The aim of this report is to define user requirements of the use of satellite cloud data in operational weather forecasting, numerical weather prediction modelling and climate research. The term satellite cloud data has been used in a wide context and is not specifically related to the anticipated CLOUDMAP products. This approach is chosen to enable an extension of the CLOUDMAP product list in the course of the project anticipating on the user requirements.

At the CLOUDMAP kick-off meeting in November 1997 KNMI presented preliminary and qualitative user requirements of satellite cloud data based on experience within KNMI and from literature.

Operational forecasters need:

• high temporal resolution
• real time use (short timeliness)
• 1 km spatial resolution
• indication of accuracy (error bars)
• 3D information (height assignment)
• data continuity (long term satellite programmes)

Climate researchers need:

• high signal-to-noise-ratio’s (in case of climate monitoring)
• global coverage
• long continuous time series of data

These requirements are confirmed and further quantified by the work done by KNMI in the first two years of the CLOUDMAP project. This work included three milestones:

Operational forecasting of clouds relies heavily on the frequent and timely distribution of cloud information. Forecasters consider a high product frequency as mandatory for cloud data. New development of future techniques should have a better performance compared not only to the present cloud products but also better than the available products foreseen in the (near) future.

As a framework for user requirements the specifications for cloud products from MSG will be described here and compared to the requirements as obtained by the questionnaire given in appendix C. The questionnaire was distributed during the CLOUDMAP workshop in Paris (May 26, 1998) and send to the meteorological services in Europe. Both requirements are presented in two tables given here below. Although the tables look very similar the reader may notice that there are differences between the two tables. In table 1 the SEVIRI pixel resolution should be interpreted as a horizontal spatial resolution of about 4 km.

There is an essential difference between table 1 and table 2. In table 1 the requirements are based on an instrument that will be launched in the beginning of the new century. These requirements are balanced between end user wishes and technical and economical feasibility.

Whereas the second table is a summary of desired cloud products by operational users without the limitation to a certain observation technique. Still the tables have very similar values in terms of Frequency, Resolution and, Timeliness. One should realise that the respondents work as operational forecasters. Forecasters are well aware of the possibilities and weaknesses of present and (near) future geostationary and polar satellites and may express their wishes according to these possibilities. It also shows that their requirements are realistic in terms of coming technologies.

Comparing table 2 to the table in the questionnaire one notices that the areal coverage category has been dropped. The opinions differed too much to capture them in a single value or range of values for spatial coverage. From the three given categories the global and regional coverage where the ones chosen predominantly.

The forecasting community considered cloud optical thickness, semitransparent flag and contrails maps as less relevant. Therefore they are omitted in table 2. Cloud top pressure and cloud top height both gave similar information. Only cloud top height is represented in table 2 together with cloud top temperature.

From the two tables we derive a table of user requirements, given in table 3. To accommodate all requirements of the forecasters the lowest boundary values are used. In some cases we compromise between the given boundary values of table 2 as we expect that these values will be applied for NWP model check. NWP model output is hourly so this sets the required frequency. NWP model check also relaxes the required timeliness. The horizontal resolution is comparable to the present resolution of the NOAA-AVHRR. Higher resolutions are of course acceptable but do not have a high priority. Higher vertical resolutions certainly have a high priority. A high vertical resolution for low clouds and fog is very important for aviation meteorology and has therefore a direct economical impact. The latter point is also reflected in the highest value for accuracy in the last column of table 3.

We are aware that there will be questions about technical and economical feasibility's but we choose to reflect the (coming) wishes from the forecast community, which in the future will rely more and more on satellite data.

Back to summary

The use of satellite data by NWP models has three applications each having its own requirements concerning the satellite data:

• use for initialisation of NWP models,
• use for improvement of physical parameterisation of models,
• use for validation of NWP analysis fields.

The last application is part of the job of operational forecasters. Therefore, the requirements concerning this use of satellite data are included in the "operational forecasting" part of this document.

Initialisation of NWP models

Numerical weather prediction is an initial-value problem. Nowadays, the quality of a NWP forecast is directly related to the accuracy of the initial analysis of the Earth-atmosphere system and thus to the availability of an adequate observation system. For many areas (oceans, sparsely populated land areas) the meteorological observation system is by far not adequate mainly due to economical reasons. In fact the situation is getting worse with the reduction of the number of observation stations in the densely populated areas (Europe and USA) in the last decade to reduce costs. With the coming of meteorological satellites the meteorological community has been started research the input of satellite data into the analysis modules of NWP models to fill in the gaps in the conventional observation network. Meteorological satellites provide excellent areal coverage, however the accuracy of the data is not at the same level as the accuracy of the conventional observations. To tackle this problem assimilation schemes and techniques are developed in order to filter satellite data sets before they are used for the model analysis. Studies on the impact of the input of satellite data on the NWP model forecasts all conclude that the level of impact is strongly related to the following aspects of the satellite data set:

• timely availability of the data (the time between the observation and the model analysis should be as short as possible),
• sufficient coverage of the model area by data values, not rejected by the model.

The accuracy of the data is of less priority. It is better to initialise a model with up-to-date inaccurate data than with very accurate but old data covering only a small part of the model area.

An example: a study done at KNMI on the impact of the assimilation of ERS-scatterometer wind fields in the HIRLAM model demonstrated the need for high data coverage of the model area. In the first part of the study when ERS-1 data were available the impact of the scatterometer winds was negligible probably due to a bad areal coverage of the ERS-1 data. Later on the launch of ERS-2 enabled during a short period the so called tandem scenario with two satellites in operation. During this period the impact of the scatterometer data use on the HIRLAM forecasts increased significantly due to the much better areal coverage.

It is very difficult to quantify the two mentioned requirements (timeliness/frequency and areal coverage) for cloud products because not much experience on this is available yet. One clue could be derived from the specifications of the MPEF cloud products to be derived from MSG and meant for NWP model initialisation:

The semi-operational short-range cloud forecasting model of KNMI (MetCast) requires satellite cloud products with the following specifications:

The physical parameterisation of NWP models has been worked on intensively in the 1980’s with the aid of coherent and consistent satellite data sets on global radiation and cloud cover. It is believed that this work is mostly finished now and that improvement of the NWP forecasts does not rely on a further improvement of physical parameterisation of models.

New parameterisation schemes might be necessary if the horizontal and vertical resolution of models will be increased in the near future (the increased power of computers allows for this need). The requirements to satellite data to be used for this purpose should focus on the need for spatial and vertical resolution better than the resolution of the model itself. E.g. currently a 5 km resolution HIRLAM is under development. For this work satellite data with a better resolution than 5 km are required. Other important requirements to satellite data in this context are accuracy and independence to additional data. Timeliness, frequency and areal coverage have less priority for this application.

In the CLOUDMAP workshop the use of satellite cloud products in the context of NWP models was addressed several times mostly in relation to the validation of NWP analysis fields.

In relation to the focus of the CLOUDMAP project (concerning the type of satellites used) a few important remarks were made during the workshop:

• The low frequency renders the CLOUDMAP products not useful for NWP initialisation.
• Contrail information is not relevant for NWP models.
• Precipitation information from satellite data is useful for the validation of the physics in models.

The CLOUDMAP questionnaire is filled in by two people having NWP model use in mind. However, it is not clear what kind of use they had in mind: model initialisation, validation or development. The answers of the two respondents differ quite significantly. In table 4 their answers are summarised (the ranges indicate the difference in answers).

3.2 Conclusions

The user requirements of NWP models to satellite data can not be defined as specific as has been done for operational forecasting. The problem lies in the fact that concerning the initialisation of models using satellite cloud products not much experience is available yet (except for the input of cloud motion vectors). Experience with other types of satellite data stresses the need for adequate timeliness, frequency and areal coverage. Accuracy is of less importance. For some products (e.g. cloud motion vectors) the accuracy of the height assignment is an important requirement of NWP models.

The added value of the use of satellite cloud products for parameterisation of the models has been proved in the past and is still valid in relation to the development of high(er) resolution models. Again it is difficult to quantify requirements to the data concerning this type of use. In research environments requirements to data are in most cases not existent. The device is that all data in any form at any time are valuable. High accuracy of the data can be a requirement but synergy with other data (by interpolation, simulation, calibration, etc.) may solve the problem of inaccurate data in many cases. Timeliness is not an issue in this matter.

Back to summary

The role of clouds in controlling the radiation budget of the Earth’s climate system is believed to be large. The investigation of this role involves the study of many complex dynamic and thermodynamic processes. These studies require global data sets of cloud observations with adequate spatial and temporal resolution. Satellites can only fulfil this requirement. This is one of the motivations of the CLOUDMAP project. The CLOUDMAP project is one of the many efforts to promote the use of satellite cloud data in climate research. Therefore, KNMI has searched literature to find information on other projects in which user requirements of climate research have been collected.

Long before the CLOUDMAP project the International Satellite Cloud Climatology Project (ISCCP) was started in 1983 in the framework of the World Climate Research Programme (WCRP). ISCCP aimed at a production of a 5-year global data set of cloud parameters derived from satellite data (METEOSAT, GOES, GMS, NOAA-AVHRR). The main object of the production of the global cloud data set was to contribute to an improved parameterisation of clouds in climate models. The ISCCP data set includes 30-day averages of cloud parameters for a 5-year period and is available on IBM tapes for scientific purposes. The ISCCP product specifications are listed in table 6. These numbers may give indications of what cloud information is needed in climate research.

In the CLOUDMAP workshop the following remarks relevant for climate research were made:

• Cirrus information is relevant for researching the impact on temperature and radiance balance.
• Contrail maps are relevant for climate monitoring (on longer time scales).
• Cloud opacity is relevant for the monitoring of long term climate changes.

The CLOUDMAP questionnaire was filled in by five respondents all aiming at the use of satellite data in climate research. Again the answers differ significantly and in table 7 the answers are summarised indicating a range of requirements.

Some parameters are omitted in accordance with the forecast table 2. Climatologists are interested in a global coverage of the earth. They are also interested in some classes of cloud particles with a distinction in size and shape.

Similar to NWP R&D (and research in general) it is difficult to define unambiguous user requirements for satellite data use in climate research. The way of using the data is determining the accents of the requirements:

• The ISCCP project mainly aimed at a better understanding of the processes in the climate system in order to improve the parameterisation of climate models. For this type of use the main requirements to satellite data focus on accuracy, global coverage and a large number of cloud parameters.
• Another application of satellite data in climate research can be the validation of climate models. Requirements for this application focus on accuracy and spatial resolution of the data.
• Finally the monitoring of the climate becomes feasible with the use of satellite data. For monitoring purposes data continuity in terms of time, space and quality is very important. For this reason long term satellite programmes using stable and time independent observation techniques are required in order to be able to discriminate the climate trend in a parameter from the noise in the observations (a high signal-to-noise ratio is needed).

A quantification of the different requirements is hardly possible. Table 7 shows the large ranges of requirements defined by climate researchers addressed until now.

The first user workshop of the CLOUDMAP project was held in Paris, France. The venue was chosen as it coincided with the AMS/EUMETSAT Conference. At this conference many anticipated end-users were brought together providing an excellent occasion for a user workshop. This first CLOUDMAP user workshop was focused on all three anticipated uses of satellite cloud data (nowcasting, numerical weather prediction and climate research). The aim of the discussion part was to get feedback from the audience on their opinion on the use of satellite cloud data in general and of CLOUDMAP products in particular.

11.30-11.35 Introduction to aims of workshop: Hans Roozekrans

11.35-12.00 Presentation of CLOUDMAP products: Jan-Peter Muller

12.00-13.00 Discussion using prepared statements

13.00-13.02 Handing out the user requirement questionnaire

As a reaction to the presentation of Jan-Peter Muller the participants raised a number of questions.

EM: What new things is CLOUDMAP bringing?

JM: The ability to characterise multi-layered clouds and to distinguish cloudy pixels contaminated by cirrus clouds. Another CLOUDMAP feature is the ability to derive cloud top parameters without the need for additional data sources (e.g. NWP data).

GP: What is the time difference between the images used for stereoscopy?

JM: 7 minutes for the extreme viewing angles and smaller than 4 min for the intermediate viewing angles.

GP: The low frequency of observation renders the products as not very useful.

FD: We have an interest in the thickness of the cloud layer.

JM: We can not obtain a cloud thickness layer from satellite data.

The discussion was further guided by eleven statements (bold printed):

General impression: Forecasters prefer operational satellites like METEOSAT Second Generation (MSG) and EPS as they guarantee data continuity.

The CLOUDMAP project partners would like to thank all participants for their valuable contributions to the workshop and the AMS/EUMETSAT Conference organisation for hosting the workshop.

The second user workshop of the CLOUDMAP project was held in Den Haag. The venue was chosen as it coincided with the EGS 24th general assembly. Part of the workshop programme was scheduled as part of Symposium AO30 with the title "Use of satellite data in climate studies". The dedicated CLOUDMAP paper session was scheduled at the end of the day.

After the paper session and a small break (with some snacks and drinks) the real part of the workshop, an interactive discussion with users, was held.

The first CLOUDMAP user workshop, held during the AMS/EUMETSAT Conference in Paris at the end of May 1998, was focused mostly on the use of satellite cloud data in nowcasting and numerical weather prediction modelling. This second workshop was more focused, in line with the setting, on the use in climate research. The aim of the paper part of the workshop was to inform the audience on the work done in the CLOUDMAP project. The aim of the discussion part was to get feedback from the audience on their opinion on the use of satellite cloud data in climate studies in general and of CLOUDMAP products in particular.

17.00 J.P. Mueller: CLOUDMAP evaluation of new macroscopic and microphysical cloud parameters from earth observation sensors for climate modelling.

17.15 A. Drescher, B. Schreiner: Application of high resolution MOMS/MIR data for CLOUDMAP validation.

The discussion was guided by a number of 12 statements given with the aim to provoke the discussion.

1. Satellite data « other data
Are satellite data of clouds more useful than other data sources for application to climate research?

2. Need for independent data
Do climate researchers require satellite cloud products which are independent of additional (e.g. NWP model, synops, etc.) data sources?

3. Operational satellites
Are the present and future operational satellite cloud data sets (METEOSAT, MSG,NOAA) sufficient for use in climate research (eg ISCCP data set)?

4. Experimental satellites
Do cloud data derived from experimental satellites (ERS, MOS, ENVISAT, EOS, ADEOS, etc.) have potential scientific value when used in climate research?

5. How to increase satellite data use
What is the priority order for increasing the use of satellite data in climate research?:

• better horizontal resolution,
• better vertical resolution,
• better temporal resolution.

6: Cloud top height

Is cloud top height (CTH) information important in climate research?
If so, for which application? Which parameter has priority?:

6. Cloud type
Is cloud type information relevant for climate research?
If so, in which way should cloud types be classified?:

8. Contrails
What is the role and potential of information on contrails for climate research?

• parameterization of climate models?
• impact of aviation induced cloudiness on the greenhouse effect?

9: Cloud particle size
Is cloud particle size distribution important for climate research?
If so, for which application(s)?

10: Cloud optical thickness/geometrical thickness
Is cloud optical depth and/or thickness important for climate research?
If so, for which application(s)?

11: Cloud top phase
Is cloud top particle phase important for climate research?
If so, for which application(s)?

12: Polar vs geostationary
Are polar orbiting satellites relevant for climate research?

Probably due to the very unfavourable time of the day, the number of climate researchers participating in the workshop discussion was rather disappointing. Nevertheless, the output of the workshop can be considered as very valuable for the CLOUDMAP project. The excellence of the participants has resulted into a few very clear conclusions on the usefulness of satellite derived cloud information in general and of the CLOUDMAP products in particular:

• For the purpose of process studies climate research badly needs long time series of unambiguous data of cloud parameters. The operational meteorological satellite programmes provide this long term data continuity. Therefore, the focus in climate studies is very much set to the use of the World Weather Watch satellites (METEOSAT, GOES, TIROS, etc.). However, the unambiguous calibration of the sensors on board of these satellites still forms a large drawback of their use in climate studies. CLOUDMAP products might help to derive better calibration of meteorological satellite data.
• The CLOUDMAP products provide helpful information for a better understanding of the role of clouds in global climate change that still is very much unclear. Independent information on cloud top height and optical thickness (the main CLOUDMAP products) is very useful and valuable to improve the cloud parameterisation in climate models.

The CLOUDMAP project partners would like to thank all participants for their valuable contributions to the workshop and the EGS Congress organisation for hosting the workshop.

The following questionnaire has been composed by KNMI and sent out to 17 national meteorological institutes in Europe with the aim to obtain information from the user community on the current use of cloud information and on requirements for the future:

Name: ……………………………………………………..............

Affiliation: …………………………………………………………

Function: …………………………………………………………..

Tel. no: +…………………………………………………..............

E-mail: ……………………………………………………............

Please, choose one option:

• You are a real "end-user" of cloud information being an operational forecaster, NWP modeller or climate researcher (filling in of this questionnaire should be as much as possible representative for your personal work situation)
• You are not an "end-user" of cloud information being a satellite data researcher, policy-maker, etc. (filling in of this questionnaire should be as much as possible representative for the meteorology/climate research community in general)

Please, choose one option (to help a clear interpretation of the information, provided by you, the further filling in of this questionnaire should be mostly representative for the field of application, you have selected):

• Operational weather forecasting
• General
• Aviation
• Maritime
• Other: ………………………………………..
• NWP modelling
• Climate research

Please, fill in roughly the percentage of importance for your chosen field of application

(the addition sum of the four percentages should be 100):

• Synops: ………. %
• NWP models: ………. %
• Satellite data: ………. %
• Other: …………………………… ………. %

Please, fill in roughly the percentage of importance for your chosen field of application

(the addition sum of the two percentages should be 100):

• Geostationary satellites (METEOSAT, GOES, etc.): ……… %
• Polar orbiting sat’s (NOAA, METEOR, ERS, etc.): ……... %

Please, fill in for each parameter relevant for your field of application as much as possible numbers.

• provide average/reasonable requirements in the middle of the range between optimal and minimal requirements
• if parameter is not relevant for your field of application, fill in ¾ in first column
• if requirement is not applicable to parameter, fill in n.a. or leave open
• if you don’t know an answer, fill in ?

Please, fill in roughly the percentage of importance for your chosen field of application.

• the addition sum of the six percentages should be 100
• the term adequate relates to the average of the numbers specified in the previous table by yourself

• Adequate areal coverage : ……….. %
• " horizontal spatial resolution: ……….. %
• " vertical resolution: ……….. %
• " repetition time (or data frequency): ……….. %
• " time of delivery: ……….. %
• " accuracy/quality of data: ……….. %

To obtain quantitative cloud information using the current meteorological satellites additional data sources (NWP, synops, etc.) are necessary in many cases (eg cloud top heights).

Is it important/helpful for your application field to obtain accurate cloud information from satellite data without the need for additional data?

• Yes
• No

Are you willing to provide further "user" input to the CLOUDMAP project in a later phase (eg evaluation of CLOUDMAP products)?

• Yes
• No

KNMI received filled in questionnaires from 19 respondents coming from 13 countries. The answers have been summarised in table 1. The answers to the large table in the questionnaire are not included in table 1 since the answers are in many cases quite ambiguous and incomplete. It is tried to summarise the table in the questionnaire into some general conclusions, which can be found in the main text of this User Requirement report.

Legend to table 1:

dnmi : Norwegian Meteorological Institute
smi: Swiss Meteorological Institute
ukmo : United Kingdom Meteorological Office
mic: Meteorological Institute of Croatia
knmi: Royal Dutch Meteorological Institute
hms : Hungarian Meteorological Service
zamg: Austrian Meteorological Institute
smhi: Swedish Meteorological Institute
dwd: German Meteorological Service
fmi: Finnish Meteorological Institute
lmd: Dynamical Meteorological Laboratory (France)
kmi: Royal Meteorological Institute (Belgium)
inm: Instituto Nacional de Meteorologia (Spain)

clim: climatologist
nwp: numerical weather prediction

DWD has put a remark on the questionnaire that it is willing to provide dedicated data for verification of CLOUDMAP products, e.g. IR and UM profiler data, radio sondes, sodar, lidar etc.

Back to main text