Periodic Reporting for period 1 - SD4SP (Stratospheric Dynamics for Seasonal Prediction)
Période du rapport: 2023-04-03 au 2025-04-02
The aim of this project, Stratospheric Dynamics for Seasonal Prediction – SD4SP, is to improve understanding of dynamical processes in the winter stratosphere to formulate more reliable seasonal forecasts in the North Atlantic-European region.
The relevance of making available trustworthy information at seasonal timescales is recognized by the WMO Global Framework for Climate Services Seasonal forecasting, which is a field with enormous potential influence in different socio-economic sectors, such as agriculture, health, water management, insurance and particularly energy. Despite the chaotic nature of the climate system, seasonal prediction is feasible because atmospheric variability on seasonal timescales is modulated by slowly-varying boundary conditions, such as sea surface temperature (SST), and can retain memory from internal processes with very slow relaxation rates, such as those in the stratosphere. These fluctuations are not noticeable in day-to-day weather conditions but become evident in seasonal averages, e.g. two or three-month means. Seasonal prediction has improved considerably in the last decades, but the tropics remain the region where seasonal forecasts are most successful.
In most of the extratropics, and in particular for the North Atlantic-European (NAE) winter, General Circulation Model (GCM)-based seasonal forecasts have generally added little information over a prediction based on climatology or persistence, due to the large internal atmospheric variability – dominated by the North Atlantic Oscillation (NAO). Several recent studies have claimed that a better representation of the stratosphere-troposphere coupling in GCMs could improve seasonal forecasts over NAE since the surface signature of both, polar stratospheric anomalies and tropical-extratropical stratospheric teleconnections project strongly on the NAO. Yet, the impact of resolving the stratosphere and simulating stratospheric processes is unclear and needs to be properly assessed in a forecast context. SD4SP will address the separate contribution of the tropical stratosphere and the polar stratosphere to the prediction skill of the NAE during winter. This will be achieved by performing sensitivity seasonal forecast experiments with suppressed variability in the two stratospheric regions, using two independent state-of-the-art GCMs.
The goal of SD4SP is not only to identify key sources of predictability but also to advance knowledge and simulation of the mechanisms responsible for that predictability. The project deals with extratropical teleconnections, via the stratosphere, associated with the two main modes of interannual variability at low latitudes: El Niño-Southern Oscillation (ENSO) in the tropical troposphere, and the Quasi-Biennial Oscillation (QBO) in the tropical stratosphere, both of which represent suitable phenomena to enhance regional predictability. SD4SP will improve prediction of the ENSO/QBO-NAE teleconnections by gaining new insight into the stratospheric dynamics at play and by identifying model biases in the relevant stratospheric processes. The outcomes will be of great interest to the climate modelling community and will also provide valuable information to forecast providers and decision makers that use seasonal forecast products.
The relevance of making available trustworthy information at seasonal timescales is recognized by the WMO Global Framework for Climate Services Seasonal forecasting, which is a field with enormous potential influence in different socio-economic sectors, such as agriculture, health, water management, insurance and particularly energy. Despite the chaotic nature of the climate system, seasonal prediction is feasible because atmospheric variability on seasonal timescales is modulated by slowly-varying boundary conditions, such as sea surface temperature (SST), and can retain memory from internal processes with very slow relaxation rates, such as those in the stratosphere. These fluctuations are not noticeable in day-to-day weather conditions but become evident in seasonal averages, e.g. two or three-month means. Seasonal prediction has improved considerably in the last decades, but the tropics remain the region where seasonal forecasts are most successful.
In most of the extratropics, and in particular for the North Atlantic-European (NAE) winter, General Circulation Model (GCM)-based seasonal forecasts have generally added little information over a prediction based on climatology or persistence, due to the large internal atmospheric variability – dominated by the North Atlantic Oscillation (NAO). Several recent studies have claimed that a better representation of the stratosphere-troposphere coupling in GCMs could improve seasonal forecasts over NAE since the surface signature of both, polar stratospheric anomalies and tropical-extratropical stratospheric teleconnections project strongly on the NAO. Yet, the impact of resolving the stratosphere and simulating stratospheric processes is unclear and needs to be properly assessed in a forecast context. SD4SP will address the separate contribution of the tropical stratosphere and the polar stratosphere to the prediction skill of the NAE during winter. This will be achieved by performing sensitivity seasonal forecast experiments with suppressed variability in the two stratospheric regions, using two independent state-of-the-art GCMs.
The goal of SD4SP is not only to identify key sources of predictability but also to advance knowledge and simulation of the mechanisms responsible for that predictability. The project deals with extratropical teleconnections, via the stratosphere, associated with the two main modes of interannual variability at low latitudes: El Niño-Southern Oscillation (ENSO) in the tropical troposphere, and the Quasi-Biennial Oscillation (QBO) in the tropical stratosphere, both of which represent suitable phenomena to enhance regional predictability. SD4SP will improve prediction of the ENSO/QBO-NAE teleconnections by gaining new insight into the stratospheric dynamics at play and by identifying model biases in the relevant stratospheric processes. The outcomes will be of great interest to the climate modelling community and will also provide valuable information to forecast providers and decision makers that use seasonal forecast products.
To separate the ENSO teleconnection through the stratosphere, a nobel analysis using 3-dimensional relative vorticity has been implemented to characterize the stratospheric polar vortex and explore its interaction with the Rossby wave train forced by ENSO in the troposphere. This work reconciles previous findings in the stratospheric pathway of ENSO, and reveals why split sudden stratospheric warmings (SSWs) are more frequent during La Niña while El Niño is associated with displacement-type SSWs.
In order to investigate the QBO teleconnection several sets of idealized experiments have been conducted using CanESM5.1.
The first integration was the control run (CTL), which includes 200 years after a 100-year spin-up in ocean-coupled mode with constant radiative conditions set to the year 2005. Since CanESM does not have exhibit a spontaneous QBO, the originally proposed NOTROP experiment is no longer required. Upon completion of the CTL run, initial conditions were used to branch off nudging experiments starting on September 1st, November 1st or December 1st, depending on the specific case. For the experiments, the main nudging strategies explored are:
a) Nudging protocol
i) Vertical extension: One of the project’s objectives is to investigate the distinct roles of the upper and lower cells of the QBO. This has been explored by permanently nudging either the upper or lower cell throughout the entire winter, as well as by nudging the cells for one month and then allowing the model to run freely for the subsequent months. Additionally, the vertical extent of the nudging—either extending to the model's top or up to 10 hPa (as specified in the QBOi protocol)—was also compared.
ii) Latitudinal extension: after testing different latitudinal extensions for the nudging, three different ranges were selected for comparison: “qboi-fit”, qboi, and “qboi-20”, listed in order of increasing extension.
iii) Strength of the nudging: this can be adjusted my modulating the nudging timescale – shorter timescales correspond to stronger nudging. After testing various durations ranging from 6 to 120 hours, 24-hours nudging period was selected.
iv) Since the nudging is applied in the spectral space, it was straightforward to nudge only the zonal-mean vorticity. This approach proved sufficient, as the resulting zonal-mean winds in the tropics matched the observed values. With this method, the meridional component of the wind and waves are allowed evolve freely.
b) Target state
i) Snapshots: As an initial approach, composites of EQBO and WQBO phases were created using reanalysis events. These “snapshots” were targeted throughout the entire year for each member branched off CTL in every September. However, the amplitude of these target states was insufficient to generate an extratropical signal, so the profiles were amplified (doubled). This amplification was sufficient to produce an extratropical response.
ii) Evolving anomalies: Anomalies from the snapshots were calculated and superimposed onto the model’s climatology, resulting in a tropical state that is closer to the model’s climatology and more realistic, as it incorporates some seasonality. For this case, the anomalies used to construct the EQBO phase were amplified, but the WQBO anomalies were sufficient to generate an extratropical signal.
iii) Initialization by nudging: Rather that applying permanent nudging, only the month of October was nudged to the target state, allowing the model to run freely from November onward. This approach has the advantage that of avoiding spurious effects at the boundaries of the nudging, although the QBO takes several weeks to decay.
iv) Nudging in December: To evaluate the impact of starting the nudging when the vortex is already mature, nudging was applied in December. Two amplitudes were tested: single and double.
While the first part of the project is more process-understanding oriented, the second part is devoted to the skill assessment of the added value of the stratosphere in the NAE predictability. For this aim, one set of free evolving hindcasts in addition to several sets of idealized hindcasts experiments in which the stratosphere has been nudged to different target states have been conducted with CanESM5.1 and will be also performed using the CMCC model. This includes experiments with no stratospheric variability (nudging towards the model's climatology) experiments with "perfect" QBO (nudging towards ERA5 only in the tropics), and experiments with "perfect" stratosphere conditions (nudging towards ERA5 in the entire stratosphere). These hindcasts span the period from 1980 to 2020, with each year consisting in 10 members, all initialized in November using different initial conditions. These initial conditions correspond to those used in the seasonal prediction systems, and are also applied in the hindcast nudged experiments. This yields a total of 400 simulated years per hindcast experiment.
In order to investigate the QBO teleconnection several sets of idealized experiments have been conducted using CanESM5.1.
The first integration was the control run (CTL), which includes 200 years after a 100-year spin-up in ocean-coupled mode with constant radiative conditions set to the year 2005. Since CanESM does not have exhibit a spontaneous QBO, the originally proposed NOTROP experiment is no longer required. Upon completion of the CTL run, initial conditions were used to branch off nudging experiments starting on September 1st, November 1st or December 1st, depending on the specific case. For the experiments, the main nudging strategies explored are:
a) Nudging protocol
i) Vertical extension: One of the project’s objectives is to investigate the distinct roles of the upper and lower cells of the QBO. This has been explored by permanently nudging either the upper or lower cell throughout the entire winter, as well as by nudging the cells for one month and then allowing the model to run freely for the subsequent months. Additionally, the vertical extent of the nudging—either extending to the model's top or up to 10 hPa (as specified in the QBOi protocol)—was also compared.
ii) Latitudinal extension: after testing different latitudinal extensions for the nudging, three different ranges were selected for comparison: “qboi-fit”, qboi, and “qboi-20”, listed in order of increasing extension.
iii) Strength of the nudging: this can be adjusted my modulating the nudging timescale – shorter timescales correspond to stronger nudging. After testing various durations ranging from 6 to 120 hours, 24-hours nudging period was selected.
iv) Since the nudging is applied in the spectral space, it was straightforward to nudge only the zonal-mean vorticity. This approach proved sufficient, as the resulting zonal-mean winds in the tropics matched the observed values. With this method, the meridional component of the wind and waves are allowed evolve freely.
b) Target state
i) Snapshots: As an initial approach, composites of EQBO and WQBO phases were created using reanalysis events. These “snapshots” were targeted throughout the entire year for each member branched off CTL in every September. However, the amplitude of these target states was insufficient to generate an extratropical signal, so the profiles were amplified (doubled). This amplification was sufficient to produce an extratropical response.
ii) Evolving anomalies: Anomalies from the snapshots were calculated and superimposed onto the model’s climatology, resulting in a tropical state that is closer to the model’s climatology and more realistic, as it incorporates some seasonality. For this case, the anomalies used to construct the EQBO phase were amplified, but the WQBO anomalies were sufficient to generate an extratropical signal.
iii) Initialization by nudging: Rather that applying permanent nudging, only the month of October was nudged to the target state, allowing the model to run freely from November onward. This approach has the advantage that of avoiding spurious effects at the boundaries of the nudging, although the QBO takes several weeks to decay.
iv) Nudging in December: To evaluate the impact of starting the nudging when the vortex is already mature, nudging was applied in December. Two amplitudes were tested: single and double.
While the first part of the project is more process-understanding oriented, the second part is devoted to the skill assessment of the added value of the stratosphere in the NAE predictability. For this aim, one set of free evolving hindcasts in addition to several sets of idealized hindcasts experiments in which the stratosphere has been nudged to different target states have been conducted with CanESM5.1 and will be also performed using the CMCC model. This includes experiments with no stratospheric variability (nudging towards the model's climatology) experiments with "perfect" QBO (nudging towards ERA5 only in the tropics), and experiments with "perfect" stratosphere conditions (nudging towards ERA5 in the entire stratosphere). These hindcasts span the period from 1980 to 2020, with each year consisting in 10 members, all initialized in November using different initial conditions. These initial conditions correspond to those used in the seasonal prediction systems, and are also applied in the hindcast nudged experiments. This yields a total of 400 simulated years per hindcast experiment.
ENSO extratropical teleconnection
For the first time, the ENSO-NAE teleconnection has been explored in a 3-Dimensional framework. Figure 3 shows the January-February 3-D structure of the stratospheric polar vortex during El Niño and La Niña using relative vorticity in the ERA5 reanalysis. The ENSO-forced wavetrain crossing North America interacts with the Canadian center of rotation of the polar vortex, that intensifies during La Niña, and weakens during El Niño. The result is the apparent displacement of the polar vortex to Eurasia during El Niño, while a more elongated vortex with two strong centers of rotation is present for La Niña. These opposite configurations over Canada from the troposphere to the middle stratosphere is key to explain the preference of split type sudden stratospheric warmings (SSWs) during La Niña versus the displacement type SSWs more prone during El Niño. More details will be found in Palmeiro et al. (2025a, to be submitted).
QBO extratropical teleconnection
(This analysis is currently ongoing)
For the first time, the ENSO-NAE teleconnection has been explored in a 3-Dimensional framework. Figure 3 shows the January-February 3-D structure of the stratospheric polar vortex during El Niño and La Niña using relative vorticity in the ERA5 reanalysis. The ENSO-forced wavetrain crossing North America interacts with the Canadian center of rotation of the polar vortex, that intensifies during La Niña, and weakens during El Niño. The result is the apparent displacement of the polar vortex to Eurasia during El Niño, while a more elongated vortex with two strong centers of rotation is present for La Niña. These opposite configurations over Canada from the troposphere to the middle stratosphere is key to explain the preference of split type sudden stratospheric warmings (SSWs) during La Niña versus the displacement type SSWs more prone during El Niño. More details will be found in Palmeiro et al. (2025a, to be submitted).
QBO extratropical teleconnection
(This analysis is currently ongoing)