Investigating the Impact of Perturbing Low Cloud-SST Feedback on ENSO Occurrence
- Xie’s group identified the strong feedback between low cloud cover and sea surface temperature (SST) in the northeast Pacific (NEP), which amplifies SST variations and influences El Niño-Southern Oscillation (ENSO) events.
- The research aims to investigate this relationship by implementing the Random Parameter version 2 scheme (RP2) in the UM atmosphere model to assess the impact of perturbing low-level cloud cover and SST interaction.
- Previous studies have shown the effectiveness of stochastic schemes like RP and a hybrid approach combining multiple schemes in capturing observed events and addressing model uncertainties.
- The experiment design includes control runs, historical runs, and variations with the RP2 scheme activated to evaluate its effects on SST bias and other indices.
- The preliminary results indicate improved SST bias in the Southern Ocean but persistent cold biases in the Northern Pacific. Other indices show minor differences. The findings will be presented at the ESM WG, acknowledging the need for further evaluation considering model design limitations and scheme outcomes.
Our research builds upon the findings of Xie’s group, which recently demonstrated the significant influence of the feedback between low cloud cover and sea surface temperature (SST) in the northeast Pacific (NEP). They discovered that this feedback mechanism, known as the wind-evaporation-SST (WES) feedback, is exceptionally strong in the NEP region, amplifying SST variations during the summer months and impacting the occurrence of El Niño-Southern Oscillation (ENSO) events as the enhanced SST variability propagates southwestward from the NEP low cloud deck (Yang et al., 2022).
To further investigate this relationship, we have implemented the Random Parameter version 2 scheme (RP2) from the Stochastic Scheme in the UM atmosphere model. Our aim is to assess the impact of perturbing the interaction between low-level cloud cover and SST, and determine if it yields similar results to Xie’s research. By activating the RP2 scheme, we can explore the potential effects and implications of modifying this relationship on atmospheric dynamics and ENSO occurrence.
The RP2 scheme utilizes an AR1 approach to perturb the “parameter,” rather than employing random perturbations:
- McCabe et al. (2016) demonstrated that perturbing parameters with the random parameter (RP) scheme in the Met Office’s convection-permitting EPS for the United Kingdom (MOGREPS-U.K.) enabled the EPS to capture observed fog events that were otherwise missed.
Based on earlier studies and a lack of sufficient physics background, the Met Office made the decision to disable the RP2 scheme and replace it with SPT (Stochastic Perturbation Tendency).:
- Additionally, the configuration replaces the RP2 scheme by SPT with conservation constraints, as the latter produces larger improvements to the ensemble spread in the Tropics and at midlatitude high levels, with no major impact on the conservation budgets.
- The combination of the SKEB2 new version and SPT with the conservation constraints makes up the final configuration of stochastic physics for current GA systems. In the mid-latitudes, it produces a similar impact to the operational configuration (SKEB2 with its operational masks plus RP2) (Christensen et al. 2015).
However, recent studies have indicated that a hybrid stochastic scheme is recommended:
- McCabe et al. (2016) also demonstrated that perturbing parameters with the random parameter (RP) scheme in the Met Office’s convection-permitting EPS for the United Kingdom (MOGREPS-U.K.) enabled the EPS to capture observed fog events that were otherwise missed (Frogner et al. 2022).
- The National Oceanic and Atmospheric Administration (NOAA) combined SPP, SPPT, and SKEB (Jankov et al., 2017) and found that the ensemble combining three stochastic schemes consistently produces the best spread–skill ratio and generally outperforms the multiphysics ensemble. This suggests that a combination of different stochastic physics schemes might have the potential to better address the model uncertainties (Xu et al. 2020).
- ACCESS-CM2 (PD-CNTL) present-day control done by mrd599 from years 1-649 and it is ongoing under project P66, (introduce the GM parameter setting after year 300)
- Restart (PD-CNTL) from year 500 to 649 but turn on the RP2. The last 50 years have been done under project LG87.
- Official ACCESS-CM2 historical run (based on recently published version: R10-HIST) and focus on the year after 1900 (P66).
- ACCESS-CM2 (PD-HIST): mrd599 recently conducted 1850-2014.
- Based on PD-HIST turn on RP2 and run from the years 1900 to 2014.
People: CSIRO ACCESS model development team
Initial conditions: PD control and Historical IC
Total KSUs required:
Total storage required:
- Yang, L., S. Xie, S. S. P. Shen, J. Liu, and Y. Hwang, 2022: Low Cloud–SST Feedback over the Subtropical Northeast Pacific and the Remote Effect on ENSO Variability. J. Climate, 36, 441–452.
- McCabe, Anne, et al. “Representing model uncertainty in the Met Office convection‐permitting ensemble prediction system and its impact on fog forecasting.” Quarterly Journal of the Royal Meteorological Society 142.700 (2016): 2897-2910.
- Christensen, H. M., I. M. Moroz, and T. N. Palmer, 2015: Stochastic and perturbed parameter representations of model uncertainty in convection parameterization. J. Atmos. Sci., 72, 2525–2544.
- Leutbecher et al., 2017: Stochastic representations of model uncertainties at ECMWF: state of the art and future vision. Q. J. R. Meteorol. Soc.
- Xu, Z., Chen, J., Jin, Z. et al. Representing Model Uncertainty by Multi-Stochastic Physics Approaches in the GRAPES Ensemble. Adv. Atmos. Sci. 37, 328–346 (2020).
- Frogner, I., U. Andrae, P. Ollinaho, A. Hally, K. Hämäläinen, J. Kauhanen, K. Ivarsson, and D. Yazgi, 2022: Model Uncertainty Representation in a Convection-Permitting Ensemble—SPP and SPPT in HarmonEPS. Mon. Wea. Rev., 150, 775–795.
Based on our research using data from Xie’s group (provided by Xie), when they decoupled the low-level cloud from the SST, there was a significant improvement in the ENSO period, shifting from 2 years to 4 years.
However, our results indicate that only the SST bias has significantly improved in the Southern Ocean, while the Northern Pacific and other locations continue to exhibit excessive cold biases. Other indices, such as the East Asia monsoon, Australia Monsoon, QNET, QNETT, TS, GS, and ENSO periodicity, show minimal differences.
Investigating how we can precisely tune the model parameters, rather than adopting Xie’s method, is still ongoing, and we welcome further suggestions from the scientific community.
The findings will be presented at the ESM WG, and it is important to note that these results are preliminary. We acknowledge that there may still be certain issues with the model design, and the PD-CNTL also has its limitations. Therefore, if the scheme demonstrates promising outcomes, we will consider its implementation or the decision to keep it deactivated based on further evaluation.