Academia.eduAcademia.edu

Outline

Title
Abstract
Introduction
Data
Discussion
Conclusions
References

Bayesian retrieval of scatterometer wind fields

Profile image of Ian T NabneyIan T Nabney

www-users.aston.ac.uk

Abstract

The retrieval of wind fields from scatterometer observations has traditional been separated into two phases; local wind vector retrieval and ambiguity removal. Operationally, a forward model relating wind vector to backscatter is inverted, typically using look up tables, to retrieve up ...

Related papers

Directional Stability and Conservation of Scattering (DiSCS)Based Directional-Ambiguity Removal Algorithm for Improving Wind Fields From Scatterometer: A QuikSCAT Example
priti sharma

IEEE Geoscience and Remote Sensing Letters, 2010

A new directional-ambiguity removal algorithm has been developed for improving the wind fields derived from a scatterometer. Exploiting the natural spatial variability of ocean surface wind directions and invoking the conservation of backscattering, a new directional stability-weighted low-pass filter has been designed to complement the circular median filtering. It improves the directional-ambiguity removal by reducing the directional noise while retaining the natural wind flow. On the basis of analysis, which will be described later, indicating the lower spatial stability (conversely, the variability) of the wind direction and the lower direction-retrieval skills of the scatterometer-derived winds at low-intensity winds (<; 3 m/s) as compared to those for winds above 3 m/s, the ambiguity removal has been performed separately for these two wind regimes. The proposed Directional Stability and Conservation of Scattering (DiSCS) algorithm is applicable to the full swath of the scatterometer. Using the DiSCS algorithm, the wind fields derived from QuikSCAT Level-2A backscatter data over global oceans for 57 orbits during July 1-4, 2005, have been compared with the winds from the National Centers for Environmental Predictions and European Centre for Medium-Range Weather Forecasts models. Similar comparisons of QuikSCAT Level-2B finished products “direction interval retrieval with thresholded nudging” winds (referred to hereinafter as FP) with the winds from these two models have been performed. The QuikSCAT-derived winds are found to be closer to the model winds as compared to those QuikSCAT FP winds.

View PDFchevron_right
On the Inversion of Wind Scatterometer Incomplete Data Sets
Maurizio Migliaccio

ERS-1/2 wind scatterometers (WSC) data sets may contain cells in which two, instead of three, measurements are available. These data sets are called incomplete. Operational inversion procedures discard such data sets and therefore no wind field is estimated. This is very limiting in semi-closed seas such as the Mediterranean Sea. In this paper we propose a new inversion procedure capable to retrieving near surface wind fields from incomplete data sets. This procedure is an enhancement of an algorithm already proposed and tested by the authors. A set of comprehensive experiments are presented and discussed. It is shown that the inversion procedure gets to remarkable results even in presence of a large number of doublets. key words: scatterometer, sea wind, wind field reconstruction * Note that semantically an incomplete data set is also a data set in which are present singlets (cells with a single measurement) and cells with no measurements. These cases are of no practical interest because theory tells that even in ideal conditions no wind field can be retrieved [12], .

View PDFchevron_right
Modeling Wind Direction from Satellite Scatterometer Data
Ian Nabney

Industrial Applications of Neural Networks, 1998

Most of the common techniques for estimating conditional probability densities are inappropriate for applications involving periodic variables. In this paper we apply two novel techniques to the problem of extracting the distribution of wind vector directions from radar scatterometer data gathered by a remote-sensing satellite.

View PDFchevron_right
Modelling wind direction from satellite scatterometer data
Ian T Nabney

1998

Most of the common techniques for estimating conditional probability densities are inappropriate for applications involving periodic variables. In this paper we apply two novel techniques to the problem of extracting the distribution of wind vector directions from radar scatterometer data gathered by a remote-sensing satellite.

View PDFchevron_right
Scatterometer wind fields: A new release over the decade 1991–2001
yves quilfen

Canadian Journal of Remote Sensing, 2002

The global surface wind fields calculated from scatterometer observations have been updated to cover the decade 1991-2001. This involves an enhancement of the remotely sensed wind retrievals, quality control, and objective method. This paper presents the improvements with respect to the first release and provides the main characteristics of the surface parameter fields. The accuracy of the resulting weekly and monthly surface wind fields is studied through comparisons with buoy and numerical model averaged wind estimates.

View PDFchevron_right
CORRECTIONS TO SCATTEROMETER WIND VECTORS - REMOVING THE EFFECTS OF RAIN
Mark Bourassa
View PDFchevron_right
Optimization of wind field retrieval procedures
Luther White

Applied Mathematics and Computation, 2005

We study the formulation of the problem to retrieve wind fields from radar data. Our formulation allows us to consider the retrieved wind fields as a function of radar locations. We examine the properties of this function with the objective of determining ''best'' locations for observations. Problems are then posed to determine radar locations to minimize certain criteria involving retrieval errors over a class of test wind fields. A numerical study is presented illustrating the theory developed.

View PDFchevron_right
Geophysical modeling error in wind scatterometry
Paul Temi Johnson GOC

1996

A scatterometer is a satellite-borne instrument designed to measure wind over the ocean. Scatterometer wind retrieval is based on the relationship between the wind over the ocean and the resulting scattering cross section of the surface; this relationship, termed the '' geophysical model function,'' maps the wind speed, relative wind direction (relative to the antenna azimuth angle), antenna incidence angle, polarization and frequency band to the scattering cross section. The sea surface temperature, salinity, long waves, wind variability within a scatterometer footprint, etc., lend variability to the backscatter. A particular observation of the wind-dependent backscatter can be viewed as a random variable with mean given by the geophysical model function and variability due to unmodelled effects and measurement errors. Little is known about the variability due to unmodelled effects, or the statistics of this variability; this paper presents some considerations and simulations to estimate the magnitude of the model function error.

View PDFchevron_right
Wind retrieval with cross-polarized SAR returns
William Perrie

2011 IEEE International Geoscience and Remote Sensing Symposium, 2011

View PDFchevron_right
Neural network wind retrieval from ERS-1 scatterometer data
Carlos Mejia

Neurocomputing, 2000

This paper presents a neural network methodology to retrieve wind vectors from ERS-1 scatterometer data. First, a neural network (NN-INVERSE) computes the most probable wind vectors. Probabilities for the estimated wind direction are given. At least 75% of the most probable wind directions are consistent with European Centre for Medium-Range Weather Forecasts winds (at _+20ø). Then the remaining ambiguities are resolved by an adapted PRESCAT method that uses the probabilities provided by NN-INVERSE. Several statistical tests are presented to evaluate the skill of the method. The good performance is mainly due to the use of a spatial context and to the probabilistic approach adopted to estimate the wind direction. Comparisons with other methods are also presented. The good performance of the neural network method suggests that a selfconsistent wind retrieval from ERS-1 scatterometer is possible. 8737 8738 RICHAUME ET AL.: NEURAL NETWORK WIND RETRIEVAL a) 2 360 ø 150 ø 210 ø 330 ø 3 o o 120 ø 240 ø 90 ø 270 ø 3 RICHAUME ET AL.: NEURAL NETWORK WIND RETRIEVAL 8739 RICHAUME ET AL.: NEURAL NETWORK WIND RETRIEVAL ET AL.: NEURAL NETWORK WIND RETRIEVAL RICHAUME ET AL.: NEURAL NETWORK WIND RETRIEVAL ET AL.: NEURAL NETWORK WIND RETRIEVAL 8751 where v is the wind speed, 0 is the wind azimuth, and V the wind vector. These numbers are computed with the residuals defined in Appendix 1 except for wind vector RMS Vrms, which is defined with the following residual: 2 __ 2 2 COS(,¾res) ' Vre s '-•/V• n t-V .... f 2VaV .... f Assuming bias free estimators, FOM = 1 if specifications of _+2 m s -• in wind speed error and _+20 ø in wind direction error are exactly fulfilled. The greater the FOM, the better the inverse wind field.

View PDFchevron_right

References (33)

  1. Abrahamsen, P., A Review of Gaussian Ran- dom Fields and Correlation Functions, Sec- ond Edition, Technical Report No. 917, Nor- wegian Computing Center, Box 114 Blindern, N-0314 Oslo, NORWAY, 1997. Available from http://www.nr.no/research/sand/articles.html
  2. Bishop, C. M., Neural Networks for Pattern Recognition, 482 pp., Oxford University Press, Oxford, 1995.
  3. Bishop, C. M., Mixture Density Networks, Technical Report NCRG/94/004, Neural Computing Research Group, Aston University, Birmingham, UK, 1994. Available from: http://www.ncrg.aston.ac.uk/ Cornford, D., I. T. Nabney, and G. Ramage, Improved Neural Network Scatterometer Forward Models, Jour- nal of Geophysical Research, submitted, 2000.
  4. Cornford, D., I. T. Nabney, and C. K. I. Williams, Adding Constrained Discontinuities to Gaussian Process Mod- els of Wind Fields, in Advances in Neural Information Processing Systems 11, M. S. Kearns, S.A. Solla and C. A. Cohn (eds), MIT Press, Cambridge, Massachusetts, 1999. Available from http://www.ncrg.aston.ac.uk/.
  5. Cornford, D., Non-Zero Mean Gaussian Process Prior Wind Field Models, Technical Report NCRG/98/020,Neural Computing Research Group, Aston University, Birmingham, U.K., 1998. Available from http://www.ncrg.aston.ac.uk/.
  6. Cornford, D., Flexible Gaussian Process Wind Field Models, Technical Report NCRG/98/017,Neural Computing Research Group, Aston Univer- sity, Birmingham, U.K., 1998. Available from http://www.ncrg.aston.ac.uk/.
  7. Cressie, N. A. C., Statistics for Spatial Data, John Wiley and Sons, New York, 1993.
  8. Daley, R., Atmospheric Data Analysis, 457 pp., Cam- bridge University Press, Cambridge, 1991.
  9. Dickinson, S. and R. A. Brown, A Study of Near-Surface Winds in Marine Cyclones Using Multiple Satellite Sensors, Journal of Applied Meteorology, 35, 769-781, 1996.
  10. Evans, D. J., D. Cornford, and I. T. Nabney, Structured Neural Network Modelling of Multi-valued Functions for Wind Vector Retrieval from Satellite Scatterome- ter Measurements, Neurocomputing Letters, accepted, 1999.
  11. Figa, J. and Stoffelen, A., On the Assimilation og Ku- Band Scatterometer Winds for Weather Analysis and Forecasting, IEEE Transactions on Geoscience and Re- mote Sensing, submitted, 1999.
  12. Gilks, W.R., S. Richardson, and D. J. Spiegelhalter, Markov Chain Monte Carlo in Practice, 486 pp., Chap- man and Hall, London, 1996.
  13. Haltiner, G. J. and R. T. Williams, Numerical Prediction and Dynamic Meteorology, 477 pp., John Wiley, Chich- ester, 1980.
  14. Hoffman, R. N., SASS Wind Ambiguity Removal by Di- rect Minimization. Part II: Use of Smoothness and Dy- namical Constraints, Monthly Weather Review, 112, 1829-1852, 1984.
  15. Homleid, M. and L-A. Breivik, Preparations for the As- similation of ERS-1 Surface Wind Observations into Numerical Weather Prediction Models, Tellus, 47A, 62-79, 1995.
  16. Levy, G., Southern-Hemisphere Low-Level Wind Circula- tion Statistics from the SeaSat Scatterometer. Annales Geophysicae -Atmospheres, Hydroshperes and Space Sciences, 12, 65-79, 1994.
  17. Long, D. G. Wind Field Model-Based Estimation of SEASAT Scatterometer Winds, Journal of Geophysi- cal Research, 98, 14651-14668, 1993.
  18. Lorenc, A. C.,, Analysis Methods for Numerical Weather Prediction, Quarterly Journal of the Royal Meteorolog- ical Society, 112, 1177-1194, 1986.
  19. Lorenc, A. C., R. S. Bell, S. J. Foreman, C. D. Hall, D. L. Harrison, M. W. Holt, D. Offiler, and S. G. Smith, The Use of ERS-1 Products in Operational Meteorology, Advances in Space Research, 13, 19-27, 1993.
  20. Morgan, N. and Boulard, H. A., Neural Networks for Sta- tistical Recognition of Continuous Speech, Proceedings of the IEEE, 83, 742-770, 1995.
  21. Offiler, D., ERS-1 wind retrieval algorithms, U.K. Mete- orological Office Memorandum No. 86, Meteorological Office, Bracknell, UK, 1987.
  22. Offiler, D., The Calibration of ERS-1 Satellite Scatterom- eter Winds, Journal of Atmospheric and Oceanic Tech- nology, 11, 1002-1017, 1994.
  23. Richaume, P., F. Badran, M. Crepon, C. Mejia, H. Roquet, S. Thiria, Neural Network Wind Retrieval from ERS-1 Scatterometer Data, Journal of Geophysi- cal Research, submitted.
  24. Royle, J. A., L. M. Berliner, C. K. Wikle and R. Milliff, A Hierarchical Spatial Model for Constructing Wind Fields from Scatterometer Data in the Labrador Sea, in Case Studies in Bayesian Statistics, (in review).
  25. Schyberg, H. and L-A. Breivik, Ambiguity Removal Algo- rithm Evaluation, Research Report No. 64, Norwegian Meteorological Institute, Oslo, Norway, 1998.
  26. Shaffer, S. J., R. S. Dunbar, S. V. Hsiao, and D. G. Long, A Median-Filter-Based Ambiguity Removal Algorithm for NSCAT, IEEE Transactions on Geoscience and Re- mote Sensing, 29, 167-174, 1991.
  27. Stoffelen, A., Scatterometry, PhD Thesis, Universiteit Utrecht, 1998.
  28. Stoffelen, A., Error Modelling and Calibration; Towards the True Near Surface Wind Speed. Journal of Geo- physical Research, 103, 7755-7766, 1998.
  29. Stoffelen, A. and D. Anderson, Scatterometer Data In- terpretation: Estimation and Validation of the Trans- fer Function CMOD4. Journal of Geophysical Research, 102, 5767-5780, 1997.
  30. Stoffelen, A. and D. Anderson, Ambiguity Removal and Assimilation of Scatterometer Data, Quarterly Journal of the Royal Meteorological Society, 123, 491-518, 1997.
  31. Tarantola, A., Inverse Problem Theory, Elsevier, London, 613pp, 1987.
  32. Williams, C. K. I., Prediction with Gaussian Processes: From Linear Regression to Linear Prediction and Be- yond, in Learning and Inference in Graphical Models, M. I. Jordan (ed.), Kluwer, London, 1998.
  33. D. Cornford and Ian T. Nabney and David J. Evans, Neural Computing Research Group, Electronic Engineering and Computer Science, Aston University, Aston Triangle, Birmingham B7 4ET, UK (e-mail: d.cornford@@aston.ac.uk, i.t.nabney@@aston.ac.uk) This preprint was prepared with AGU's L A T E X macros v5.01. File fwd˙ncrg formatted July 2, 2007.

Related papers

Bayesian Analysis of the Scatterometer Wind Retrieval Inverse Problem: Some New Approaches
Dan Cornford

Journal of the Royal Statistical Society Series B: Statistical Methodology, 2004

SummaryThe retrieval of wind vectors from satellite scatterometer observations is a non-linear inverse problem. A common approach to solving inverse problems is to adopt a Bayesian framework and to infer the posterior distribution of the parameters of interest given the observations by using a likelihood model relating the observations to the parameters, and a prior distribution over the parameters. We show how Gaussian process priors can be used efficiently with a variety of likelihood models, using local forward (observation) models and direct inverse models for the scatterometer. We present an enhanced Markov chain Monte Carlo method to sample from the resulting multimodal posterior distribution. We go on to show how the computational complexity of the inference can be controlled by using a sparse, sequential Bayes algorithm for estimation with Gaussian processes. This helps to overcome the most serious barrier to the use of probabilistic, Gaussian process methods in remote sensi...

View PDFchevron_right
A New Algorithm for Wind-Vector Retrieval From Scatterometers
Vijay Agarwal

IEEE Geoscience and Remote Sensing Letters, 2008

A new efficient algorithm for retrieving wind-vector solutions from scatterometers is developed based on a criterion of minimum normalized standard deviation (NSD) of wind speed derived from backscatter measurements using a geophysical model function (GMF). Its performance has been evaluated through simulations using QSCAT-1 GMF and the QuikSCAT observational geometry. The present algorithm, named the NSD algorithm, is found to be computationally more efficient (two to three times) besides being at par with the maximum-likelihood estimator (MLE) algorithm in terms of retrieval skill, retrieval errors, and distribution of solutions, on the basis of simulations as well as comparison of limited QuikSCAT-data-derived winds with National Centers for Environmental Prediction and European Centre for Medium-Range Weather Forecasts model winds. Simulation results and analysis of sample QuikSCAT data are presented.

View PDFchevron_right
On bayesian scatterometer wind inversion
M. Portabella, Ad Stoffelen

2006

In a quest for a generic unbiased scatterometer wind inversion method, the different inversion procedures currently in use are revisited in this paper. A careful examination of both the errors in the wind and in the measurement domain, combined with the non-linear shape of the Geophysical Model Function, GMF, leads to a generic and novel Bayesian wind retrieval approach in the measurement domain. In this approach the shape of the GMF solution manifold in measurement space is more important than the specified noise. This shape is related to the system wind direction sensitivity, and when this sensitivity is uniform, realistic and precise wind direction distributions are retrieved, even when measurements lie far from the GMF manifold. A simplified measurement space transformation that produces such uniform sensitivity for the European Remote Sensing Satellite (ERS) scatterometer is presented and shown to have reduced wind direction bias compared to the more traditional 1 (measurement-noise normalized) inversion for ERS. Moreover, the simplified wind inversion reveals a similar performance to the current operational ERS wind inversion, but is potentially more generally applicable. The simplified method is then applied to SeaWinds but is ineffective. In this case the instrument geometry results in a low sensitivity to wind direction at a few specific directions. As a consequence, certain wind direction solutions remain favoured in the SeaWinds inversion.

View PDFchevron_right
A method for including mesoscale and synoptic-scale information in scatterometer wind retrievals
Ralph Foster

Journal of Geophysical Research, 2010

A method for improving scatterometer wind retrievals based on the mesoscale and synoptic-scale structure of the flow field is presented and evaluated. The large-scale structure of the flow is inferred from the sea-level pressure (SLP) field derived from the scatterometer winds using a planetary boundary layer model. The wind vectors derived from this SLP field can be used to either (1) inform the ambiguity selection, (2) correct the direction of the scatterometer wind vectors, in all cases or above a certain threshold, or (3) replace the scatterometer winds. The methodology is demonstrated with QuikSCAT (QS) scatterometer wind vectors. The new wind vector set is evaluated statistically by comparison with buoy measurements, with numerical weather prediction model analyses, and by spectral analysis. It is found that the new wind vectors are particularly valuable at nadir and in rain-contaminated areas, and that their spectral behavior is closer to a power law than are the uncorrected QS winds.

View PDFchevron_right
Bayesian inference for wind field retrieval
Ian T Nabney

Neurocomputing, 2000

In many problems in spatial statistics it is necessary to infer a global problem solution by combining local models. A principled approach to this problem is to develop a global probabilistic model for the relationships between local variables and to use this as the prior in a Bayesian inference procedure. We show how a Gaussian process with hyper-parameters estimated from Numerical Weather Prediction Models yields meteorologically convincing wind elds. We use neural networks to make local estimates of wind vector probabilities. The resulting inference problem cannot be solved analytically, but Markov Chain Monte Carlo methods allow us to retrieve accurate wind elds.

View PDFchevron_right
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved