast data (randomly chosen actual values of yield fluctuations that occurred in the past and are regarded as predicted values).
Fig. 3. Countries where variations in national average yield could be predicted
three months before harvesting
As a result of aggregating the result of yield prediction for every 120 km mesh to the country level as shown in Fig. 2, country where the fluctuation of the actual national average yield could be predicted (orange) and the country which was unpredictable (black) can be found. In the evaluation by country , crop yield fluctuation can be predicted where the predictive skill is significantly higher than the forecasts for 32 years (1984-2015) which is not based on seasonal forecast data (randomly chosen actual values of yield fluctuations that occurred in the past and are regarded as predicted values).
Fig. 4. Variations in average yield in major producer countries predicted
three months before harvesting
Actual yield fluctuation (calculated from FAO statistics, black) and prediction at 3 months before harvesting (orange) for some representative countries.
Fig. 5. Share of maize, soybean, rice, and wheat exports per country in 2016
Data from statistical databases of the Food and Agriculture Organization of the United Nations (FAOSTAT).
