1 of 1

High-Order Remapping from MPAS-Ocean to ROMS �With the SEAHORCE Partnership (BER)

Scientific Achievement

Oceanographic data are accurately transferred from a coarse global climate simulation (MPAS-Ocean) to a high-fidelity regionally-refined simulation (ROMS) through the use of Multivariate Functional Approximation (MFA).

Significance and Impact

MFA provides a high-order intermediate representation to smoothly transform 3D fields between two simulations, producing consistent, high-order accurate, and smooth downsampled fields from coarse MPAS-O to 10X finer ROMS meshes. The continuous MFA model can enable multiscale coupling while preventing mesh artifacts that result from intersection-based conservative L₂ projections.

Top row: Data produced by MPAS-Ocean (left) on a coarse (green) mesh are remapped to a fine (blue) mesh that can be used as input to ROMS (right).

Bottom row: Controlling the degree of smoothness using MFA. Bathymetry in MPAS is remapped to ROMS, with the polynomial degree=1 (left) compared with degree=3 (right). The scientist can accentuate high-frequency details or low-frequency overall trends by selecting the MFA degree. The MFA library was developed under ASCR ECRP and is maintained under the SciDAC RAPIDS Institute.

Technical Approach

MFA is a meshless data representation built on a B-spline functional model.
Any mesh-based data set may be converted into an MFA model, and subsequently converted back to a different mesh structure.
Field remapping is achieved by representing coarse mesh data produced by MPAS-Ocean as an MFA model, then resampling the model on the fine mesh used by ROMS.

PI(s)/Facility Lead(s): Tom Peterka, David Lenz, Vijay Mahadevan (ANL)

Collaborating Institutions: Argonne National Laboratory

ASCR Program: SciDAC

ASCR PM: Kalyan Perumalla

Publication(s) for this work: Lenz et al. Customizable Adaptive Regularization Techniques for B-Spline Modeling. Journal of Computational Science, 71, 2023.

Code Developed: https://github.com/tpeterka/mfa

Degree = 1 Degree = 3

MPAS-Ocean Remap ROMS

LOCAL LAB POC:

Tom Peterka

TALKING POINTS:

MFA is a meshless data representation built on a B-spline functional model.
Any mesh-based data set may be converted into an MFA model, and subsequently converted back to a mesh-based structure with different mesh connectivity.
Physics remapping is achieved by representing coarse mesh data produced by MPAS-Ocean as an MFA model, then resampling the model on the fine mesh used by ROMS.
The degree of smoothness in the remapped model may be selected by the user at run time.
Accurate remapping from MPAS-Ocean to ROMS has been demonstrated for 2D and 3D datasets

METADATA:

Name of the associated awarded project: SciDAC RAPIDS Institute

PI name(s): The RAPIDS lead PIs are Rob Ross and Lenny Oliker. The RAPIDS investigators involved in this work are Tom Peterka and David Lenz (ANL). The SciDAC SEAHORCE partnership lead PI is Robert Hetland (PNNL), and the co-PI involved in this work is Vijay Mahadevan (ANL).

Name of the program manager: SciDAC managed by Kalyan Perumalla. Supporting technology managed by ASCR research (Margaret Lentz and Hal Finkel)

CITATIONS:

Peterka, Tom, David Lenz, Iulian Grindeanu, and Vijay S. Mahadevan. "Towards Adaptive Refinement for Multivariate Functional Approximation of Scientific Data." In 2023 IEEE 13th Symposium on Large Data Analysis and Visualization (LDAV), pp. 32-41. IEEE, 2023.
Lenz, D., Yeh, R., Mahadevan, V., Grindeanu, I., Peterka, T.: Customizable Adaptive Regularization Techniques for B-Spline Modeling. Journal of Computational Science, 71, 2023.
Mahadevan, V., Lenz, D., Grindeanu, I., Peterka, T.:Accelerating Multivariate Functional Approximation Computation with Domain Decomposition Techniques. Journal of Computational Science (JCS) 2024.
Sun,J., Lenz,D., Yu,H., Peterka,T.: Scalable Volume Visualization for Big Scientific Data Modeled by Functional Approximation. Proceedings of IEEE International Conference on Big Data, 2023.
Sun, J., Lenz, D., Yu, H., Peterka, T.: MFA-DVR: Direct Volume Rendering of MFA Models. Journal of Visualization, 2023.