Probability of the formation of a
thunderstorm depends on evolution of local condition and on advection
of clouds. For this second point, CI is unfortunately too scarce for
a full object-approach that allows a good following of meteorological
systems. CI is a pixel product
This preliminary step requires in optimum configuration
Then, NWP data are used for a guidance when available, to eliminate stable areas and focus on more unstable pixels
NWCSAF cloud type product is used to eliminate non cloudy areas and focus on cloudy pixels
![]() CT product |
![]() Non cloudy pixels identification |
![]() mask for stretched pixels (backup file) |
![]() stretched pixels + non cloudy pixels |
![]() IR10.8 image |
![]() Working image for both object and pixel approach for CI |
Thus, large areas are ignored in the
following processes, which may focus on a restricted set of pixels to
be
analyzed.
A 2D movement field is estimated in optimum configuration with blending NWP wind field in the low level (850hPa) and last available HRW wind observations, remapped on grid-field and selected versus the corresponding pixel’s brightness temperature. Priority in the blending is given to HRW wind observations, affected to a 9-pixels-box centred on the corresponding pixel.
This blended field is
at first used as
guess movement for initialization in the tracking process of object
analysis (“cold start” cases, or orphan cells).
Then this field is updated with objects movement vectors, to finally be
considered as a pixel tracker for trends calculations.
An object analysis process is undertaken like in RDT-CW software, but has been adapted to focus on warm cloud cells from lowest layers. The objectives of this step are
To take benefit from techniques allowing to catch cloud cells movement
To access cloud cells’ parameters variations along its trajectory
More details about tracking can be found in RDT-CW algorithm description.
CI-specificities rely on
Specific Cold limit for adaptative thresholding : the limit has been set to -25°C (instead of -75°C for RDT-CW) in order to limit the analysis to lowest levels
Minimum vertical extension of objects, which has been set to 3° (instead of 6° for RDT-CW) to focus on lower extended cloud systems
This step takes benefit from movement guess field as input to increase cell’s speed reliability, and on the other hand delivers as output an updated movement field with the analyzed objects’ speeds. All pixels belonging to a tracked cloud system are affected the corresponding movement speed instead of previous pixel’s movement values.
This final blended movement field is a key point for further relevant trends calculations
Brightness Temperature Differences are processed for each eligible-CI pixel from various available channels, for current data and data from previous slot.
BTDs taken into account are
WV6.2-WV7.3,
WV6.2-IR10.8
IR10.8-IR8.7
IR12.0-IR10.8,
IR13.4-IR10.8
Then, BT (IR10.8) and BTD trends are calculated for each eligible-CI pixel using the speed and direction of updated 2D movement field as guidance for identifying pairs of current and corresponding pixels in previous image.
When tracking of aggregated pixels (belonging to a tracked cloud system as object) is available, corresponding trends are used for some parameters instead of single pixel-trends, and should be able to provide trends over longer depth.
Each eligible-CI has then a list of BT and BTD values and trends. According to previous studies about convection initiation, parameters are grouped as:
Representative for Cloud-top Glaciation
IR10.8 Brightness temperature, Time spent below freezing level, IR10.8-IR8.7 BTD
Representative for Cloud depth / vertical extension
WV6.2-IR10.8 BTD , IR13.4-IR10.8 BTD , IR12.0-IR10.8 BTD
Representative for Cloud growth (updraft)
All BTDs trends , IR10.8 BT trends
Parameter name |
Relevant value |
Meaning |
BT IR10.8 |
> -25° |
Brightness temperature (glaciation) |
BTZG |
Within 30min |
Time since crossing 0°C (glaciation) |
BTD4 |
]-10° , 0°C[ |
IR10.8-IR8.7 (glaciation) |
BTD |
]-35° , -10°C[ |
WV6.2-IR10.8 (height) |
BTD6 |
]-25° , -5°C[ |
IR13.4-IR10.8 (height) |
BTD5 |
]-3° , 0°C[ |
IR12.0-IR10.8 (height) |
TxBT 15’ |
] -4°/15’ , -50°/15’[ |
Temperature change rate (growth) |
TxBT 30’ |
] -4°/15’ , -50°/15’[ |
Temperature change rate (growth) |
TxBTD 15’ |
> 3°/15’ |
BTD 15 Trend (growth) |
TxBTD 30’ |
|
BTD 30 Trend (growth) |
TxBTD4 15’ |
]0°/15’ , 10°/15’[ |
BTD 15 Trend (growth) |
TxBTD5 15’ |
]0°/15’ , 10°/15’[ |
BTD 15 Trend (growth) |
TxBTD6 15’ |
> 3°/15’ |
BTD 15 Trend (growth) |
CI-diagnosis should be derived from statistical models using Interest fields’ values of pre-CI pixels. Those models will rely on a specific ground truth (high reflectivity from radar data, convective cells from RDT-CW, lightning data). In logistic regression a pixel that belongs to a convective path during a given period will be considered as ground truth. Such a tuning has been postponed to the next release.
Current
CI output is
estimated with empirical rules defined by count of relevant criteria.
The v
Nb of Growth relevant parameters (over 3) |
Nb of Glaciation relevant parameters (over 3) |
Nb of Height relevant parameters (over 4) |
Result
|
>or= 2 |
>or= 3 |
>or= 4 |
HIGHPROB |
|
|
>or= 3 |
MODPROB |
|
|
< 3 |
LOWPROB |
|
>or= 2 |
>or= 4 |
MODPROB |
|
|
>or= 3 |
LOWPROB |
|
|
<3 |
VLOWPROB |
>or= 1 |
>or= 3 |
>or= 4 |
MODPROB |
|
|
< 4 |
LOWPROB |
|
>or= 2 |
>or= 4 |
LOWPROB |
|
|
>or= 3 |
VLOWPROB |
0 |
>or= 3 |
>or= 3 |
LOWPROB |
|
|
<3 |
VLOWPROB |
Other cases |
0 |
Empirical rules for CI-diagnosis. HIGHPROB means between 75 and 100%, MODPROB between 50 and 75%, LOWPROB between 25 and 50%, VLOWPROB between 0 and 25%
|
![]() Example of CI probability for next 30min. Here a value of 3 stands for a probability of 75% |