Cluster variable aperture (CVA) simulated annealing has been used as an inv
ersion technique to construct fluid flow models of fractured formations bas
ed on transient pressure data from hydraulic tests. A two-dimensional fract
ure network system is represented as a filled regular lattice of fracture e
lements. The algorithm iteratively changes element apertures for a cluster
of fracture elements in order to improve the match to observed pressure tra
nsients. Aperture size is chosen randomly from a list of discrete apertures
. The cluster size is held constant throughout the iterations. Since hydrau
lic inversion is inherently nonunique, it is important to use additional in
formation. We investigated the relationship between the scale of heterogene
ity and the optimal cluster size and shape to enhance convergence of the in
version and improve the results. In a spatially correlated transmissivity f
ield, a cluster size corresponding to about 20% to 40% of the practical ran
ge of the spatial correlation is optimal. Inversion results of the Raymond
test site data are also presented and based on an optimal cluster size; the
practical range of the spatial correlation is estimated to be 5 to 10 m.