9:45 AM - 10:00 AM
[AAS10-04] Global mean temperature and its seasonal predictability for 2023 and 2024
★Invited Papers
Keywords:Global mean temperature, seasonal predictability
The global mean annual temperature reached a record high in 2023, and the record was further broken in 2024 (WMOPress release 12Jan2024). Could such an abnormally high global temperatures in 2023 and 2024 have been predicted on the time scale of seasonal forecasts?
The causes of the abnormally high temperatures in 2023 are thought to be global warming, as well as the rise in the temperature of very shallow tropical ocean surface waters due to the El Niño phenomenon that occurred in the spring, the sustained high ocean temperatures in mid-latitudes such as the Kuroshio Extension region due to the La Niña phenomenon that lasted for almost three years until the winter of 2022/23, and the high sea surface temperatures in the tropical North Atlantic. In addition, anomalies have been observed in the global mean net energy balance at the top of the atmosphere that are thought to be caused by a decrease in albedo (Minobe et al. 2024), and it has been pointed out that a decrease in low clouds may be the cause (Goessling et al. 2025).
If the seasonal ensemble forecast system using the coupled atmosphere-ocean model currently in operation could properly simulate and predict the effects of global warming, the above-mentioned oceanic variations, or the coupled atmosphere-ocean variations, it may have been able to predict the record high global mean temperatures in 2023 and 2024. Global mean temperatures are expected to be more predictable than forecasts for smaller areas because unpredictable components due to the chaotic nature of the atmosphere are reduced by averaging them globally. Meanwhile, as the effects of global warming become apparent, seasonal forecast information on global mean temperatures could be useful information for promoting mitigation and adaptation measures for global warming. Therefore, we investigated how the seasonal ensemble forecast system of the Japan Meteorological Agency predicted the global mean temperatures.
The seasonal ensemble forecast system (Coupled Prediction System CPS3, Hirahara et al. 2023), which is used for the Japan Meteorological Agency's three-month forecast, performs a five-member seven-month forecast with 00 UTC as the initial value every day. In addition, to obtain the systematic error of the model, we also performed forecast calculations in advance using the mid- and end-of-month data for each month from 1991 to 2020 as the initial values. The predicted global mean 2m temperature were compared with the 2m temperature from JRA-3Q (Kosaka et al. 2024).
The figure shows the time evolution of the annual mean global average 2m temperature anomaly. The CPS3 ensemble mean deviation forecast with June as the initial time closely simulated the interannual variation shown in JRA-3Q, and the standard deviation of the prediction error obtained from the hindcast period was 0.04°C. However, the errors for 2023 and 2004 were both -0.09°C, which was larger than the standard deviation. The errors for 2023 and 2004 were both -0.09°C, which was larger than the standard deviation. However, the ensemble mean forecast with June as the initial time predicted that the annual average would exceed the highest value on record, and although the deviation was not large enough, the annual mean temperature was predicted to be the highest on record as of July.
We found that the record high temperatures of the global average temperature for 2023 and 2004 were predicted to some extent in July by CPS3. However, the error range was larger than the predictions for the hindcast period. We would like to consider the reason for this in the future. We would also like to verify what kind of predictions the ensemble members are making.
The causes of the abnormally high temperatures in 2023 are thought to be global warming, as well as the rise in the temperature of very shallow tropical ocean surface waters due to the El Niño phenomenon that occurred in the spring, the sustained high ocean temperatures in mid-latitudes such as the Kuroshio Extension region due to the La Niña phenomenon that lasted for almost three years until the winter of 2022/23, and the high sea surface temperatures in the tropical North Atlantic. In addition, anomalies have been observed in the global mean net energy balance at the top of the atmosphere that are thought to be caused by a decrease in albedo (Minobe et al. 2024), and it has been pointed out that a decrease in low clouds may be the cause (Goessling et al. 2025).
If the seasonal ensemble forecast system using the coupled atmosphere-ocean model currently in operation could properly simulate and predict the effects of global warming, the above-mentioned oceanic variations, or the coupled atmosphere-ocean variations, it may have been able to predict the record high global mean temperatures in 2023 and 2024. Global mean temperatures are expected to be more predictable than forecasts for smaller areas because unpredictable components due to the chaotic nature of the atmosphere are reduced by averaging them globally. Meanwhile, as the effects of global warming become apparent, seasonal forecast information on global mean temperatures could be useful information for promoting mitigation and adaptation measures for global warming. Therefore, we investigated how the seasonal ensemble forecast system of the Japan Meteorological Agency predicted the global mean temperatures.
The seasonal ensemble forecast system (Coupled Prediction System CPS3, Hirahara et al. 2023), which is used for the Japan Meteorological Agency's three-month forecast, performs a five-member seven-month forecast with 00 UTC as the initial value every day. In addition, to obtain the systematic error of the model, we also performed forecast calculations in advance using the mid- and end-of-month data for each month from 1991 to 2020 as the initial values. The predicted global mean 2m temperature were compared with the 2m temperature from JRA-3Q (Kosaka et al. 2024).
The figure shows the time evolution of the annual mean global average 2m temperature anomaly. The CPS3 ensemble mean deviation forecast with June as the initial time closely simulated the interannual variation shown in JRA-3Q, and the standard deviation of the prediction error obtained from the hindcast period was 0.04°C. However, the errors for 2023 and 2004 were both -0.09°C, which was larger than the standard deviation. The errors for 2023 and 2004 were both -0.09°C, which was larger than the standard deviation. However, the ensemble mean forecast with June as the initial time predicted that the annual average would exceed the highest value on record, and although the deviation was not large enough, the annual mean temperature was predicted to be the highest on record as of July.
We found that the record high temperatures of the global average temperature for 2023 and 2004 were predicted to some extent in July by CPS3. However, the error range was larger than the predictions for the hindcast period. We would like to consider the reason for this in the future. We would also like to verify what kind of predictions the ensemble members are making.