Improved multi-beam neural network scatterometer forward models

Dan Cornford, Ian T. Nabney, Guillaume Ramage

Research output: Working paperTechnical report

Abstract

Current methods for retrieving near surface winds from scatterometer observations over the ocean surface require a foward sensor model which maps the wind vector to the measured backscatter. This paper develops a hybrid neural network forward model, which retains the physical understanding embodied in ¸mod, but incorporates greater flexibility, allowing a better fit to the observations. By introducing a separate model for the mid-beam and using a common model for the fore- and aft-beams, we show a significant improvement in local wind vector retrieval. The hybrid model also fits the scatterometer observations more closely. The model is trained in a Bayesian framework, accounting for the noise on the wind vector inputs. We show that adding more high wind speed observations in the training set improves wind vector retrieval at high wind speeds without compromising performance at medium or low wind speeds.
Original languageEnglish
Place of PublicationBirmingham, UK
PublisherAston University
Number of pages18
ISBN (Print)NCRG/2000/008
Publication statusPublished - 2000

Fingerprint

scatterometer
wind velocity
surface wind
backscatter
sea surface
sensor

Keywords

  • surface winds
  • scatterometer observations
  • foward sensor model
  • wind vector
  • measured backscatter
  • neural network forward model
  • Bayesian framework
  • wind speeds

Cite this

Cornford, D., Nabney, I. T., & Ramage, G. (2000). Improved multi-beam neural network scatterometer forward models. Birmingham, UK: Aston University.
Cornford, Dan ; Nabney, Ian T. ; Ramage, Guillaume. / Improved multi-beam neural network scatterometer forward models. Birmingham, UK : Aston University, 2000.
@techreport{4b377f633b7149bba5c97bf19a79e7bd,
title = "Improved multi-beam neural network scatterometer forward models",
abstract = "Current methods for retrieving near surface winds from scatterometer observations over the ocean surface require a foward sensor model which maps the wind vector to the measured backscatter. This paper develops a hybrid neural network forward model, which retains the physical understanding embodied in ¸mod, but incorporates greater flexibility, allowing a better fit to the observations. By introducing a separate model for the mid-beam and using a common model for the fore- and aft-beams, we show a significant improvement in local wind vector retrieval. The hybrid model also fits the scatterometer observations more closely. The model is trained in a Bayesian framework, accounting for the noise on the wind vector inputs. We show that adding more high wind speed observations in the training set improves wind vector retrieval at high wind speeds without compromising performance at medium or low wind speeds.",
keywords = "surface winds, scatterometer observations, foward sensor model, wind vector, measured backscatter, neural network forward model, Bayesian framework, wind speeds",
author = "Dan Cornford and Nabney, {Ian T.} and Guillaume Ramage",
year = "2000",
language = "English",
isbn = "NCRG/2000/008",
publisher = "Aston University",
type = "WorkingPaper",
institution = "Aston University",

}

Cornford, D, Nabney, IT & Ramage, G 2000 'Improved multi-beam neural network scatterometer forward models' Aston University, Birmingham, UK.

Improved multi-beam neural network scatterometer forward models. / Cornford, Dan; Nabney, Ian T.; Ramage, Guillaume.

Birmingham, UK : Aston University, 2000.

Research output: Working paperTechnical report

TY - UNPB

T1 - Improved multi-beam neural network scatterometer forward models

AU - Cornford, Dan

AU - Nabney, Ian T.

AU - Ramage, Guillaume

PY - 2000

Y1 - 2000

N2 - Current methods for retrieving near surface winds from scatterometer observations over the ocean surface require a foward sensor model which maps the wind vector to the measured backscatter. This paper develops a hybrid neural network forward model, which retains the physical understanding embodied in ¸mod, but incorporates greater flexibility, allowing a better fit to the observations. By introducing a separate model for the mid-beam and using a common model for the fore- and aft-beams, we show a significant improvement in local wind vector retrieval. The hybrid model also fits the scatterometer observations more closely. The model is trained in a Bayesian framework, accounting for the noise on the wind vector inputs. We show that adding more high wind speed observations in the training set improves wind vector retrieval at high wind speeds without compromising performance at medium or low wind speeds.

AB - Current methods for retrieving near surface winds from scatterometer observations over the ocean surface require a foward sensor model which maps the wind vector to the measured backscatter. This paper develops a hybrid neural network forward model, which retains the physical understanding embodied in ¸mod, but incorporates greater flexibility, allowing a better fit to the observations. By introducing a separate model for the mid-beam and using a common model for the fore- and aft-beams, we show a significant improvement in local wind vector retrieval. The hybrid model also fits the scatterometer observations more closely. The model is trained in a Bayesian framework, accounting for the noise on the wind vector inputs. We show that adding more high wind speed observations in the training set improves wind vector retrieval at high wind speeds without compromising performance at medium or low wind speeds.

KW - surface winds

KW - scatterometer observations

KW - foward sensor model

KW - wind vector

KW - measured backscatter

KW - neural network forward model

KW - Bayesian framework

KW - wind speeds

M3 - Technical report

SN - NCRG/2000/008

BT - Improved multi-beam neural network scatterometer forward models

PB - Aston University

CY - Birmingham, UK

ER -

Cornford D, Nabney IT, Ramage G. Improved multi-beam neural network scatterometer forward models. Birmingham, UK: Aston University. 2000.