In December of 1994 a fluid injection experiment which triggered several hu
ndreds of microearthquakes was conducted at the KTB main borehole (Oberpfal
z, Germany). These events were recorded with a temporal seismic network at
the surface. Out of the complete data set, a cluster of five events recorde
d at four mini-arrays consisting of eight or nine stations was used to inve
stigate the crustal scattering properties in the vicinity of the KTB. For t
his purpose, the 'Double Beam Method' (DBM; Kruger et nl. 1993, 1995, 1996)
and the 'Double Beam Imaging Method' (Scherbaum, Kruger & Weber 1997) were
extended to curved wave fronts to drop the restriction of planewave propag
ation. This technique is used for imaging the crustal scattering strength u
sing earthquake clusters recorded at close-by mini-arrays. The results of t
he array analysis show that the composition of the P coda is mainly affecte
d by the site location of the arrays. Near-surface and deeper crustal scatt
ering contribute in a very complicated pattern. Furthermore, with the prese
nt data set it was possible to identify reflections from the top of the Erb
endorf Body. This is a very pronounced arrival in mast of the recorded trac
es. In one of the arrays its amplitudes are even greater than the direct P
phases. Five to eight coherent phases could be identified by the mini-array
s. Using only these phases, synthetic P-coda traces were constructed, which
only contain the coherent part of the observed wavefield. By subtracting t
he synthetic coherent wavefield from the original traces we achieve a varia
nce reduction in the P coda of up to 37 per cent. This leads to the conclus
ion that a large amount of the P coda at the KTB can be modelled by a simpl
e deterministic single-scattering model using a small number of individual
scatterers.