JAMSTEC, 24 Jan 2019
In a collaborative effort, Takeshi Doi of the Application Laboratory of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) and his colleagues have developed a system capable of accurately predicting the occurrence of extremely strong El Niño/Southern Oscillation (ENSO; *1) and Indian Ocean Dipole phenomena (*2) several months in advance. When these phenomena occur, they produce abnormal seasonal weather (e.g., heat waves and warm winters) around the world, including in Japan. Therefore, extremely strong occurrences of these phenomena merit close monitoring, and the generation of accurate predictions as to when they will occur is of extreme importance from both societal and economic standpoints.
An integrated atmosphere-ocean-land-sea ice systemic circulation model was used by JAMSTEC Application Lab scientists as the basis for developing the SINTEX-F Seasonal Climate Prediction System (*3). The system was designed for performing predictive research on the occurrence of seasonal abnormalities, ENSO phenomena, and other such events that will occur several months in the future. It has traditionally been difficult to understand and predict the scale of these events; despite this fact, the SINTEX-F Seasonal Climate Prediction System was developed and the number of parallel worlds used in its predictive simulations was increased from approximately 10 to approximately 100. As a result, though their frequency of occurrence was rare, the system succeeded in accurately predicting the occurrences of extremely strong ENSO (Fig. 1) and Indian Ocean Dipole phenomena (Fig. 2) several months in advance. Additionally, when the system’s accuracy in predicting extreme drought events across the globe nearly six months in advance was examined, the results showed that the system improved the accuracy of predictions (Fig. 3). This is the first experiment yet to have examined the accuracy of a seasonal climate prediction system based on a single climate model and featuring the generation of approximately 100 parallel worlds for past climatic events.
The vast calculations required in this study were made possible by the free use of the functions of a global simulator. The knowledge gained by these simulations is expected to lead to the strategic development of numerous seasonal climate prediction systems in the future. Our goals are for this success to spark the comprehensive analysis of prediction results featuring numerous parallel worlds, the discovery of new processes and the identification of new prediction signals, the development of climate prediction research and expansion of applied research based on predictive information for extreme climatic events (e.g., crop and infectious disease forecasts), and to make specific contributions to societal initiatives aimed at ensuring the safety and security of the public.
The above results were published in the Journal of Climate issued by American Meteorological Society on January 22, 2019 (JST).
Title: Merits of a 108-member ensemble system in ENSO and IOD predictions
Authors: Takeshi Doi1, Swadhin K. Behera1, and Toshio Yamagata1
1. Application Laboratory, JAMSTEC
*1 El Niño/Southern Oscillation (ENSO Phenomenon): El Niño is a climate variation phenomenon observed in the tropical Pacific Ocean that occurs once every several years. When an El Niño event occurs, sea surface temperatures are higher than during an average year in the eastern-central tropical Pacific and lower than in an average year in the west. These changes in ocean temperatures cause the vigorous convection that usually occurs in the western part of the tropical Pacific to move eastward, resulting in warmer temperatures and less rain in Indonesia and northern South America than in an average year. Additionally, the atmospheric anomalies in the tropics tend to produce cooler summers and warmer winters in Japan. When focusing on the ocean, such events are called “El Niño;” when focusing on the atmosphere, they are called “Southern Oscillation.” However, the two phenomena are closely connected; collectively, they are called the “El Niño/Southern Oscillation,” or the “ENSO phenomenon.” Meanwhile, the “La Niña phenomenon” is considered to be the reverse of El Niño, wherein sea surface temperatures in the western tropical Pacific are warmer than in an average year and they are cooler than in an average year in the eastern-central tropical Pacific. Such variations in ocean temperatures cause convection in the western tropical Pacific to be more intense, and Indonesia consequently receives more rain than in an average year. These atmospheric changes from the tropics also tend to produce extremely hot summers and colder winters in Japan.
*2 Indian Ocean Dipole: The Indian Ocean Dipole is a climate variation phenomenon that is observed in the tropical Indian Ocean once every several years from summer to autumn. This phenomenon has both positive and negative phases. When a positive Indian Ocean Dipole occurs, sea surface temperatures become cooler than in an average year in the southeastern tropical Indian Ocean and warmer than in an average year in the west. These variations in ocean temperatures cause the vigorous convection that usually occurs in the eastern Indian Ocean to move westward, and East Africa receives more rain, while Indonesia receives less. These variations also tend to result in less rain and higher temperatures in Japan. Conversely, when a negative Indian Ocean Dipole occurs, sea surface temperatures are warmer than in an average year in the southeastern tropical Indian Ocean and colder than in an average year in the west, causing convection in the eastern Indian Ocean to be more intense than usual. During such events more rain falls in Indonesia and Australia, and there is generally more rain and temperatures are lower in Japan.
*3 SINTEX-F Seasonal Climate Prediction System: In order to gain greater insights into and to predict climate variation phenomena that are generated on a scale of several months to several years, JAMSTEC Application Laboratory scientists used the Earth Simulator previously developed through Japanese and European collaboration to develop and improve a dynamical, seasonal climate prediction system, based on the SINTEX-F climate model. A climate model was created using a group of differential equations pertaining to the physics of the atmosphere, ocean, land, and sea ice. The globe was divided into a three-dimensional grid, and a group of programs that integrates formulae over time was assigned to each square. Current observation data were incorporated into the climate model; for seasonal climate predictions, data on water temperature abnormalities in the oceans, which are massive heat sinks, are particularly important. A supercomputer was used to calculate how these data developed over time, making it possible to predict seasonal climatic abnormalities (i.e., divergences from an average year) several months in advance. JAMSTEC has a world-class climate simulating supercomputer and is working to develop and expand its oceanographic observation network. This two-pronged approach of observations and numerical calculations is capable of producing highly accurate advance predictions of phenomena that cause natural disasters. Seasonal climate prediction research makes a vital contribution to public safety and security.
Figure 1. Variation over time in the ENSO index Nino3.4 (i.e., the divergence from annual mean values in sea surface temperatures averaged over the area of the eastern-central tropics of the Pacific Ocean, from 190º-240ºE, 5ºS-5ºN, ºC).Thick black line, observed values; light blue lines, individual ensemble (parallel world) prediction values from 6/1/1988 onward using 12 existing ensembles from SINTEX-F. Thick blue line, average prediction value from 6/1/1988 onward using 12 existing ensembles from SINTEX-F; orange line, 96 prediction ensemble members newly added for this study; thick red line, average value for 108 ensembles (i.e., the existing 12 ensembles and the 96 added ensembles).
Figure 2. Variation over time in the Indian Ocean Dipole index DMI (i.e., the difference between sea surface temperatures anomalies in the eastern and western tropics of the Indian Ocean, ºC). The colored lines are defined as in Fig. 1.
Figure 3. SEDI skill scores for the predictions made each year, starting in early June, for the global distribution of average precipitation from November to December of 1983-2015. For extreme drought events that have a less than 15% chance of occurring from a climatic standpoint, if their likelihood of occurrence is predicted to be approximately 20% greater than it is in an average year by the SINTEX-F system, then the simulation predicts that the event will occur, and a skill score is calculated to determine the accuracy of the prediction data. The results indicate that predictions using 108 ensembles (b) are more accurate than those using only 12 ensembles (a) for Mexico, northern Brazil, southern China, and other regions (c).