ISRO / UKSA Joint Working Group University of Reading 29 March 2011 Use of observational information for convective-scale predictions Stefano Migliorini University of Reading [email protected]with contributions from: Ross Bannister, Mark Dixon, Stephen English, Graeme Kelly, Cristina Prates, Randhir Singh and Robert Tubbs
15
Embed
ISRO / UKSA Joint Working Group University of Reading29 March 2011 Use of observational information for convective-scale predictions Stefano Migliorini.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ISRO / UKSA Joint Working Group University of Reading29 March 2011
with contributions from: Ross Bannister, Mark Dixon, Stephen English, Graeme Kelly, Cristina Prates, Randhir
Singh and Robert Tubbs
2
Contents
• Probabilistic forecasting at convective-scale
• Operational radiative transfer modelling for NWP
• Assimilation of cloudy radiances
• Advanced wind measurements from space
High-resolution numerical weather predictions
• Increase of computer power allows met agencies to run NWP forecasts at increasing resolution: deep-convection permitting models (~1 km res)
• Less need for parameterizations (e.g. convection) and better topography and surface fields
• Better use of high-res observation
• Aim is to improve forecasts of small-scale processes, storms, hazardous weather
3
Predictability limits• Error growth timescales decrease with horizontal scale. It
is estimated that a 20-km wavelength (e.g. thunderstorm) has a predictability horizon of ~1-2h
• Errors are always present: initial and boundary conditions, model errors (e.g. approximations, resolution, parameterizations)
4
Ensemble prediction at convective scale
• Error growth at convective scale stronger in the presence of convective instability: need for accurate representation of ongoing convection at initial time (e.g., from radar), particularly for nowcasting (0-6 hours).
• Large sensitivity to initial conditions and to model error motivates the need for probabilistic forecasts at convective scale.
• Initial-condition uncertainty for nowcasting should depend on available observational information content: ensemble filtering
• We also obtain flow-dependent (but low-rank) description of forecast uncertainty that can be used for high resolution data assimilation
5
The Met Office EPS
6
• Met Office Global and Regional Ensemble Prediction System (MOGREPS)
• BC from the global ensemble
• IC calculated using an ETKF
to generate 23 increments
Ensembles at convective scale
• 1.5 km resolution over 70 vertical levels over southern UK
• Ensemble of 23 forecast members + control
• Initial 1.5 km perts interpolated from NAE EPS fields at 24 km and 1.5 km operational analysis
7
Case study: 26 July 2007• active front, line convection, strong dynamical forcing
9
• T+1 forecast valid at 12 UTC, with LH nudging acting between 10.30 and 11.30.
• Substantial variability that “compensate” shortcomings in deterministic forecast (member 0): heavy rainfall over the north coast of South West England captured by members 2, 3, 4, 6 and 18; light precip in the East by several member (e.g., 11)
Case study: forecasts of precipitation
Sensing the Earth from space
• Line-by-line models calculates absorption coefficients for each spectral line for all relevant molecules over layers
• Radiative transfer equation solved; top-of-atmosphere monochromatic radiance convolved with instrument response; computationally expensive
• “fast” RT codes (e.g. RTTOV) determine transmittance from optical depth calculated as regression of set of predictors (T, WV)
10
2
1
22 1 1 2
( , )( ) ( ) ( , ) ( ( ))
s
s
d s sL s L s s s B T s ds
ds
2
1
1 20
( , ) exp( ( ( )) ( ) )abs
s N
m mms
s s s s ds
• A fast radiative transfer model (RTM) to compute emitted infrared radiances for a very high resolution radiometer (VHRR), onboard the operational Indian geostationary satellite Kalpana has been developed and verified (Singh et al., 2009) . This work is a step towards the assimilation of Kalpana water vapor (WV) radiances into numerical weather prediction models.
• Satellite observations regarding clouds can be assimilated as either raw radiances or as retrieved parameters.
• RTMs used in assimilation assume a simple cloud model: a single layer of geometric thin grey cloud
12
Cld Cl Opc c cL ( ) (1 N )L ( ) N L ( ,p )
Cloud Parameters: Effective Cloud Fraction (ECF) and Cloud Top pressure (CTP)
Cloud-free radiance Radiance for 100% opaque cloud cover
• Statistics of 1D-Var retrieval obtained from a set of 200 pseudo IASI observations where a layer of cirrus cloud type is placed above a layer of cumulus cloud. The true cloud top is represented in orange and the retrieved in blue.
• New two-layer cloud model to increase the number of assimilated cloud-affected radiances
13
cld clr op ope1 e2 e1 1 e2 e1 2L ( ) (1 N )(1 N )L ( ) N L (p , ) N (1 N )L (p , )
Assimilation of cloudy radiances (2/2)
Thanks to Cristina Prates
Atmospheric motion vectors for NWP
• Suitable features will be identified on forecasts at different lead times to determine a set of simulated AMVs.
• The relationship between estimated AMVs from simulated SEVIRI data and “true” atmospheric winds will be investigated, to improve the expression of the observation operator used for data assimilation.
• Experiments will be performed with model data at different horizontal resolutions, ranging from 1.5 km to 500 m.
14
Court
esy
Met
Offi
ce
Conclusions
15
• Accurate weather prediction relies on accurate observations and reliable models
• Uncertainty is inevitable (finite predictability) and needs to be quantified for meaningful predictions
• Satellite information is essential to guarantee global coverage
• Still a lot of research needs to be done to improve predictions at high-resolution (clouds, storms, high-impact weather)