Nowcasting Rainfall from Meteosat Data for Africa

Abstract:

There is an urgent need for improved predictions of high-impact weather
in Africa, particularly of the heavy rain and floods that can result
from intense, large or slow-moving systems of deep moist convection.
Numerical weather prediction is inherently challenging in the tropics,
and in Africa this is compounded by a lack of in-situ observations for
model initialisation, and as a result skill of NWP for convective
rainfall is very often very low, even at short lead times. Nowcasting
provides a complementary approach to NWP, with the often large and
long-lived nature of the high-impact convective events making forward
extrapolation of observed systems valuable. Outside of South Africa
there is minimal coverage by meteorological radars, but excellent
spatial and temporal coverage from geostationary satellite. Here we
apply optical flow algorithms to rain rates retrieved from Meteosat
data, showing that there can be skill for many hours, and investigating
skill as a function of the meteorology. This work is taking place within
the GCRF Africa SWIFT project, and is running alongside developing new
AI techniques for probabilistic nowcasts. We will highlight strengths
and weaknesses of each approach in this context.