Ensemble Coupled Assimilation with MPI Online Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ensemble coupled assimilation systems face inefficiencies due to high computational overhead from data file reading and writing, poor parallel scalability, and limitations in supporting coupled models, especially in online interaction methods, which affect the timeliness and operability of numerical prediction.
Innovation Solution
An ensemble coupled assimilation system that enables collaborative operation of all ensemble members in the same MPI task, supports flexible use of different assimilation algorithms, and facilitates efficient online interaction without data files, allowing parallel operation and data interpolation between grids.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If offline interaction based on file reading and writing is used, then the independence of operation between ensemble members and assimilation module is maintained, but computational efficiency decreases due to high overhead of data file reading and writing operations
Solution Approach 1:
The patent merges the model and assimilation module into a single online interaction framework where they share the same memory space and execute simultaneously. This eliminates file reading/writing overhead while maintaining operational independence through modular design. The coupled model and assimilation algorithm communicate via shared memory variables rather than file I/O, resolving the contradiction between operational independence and computational efficiency.
2Ease of manufacture
If offline interaction with software restarting is used, then code changes are minimized, but the operating speed of the assimilation system continuously decreases
Solution Approach 1:
The patent performs preliminary integration of the assimilation algorithm code into the coupled model framework during the compilation stage. By pre-configuring the online interaction interface and shared memory structures before execution, the system eliminates the need for repeated software restarting while maintaining code modularity. This preliminary setup enables continuous execution at high speed without sacrificing the ease of code integration.
3Productivity
If PDAF online interaction method is used, then data file reading and writing overhead is avoided, but coupled models cannot be supported because all MPI processes must be used by the assimilation algorithm
Solution Approach 1:
The patent segments the MPI process usage between the coupled model and assimilation algorithm. The coupled model retains control over its own MPI processes for forward integration, while the assimilation algorithm uses a separate process or thread that accesses model data through shared memory. This segmentation allows online interaction efficiency while supporting coupled models with multiple component models, each managing their own computational resources independently.
4Device complexity
If PDAF requires one assimilation algorithm for all component models, then the system is simpler to implement, but flexibility for using different assimilation algorithms is lost
Solution Approach 1:
The patent implements a dynamic assimilation algorithm selection mechanism where different assimilation algorithms can be assigned to different component models based on their specific requirements. The system architecture allows runtime configuration of assimilation algorithms for each model component, enabling flexibility while maintaining a unified online interaction framework. This dynamic approach balances system simplicity with algorithmic versatility.
Data Source
AI summary
Disclosed is an ensemble coupled assimilation system for numerical prediction and the method thereof. The ensemble coupled assimilation system includes a module for integrating coupled models and managing collaborative ensemble operation, an assimilation algorithms integration module, a module for configuring ensemble coupled assimilation experiment, and an online interaction module for ensemble coupled assimilation.