Spring flooding of north Italian rice fields

By HansPeter Roesli (Switzerland)

In 2001 the case study Detection of flooded rice fields in Italy documented the flooding of the north Italian rice fields in spring with imagery from Meteosat 7, NOAA-12, Terra/Aqua and Landsat.

Advances in modern satellite technology means revisiting that area, covering the evolution in spring 2022, would show bettter details. The Meteosat-11 SEVIRI High Resolution Visible (HRV) band has a 2.5-fold higher spatial resolution than the VIS0.7 band of Meteosat-7, and the European polar satellites Sentinel-2/3 offer multi-spectral capabilities at 10m/300m spatial resolution, respectively.

The region inspected was the Po Valley west of Milan. On the May Sentinel-3 OLCI Terrestrial Chlorophyll Index (OTCI) image (Figure 1) an area full of blue speckles, framed with a black polygon, stands out against a differently coloured background. Essentially, it is situated in the Piedmont region, west of the river Ticino and north of the river Po, with Novara and Vercelli as the largest towns (for the OTCI algorithm used to compose this image, see further down).

Figure 2 shows before (3 March — top row) and after (10 May — bottom row) the flooding with data from Meteosat-11 SEVIRI and Sentinel-3A OLCI. They compare the HRV band (left images) to the OTCI (right).

OTCI images are displayed in the EO browser of the Sentinel Hub; the algorithm and colour scheme are described at the repository of custom scripts. Blue colours indicate extremely low OTCI indices, like water, sand or snow. Extremely high indices, displayed in white, also suggest the absence of chlorophyll, but over bare ground, rock or clouds. For indices in between these extreme values the specific colour table has shades from red (low chlorophyll values) over yellow and to dark green (high chlorophyll values); these shades render vegetation health. In the context here the blue (dark) shades are of most interest: they highlight the flooded paddies, and cannot be confused with equally blue-coloured snowy areas that appear off the Po Valley over the Pennine Alps in the north-west corner of the overview image above (Figure 1).

The OTCI images on the right side of Figure 2 show a dramatic change from red-dominated chlorophyll-poor fields in March (after the extremely dry winter 2021/22) to yellow-green-coloured chlorophyll-rich ones in May. In contrast the central part in the May image was dominated with blue-speckles. The speckles are due to the water surface of the paddies that in March were still dry and slightly vegetated (red to yellow shades).

The HRV couple on the left side appear in uniformly light grey in March (almost bare fields with moderately high reflectivity); only the major river valleys (Po at the bottom and its main tributaries Dora-Baltea, Sesia and Ticino from west to east) remain partially darker grey. In May fuzzy middle-grey patches (lower reflectivity) appear, in particular between the towns of Saluggia and Vercelli, as well as around Novara. Beside the considerable difference in spatial resolution, these patches compare well to the blue-stained areas in the corresponding OTCI image, confirming that they are due to the low reflectivity of the flooded paddies.

The decametric spatial resolution of Sentinel-2 MSI is able to resolve individual flooded paddies. Images downloaded from the EO browser covered an area corresponding to the area inside the black rectangle on Figure 3.

The Sentinel-2 imagery (Figure 4) again compares the situation before (25 March — top row) and after (11 May — bottom row) the flooding, using two different algorithms for MSI bands.

The left side shows what in the browser is called the ‘false-colour RGB composite’ (FCOL). The FCOL RGB composite scheme puts the near infrared (B8), red (B4) and green (B3) band on the red, green and blue beam, respectively. Since plants reflect near infrared and green light while they absorb red, dense vegetation is blazing red, e.g. the forested larger patch near the bottom of the images (red arrow). Due to water being a bad reflector in all three bands, flooded areas appear in dark shades (e.g. white arrows). The right side shows the normalised difference snow index (NDSI). NDSI is the normalised difference of two bands: one in the green (band 3 – 0.56μm) and one in the short-wave infrared region (band 11 – 1.61μm). While NDSI is usually used to differentiate between cloud and snow, it also highlights water surfaces. Like snow water surfaces get rendered in a vivid blue. Like with the OTCI images the spring flooding is well highlighted with this algorithm. Compared to the companion FCOL image on the left, even the network of dams/paths between the paddies stays out prominently.

A time sequence of the OTCI images has to be limited to essentially cloud-free overpasses and Sentinel-3A OLCI data. The OTCI algorithm is very sensitive to spectral differences of the bands between OLCIs on the two Sentinel-3 satellites, and even thin cloud veils impact heavily on the result. Going around clockwise through the images in Figure 5 reveals that the flooding was essentially occurring during the three weeks between the second part of April and beginning of May, essentially given by the difference between top and bottom row.

In the future, the Metop-SG satellites will provide more information on precipitation and flooding. Together with METImage, which can offer 500m resolution images for the same channels as in SEVIRI, the passive microwave imager MWI can distinguish flooded areas due to its different spectral signature over open water and flooded land areas. The advantage of microwave imagers is that they can provide information on flooded areas under cloudy conditions although they have a lower resolution compared to IR imagers.