Monitoring African floods using the MSG IR3.9 channel

by Rogerio Bonifacio (UN World Food Programme, Sudan) and Jochen Kerkmann (EUMETSAT)

The Bulozi floodplain in Western Zambia plays an important role in local agriculture and economy. This vast area of wetland is one of the most important areas for the production of fish and cattle in southern Africa. It has an interesting cycle where populations change their activities during the year in accordance with the flood levels (recessional rice planting, livestock pasture, etc).

The Bulozi floods are seasonal and repeat themselves every year with peak extent in April-May. Traditionally, NOAA AVHRR data has been used to classify the floodplain and determine floodplain extent. An example of an AVHRR image of the floodplain from April 1999 is given in the following image (NOAA AVHRR, IR3.7 channel).

Recently, MSG has also been used to monitor the floods. An example from February 2004 is briefly described in the case study Floods over Eastern Angola and Western Zambia.

In Sudan there is a similar (larger) system in the south of the country, the Sudd Marshes formed by the spread of the Nile across a wide flat area (see NDVI map from January 2006). The UN World Food Programme is very interested in monitoring this seasonal flood cycle as it is crucial to the livelihood of a large population.

Peak extent/spread of the Sudd Marshes is usually in November-December, precisely when the ITCZ recedes and the wet weather moves away. Dry clear conditions are maintained until at least (late) March. This period of November–March is of interest as the flood recession dynamics are a factor in population well being (livestock, agriculture, fishing) though there isn't a lot of hard information since it is a fairly remote area.

As regards the Sudd Marshes and the Bulozi floodplain, monitoring of the flood extent from satellites is quite difficult with visible channels as there is a lot of swamp vegetation growing on water. In this situation, the reflectance signal may arise principally from the canopy rather from the water underneath.

It has been shown that the thermal data provided by the IR3.7 (AVHRR) or IR3.9 (MSG) channels may better represent the water surface (canopy temperature probably similar to that of the water). For example, on the Zambian floodplain the vegetation index data (NDVI) was relatively high over areas that the thermal data indicated as wet. A NDVI-thermal joint analysis would probably allow identification of 'vegetation on water', 'open water' and 'ground' classes.

The Meteosat-8 IR3.9 night-time images below show the situation of the floods in the Sudd Marshes during the season November 2005 to January 2006. In the first image (upper left) the area is largely covered by clouds and, thus, the flood extent is not detectable.

About one month later (upper right and lower left images) the thermal signature from the warmer water surfaces is clearly visible. Some hot spots (fires) and another flooded area in the region of the Sudan-Kenya border can also be detected. Finally, on 18 January 2006 the warmer (flooded) area has disappeared in the Sudd area. Some pictures showing the Channels of the River Nile and the swampy ground in the Sudd area are given below the satellite images.

Meteosat-8 IR3.9 images (night-time)

Photos from the Sudd Area

All photos supplied by WFP Sudan VAM Unit.