Meteorological product extraction: Making use of MSG imagery

Transcription

1 Meteorological product extraction: Making use of MSG imagery Kenneth Holmlund, Simon Elliott, Leo van de Berg, Stephen Tjemkes* Meteorological Operations Division *Meteorological Division EUMETSAT Am Kavalleriesand 31, D Darmstadt, Germany ABSTRACT The generation of meteorological products from satellite imagery data is a key element of the EUMETSAT satellite programs. After the successful launch of Meteosat Second Generation (MSG) in August 2002, the satellite is currently under commissioning. After successful commissioning the operational products derived with current Meteosat imagery data will be enhanced with new products derived with the MSG Meteorological Product Extraction Facility (MSG MPEF). The improved capability of the MSG image data as compared to Meteosat enables more accurate cloud detection and classification. The higher sampling in time and space provides further improvements that will directly increase the quality of several of the existing key products like Clear Sky Radiance (CSR) and Atmospheric Motion Vector (AMV) products. The multitude of channels will enable the estimation of cloud top heights with higher accuracy than before as well as the derivation of completely new products like Total Ozone (TOZ) and Global Instability Index (GII). This paper will discuss the improved capabilities offered by the MSG imager and how these are utilised in the MSG MPEF. The current status of the product commissioning will also be presented as well as how they compare to the products with Meteosat imagery data. Finally an overview of future improvements currently being explored and developed at EUMETSAT will be given. 1 INTRODUCTION Currently, satellite data is the main source of information in Numerical Weather Prediction (Szyndel, 2003). The generation of meteorological products from satellite imagery data is a key element of the EUMETSAT satellite programs. The derivation of products from the Meteosat satellites started already with Meteosat-1 in 1979 and major improvements in the algorithms during the 80 s established the derived products as a major source of information for global NWP. The Meteosat satellites were the first to provide observation in the water vapour band from geostationary orbit with a 5 km resolution every 30 min. The water vapour data combined with the infrared window measurements enabled the derivation of many important products and the derivation of high quality cloud height measurements. Accurate cloud height information is mandatory for many applications like the Atmospheric Motion Vectors (AMVs) derived by tracking cloud and moisture features in consecutive images. The successful Meteosat era is now being complemented and continued by Meteosat Second Generation (MSG). MSG was launch in August 2002 and the satellite is currently under commissioning. After successful commissioning the operational products derived with current Meteosat imagery data will be enhanced with new products and algorithms derived with the MSG Meteorological Product Extraction Facility (MSG MPEF). The improved capability of

2 the MSG image data as compared to Meteosat enables more accurate cloud detection and classification.. Especially the higher number of channels in the infrared and near- and also the two channels in the visible provide new data that will improve the detection and classification of clouds and other atmospheric features. The SEVIRI (Spinning Enhanced Visible and Infra- Red Imager) instrument will also have the capability to detect the cloud phase and provides the potential to derive completely new products like Total Ozone (TOZ) and Global Instability Index (GII). The higher sampling in time and space provides further improvements that will directly increase the quality of several of the existing key products like Clear Sky Radiance (CSR) and Atmospheric Motion Vector (AMV) products. The multitude of channels will also enable the estimation of cloud top heights with higher accuracy than before. 2 OVERVIEW OF SEVIRI The Meteosat Second Generation (MSG) satellite is a spin-stabilised satellite rotating at 100 rpm like the first generation Meteosat satellites. The SEVIRI instrument is scanning in each spectral band the Earth with three detectors instead of one as is done with the imager on Meteosat. This enables a 3 km sampling distance between individual measurements (1 km for High Resolution Visible (HRV)) and a full field of view coverage of 15 min instead of 30 min as provided by the first generation satellites. The absolute pointing accuracy is for SEVIRI within 3 km RMS for a single image and 1.2 km for image to image registration. The relative accuracy for image sub-areas is even higher with 1.2 km for 500 pixels and 0.75 km over 16 km. The co-registration of the channels is for the HRV, 0.6, 0.8 and 1.6 µm 0.6 km and for the other channels 0.75 km. For details on the performance of SEVIRI see Müller (2003). The SEVIRI instruments provides observations with 12 channels, three visible channels (HRV, 0.6, 0.8 µm) a near-infrared channel (1.6 µm) and 8 infrared channels (3.9, 6.2, 7.3, 8.7, 9.7, 10.8, 12.0, 13.4 µm). Figure 1 presents the spectral response functions of the 8 infrared channels together with the atmospheric transmission for mid-latitude summer (nadirview) and the corresponding weighting functions for the various channels. Figure 1. The SEVIRI spectral response functions of the 8 infrared channels together with the atmospheric transmission for mid-latitude summer at nadir (left) and the corresponding weighting functions for the various channels (right). The SEVIRI channels will continue to provide global (i.e. MSG field of view) coverage in the same spectral bands as provided with the current Meteosat satellites, i.e. visible, infrared window and water vapour absorption channels. However, there is an abundance of new information provided by the multitude of new channels. The potential of the new additional channels is discussed in detail in various presentations and posters related to this user

3 conference (EUMETSAT, 2003) and hence a detailed discussion is omitted here. Table 1 gives a summary of the potential capability of the various channels. It is worthwhile to highlight the capabilities of some of the channels e.g. the potential of the 3.9 µm can for fire detection. Furthermore the 8.7 µm channel can be used for the discrimination of cloud phase and the 9.7 µm channel for the derivation of total ozone. The 13.4 µm channel in the CO 2 absorption band can be used in the so-called CO 2 ratioing to correct for semi-transparency effects in cloud height assignment. Finally it should be noted the multitude of possible channel combinations for the detection low cloud, stratus and fog. For further details on the SEVIRI capabilities please look into other articles in these proceedings (EUMETSAT, 2003). 3 THE METEOSAT SECOND GENERATION METEOROLOGICAL PRODUCT EXTRACTION FACILITY 3.1 Overview The EUMETSAT MSG Meteorological Product Extraction facility (MPEF) is a near-real time system (Figure 2). The product generation is controlled with a schedule that starts all the activities from data acceptance, through product generation, quality control, product distribution to product verification (Figure 2). The input data currently used are rectified and calibrated level 1.5 image data, ECMWF (European Centre for Medium Range Weather Forecasts) model data and meteorological observation like radiosonde data. All activities are configurable with set-up parameters and the processes are continuously monitored. The system also produces automated statistical reports for various purposes like monitoring, validation, verification and reporting. The generated products are distributed to the users via the Global Telecommunication System (GTS), via satellite or are available via the U-MARF (Unified Meteosat Archiving and Retrieval Facility). The required availability is 98.5 %, but figures over 99% are expected. The following sections will give an overview of the various tasks and processes currently foreseen for the MSG MPEF. Figure 2. The EUMETSAT MSG MPEF flow diagram.

4 3.2 The MSG MPEF Radiative Transfer Model (RTM) The MSG MPEF Radiative Transfer Model (RTM) is a course resolution line-by-line model that performs monochromatic calculations at roughly 1 inverse cm resolution (Tjemkes and Schmetz, 1997). It currently uses HITRAN 92 with CDK 2.0 line parameters as static input data as well as static surface emissivity. The dynamic data is provided by ECMWF model data at 31 levels and radiosonde data. The accuracy of the model has been verified against LBL-RTM and is 1 2 %. The model output is used for scenes and cloud analysis, height assignment, atmospheric correction tables, tropospheric humidity and calibration monitoring. 3.3 Calibration Monitoring One of the most important aspects of satellite data is the availability of high quality calibration. The SEVIRI satellite performs for the infrared channels a black-body calibration. This calibration is monitored by the vicarious calibration in the MPEF (van de Berg, 2003). The vicarious calibration used is similar to that used with the Meteosat satellites but it has been further improved, expanded and modified for MSG. Additionally to monitoring it can also supports the selection and tuning of the blackbody models. Figure 3 shows the monitoring results for the first week in October 2003 for the infrared split window channels. The diurnal variation is of the order of 0.5 % and currently the difference to the black-body calibration is of the order of 0.5 %. For visible channels only a vicarious calibration approach is used (Govaerts, 2003). Figure 3. The results form the vicarious calibration ion early October On the left the results for the IR 10.8 micron channel and on the right the corresponding results for the 12.0 micron channel. 3.4 The MPEF baseline products One of the main objectives of MSG is to provide continuity to the already well established products derived with current Meteosat data. Additionally the SEVIRI data will provide the opportunity to derive new and improved products. At EUMETSAT the SEVIRI capabilities are used to improve the quality of the derived products like the Atmospheric Motion Vectors, cloud analysis and clear sky radiance products. New products like mid-tropospheric humidity, Global Instability Index and Total Ozone are to also produced. Table 1. Present the current

5 baseline products, distribution means, validation status in December (RORR = Routine Operations Readiness Review) and the status at the Image and Product Validation Review (IPVR) phase-out. Table 1. The MSG MPEF products. Comments: 1) The products are validated against other observations, but validation for all seasons will be completed later. 2) Internal MPEF products 3) Only available via U-MRF, 4) Implementation in progress 5) For information, the following OSI SAF products will also be added to the EUMETCAST (LRIT) distribution when ready: Sea Ice Products (High latitudes) with 3 sub-products edge, type and cover and the Atlantic Sea Surface Temperature Acrony UMARF GTS EUMETCas Validated Validated at Products m archive t (LRIT) (5) at RORR IPVR close out Atmospheric Motion Vectors AMV Yes Yes Yes Partial (1) Partial (1) Cloud Analysis CLA Yes Yes Yes Yes Yes Cloud Analysis Image CLAI Yes No Yes No Yes Cloud Mask CLM Yes (4) No Yes No Yes Cloud Top Height CTH Yes No Yes No Yes Clear Sky Radiance CSR Yes Yes No Yes Yes Climate Data Set CDS Yes (3) No No Yes Yes High Res. Precipitation Index HPI Yes (3) No No No Yes ISCCP Data Set AC, B1 & B2 IDS Yes (3) No No No Yes Tropospheric Humidity TH Yes Yes Yes Partial (1) Partial (1) Total Ozone TOZ Yes Yes Yes No Yes Sea Surface Temperature (2) SST No No No Yes Yes Scenes Analysis (2) SCE No No No Yes Yes Radiative Transfer Model (2) RTM No No No Yes Yes Calibration Support CAL Yes (3) No No Yes Yes Global Instability GII Yes (4) No Yes No Yes The following section will give examples of some of the currently derived products. These products have been derived during the commissioning phase of MSG-1 and do hence not necessarily present the final expected quality of the products during routine operations, but give a flavour of the scope of the products. 3.5 Current status of the MSG MPEF products Scenes and cloud analysis The algorithm used for the scenes and cloud analysis in the MSG MPEF is a pixel-based thresholding technique based on the work by Saunders and Kriebel (1988). Their approach has been further refined, enhanced and adapted to MSG data by Lutz (2003). The scenes and cloud analysis is a pre-requisite for further processing and contributes crucially to the derivation of clear sky radiances and AMVs as well as most other products. The current performance of the scenes and cloud analysis is demonstrated in Figure 4. In the cloud mask the clouds are identified in white and the clear sky areas are coded according to information from the static scene type file. For the cloud analysis different cloud types are identified in different shades of grey, the surface information is again derived from the surface scene type. Additionally the computed sun glint area over clear-sky sea is indicated in yellow.

6 Figure 4. The cloud mask derived by the MSG MPEF (left) and the corresponding cloud analysis The Tropospheric Humidity and Clear Sky Radiance products The derivation of the clear sky radiance product is perform in all 8 infrared channels. The basic derivation is on a pixel basis and the final product is produced on a segment size of 16*16 pixels. The main provessing factors are the need for a proper cloud clearance and good calibration. In the clear sky areas the tropospheric humidity is also derived. The relationship between observed radiance in the 6.2 and 7.3 µm channels and the corresponding mid- and upper-tropopheric humidity is based on the regression approach by Soden and Bretherton (1993). Also the tropospheric humidity results are currently distributed to the user in segmented form. Figure 5 present an example of the upper tropospheric humidity product and the clear sky radiance product for the 10.8 µm channel. Figure 5. The Upper Tropospheric Humidity (UTH) field derived with the MSG MPEF (left) and the Clear Sky radiance (CSR) product for the IR channel (right).

7 3.5.3 The Global instability product and the Total ozone Product The Global Instability Index product is operationally derived with a statistical regression model for which the parameters have been derived with a neural network approach (König, 2003). The total ozone product is currently derived with a statistical/physical regression model (Tjemkes, 2003). Figure 6 gives an example of the buoyancy derived with the GII algorithm and the total ozone product. Figure 6. A example of the Global Instability Index product (left, here the buoyancy) and the total ozone product (right) The Atmospheric Motion Vectors The main methodologies for the derivation of Atmospheric Motion Vectors is based on the approach already used with the current geostationary satellites. For details see (Holmlund, 2000). One of the main difference to the current operational approach is the dynamic target selection scheme, which enables not only a better selection of features for tracking, but also the flexibility to change the resolution of the product according to user requirements. Another important feature is the multitude of height assignment methods that can be used in parallel. Some of the methods are already used with the current system, but some are significantly improved (e.g. IR/WV semi-transparency correction) or completely new (e.g. CO2-ratioing). Furthermore, the existence of two water vapour channels enable a derivation of cloud heights that is completely independent of forecast data.. Figure 7 present this approach where two regression lines are defined using the WV 6.2 and 7.3 µm channels. The lines are defined by using parts of the semi-transparent clouds where the semi-transparency is changing. The semi-transparency corrected height is defined by the intersection of the two lines. For illustration the computed opaque cloud curve as used in the IR/WV semi-transparency scheme is also presented. In the single WV-channel approach the corrected height is defined by the intersection of the opaque cloud curve and the regression line. In the example presented the two different correction scheme provide a different results. This difference may be attributed to calibration problems, inaccuracy s in the RTM calculations, errors in the used model data or errors in the definition of the regression lines. The potential of deriving a NWP model independent height assignment is especially attractive as the problem of correlated errors is now considered as one of the biggest problems in the AMVs (Bormann et. al., 2003) and may be related to the correlated errors in the NWP temperature fields.

8 Figure 7. The two WV-channel/IR window channel height assignment approach. The derived cloud height is at the intersection of the two regression lines. The difference to the intersection with the opaque cloud curve represents an error in either calibration, derivation of the regression lines, model data or radiative transfer calculations. The MSG data will already at the stage of Day-1 of operations provide a factor of 10 more winds than the current operational system (MTP, Meteosat Transition Programme). Table 2 presents a summary of the performance of the current system, MSG Day-1 and an indication of future evolution of the MSG AMVs. Ave no. of AMVs/ Channel/product MTP 3 160/80 km MSG Day-1 Day km 50 km Table 2. A comparison of the performance of the MTP AMVs vs. MSG AMVs for Day-1, Day-1.5 and Day-2. No of channels Average product No. semi- density transparency correction methods 1 4 (12) The high density of the MSG AMV fields is illustrated in Figure 8 that present a full MSG field of view with infrared AMVs and a close-up of AMVs derived with the visible image data over the North Atlantic.

9 Figure 7. The Atmosppheric Motion Vector (AMV) field (left) derived from cloudy targets with the IR channel (left) and a subsection of the AMV field derived with the VIS channel (right). 4 FUTURE IMPROVEMENTS During commissioning of MSG-1 the main objective of the product validation is to verify the implementation of the scientific algorithms and to validate the baseline products. Many of the capabilities of the SEVIRI instrument are not yet fully exploited. Further improvements in cloud height assignment as well as in the resolution of products is expected. Also the use of new and better algorithms, e.g. a physical retrieval of the Global Instability Index and the use of optimal estimation for the derivation of total ozone are currently being explored. The current scenes and cloud analysis scheme does also not fully utilise the multi-spectral information and could be further enhanced to derive cloud properties like optical depth. This information may turn out to be crucial for several products e.g. for the AMVs in order to properly establish the level at which the clouds travel. Many improved algorithms and products are developed and distributed by the EUMETSAT Satellite Application Facilities (SAFs). For some further details on the products and algorithms planned within the SAF context see these proceedings (EUMETSAT, 2003) or consult the EUMETSAT WEB pages 5 CONLUSIONS The MSG satellite is currently going through its final phase of commissioning, i.e. image and product commissioning. The initial image and product validation results presented show that the expectations on the new satellite will be fulfilled. The MSG satellite will become operational early 2004.

10 References Bormann, N., Saarinen S., J-N. Thepaut and G. Kelly, 2002: The Spatial Structure of Observation Errors in Atmospheric Motion Vectors. Proc. Sifth International Winds Workshop, Wisconsin, Madison, USA, EUMETSAT, EUM P 35, Bretherton and Soden, 1993: Upper Tropospheric Relatative Humidity From the GOES 6.7 um Channel: Method and Climatology for July 1987, J. Geophys. Res, Vol.98, NO D9, , September 20, 1993 EUMETSAT, 2003: Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Goaverts, Y., 2003: Vicarious Calibration of MSG/SEVIRI Channels. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Holmlund, K., 2000: The Atmospheric Motion Vector retrieval Scheme for Meteosat Second Generation. Proc. Fifth International Winds Workshop, Lorne, Australia, EUMETSAT, EUM P 28, Müller, J., 2003: Meteosat Second generation: SEVIRI Radiometric Performance Results from the MSG-1 Commissioning Phase. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Van de Berg, L, 2003: The Calibration of the Infrared Channels On-board of Meteosat First and Second Generation Spacecraft. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. König, M., 2003: Atmospheric Instability Parameters Derived from MSG SEVIRI Observations. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Lutz, H-J, 2003:: Scenes and Cloud Analysis from Meteosat Second Generation (MSG) Observations. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Saunders R.W. and Kriebel K.T., 1988: An improved method for detecting clear sky and cloudy radiances from AVHRR data. Int. J. of Remote Sensing, Vol. 9, pp.123 Szyndel, M., 2003: Initial Experiences of SEVIRI Clear Sky Radiances at ECMWF. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Tjemkes, S., 2003: Total Ozone from Meteosat Second Generation. Proc. The 2003 EUMETSAT Meteorological Satellite Conference, Weimar, Germany, EUMETSAT, EUM P39. Tjemkes S. and J. Schmetz, 1997: Synthetic Satellite radiances Using the Radiance Sampling Method. J. Geophys. Res., Vol 102 (D2),