Report on




held on October 7th, 2003 during


the 25th EWGLAM and 10th SRNWP Meetings in Lisbon, Portugal












Jean Quiby (MeteoSwiss)

Jean-Marie Bettems (MeteoSwiss)

Hans Roozekrans (KNMI)

1. Introduction

The main objective of the CLOUDMAP2 project is to develop and to introduce new satellite derived cloud products in the meteorological and climate research communities. The essential success factor to meet this objective is to directly communicate with the anticipated end-users. As already proven in the CLOUDMAP1 project the concept of a “roadshow” along relevant user institutes / communities is a very effective way to promote the project results and to obtain feedback. This roadshow is organised by KNMI and embraces a total of five workshops This second roadshow workshop was an integrated part of the 25th EWGLAM and 10th SRNWP Meetings in Lisbon, Portugal.

The focus of this second roadshow workshop was to evaluate the potential use of cloud observations in general and CLOUDMAP2 products in particular in numerical weather prediction (NWP). The yearly EWGLAM and SRNWP meetings are attended by the NWP community working in the field of (mostly regional) NWP modeling. This meeting was therefore an excellent opportunity for this roadshow meeting.  Around 80 people attended this workshop (the names are listed in the appendix). The workshop started at 13.30 AM and ended at 15.30 AM.

The CLOUDMAP2 project team was represented by:

Hans Roozekrans (KNMI, The Netherlands)

Jean Quiby (MeteoSwiss, Switzerland)



2. Presentation on CLOUDMAP2 project and products

Hans Roozekrans of KNMI started the workshop with a half-hour presentation of the C2 project and products. The presentation included an overview of the project partners, a description of the objectives and the deviations from the C1 project. Also NWP work done at SMHI and KNMI in the framework of CLOUDMAP2 was presented. The C2 Web site was named as a valuable reference for further information (



3. Discussion part

After this presentation Hans Roozekrans started the discussion part by presenting the focus of this workshop, namely the potential use of cloud observations in NWP:

·         Assimilation of cloud observations in NWP models

·         Use of cloud observations for NWP model monitoring

·         Use of cloud observations for parameterisation work

A total of 8 q

uestions/topics were prepared to lead the discussion. Questions and a summary of the answers are reported below. The minutes of the discussion were kindly written by Jean Quiby and Jean-Marie Bettems of MeteoSwiss.


1. What are the most relevant EO derived observations for NWP?

How important are cloud observations for NWP?

The most important satellite data today are the conventional ones or, in other terms, the ones used operationally. Primarily the temperature and water vapour profiles derived from the ATOVS sounders of the NOAA satellites. Secondarily the METEOSAT cloud winds. The future of these data has to be assured.

From the non-conventional satellite data, the most important for the global models are or will be the surface winds derived from the QuickSCAT satellites over seas and oceans.

For the Limited Area Models (LAM) the most relevant EO derived data will be the ZTD (Zenital Total Delays) deduced from the GPS (Global Positioning System) satellites. From these delays, the vertically integrated water vapour can be computed.

Really important today is only the cloudiness of the geostationary satellites when it is used to derived winds. At some national weather services (NWS), the model cloudiness is operationally verified with METEOSAT images.


2. Will radiance-based data-assimilation be the future trend for all NWP models (global and regional)?

A look into the past will strongly influence the answer to be given to this question.

Let us consider what happened with the SATEMP data: this product (temperature profiles) has been largely used at the beginning until the NWP satellite specialists proved that the direct assimilation of the radiances gives better results. Thus the answer to this question is “yes” for the global models as well as for the regional ones. Assimilation of radiances will remain the most elegant way to use the satellite information.


3. If so, is there still a role for EO products (retrieved geophysical parameters) in NWP (e.g. for verification purposes)?

It is presumably correct to anticipate that these products will have the same destiny as the SATEM data. In the first years of their existence, they will be used. Then the NWP groups will prefer to have the primary data (for example radiances) to develop these products themselves. The trend to use primary data is already visible for the GPS data: several NWP centres already prefer to compute the IWV (integrated water vapour) themselves directly from the ZTD (Zenital Total Delay) instead of receiving from the GPS processing centres the IWV.

What about the MODIS derived products?

From the MODIS radiances, many products are now operationally derived in the framework of the CLOUDMAP2 project. Is the NWP community interested in these products?:

= CTP (Cloud Top Pressure), CTH (Cloud Top Height) and CTT (Cloud Top Temperature): It is too early today to say whether these parameters will be assimilated in the future. The only use that can be foreseen in the near future is for model validation. Example: has the model a tendency to have cirrus cloud too high or too low?

= Cloud water: From the NWP point of view, cloud water can only be assimilated if the other parameters like the vertical wind component or the relative humidity are correct, otherwise the cloud will disappear (in case of downward wind in the model) or evaporate (if the relative humidity is too low).

It has nevertheless been stressed that we should not wait till we have "perfect" analyses to introduce these new parameters. Even today new parameters can still bring some improvement, at least in the first hours of the forecasts.


4. The current EO observation resolution is 1 km. Is there a potential use for such detailed data?

Contradictory opinions have been expressed here.

Arguments against:

- Already today we do not use for the global models the ATOVS data with their full horizontal resolution. Only a subset of the possible profiles are used. No averaging or super-observations are made today.

- The amount of data will become gigantic and give a real problem to the National Weather Services: the overload in communications, processing and archiving will become disproportionate in the resources necessary for the production of numerical forecasts.

Arguments in favour:

- the horizontal resolution of the NWP models is steadily increasing. Consequently the spatial density of the observation must also increase in order to have the small-scale features in the analysis. When the regional models will have an operational resolution of 1 to 3 km they will need an EO observation resolution of 1 km.

- Very high resolution maybe useful for model validation.


5. CLOUDMAP2 aims to provide independent cloud observations. How valuable/important is this for NWP?

It is definitely important. In order to determine the heights of the winds derived from cloud motion as yielded by the geostationary satellites, information from global models are used. But these models assimilate the same satellites winds. This feedback is a negative feature in this process. A model independent procedure for the determination of the cloud wind heights would be a great advantage. This is theoretically possible with the use of the stereo-matching technique with sensors like MISR or ATSR2.


6. What is the role for ground based RS cloud data in NWP, e.g. the valuable ARM sites around Europe?

Very few comments have been made. It is difficult to answer this question for the future. For the present time, no ground based RS data are used operationally. Maybe the NWP community will start to use them for parameterisation development work and validation purposes.


7. Error characteristics of observations are important for NWP. Can satellite observations contribute to derive theses?

Yes. But this property is not restricted to the satellite observations. Any observing system can help to derive the error characteristics of another observing system. The specification of the error characteristics of a new observing system can be best derived when it is confronted to several other observing systems.


8. Is more ocean information required, as land is better covered by synops/surface observations?

Oceans are poorly observed. The only observing system able to fill these large data sparse areas is the space observing system. Both the sounders and the imagers can contribute, as well as active sensors like the ones mounted on QuickSCAT. But many satellite observations, like the profiles derived from the sounders, are best used when they can be “anchored” to a surface pressure or a geopotential surface height. Thus in situ measurements remain necessary. The great advantage of the oceans when compared to the land areas is the radiative homogeneity of the surface.

In relation to this what will be the value of satellite observations in the context of a degrading “conventional” observation network?

Or: will the satellite be able to replace the conventional observations?

The number of the classical observing stations, surface (SYNOP) and upper-air (TEMP), has already diminished in Europe and worldwide and will continue to diminish (cf. the EUCOS plan). Will the satellite be able to fill the gaps?

To fill the gaps, the NWP community has no choice: the use of the satellite information in the determination of the initial conditions of their models will increase where it is possible. With the rarefaction of the upper air sounding stations, an effort should be made for a stronger use of the AMSU B or AIRS data as a replacement for the loss of the humidity profiles. The use of the SSMI, MODIS or GPS data, which allow a determination of the vertically integrated water vapour, should also be strongly encouraged.

Important statement: The European short-range NWP Community does not support the thinning of the conventional observing network, as it will - with the increase of the model resolution - further aggravate the ratio "number of classical observations/number of grid points".