Sub Seasonal to Seasonal (S2S) Prediction & Seasonal to Decadal (S2D) Prediction Conference
Sub Seasonal to Seasonal (S2S) Prediction & Seasonal to Decadal (S2D) Prediction Conference, Boulder, Colorado
On the week of the 17th until the 21st of September, 10 team members from the Decadal Climate Forecasting Project travelled to Boulder in Colorado to attend the International Conference on Sub Seasonal to Decadal Prediction.
About the conference
Weather and climate variations on sub seasonal to decadal timescales can have enormous social, economic, and environmental impacts, making skilful predictions on these timescales a valuable tool for policymakers and industries. As such, there is a growing interest in the scientific, operational, and applications communities in developing forecasts to improve our foreknowledge of events. On S2S timescales, these include high-impact meteorological events such as tropical cyclones, floods, droughts, and heat and cold weaves. The S2D timescales broadly focus on similar events, but it needs to decipher the roles of internal and externally-forced variability of forecasts. The S2S and S2D communities share very similar resolution and biases; atmosphere-ocean coupling; forecast initialization; ensemble generation, bias correction, calibration, and skill assessment; and linking research, operations and users.
The conference provided an important perspective on the status and future effort required for S2S and S2D climate forecasting. A key message to emerge from the meeting is that climate forecasting is challenging and we are making slow progress to improve the forecasts. A consensus thought for accelerating our progress is to closely integrate the climate model development, data assimilation, forecasting and application efforts. This is how the CSIRO DCFP is structured, so it confirms our approach. Further, the SCFP forecasting system (CAFE), the process understanding and the application research is world class and our approaches are novel and worth pursuing.
The DCFP received good exposure at the conference with 4 oral presentations and 7 poster presentations. The oral and poster presentations are listed below with a link to the presentation and a short summary of the feedback received at the meeting.
Poster Presentations
Exploration of strong and weakly coupled ocean-atmosphere data assimilation in climate reanalysis and forecasting presented by Dr Paul Sandery
There was a range of interest in our work on coupled data assimilation. Feedback was we have made significant progress in the assimilation of a comprehensive array of ocean observations and atmospheric data. The conference outlined the need for data assimilation, ensemble forecasting, consistent model based analysis, forecast initialisation, bias detection and correction. These are already features of the CAFE-EnKF system. Many groups, such as ECMWF, NCAR, NASA, NERSC, GFDL and the UKMet Office have and continue to work on coupled data assimilation. Our results on the sensitivity of the cross-domain error-covariance’s in weak and strongly coupled data assimilation are in general agreement with the findings of others and are providing guidance for the upcoming CAFE-88 coupled reanalysis.
In-Situ ocean observations: Building a research quality library for data assimilation in a climate forecasting system,
Presented by Dr Thomas Moore
The experience attending the Second International Conference on Seasonal to Decadal Prediction (S2D) showed at least 20% of this research community was focused on the importance of sub-surface ocean observations. My conversations with our international colleagues – including Japanese, Norwegian and UK representatives – made it clear that the approach our team is taking to the assimilation of sub-surface ocean data is on par with state-of-the-art effort worldwide.
Stochastic subgrid turbulence parameterisation maintaining resolution independent statistics for all spatial scales over decadal timescales,
Presented by Dr Vassili Kitsios
There was interest in how this data centric parameterisation approach can be used to improve the accuracy and computational efficiency of our couple climate models.
Diagnosing the atmospheric mechanisms that influence forecast skill of rainfall extremes
Presented by Dr Carly Tozer
Attending the conference highlighted to me that we are a relatively unique research group in that we cover the scope of climate observations, model development, model verification and applications in the one group. The conference provided a great opportunity to see where our research group, and the activities we’re undertaking, fit within the broader climate prediction community. I appreciated the opportunity to meet researchers tackling similar issues to us.
Verifying and communicating climate forecasts
Presented by Dr James Risbey
Harnessing dynamical seasonal climate forecasts for agricultural applications in Australia
Presented by Dr Andrew Schepen
Assessing ENSO predictability the associated variability of biogeochemical fields in the ocean
Presented by Dr Richard Matear
My poster was well received. In agreement with the poster results, several other groups confirmed that their tropical Pacific Net Primary Productivity was less predictable than SST. In a subsequent workshop on ocean carbon forecasting it was agreed that international collaboration on potential predictability experiments would be a useful way for the community to work together. Also, for many groups the way they assimilated physical ocean observations into the climate model create spurious behaviour that corrupted the biogeochemical fields. This is an important problem to solve and it appears the way we do data assimilation is one possible solution.
Oral Presentations
Coupled data assimilation and ensemble initalization with application to multi-year ENSO prediction
Presented by Dr Terry O’Kane
Skill assessment of the CSIRO multi-year climate analysis forecast ensemble system
Presented by Dr Dougie Squire
One focus of my presentation was on doppyo, software our team is developing to perform climate diagnostics and verification. We received substantial interest in the package and are establishing related collaborative efforts with a number of international institutions. Attending S2D confirmed that the community recognises many of the current approaches to data processing don’t suit the increasingly large datasets being generated. There is considerable interest in scalable computational methods, such as those provided within doppyo.
Application of normal mode functions for the improved balance in the CAFE data assimilation system and characterisation of modes of variability
Presented by Dr Vassili Kitsios
There was interest in how normal modes can be used to filter the atmospheric fields in order to improve forecast skill.
Subsurface Variability and Teleconnections in the Indian Ocean
Presented by Dr Chris Chapman