Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and...

8
© Crown copyright 2005 Page 1 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data

description

Page 3© Crown copyright 2005 Mid-level mean cloud fractions  The mid-level cloud fractions are much reduced compared to observations  Direct modification of the diagnostic ice cloud fraction to better improve the mid-level cloud fractions, as in HadGAM

Transcript of Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and...

Page 1: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 1

Damian Wilson, 12th October 2005

Assessment of model performance and potential improvements using CloudNet data

Page 2: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 2

Summary

CloudNet data has allowed a long period comparison of cloud properties between models and observations

There is a generally good agreement between the models (including the Met Office) and observations, better than is generally thought

However, there are some specific differences which we can address (for the Met Office)

Microphysical information is also available which can guide future parametrization development

Page 3: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 3

Mid-level mean cloud fractions

The mid-level cloud fractions are much reduced compared to observationsDirect modification of the diagnostic ice cloud

fraction to better improve the mid-level cloud fractions, as in HadGAM

Page 4: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 4

Low level cloud

The low level cloud peak is at too low an altitudeA boundary layer problem?

The model has a large amount of fog and very low cloud in stable boundary layersPossibly due to inaccuracies in the stable boundary

layer turbulent flux profiles JJA: Downwards, stable

HadGAM

PC2

m

Observations

Page 5: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 5

Thick liquid water contents

There are not enough of the highest liquid water contentsPossibly from not including the convective cloud.

The representation of liquid water in convection schemes should be investigated

Possibly a poor representation of the drizzle process, with too ready autoconversion

Page 6: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 6

Supercooled liquid water content

The mesoscale model (and to a lesser extent, the global model) has significantly less supercooled liquid water than the observationsPossible adjustment of the overlap

between liquid and ice in mixed phase clouds

Page 7: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 7

Cloud water PDFs

The histogram of cloud fraction and liquid water contents for low level cloud suggests not enough cloud fractions of 1Tuning the PDF shape and critical relative humidity

values should improve the model

Page 8: Page 1© Crown copyright 2005 Damian Wilson, 12 th October 2005 Assessment of model performance and potential improvements using CloudNet data.

© Crown copyright 2005 Page 8

Other uses of CloudNet data

CloudNet has provided more detailed information which could be used in developing better parametrizations in the future. These include: Ice particle size distributions Inhomogenerity information and cloud overlaps to

inform the radiation scheme assumptionsDiagnostic area cloud fraction representationDrizzle parametrization improvements