| There are various quality problems associated with radar rainfall data viewed
in images that include ground clutter, beam blocking and anomalous propagation, to name a few.
To obtain the best rainfall estimate possible, techniques for removing ground clutter
(non-meteorological echoes that influence radar data quality) on 2-D radar rainfall image data
sets are presented here. These techniques concentrate on repairing the images in both a
computationally fast and accurate manner, and are nearest neighbour techniques of two sub-types:
Individual Target and Border Tracing. The contaminated data is estimated through Kriging,
considered the optimal technique for the spatial interpolation of Gaussian data, where the
"screening effect" that occurs with the Kriging weighting distribution around target points is
exploited to ensure computational efficiency. Matrix rank reduction techniques in combination
with Singular Value Decomposition (SVD) are also suggested for finding an efficient solution
to the Kriging Equations which can cope with near singular systems. Rainfall estimation at
ground level from radar rainfall volume scan data is of interest and importance in earth
bound applications such as hydrology and agriculture. As an extension of the above, Ordinary
Kriging is applied to three-dimensional radar rainfall data to estimate rainfall rate at
ground level.
Keywords: ground clutter, data infilling, Ordinary Kriging, nearest neighbours,
Singular Value Decomposition, border tracing, computation time, ground level rainfall
estimation |