Quantitative Assessment of State-Dependent Atmospheric Motion Vector Uncertainties

Posselt, Derek J.; Wu, Longtao; Mueller, K. J.; Huang, Lei; Irion, F. W.; Brown, Shannon; Su, Hui; Santek, David; Velden, Christopher S.

doi:10.1175/jamc-d-19-0166.1

Cited by 22 publications

(32 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, when we analyze this error and its relationship to water vapor, we see that 'unskilled' regime correlates highly with areas of low water vapor in Figure 4. This matches the error patterns discussed in Posselt et al (2019).…”

Section: Error Regimesupporting

confidence: 89%

“…While our methodology in principle could be used to quantify uncertainties in any measurements used in data assimilation, in this paper we devote special emphasis to the use case of wind-tracking algorithms. In particular, we trained our model on the simulated data used by Posselt et al (2019), in which they applied an AMV algorithm to outputs from the NASA Goddard Space Flight Center (GSFC) Global Modeling and Assimilation Office (GMAO) GEOS-5 Nature Run (G5NR; Putman et al 2014). The Nature Run is a global dataset with ~7 km horizontal grid spacing that includes, among other quantities, three-dimensional fields of wind, water vapor concentration, clouds, and temperature.…”

Section: Simulation and Feature-tracking Algorithmmentioning

confidence: 99%

“…The algorithm searches a preset number of nearby pixels to minimize the sum-of-absolute-differences between aggregated water vapor values across the pixels. Posselt et al (2019) describes the sensitivity of the tracking algorithm and the dependency of the tracked winds on atmospheric states in detail.…”

Section: Simulation and Feature-tracking Algorithmmentioning

confidence: 99%

“…In this study, we detail a data-driven tool for building an AMV uncertainty model using observing system simulation experiment (OSSE) data. We build on the work by Posselt et al (2019) in which a water vapor feature-tracking AMV algorithm was applied to a high-resolution numerical simulation, thus providing a global set of AMV estimates which can be compared to the reference winds produced by the simulation. In this case, a synthetic "true" state is available with which AMVs can be compared and errors are quantified, and it is shown that tracking errors in AMV estimates are state dependent.…”

mentioning

confidence: 99%

“…In this case, a synthetic "true" state is available with which AMVs can be compared and errors are quantified, and it is shown that tracking errors in AMV estimates are state dependent. Our approach will use a conjunction of machine learning (random forest) and unsupervised parametric clustering (Gaussian mixture models) to build a model for the uncertainty structures found by Posselt et al (2019). The realism and robustness of the resulting uncertainty estimates depend on the realism and representativeness of the reference dataset.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Uncertainty Quantification for Atmospheric Motion Vectors with Machine Learning

Teixeira

Nguyen

Posselt

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. Wind-tracking algorithms produce Atmospheric Motion Vectors (AMVs) by tracking clouds or water vapor across spatial-temporal fields. Thorough error characterization (also known as uncertainty quantification) of wind-tracking algorithms is critical in properly assimilating AMVs into weather forecast models and climate reanalysis datasets. Uncertainty quantification should yield estimates of two key quantities of interest: bias, the systematic difference between a measurement and the true value, and standard error, a measure of variability of the measurement. The current process of specification of the errors input into inverse modelling is often cursory and commonly consists of a mixture of model fidelity, expert knowledge, and need for expediency. The methods presented in this paper supplement existing approaches to error specification by providing an error-characterization module that is purely data-driven and requires few tuning parameters. This paper proposes an error-characterization method that combines the flexibility of machine learning (random forest) with the robust error estimates of unsupervised parametric clustering (using a Gaussian Mixture Model). Traditional techniques for uncertainty quantification through machine learning have focused on characterizing bias, but often struggle when estimating standard error. In contrast, model-based approaches such as k-means or Gaussian mixture modelling can provide reasonable estimates of both bias and standard error, but they are often limited in complexity due to reliance on linear or Gaussian assumptions. In this paper, a methodology is developed and applied to characterize error in tracked-wind using a high-resolution global model simulation, and it is shown to adequately capture the error features of the tracked wind.

show abstract

Section: Error Regimesupporting

confidence: 89%

Section: Simulation and Feature-tracking Algorithmmentioning

confidence: 99%

Section: Simulation and Feature-tracking Algorithmmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Uncertainty Quantification for Atmospheric Motion Vectors with Machine Learning

Teixeira

Nguyen

Posselt

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

A collocation study of atmospheric motion vectors (AMVs) compared to Aeolus wind profiles with a feature track correction (FTC) observation operator

Hoffman

Lukens

Ide

et al. 2021

Quart J Royal Meteoro Soc

View full text Add to dashboard Cite

A method to apply an empirical feature track correction (FTC) in a new observation operator for atmospheric motion vectors (AMVs) is proposed. The FTC AMV observation operator determines the background estimate of the observed AMV vector wind, adjusting the background profile by determining an optimal height adjustment, averaging the profile over a layer of optimal thickness, and applying a linear correction to the averaged profile wind. The FTC observation operator is tested in the context of a collocation study between AMVs projected onto the collocated Aeolus horizontal line-of-sight (HLOS) and the Aeolus HLOS wind profiles. This study is a prototype for a variational FTC for numerical weather prediction data assimilation systems in which the Aeolus wind profiles take the place of the background in the FTC observation operator. Compared to a collocation where the Aeolus profile is interpolated linearly in height to the AMV height, a simple ad hoc averaging approach and the FTC approach reduce the mean square difference between the AMV observation and the Aeolus estimated AMV observation by 38% and 43%, respectively.

show abstract

Global Three‐Dimensional Water Vapor Feature‐Tracking for Horizontal Winds Using Hyperspectral Infrared Sounder Data From Overlapped Tracks of Two Satellites

Ouyed

Smith

Zeng

et al. 2023

Geophysical Research Letters

Self Cite

View full text Add to dashboard Cite

The three-dimensional (3D) structure of the global horizontal wind field remains largely unobserved . Atmospheric Motion Vectors (AMVs) based on cloud tracking have been used since the 1960s to fill some of the gaps in global wind fields observation (Eyre et al., 2020). In this type of AMVs, features in clouds are tracked across sequential images. These cloud-tracking AMVs are derived from geostationary (GEO) satellite and Low Earth Orbiting (LEO) satellite measurements (Key et al., 2003;.However, all these cloud-tracking AMVs face the same problem: clouds in general are sparse in the 850-500 hPa layers, leaving most of the vertical structure in the middle troposphere unobserved . Furthermore, each cloud-tracking AMV derived from an image sequence is assigned to a single level of cloud top at a given horizontal coordinate thus, dynamic quantities such as wind shear, vorticity, and divergence cannot be obtained. Finally, in the case of cloud-tracking winds, height assignment forms the largest source of error (Salonen et al., 2015).An alternative is AMVs retrieved from radiances of water vapor bands (Velden et al., 1997(Velden et al., , 2005. Indeed, these AMVs are widely used (e.g., http://tropic.ssec.wisc.edu/misc/winds/info.html). However, these water vapor AMVs have poor vertical resolution, since they are derived from instruments with limited water vapor bands. Other relevant observations to our work are the wind speeds derived by a wind lidar carried by a LEO satellite (Aeolus) that calculates the wind speed in its line of sight (LOS) with high vertical resolution (e.g., 0.25 km near surface; Stoffelen et al., 2020). Assimilation of Aeolus winds in numerical weather prediction models causes statistically significant increases in forecast skill (Rennie et al., 2021). Recent work shows that the LOS wind component derived by Aeolus is around 5 m s −1 for Rayleigh scattering and 4 m s −1 for Mie scattering (these values

show abstract

Quantitative Assessment of State-Dependent Atmospheric Motion Vector Uncertainties

Cited by 22 publications

References 35 publications

Uncertainty Quantification for Atmospheric Motion Vectors with Machine Learning

Uncertainty Quantification for Atmospheric Motion Vectors with Machine Learning

A collocation study of atmospheric motion vectors (AMVs) compared to Aeolus wind profiles with a feature track correction (FTC) observation operator

Global Three‐Dimensional Water Vapor Feature‐Tracking for Horizontal Winds Using Hyperspectral Infrared Sounder Data From Overlapped Tracks of Two Satellites

Contact Info

Product

Resources

About