Reveal the Role of Butterfly Effects and Multiscale Processes in Predictability using Advanced Concurrent Visualization and Multiscale Analysis (PEEMD) Methods
Abstract
Supercomputing technology and global weather/climate modeling systems have advanced rapidly in recent years. The advances provide opportunities for applying high-dimensional Lorenz models to improve our understanding of chaos theory and, thus, revisiting predictability issues for (high-impact) weather systems. In this study, we highlight the role of butterfly effects and multiscale processes in model predictability using concurrent visualization (CV) and parallel ensemble empirical mode decomposition (PEEMD) methods.
We demonstrate the enabling role of the aforementioned technology in producing remarkable simulations of large-scale tropical waves and hurricanes at extended-range time scales (Shen et al., 2010, 2013a). To efficiently analyze and visualize multiscale processes in big data sets, we have developed the CV technology and the PEEMD method. Using the CV, a simulation code is instrumented such that its data can be extracted for analysis while the simulation is running. Thus, the CV enables model simulations at much higher temporal resolution than traditional methods. We apply the CV to illustrate a potential role of butterfly effects and/or multiscale interactions in model predictions (Shen et al., 2013b, 2016). The PEEMD package was deployed to study multiscale interactions by implementing three-level parallelism into the ensemble Empirical Mode Decomposition (EMD), achieving a scaled performance of 5,000 cores (Shen et al., 2017). The PEEMD was used in a 10-year data analysis (Wu and Shen, 2016) to examine a conceptual model that focuses on the impact of multiscale processes (Shen et al., 2010) in the formation of tropical cyclones. Based on the aforementioned studies and recent studies (e.g., Shen, 2014), we first identify ingredients for sensitive dependence on initial conditions and discuss their relevance to the predictions of diverged hurricane tracks. Secondly, we suggest the view: "weather is a superset that consists of chaotic and non-chaotic processes (Shen et al., 2018). Lastly, we propose an integrated system that consists of the above technology and newly deployed methods (e.g., recurrence analysis) for processing various global model data and satellite data, monitoring the time evolution of identified atmospheric factors, and providing reliable predictions.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFMIN43A..13S
- Keywords:
-
- 0399 General or miscellaneous;
- ATMOSPHERIC COMPOSITION AND STRUCTUREDE: 3399 General or miscellaneous;
- ATMOSPHERIC PROCESSESDE: 1899 General or miscellaneous;
- HYDROLOGYDE: 1996 Web Services;
- INFORMATICS