Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic Basis Function Interpolation for Adaptive In Situ Data Integration in Ocean Modeling

Derek DeSantis, Ayan Biswas, Earl Lawrence, Phillip Wolfram

TL;DR

This paper develops Dynamic Basis Function Interpolation (DBFI) to fuse in situ buoy data with the MPAS-O ocean model, yielding bias-corrected surface temperatures that preserve seasonal and multi-timescale variability. DBFI builds time-aware basis functions from dynamic mode decompositions and combines them with a fast Moore–Penrose regression to interpolate $F(x,t)=\sum_{j=1}^M a_j\phi_j(x,t)$ using global buoy observations. Among static and dynamic variants, the DMD_dyn model, which weights spatial modes by their temporal dynamics $\phi_j(x,t)=V_{t,j}U_{\hat{x},j}$, provides the best predictive skill and reduces latitudinal biases without requiring Gaussian error assumptions or costly data assimilation. The approach offers a scalable, low-cost alternative to Kalman-filter-based methods, with potential for online, horizon-spanning bias correction in ocean simulations.

Abstract

We propose a new method for combining in situ buoy measurements with Earth system models (ESMs) to improve the accuracy of temperature predictions in the ocean. The technique utilizes the dynamics \textit{and} modes identified in ESMs alongside buoy measurements to improve accuracy while preserving features such as seasonality. We use this technique, which we call Dynamic Basis Function Interpolation, to correct errors in localized temperature predictions made by the Model for Prediction Across Scales Ocean component (MPAS-O) with the Global Drifter Program's in situ ocean buoy dataset.

Dynamic Basis Function Interpolation for Adaptive In Situ Data Integration in Ocean Modeling

TL;DR

This paper develops Dynamic Basis Function Interpolation (DBFI) to fuse in situ buoy data with the MPAS-O ocean model, yielding bias-corrected surface temperatures that preserve seasonal and multi-timescale variability. DBFI builds time-aware basis functions from dynamic mode decompositions and combines them with a fast Moore–Penrose regression to interpolate using global buoy observations. Among static and dynamic variants, the DMD_dyn model, which weights spatial modes by their temporal dynamics , provides the best predictive skill and reduces latitudinal biases without requiring Gaussian error assumptions or costly data assimilation. The approach offers a scalable, low-cost alternative to Kalman-filter-based methods, with potential for online, horizon-spanning bias correction in ocean simulations.

Abstract

We propose a new method for combining in situ buoy measurements with Earth system models (ESMs) to improve the accuracy of temperature predictions in the ocean. The technique utilizes the dynamics \textit{and} modes identified in ESMs alongside buoy measurements to improve accuracy while preserving features such as seasonality. We use this technique, which we call Dynamic Basis Function Interpolation, to correct errors in localized temperature predictions made by the Model for Prediction Across Scales Ocean component (MPAS-O) with the Global Drifter Program's in situ ocean buoy dataset.
Paper Structure (11 sections, 14 equations, 4 figures, 1 table)

This paper contains 11 sections, 14 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Basis Function Interpolation and Data Integration
  4. Static and Dynamic Basis Function Interpolation
  5. Model Summary
  6. Results
  7. Discussion
  8. Conclusion
  9. Appendix
  10. Time Filtering GDD Data to MPAS-O
  11. DMD Short Summary

Figures (4)

  • Figure 1: Bias error distribution of models in Table \ref{['tab:models']}. Each model is fit on $80\%$ of the data and tested on the remaining $20\%$ to produce an error measurement. The distribution of errors for each model is created by choosing different starting times $t_1$ and nowcast time horizions $T$ (in months).
  • Figure 2: Bias errors of models for time period Jan. 1st 2008-Jan 1st 2009. Plots (a) and (c) display error distributions for different models. Plots (b) and (d) compute the average error on each of the five day time slices to explore inability to capture seasonality/baises.
  • Figure 3: Spatial comparison of DMD_dyn and MPAS.
  • Figure 4: DMD Mode plots.