Fast Fourier Transform Ensemble Kalman Filter with Application to a Coupled Atmosphere-Wildland Fire Model

TL;DR

This paper introduces a novel FFT-based Ensemble Kalman Filter that efficiently estimates covariance with small ensembles and corrects both position and amplitude errors, demonstrated on a coupled weather-fire model.

Contribution

It presents a new FFT EnKF method that improves covariance estimation and analysis speed, combined with morphing EnKF for position error correction in coupled models.

Findings

Efficient covariance estimation with small ensembles using FFT.

Successful correction of position and amplitude errors in coupled models.

Application to WRF-Fire demonstrates practical effectiveness.

Abstract

We propose a new type of the Ensemble Kalman Filter (EnKF), which uses the Fast Fourier Transform (FFT) for covariance estimation from a very small ensemble with automatic tapering, and for a fast computation of the analysis ensemble by convolution, avoiding the need to solve a sparse system with the tapered matrix. The FFT EnKF is combined with the morphing EnKF to enable the correction of position errors, in addition to amplitude errors, and demonstrated on WRF-Fire, the Weather Research Forecasting (WRF) model coupled with a fire spread model implemented by the level set method.