Figure - available from: Inverse Problems
This content is subject to copyright. Terms and conditions apply.
Experiment #1. Influence of the temperature ζ (left) and of the preconditioning parameter H (right) on the evolution of the endpoint error criteria (2.7) with (2.8) for p = 2.
Source publication
Atmospheric motion vectors (AMVs) extracted from satellite imagery are the only wind observations with good global coverage. They are important features for feeding numerical weather prediction (NWP) models. Several Bayesian models have been proposed to estimate AMVs. Although critical for correct assimilation into NWP models, very few methods prov...
Citations
... Synergistic use of both methods is adopted by NOAA's operational algorithm for AMV production. It is worth noting that sampling approaches (Héas et al. 2023a) for estimating AMVs together with their errors are important for quantitative applications such as improving forecasts through assimilating these winds into NWP models, and such approaches should be further explored and applied to wind estimation. ...
Weather satellites not only provide atmospheric thermodynamic and hydrometric information, but also important dynamic information when high temporal data are used. Atmospheric motion vectors (AMVs) have been routinely derived from global geostationary satellite imagers over tropical and mid-latitude regions, and polar orbiting satellite imagers over high-latitude regions since the 1990s and have been widely used in numerical weather prediction (NWP). These AMVs result from tracking clouds and moisture primarily from the infrared and visible bands. While the coverage is good where there are clouds and moisture targets, AMVs can only provide winds for a few tropospheric layers and thus lack vertical information. Recently, active remote sensing technologies have been developed for vertical wind profiling from satellites. Those wind estimates have good accuracy but limited spatial coverage. Expanding wind estimates from two-dimensional (2D) to three-dimensional (3D) over expansive domains is important for improving nowcasting and NWP applications. Hyperspectral infrared sounder observations from polar orbiting satellites have been used for 3D wind exploration, but lack the temporal resolution needed for 3D winds over tropical and middle latitude regions. The feasibility of 3D winds using geostationary hyperspectral infrared sounders has also been demonstrated and validated using 15-minute Geostationary Interferometric Infrared Sounder observations. Tropospheric 3D winds will be better achieved through combining both active and passive observations in the future. This paper provides an overview on tracking features from satellites for obtaining tropospheric winds and the evolution from 2D to 3D coverage, along with their potential applications, challenges and future perspectives.
... Advanced methods for Bayesian sampling of (1.1) are gradient-based Markov chain Monte Carlo (MCMC) methods such as Hamiltonian Monte Carlo [6] and preconditioning (of gradients) techniques [14]. Estimation with such advanced MCMC techniques can still exhibit slow convergence rates, as the simulation may need a large time to explore the target distribution, especially in high-dimension [22]. Sequential Monte Carlo (SMC) methods or other methods based on interacting clones/particles are an interesting alternative for Bayesian sampling, and have shown to be stable in certain high-dimensional context [2], in particular adaptive SMC simulation [3]. ...
This work proposes an adaptive sequential Monte Carlo sampling algorithm for solving inverse Bayesian problems in a context where a (costly) likelihood evaluation can be approximated by a surrogate, constructed from previous evaluations of the true likelihood. A rough error estimation of the obtained surrogates is required. The method is based on an adaptive sequential Monte-Carlo (SMC) simulation that jointly adapts the likelihood approximations and a standard tempering scheme of the target posterior distribution. This algorithm is well-suited to cases where the posterior is concentrated in a rare and unknown region of the prior. It is also suitable for solving low-temperature and rare-event simulation problems. The main contribution is to propose an entropy criteria that associates to the accuracy of the current surrogate a maximum inverse temperature for the likelihood approximation. The latter is used to sample a so-called snapshot, perform an exact likelihood evaluation, and update the surrogate and its error quantification. Some consistency results are presented in an idealized framework of the proposed algorithm. Our numerical experiments use in particular a reduced basis approach to construct approximate parametric solutions of a partially observed solution of an elliptic Partial Differential Equation. They demonstrate the convergence of the algorithm and show a significant cost reduction (close to a factor 10) for comparable accuracy.
