CLOUDMAP - Product specification report

Authors: Hans Roozekrans and Paul de Valk (KNMI)

1. Introduction

Objective of report

To describe the current use and data sources of cloud information in operational forecasting, in Numerical Weather Prediction (NWP) and in climate research.

Sources of information

2. Description of the main data sources of cloud information as used in meteorology and climatology

Satellite data

The most commonly used satellites in operational meteorology is METEOSAT (or GOES or CMS in outside Europe) and NOAA-AVHRR.

METEOSAT is a spin stabilised satellite located in a geostationary orbit at 36.000 km above the Earth at the crossing of the equator and the 0-meridian. METEOSAT is operated by the inter-governmental organisation EUMETSAT financially supported by seventeen European countries. METEOSAT’s primary mission is to observe the evolution of cloud formations for the purpose of operational meteorology. To do this it generates images of the full Earth disk (as seen by the satellite) each 30 minutes. The METEOSAT spin scan radiometer operates in three spectral bands:

The radiometer scans the Earth from south to north, which takes about 25 minutes. The spatial resolution of the WV and IR images is about 5 km sub-satellite point (at the equator/0-meridian crossing ). The resolution of the VIS images is 2.5 km sub-satellite point.

The raw METEOSAT images are centrally processed (calibrated and navigated) at EUMETSAT in Darmstadt and after that re-transmitted to the METEOSAT satellite and then transmitted in real time to a large community of users at any location within the field of view of METEOSAT.

Images of GOES and CMS are transmitted via METEOSAT to the users.

At EUMETSAT headquarters in Darmstadt a central processing facility called MPEF (Meteorological Products Extraction Facility) provides a list of high level satellite products mainly for application in NWP models and in climate research. MPEF produces the following cloud products:

The CMW, HRV and CLA products are distributed in real time to the end-users via the GTS. The CTH product is transmitted via METEOSAT in the form of a WEFAX picture (the product is primarily meant for aviation meteorology).

The polar orbiting meteorological NOAA-satellites are operated by the National Oceanographic and Atmospheric Administration (NOAA) of the USA. A pair of two satellites is operational continuously since 1980. The primary mission of the NOAA-satellites is similar to the mission of METEOSAT: imaging clouds. The imaging sensor on board of the NOAA-satellites is the Advanced Very High Resolution Radiometer (AVHRR). The characteristics of the NOAA-AVHRR satellite system differ considerably from METEOSAT. METEOSAT provide images with a much better temporal resolution than the NOAA-AVHRR: every half hour against every six hours. On the other hand the spatial and spectral resolution of the AVHRR is by far superior to METEOSAT:

Synops data

Synop data are values of meteorological parameters observed or measured at the ground. To derive standardisation (in definition and quality of the parameters) of measurements or observations the World Meteorological Organisation (WMO) has defined strict procedures on how parameters should be measured or observed (methodology, location, time, etc.). Each national weather service is responsible for a network of synop stations. Observations and measurements are done regularly at fixed times (hourly or 3-hourly). In the Netherlands 15 synop stations at land are operated by KNMI. In other European countries the synops networks are less dense. At sea the situation is even less favourable. From a few platforms in the North Sea synops are observed. Also synops from "selected ships" sailing the seas and oceans all over the world are available. Synops are in real time available for the weather forecasters via the Global Telecommunication System (GTS) of WMO. Most national weather services (like KNMI) archive all synops observed around the world.

The observation of cloud parameters (type, fraction, height) at synop stations is still done by human observers. Although the WMO has defined strict requirements concerning the quality of the observations the involvement of human beings and their experience has impact on the quality. Most national weather services are currently investigating the possibility to automate the observation of cloud parameters at synop stations. Besides the need for more objective observations is the reduction of costs the main drive for the automation of cloud observations.


At the main synop stations radiosondes (measuring device attached to a gas filled balloon) are released every six hours. Radiosondes measure the temperature, pressure and humidity at different heights in the atmosphere and from the horizontal displacement of the balloon the wind speed and direction at different heights is measured. The humidity profile of the atmosphere measured by the radiosonde provides information on the height of the cloud base and top.

The density of radiosonde stations is low (due to costs of radiosonde measurements). In Europe and North America the situation compared to other parts of the world is quite good. At land stations are separated 100 to 150 km from each other. At sea the density is much lower (a few platforms in the North Sea and one or two weather ships in the North Atlantic).

Miscellaneous ground based sensors

At most airports accurate measurements of cloud base height are done by using LIDAR (light detection and range) or SODAR (sound detection and range) techniques (in so called ceilometers). LIDAR ceilometers sent out a laser pulse to the bottom side of a cloud and receive the reflectance of the cloud. The travel time of the pulse is related directly to the height of the cloud. The maximum height of clouds, which can be measured by ceilometers ranges from 4 to 7 kilometres.

Occasionally ground based IR radiometers are used at airports to measure the temperature of the cloud base.

Ground based weather radars are used to detect rain bearing clouds. Most European countries operate one or more radar systems. Western and southern Europe is almost completely covered by weather radars. Radar data are exchanged between the national services in order to obtain European wide images of rain complexes.

NWP model output

Most NWP models are able to produce cloud cover and cloud height analysis and forecast fields who are used by the forecasters. At KNMI mainly two NWP models are in use, both having its own application: the global model of the European Centre for Medium range Weather Forecasting (ECMWF) and the regional HIgh Resolution Limited Area Model (HIRLAM).

The ECMWF model has a global coverage, is run every 6 hours and has horizontal resolution of 150 km.

The HIRLAM model is developed at the national weather services of Denmark, Finland, Iceland, Ireland, Sweden, Norway and the Netherlands. HIRLAM as implemented at KNMI covers Europe and the North Atlantic. The horizontal grid point distance is 0.5 degrees, which corresponds to 55 km. HIRLAM is run every 3 hours. In preparation of the forecasting work HIRLAM generates analysis fields of observations (e.g. synops) taking into account surface type and orography.

3. Cloud information used in operational forecasting

Main data sources used by forecasters (source: CLOUDMAP questionnaire):

It is remarkable that nowadays in operational weather forecasting satellites have taken the lead as most important data source to obtain adequate cloud information.

The use of the geostationairy METEOSAT (79 %) dominates the use of polar orbiting NOAA satellites (21 %). However, in the Scandinavian countries it is the opposite due to the good temporal coverage of NOAA and the very oblique view angle of METEOSAT for these areas.

Qualitative cloud information:

Nowadays satellite images are indispensable for the forecaster to obtain a synoptic view of the distribution of clouds and cloud structures over large areas. The polar orbiting NOAA satellites provide spatial detailed views with a low temporal resolution. The geostationairy METEOSAT satellites provide images with high temporal resolution (every half-hour) enabling the monitoring of the dynamics of clouds. The METEOSAT image loop is always in the view of the forecaster during his shift.

The use of false colour composite images by the forecasters becomes more and more familiar. In those images several bands are combined into one image, each band is assigned to a different colour. E.g. NOAA-AVHRR band 1 (visible) is assigned to red, band 2 (near infrared) to green and band 4 (thermal IR) to blue. This colour composite image enables the forecaster to distinguish at a glance between high and low clouds and between thin and thick clouds.

With the introduction of meteorological workstations satellite images form a truthful data source to validate the performance of NWP models. The cloud patterns and structures in the image, being a result of typical physical processes in the atmosphere, are used to check the structures in the analysis (e.g. pressure) fields of NWP models. In case an analysis field does not match with the cloud patterns in the satellite image a forecaster can be suspicious about the validity of the NWP model forecast.

Quantitative cloud information:

  • Cloud type
  • Application:

    - nowcasting

    - aviation meteorology

    Data sources:

    - synops

    - NOAA-AVHRR imagery


    Human observers at synop stations classify the clouds in their view. In case of multi-layer clouds off course only the lowest clouds are visible.

    Several algorithms (at MeteoFrance and SMHI, Sweden) based on threshold techniques have been developed to classify cloud types in NOAA-AVHRR images (the 3 bands of METEOSAT are not sufficient to perform this task). SMHI is the only weather service where the satellite based cloud type classifications are operationally used by the forecasters. In case of multi-layer clouds synops and satellite images can be complementary (satellites see the upper clouds).


    Professional human cloud observers are trained to classify cloud types very accurately. However, at night the lack of sunlight hampers an accurate cloud type classification.

    SMHI has validated their AVHRR algorithm for Sweden and claims an accuracy of 90% correctly classified pixels in daytime images (for night-time images no figures are available, but the lack of visible channels cause a lower accuracy).

  • Cloud top temperature/height
  • Application:

    - nowcasting of severe weather

    - aviation meteorology

    Data sources


    - radiosonde data

    - NWP model analyses


    Satellite imagery provides the brightness temperature of the cloud tops. Radiosonde data provide real cloud top temperatures. However, radiosonde data sets have a limited spatial and temporal resolution and distribution (about 50 stations spread over Europe) and have a time resolution of 6 hours. Therefore, forecasters use radiosondes in combination with satellite imagery in such a way that radiosonde temperature profiles are used to quantify the satellite images towards real cloud top temperatures.

    For the conversion the satellite cloud top temperatures to cloud top heights (CTH) NWP model analysis fields are used (in which the radiosonde data are included).

    Aviation forecasters use quite often the MPEF/CTH product of METEOSAT.

    Accuracy (of CTH):

    Radiosondes/NWP analysis: 20 m

    Satellite imagery: For opaque clouds the accuracy is about 500 m. For semi-transparent clouds the determination of a true CTH is more complicated and less accurate. At KNMI an algorithm has been developed to correct the cloud top temperature (and thus height) of semi-transparent clouds making use of the optical band(s) of the satellite.

  • Cloud base temperature/height
  • Application:

    - nowcasting

    - aviation meteorology

    Data sources:

    - synops

    - radiosondes

    - sodar/lidar


    The measurement of cloud base temperatures can not be done directly by satellite. Cloud information in synops is collected by human observers.

    At most airports sodar/lidar systems are placed to measure the cloud base height in an automatic way. These data are available for all forecasters in real time (via the GTS).


    Synops: the accuracy depends very much on the availability of daylight (at night it is very difficult for an observer to estimate the height of a cloud) and on the vertical visibility.

    Sodar/lidar: very accurate (1 or 2 m).

  • Cloud fraction
  • Application:

    - aviation (especially the cloud fraction of low clouds is a very important parameter)

    Data sources:

    - synops

    . - satellite data


    Cloud fraction in synops is observed by qualified human observers using strict procedures. An observer views the total sky and estimates the percentage of the sky, which is covered by clouds. This percentage is converted to a measure called okta. 1 okta means 1/8 part of the sky visible for the observer is covered with clouds. 0 okta means the sky is completely cloud free. 8 octa means the sky is completely cloud covered. The area for which the observation is valid is related to the atmospheric conditions (visibility), the amount of obstacles around the observer (mountains, trees, buildings, etc.) and the height of the clouds (the higher the larger the area). At an average it can be assumed that the synops area is around 30 x 30 km² . At sea the area is much larger (50 x 50 km² ).

    Satellite imagery provides in first instance a qualitative measure of the cloud fraction. Algorithms for quantification of cloud fractions in satellite imagery are already operational (for NOAA/AVHRR imagery) or under development (METEOSAT). The spatial resolution of the sensors limits the potential accuracy of the cloud fractions derived from satellite images.


    Although the WMO has set very strict quality demands to the synop measurements in practice the quality of cloud fraction synops varies considerably. The estimation of cloud fraction in the range of 2 until 6 okta appears to be a very difficult task and the result is very much related to the experience and education of the observer. In general the accuracy of cloud fraction synops is about 1 okta (in the range of 2 until 6). At night the accuracy can be assumed to be 2 okta’s.

  • Cloud phase
  • Application:

    - aviation (icing of planes)

    - forecasting of precipitation

    Data sources:

    - radiosondes

    - NWP model output

    - satellite imagery


    From radiosonde data the phase of the cloud can be determined directly. NWP models provide parameter fields, which are directly or indirectly related to cloud phase. With the use of NOAA-AVHRR channel 3 (3.7 m m) images it is possible to recognise water and ice clouds.

    The combination of knowledge on cloud phase and cloud top temperature leads to the recognition of icing conditions for the aviation.


    Icing conditions can be very local and therefore radiosondes and NWP data do not provide the required spatial resolution, although the cloud phase information on itself is very accurate.

    Satellite images provide the adequate spatial resolution but the cloud phase information is not so accurate (more qualitative).

  • Precipitation
  • Application:

    - nowcasting

    Data sources:

    - weather radar

    - satellite data (METEOSAT, SSM/I)

    - synops


    The weather radar is the most important data source for the forecaster to obtain information on precipitation distribution. The radars provide information on where the rain systems are and a certain measure of the intensity of the rain. Satellite images put the radar information in a wider context (larger area and visibility of other clouds). Synops provide information on the exact amount of precipitation and also the type of precipitation (can not be detected from radar data). SSM/I satellite data are used more and more in operational forecasting providing quite accurate information on precipitation distribution and intensity over large areas (especially above sea/oceans where no radars are available).


    See methodology.

  • Cloud optical thickness
  • This cloud parameter is hardly used by operational forecasters.

  • Cloud motion wind vectors
  • Operational forecasters hardly directly use this product based on geostationairy satellite data.

    Back to introduction

    4. Cloud information used by operational NWP models

    Main data sources used:

    The CLOUDMAP questionnaire was responded by only two people having in mind NWP models as end-user of cloud information. Both indicated that satellite data and synops (including radiosondes) are used on an equal level of importance as data source for NWP models.

    Main cloud parameters used:

  • Synops:
  • The fraction, temperature, pressure and phase of clouds are parameters, which are assimilated into NWP models.

  • Satellite data:
  • The satellite derived cloud parameter most commonly used by NWP models is cloud motion wind (CMW) product (eg as produced by EUMETSAT within MPEF). All major NWP models in Europe (ECMWF, UKMO, MeteoFrance, DWD) except the HIRLAM model are using the CMW’s operationally. The impact of the CMW’s is most positive in area’s with low density of other observations (e.g. the southern hemisphere). The efficiency of CMW’s is hampered by the lack of an accurate height assignment to the vectors and of wind vectors in cloud free areas. Research is going on (e.g. at EUMETSAT) to remove these limitations by using all channels, signals are sensitive for clouds at different heights. Especially the use of the WV channel of METEOSAT looks very promising to obtain wind vectors in cloud free areas.

    The Cloud Analysis (CA) product of MPEF is not yet used in NWP models.

    At KNMI a short range cloud forecasting model called MetCast is developed which uses METEOSAT data (cloud fraction and temperature) for initialisation of the model. The model currently is semi-operational used by the forecasters.

    Accuracy of cloud parameters in relation to use in NWP models:

    The requirements of NWP models concerning the quality/accuracy of input data are implicitly included in the assimilation scheme of the model. The assimilation scheme itself checks the quality of the data value using several tests (time series analysis, spatial context analysis, etc.) and decides if the input data is used or rejected. Tuning of the assimilation scheme is a matter of studying the impact of input data on the model result by changing thresholds in the tests. Off course it is clear that the more accurate the input data will be the more impact the use of the data in the model will have.

    Back to introduction

    5. Cloud information used in climate research

    Main data sources used:

    The CLOUDMAP questionnaire was responded by only five people having in mind climate research as end-user of cloud information. The average answer to which data sources are used in climate research looks as follows:

  • Synops: 21 %
  • NWP output: 11 %
  • Satellite data: 64 % (83% geostationairy sats. and 17% polar orbiting sats.)
  • Other: 4 %
  • As expected satellite data is by far the most used data source in climate research. Climate change is a process on a global scale. Satellite data provide the global coverage needed to study and monitor climate change. The climate research community is convinced that satellite data are essential for their tasks.

    Activities in climate research can be divided into two streams:

    Both activities have their own needs for data/observations. The monitoring task needs long time series of consistent and accurate data sets. Global coverage is not always a necessity. Data continuity is the most important requirement in this context. Synops data (time series of 100 to 150 years are available) is the main data source for this purpose. The availability of meteorological satellites for more than 30 years now makes satellite data sets more and more suitable for use in monitoring experiments.

    The second stream of activities aiming at understanding the key processes in the climate system uses a large variety of data sources. Many field campaigns involving all sorts of instruments (groundbased, aircrafts, satellites) are held in the past or planned for the future. Most of them are related to large international climate research projects. The climate research group of KNMI is participating in several campaigns. The following URL address provides detailed information on the instruments and data sources, which are used for this purpose by KNMI: (click "Climate research" and "Atmospheric research")

    A very important project, which must be mentioned in this report, is the International Satellite Cloud Climatology Project (ISCCP) being an element of the World Climate Research Project (WCRP). The project was started in 1983 with the aim to produce a 5-year global data set of cloud parameters derived from satellite data. This global data set was meant to contribute to basic research aiming at a better understanding of the Earth’s radiation budget and hydrological cycle and of the role of clouds in both systems. This better understanding should lead to an improved parameterisation of clouds in climate models. The main data sources used in the ISCCP project are the geostationairy satellites METEOSAT, GOES and GMS. The primary data are from the two standard visible (0.6 m m) and infrared (11m m) channels common to all of the satellites. The NOAA-AVHRR satellites are used in addition to obtain coverage in the polar regions and as a basis for normalisation of the radiance's observed by the different geostationairy satellites. Another motivation for using AVHRR is the use of multi-spectral observations for the purpose of discriminating cloud properties not derivable from the primary two-channel data. The ISCCP data set includes 30-day averages of cloud parameters (spatial averaged over 250 km x 250 km boxes) for a 5 year period and is available on IBM tapes for scientific purposes.

    Main cloud parameters used:

    Climate research is wide field of research covering many aspects of the climate system. Therefore it is difficult to list all cloud parameters being important for climate research. It might be adequate to use the cloud product list as produced in the ISCCP project as a guide to what are key cloud parameters in climate research:

    Back to introduction