Layer Precipitable Water PGE07 NWCSAF    Table of contents 1.- Goal of LPW product  2.- LPW algorithm summary description  3.- List of inputs for LPW  4.- Coverage and resolution 5.- Description of LPW outputs 6.- Example of LPW visualisation Access to "Algorithm Theoretical Basis Document for Clear Air Products” for a more detailed description.

1.- Goal of LPW product

The LPW, developed within the SAF NWC context, aims to support nowcasting applications. The LPW allows measures of the precipitable water in different layers in cloud free areas to be obtained.

The central aim of the LPW is therefore to provide information on the water vapor contained in a vertical column of unit cross-section area in three layers in the troposphere:

• Boundary Layer (BL): 1013hPa-840 hPa
• Middle Layer (ML): 840hPa-437hPa
• High Layer (HL):  <437hPa

Besides, the PGE06 product provides the total precipitable water parameter obtained with a similar neural network algorithm. A conversion from mm to grey levels or digital count is necessary for writing BL, ML and HL outputs in  8 bits for each one.

In order to assign the quality of these three parameters, one-year data was tested, and the statistical parameters obtained from the comparison to ECMWF profiles and radiosonde were presented in the 2007 Eumetsat Meteorological Satellite Conference.

2.- LPW algorithm summary description

The implementation of LPW algorithms are mainly based on simulated radiances (RTTOV-7 and 60L-SD database). LPW's main algorithms have been devised as one neural network for each precipitable water layer. Therefore, the software has been designed to change the topology and weights only by changing the name of the new neural networks in the configuration file. When the PGE07 is started, all the names of the files with the neural networks are read from the configuration file and the topology and weights are allocated and loaded.

The CMa output is used to identify clear air pixels. Only if the pixel is labelled as clear air and the satellite zenith angle of this pixel lies within the configurable threshold, the pre-processing and processing are calculated in this pixel.

The whole process includes:
 

• Pre-processing:
  • Global correction. Two parameters (scale and offset) are obtained with a robust regression for each IR band in order to pass from effective radiances Meteosat-8 to spectral radiances Meteosat-8 and they are included in the LPW MSG1 configuration files. Other two parameters (scale and offset) are obtained for each IR band to pass from effective radiances Meteosat-9 to spectral radiances Meteosat-8 and are included in the LPW default configuration file.
  • Local bias correction of SEVIRI radiances (optional)
  • Smoothing of SEVIRI radiances (optional) performing a median o mean/average (configurable in the model configuration file) over a window centred at the processing pixel. Boundaries of the region are not considered.
  • Normalisation of SEVIRI radiances.
  • Correction of SEVIRI radiances to 45º MSG zenith angle. This step will use a dedicated neural network, whose inputs are the SEVIRI radiances and the satellite zenith angle.

• Processing:

  The calculus of each Layer precipitable water parameter will use a dedicated Neural Networks (NN). The NNs will be fed with corrected normalized IR SEVIRI radiance values (WV6.2, WV6.3, IR8.7, IR9.7, IR10.8, IR12.0 & IR13.4m m), and each one will provide as output a precipitable water parameter.
  For land pixels the normalised altitude is also needed as input for the neural network and the neural networks weights that correspond to IR8.7 radiances are fixed to zero.           

  Denormalisation of the precipitable water to mm, using the thresholds defined in the configuration file.

   

• Quality control:

  To check the product quality, different flags have been added to the output products.

  • Out of radiance flag: It indicates how many bands are out of range for product confidence purposes.
  • Number of cloudy pixels in the largest smoothing processing window: indicating the confidence of the smoothing pre-processing step.
  • Quality flag: it checks the difference between the TPW (PGE06) and the sum of the three layers in each pixel.
• Post-processing:

  The post-processing step will include:

  • Local bias adjustment of each parameter (optional)
  • Smoothing of each layer when it is required in the model configuration file, given the proper parameters such as method to use (mean/average or median), length of the processing box and minimum number of cloud free pixels in the window to perform the smoothing. The process consists of making a mean/average or a median of the parameter over a window centred at the pixel.
  • Conversion of precipitable water for each layer from mm to counts.
  • Over cloudy pixel, a configurable IR channel is shown.

  Note: To use this algorithm with MSG1 in rapid scan mode located at 9.5º E, it is necessary to change the configuration file in order to include a calibration correction for each IR radiance.  The scale and offset for each IR band are activated as a global bias correction in the configuration file (safnwc_pge07_msg1.cfm), and they are applied during the pre-processing step. The local bias correction files for 9.5ºE are also called from this configuration file, which is supplied separately by the developers.

   

3.- List of inputs for LPW

• MSG radiance values: The following SEVIRI radiance values are needed at full IR spatial resolution:

R13.4 m	R12.0 m	R10.8m	R9.7m	R8.7m	R7.3m	R6.2m
Mandatory	Mandatory	Mandatory	Mandatory	Mandatory	Mandatory	Mandatory

The LPW system processing checks the availability of SEVIRI channels for each pixel; no results are produced for pixels where even only one channel is missing.
The SEVIRI channels are input by the user in HRIT format, and extracted on the processed region by SAFNWC software package.
 

• NWCSAF/MSG Products used as inputs:

• PGE01(Cma): The CMa is mandatory. After the CMa process is executed, the LPW system processing is computed.
• PGE06(TPW): The TPW is not mandatory. After the TPW process is executed, the LPW system processing is computed.

• Satellite zenith angles associated to selected region: This information is mandatory. It is computed by the LPW software itself, using a function available in the NWCLIB.
• Ancillary data sets: The following ancillary data, remapped onto satellite images, are mandatory:
  • Land/sea mask
  • Elevation mask

These ancillary data are available in the SAFNWC software package on MSG full disk in the default satellite projection at full IR resolution. They are extracted on the processed region by the LPW software itself.
Input Files: Region Configuration Files, Neuronal Network Files, PGE07 Model Configuration Files, are available in the SAFNWC software package, and are needed by the LPW software.

4.- Coverage and resolution

The software to extract Layer Precipitable Water (LPW) from MSG SEVIRI Imagery over MSG N (Europe, North Africa and adjacent seas) has been designed within the EUMETSAT SAF NWC.  The validity of this product for other areas will not be guaranteed. However, the processing of the  full MSG SEVIRI image is possible. The selected sub area must be a rectangular area defined in satellite projection of numlin x numele SEVIRI IR pixels.

This product will be derived for every slot at SEVIRI IR pixel horizontal resolution (15 minutes for nominal mode or 5 minutes for rapid scan mode).

5.-Description of LPW outputs
 

• Description of the conversion from mm to brightness values

The units of precipitable water are mm. To allow for a resolution better than 1 mm in BL, ML and HL values, it is necessary to use float values or to convert from mm to counts. The conversion to counts was selected since the final output size is smaller.

Each LPW parameter output is written in 8 bits (0:255). The first 8 values are reserved values, the next 120 values are precipitable water information and the last 128 are the values of the configurable IR band degrade to 7 bits.

The maximum and minimum used in the LPW parameters are the following 

Parameter	Minimum (mm)	Maximum (mm)	Range(mm)
BL	0	35	35
ML	0	45	45
HL	0	8	8

There are 120 grey levels for each LPW parameter, therefore the scales and offsets for the parameters are the following:

Parameter	Scale	Offset
BL	35/119	-8*35/119
ML	45/119	-8*45/119
HL	8/119	-8*8/119

 These parameters appear in the HDF files.

 

• Description of LPW output format

The format will be SAFNWC, the content of the LPW is HDF-5 format:

• 8 bits to show the BL:     BL clear air pixels.     Configurable IR channel cloudy pixels.

• 8 bits to show the ML:    ML clear air pixels.   Configurable IR channel cloudy pixels.

• 8 bits to show the HL:     HL clear air pixels.     Configurable IR channel cloudy pixels.

• 1 to describe clear sky flag:

   0     Air clear pixel. LPW is calculated
   1     Cloudy pixel. IR value is given

• 2 bits for Number of out of range radiances: Number of bands whose radiances are out of a range specified in the configuration file. The ranges of the radiances for the seven different channels are specified in the configuration file. Possible values are:

  0     All bands within the range
  1     One band out of range
  2     Between 2 and  4 bands out of range
  3     Between 5 and  7 bands out of range

• 2 bits to label number of cloudy pixels in the biggest window: This label presents information about the neighbours of the processed pixel depending on the cloudy pixels in the biggest window used to perform the spatial pre-processing of bands. The window width is a configurable parameter (normally: none, 3x3 or 5x5 pixels). Possible values are:

  0     [0, 1/8] cloudy pixels.   (If window 3x3: [0, 1] pixel     &     if window 5x5: [0, 3] pixels)
  1     (1/8, 1/4] cloudy pixels (If window 3x3:    [2] pixels     &     if window 5x5: [4, 6] pixels)
  2    (1/4, 1/2] cloudy pixels  (If window 3x3: [3, 4] pixels    &     if window 5x5: [7, 12] pixels)
  3    (1/2, 1] cloudy pixels     (If window 3x3: [5, 8] pixels    &     if window 5x5: [13, 24] pixels)

• 2 bits the quality flag: This flag checks the difference between the TPW (PGE06) and the sum of the three layers in each pixel. It depends on the % of the differences.

  0     High quality [0,10]%
  1     Medium quality (10,30]%
  2     Low quality (30,50]%
  3     Bad quality (<50%) or cloudy pixels or MSG zenith angle > threshold.

   

6.- Example of LPW visualisation

We present here as an example the images corresponding to the Gordon hurricane approaching the Azores Islands on the19th of September, 2006. The Layer Precipitable Water images have been derived under version 2.0.

Figure 1: Layer Precipitable Water output examples:(1) Boundary Layer (2) Middle Layer (3)  High Layer 

• Description of the example

The products of LPW have been generated from SEVIRI using the statistical retrieval method coded in version  2.0 of the 2008 SAFNWC package delivery. The products displayed are the three layers of atmospheric precipitable water vapour values: boundary Layer precipitable water vapor value (1), the middle layer precipitable water vapor value (2) and  the high layer precipitable water vapor value (3). (Precipitable water is the amount of liquid water, in millimeters, if all the atmospheric water vapor in the column is condensed). The value is color-coded with browns being the driest and reds being the moistest. The color scale is similar to the CIMSS TPW color scale. Clouds are represented as a grey color (see the color bar and the mm values at the bottom of the images).

One of the 2006 hurricanes became a special case, due to the change in its direction (away from the tropical waters where it was formed towards the open Atlantic – and the European landmasses). Gordon reached its secondary peak of 105 mph (165 km/h) on the morning of September 19th, before weakening slightly . By observing the images of the figure 1, it can be appreciated that the highest values for all layers (especially for low levels) are associated with the secondary peak of the hurricane.

 References

Miguel A. Martinez, Mercedes Velazquez, Gabriela Cuevas, Juan Ruiz, 2007. Improvements to the neural network retrieval of Layer Precipitable Water including an IR SEVIRI local radiance-bias correction. Joint 2007 EUMETSAT Meteorological Satellite Conference and the 15th Satellite Meteorology & Oceanography Conference of the American Meteorological Society. Amsterdam, The Netherlands.

Martinez, M.A., M. Velazquez, M. Manso, I. Mas. Application of LPW and SAI SAFNWC/MSG satellite products in pre-convective environments. Atmospheric Research, Volume 83, Issues 2-4, February 2007, Pages 366-379