Efficient data assimilation to correct non-Gaussian forecast errors in numerical weather predictionUniversity of Exeter Ref: 1203
Location: Streatham Campus, University of Exeter, EX4 4QJ
Primary supervisor: Dr Frank Kwasniok, University of Exeter
Secondary supervisor: Dr Gordon I nverarity, Met Office, Exeter
Data assimilation algorithms such as three- and four-dimensional variational data assimilation (3D- and 4D-Var, respectively) perform a least-squares minimisation of a cost function that includes terms penalising the forecast error at the start of a window containing observations to be assimilated and the combined observation and representation error for each observation. A key assumption is that the forecast errors to be corrected are Gaussian.
However, this assumption becomes increasingly invalid as the time between assimilation cycles increases and as the forecast model becomes more nonlinear at higher resolutions. The logarithm of total aerosol is one such quantity whose training data used to estimate the forecast-error covariance matrix can be shown to be non-Gaussian. This quantity is required for visibility assimilation, which is an important element of fog and air quality forecasting. An alternative probability transformation of the control variable that makes the forecast error more Gaussian has the potential to improve the accuracy of these forecasts.
This project could also yield computational cost savings for the Met Office. The observation operator that converts the logarithm of total aerosol into simulated visibility observations for comparison with actual observations is strongly nonlinear. Efficient minimisation methods that are used for global data assimilation, where visibility is not assimilated, could also be used for UK data assimilation, where visibility is assimilated, if the observation operator associated with a better probability transformation was only weakly nonlinear. This would be a significant factor in helping to deliver computationally affordable convective-scale 4D-Var (the Met Office currently uses 3D-Var for its operational UK data assimilation).
This project aims to investigate how non-Gaussian forecast errors can be better handled in 3D- and 4D-Var. More specifically, it aims to improve the assimilation of visibility data in numerical weather prediction and help implement better techniques in the Met Office's operational data assimilation system which significantly improves the efficiency and accuracy of the computationally expensive minimisation process.
Application criteriaApplicants should have or expect to achieve at least a 2:1 Honours degree, or equivalent, in mathematics, statistics, physics or meteorology. Candidates should have a keen interest in the application of mathematics and statistics in weather and climate science.
Primary supervisor: Dr Frank Kwasniok, University of Exeter
Secondary supervisor: Dr Gordon I nverarity, Met Office, Exeter
Data assimilation algorithms such as three- and four-dimensional variational data assimilation (3D- and 4D-Var, respectively) perform a least-squares minimisation of a cost function that includes terms penalising the forecast error at the start of a window containing observations to be assimilated and the combined observation and representation error for each observation. A key assumption is that the forecast errors to be corrected are Gaussian.
However, this assumption becomes increasingly invalid as the time between assimilation cycles increases and as the forecast model becomes more nonlinear at higher resolutions. The logarithm of total aerosol is one such quantity whose training data used to estimate the forecast-error covariance matrix can be shown to be non-Gaussian. This quantity is required for visibility assimilation, which is an important element of fog and air quality forecasting. An alternative probability transformation of the control variable that makes the forecast error more Gaussian has the potential to improve the accuracy of these forecasts.
This project could also yield computational cost savings for the Met Office. The observation operator that converts the logarithm of total aerosol into simulated visibility observations for comparison with actual observations is strongly nonlinear. Efficient minimisation methods that are used for global data assimilation, where visibility is not assimilated, could also be used for UK data assimilation, where visibility is assimilated, if the observation operator associated with a better probability transformation was only weakly nonlinear. This would be a significant factor in helping to deliver computationally affordable convective-scale 4D-Var (the Met Office currently uses 3D-Var for its operational UK data assimilation).
This project aims to investigate how non-Gaussian forecast errors can be better handled in 3D- and 4D-Var. More specifically, it aims to improve the assimilation of visibility data in numerical weather prediction and help implement better techniques in the Met Office's operational data assimilation system which significantly improves the efficiency and accuracy of the computationally expensive minimisation process.
Application criteriaApplicants should have or expect to achieve at least a 2:1 Honours degree, or equivalent, in mathematics, statistics, physics or meteorology. Candidates should have a keen interest in the application of mathematics and statistics in weather and climate science.
Summary Application deadline: 19th April 2013
Number of awards: 1
Value: Three-and-a-half Year
Studentship: Tuition fees (UK/EU) and an annual maintenance allowance at
current research council rate.
Duration of award: per year
Contact: Liz Roberts emps-pgr@exeter.ac.uk
How to ApplyTo apply, you must click 'Apply' below. You will be asked to submit some personal details and upload a full CV, covering letter and details of two academic referees. Your covering letter should outline your academic interests, prior research experience and reasons for wishing to undertake this project.
For general enquiries please contact Liz Roberts at emps-pgr@exeter.ac.uk
´
More information: http://www.jobs.ac.uk/job/AGE485/mathematics-phd-studentship-in-weather-and-climate-science/
For general enquiries please contact Liz Roberts at emps-pgr@exeter.ac.uk
´
More information: http://www.jobs.ac.uk/job/AGE485/mathematics-phd-studentship-in-weather-and-climate-science/
No hay comentarios:
Publicar un comentario