Using Meteosat-11 SEVIRI infrared and near-infrared RGB imagery it is possible to see differences in the seasons of 2017 and 2018.
By Jose Prieto (EUMETSAT)
Image composition based on the RGB scheme is usually applied to simultaneous colour sources, like the different channels from an instrument for a particular time. RGB is used to produce colour from objects based on multiple (often triple) observations.
The following two sets of imagery compare the periods January-September (nine months) in 2017 and 2018. There are two 'flavours' for the products: infrared 10.8 µm (IR, for short) and 0.8 µm (NIR). All the images are based on data filters to remove cloud, like the maximum (for IR) or minimum functions (for NIR) of brightness temperature and albedos respectively for periods of three months, with one image taken around 12 UTC to represent each channel.
In summary, the rules below apply to the image interpretation
The infrared evolution RGB reflects how COLD the first nine months in the year were in any area.
Bright areas were cold. If blue predominates, cold anomalies concentrated in the winter months, if green predominates, only the spring was cold. If red, only the summer. For instance, in France, 2017 appears red, indicating a cold summer thermal anomaly compared with 2018.
The near-infrared evolution product shows how DRY the same first nine months were.
Black areas had more rain and vegetation than other years. Bright areas were drier. If blue predominates, the winter was either snow-covered or dry, but spring and summer were more wet. If green, only spring was dry. If reddish, the summer was dry. For instance, 2018 autumn in Finland and the Baltic countries and down as far south as Ukraine was drier in 2018 than the previous year, and appears more in red.
See here for example, the precipitation comparison from FMI for both summers. Central-eastern Finland does not show much difference between the 2017 and 2018 RGBs, as is the case in the precipitation ground truth chart.
A limitation of the evolution RGBs is that one of the four seasons does not fit in the scheme — in this case the months October, November and December were left out. A possible improvement would be to consider meteorological seasons, rather than the natural quarters in the year.
Note:This case study is experimental! Constructive inputs for discussion are welcome.How to contact us User Service Helpdesk.
These RGB products are based on only three time points in the year, so they do not reflect average conditions, rather anomalies. For instance, a sudden snow event in a small area which is usually snow free will noticeably change the hue in that region, even if the snow was of little environmental or financial impact. They are useful for regions of the size of Austria or Italy, or bigger.