Automatic Satellite Image Interpretation-Next Generation 


Table of contents

1. Goal of ASII-NG product  
2. ASII-NG algorithm summary description  
3. List of inputs for ASII-NG  
4. Coverage and resolution 
5. Description of ASII-NG outputs 
6. Example of ASII-NG visualisation

Access to "Algorithm Theoretical Basis Document for the "Automatic Satellite Image Interpretation" product (ASII-NG PGE17, v1.1)" for a more detailed description.



1. Goal of ASII-NG product

ASII-NG aims at detecting atmospheric features which are of interest to meteorologists and other users. In contrast to the ASII product (further development is frozen) where the identification of conceptual models was in the center of interest, the ASII-NG product identifies Clear Air Turbulence (CAT) which is directly relevant for meteorologists and e.g. aviation end users. 
Clear-air turbulence is non-convective turbulence outside the planetary boundary layer, often in the upper troposphere. CAT typically has a patchy structure and horizontal dimensions of 80-500 km in the along-wind direction and 20-100 km in the across-wind direction. Vertical dimensions are 500-1000 m, and the lifespan of CAT is between half an hour and a day (Overeem 2002). As CAT involves physical processes with scales usually smaller than the resolution of numerical weather prediction (NWP) models, forecasts of CAT with NWP are difficult to perform. Therefore, it is of interest to identify areas with risk of CAT from satellite observations. CAT is preferentially triggered by:

  • Tropopause folds
  • Gravity waves (e.g. lee waves)
  • Air mass boundaries (e.g. fronts)
  • Wind shear (e.g. jets)

At this early development stage, the feature detection is limited to tropopause folds and gravity waves (prototype, not delivered with NWC GEO v2016).


2. ASII-NG algorithm summary description

Tropopause folds

A tropopause fold describes the downward intrusion of stratospheric air into the troposphere which results in a folding of the tropopause, as schematically illustrated in Figure 1. Typically at a tropopause fold there is the vertical shearing at the jet stream combined with the ageostrophic convergence of polar, subtropical, and stratospheric air masses. Tropopause folds mark the change in the height of the tropopause and are characterized by the occurrence of strong turbulence. Stratospheric air, which characteristically has a low moisture content and a high potential vorticity can protrude down to the mid or even the lower troposphere. Consequently, tropopause folds can be located by their association with gradients in upper level moisture, which are evident in the SEVIRI 6.2 µm channel sensitive to upper tropospheric water vapour. The tropopause height can be estimated from NWP parameters such as the temperature, specific humidity or potential vorticity. Additionally, the ozone content derived from SEVIRI IR channel at 9.7 µm is a good indicator for the presence of a tropopause fold: the lower the tropopause the more ozone is sensed by the instrument. 
To derive the probability of a tropopause fold, a logistic regression relation is applied (input parameters given in section 3) .


Figure 1: Illustration of a tropopause fold. (© Feltz and Wimmers 2010)


3. List of inputs for ASII-NG

Tropopause folds

Table 1 shows the input parameters required for the detection of turbulence related to tropopause folds. Two heights of the tropopause [hPa] are calculated from the NWP data, one based on specific humidity, and the other based on potential vorticity. For both tropopause heights, it is the gradient field that serves as input into the logistic regression. It is recommended that NWP data be provided up to the 50 hPa level to ensure that the tropopause is captured. Brightness temperature from channels IR 9.7 µm and IR 10.8 µm directly serve as input, yet there are also some post-processed satellite data involved: A smoothing operator is applied twice to each of WV 6.2 µm, IR 10.8 µm and the channel difference (IR 9.7 µm – IR 10.8 µm) before the gradient fields are calculated. These gradient fields are then input into the logistic regression relation.

Table 1: Input to logistic regression for tropopause folds

NWP parameter

Satellite data

specific humidity

WV 6.2 µm

potential vorticity

IR 9.7 µm

shear vorticity at 300 hPa

IR 10.8 µm

wind speed at 300 hPa



4. Coverage and resolution

  • The product is extracted in full SEVIRI pixel resolution using the satellite projection.
  • The region for processing is user defined.


5. Description of ASII outputs

The ASII-NG product is encoded in a standard NWCSAF netCDF output file. Apart from the standard fields, the netCDF file holds the derived probability for occurrence of tropopause folding. For each pixel a value between 0 and 100% is attributed. Additionally, status flags are provided to give details on processing issues.


6. Example of ASII visualisation

Figure 2: Probability of tropopause fold (highest probabilities in red; black lines encircle subjectively analysed tropopause folds of this case).