Floods around central Mediterranean region

Europe

Sentinel-3

SLSTR

Natural Colour RGB, Night Microphysics RGB

02 June 2023

01 June 2023

By Ivan Smiljanic & Djordje Gencic

For a number of days in May there was extreme rainfall across northern Italy and central Europe, including parts of Austria, Bosnia and Herzegovina, Croatia and Slovenia. Accumulated rainfall amounts, that in some places reached historical levels, were responsible for widespread floods and many landslides across a large area (Figure 1).

The World Health Organization reported that at least 13 people died in Italy and thousands had to flee their homes.

Numerous low-pressure systems passed the wider Mediterranean area in May, leaving behind more and more saturated soil, and river beds at the higher water levels. Saturated soil contributed also to many of smaller-scale flash floods and land-slides, triggered by localised convective rainfall events.

The final cyclone in a stream of those that passed over Europe, tipped the balance — producing widespread hazards. The evolution of this system, together with number of preceding ones is captured in the 10-day Meteosat-10 Airmass RGB loop (Figure 2).

In Figure 3 one can see a comparison of VIS 0.8µm images for 5 May, when the rivers were at normal levels, and 21 May after significant and widespread flooding that occurred in Croatia.

Figure 3: Comparison of VIS 0.8um imagery for normal river levels (left) and widespread flooding (right)

Since the VIS0.8 um channel in Sentinel-3 SLSTR is at 500m resolution, very fine details and meandering of rivers Sava and Kupa are shown clearly and especially in the right hand side image (21 May). It should also be mentioned that 0.8µm or vegetation channel is particularly sensitive to chlorophyll, which typically shows high reflection (in greyscale display — bright grey), whereas water surfaces reflect very little (black or dark grey). These two facts make 0.8µm channel very useful for detecting fluvial flooding, and for discrimination of land/water surfaces. In our example a very noticeable increase in river size can be seen between 5 and 21 May.

Higher resolution imagery from SLSTR reveals the large flooded areas, found mostly along major rivers in the domain. The Natural Colour RGB, that utilises the solar bands of SLSTR at 500m resolution, nicely contrast the flooded regions (in black/deep blue) against the vegetated areas (in green). This contrast comes from strong reflection from vegetation compared to poor reflection from water bodies (especially if deep and non-turbid).

A comparison before/after flooding is provided in Figure 4.

Figure 4: Before-after comparison of the flooded areas using Sentinel-3 SLSTR Natual Colour RGB, 5 May v 21 May

Aside from the daytime 'visual' reference on the flooded area, there is a way to detect flooded areas both during the day and at night, using imager instruments such as SLSTR, SEVIRI or FCI. It relies on the brightness temperature difference between IR channels around 10.5µm and 3.8µm, which shows relatively larger positive values (ca. 5K) compared to surrounding soil/vegetation. Of course, for floods of this size, relatively high resolution of these IR bands is required — in this case imagery at 1km shows some signal of floods even around the smaller rivers in the observed domain. Having these two IR bands at 1km, in the FCI instrument onboard MTG-I satellites, will potentially have this 24-h capability for detection of smaller scale floods around Europe (and elsewhere).

Figure 5 shows comparison of SLSTR Natural Colour RGB and Night Microphysics RGB (this RGB contains aforementioned channel difference on the green beam, revealing water bodies in the cyan hues). Note that the FCI instrument is the first of this kind in the geostationary orbit that has these two IR channels at resolution of 1km (others have 2 km or more).

Figure 5: Comparison of the flooded area view between Sentinel-3 SLSTR Natural Colour and Night Microphysics RGB, 21 May