We present the results of a seismic analysis of two hydrofractures spanning the entire diatomite column(1110-1910 ft or 338-582 m) in Shell's Phase II steamdrive pilot in South Belridge, California, These hydrofractures were induced at two depths (1110-1460 and 1560-1910 ft) and imaged passively using the seismic energy released during fracturing. The arrivals of shear waves from the cracking rock (''microseismic events'') were recorded at a 1 ms sampling rate by 56 geophones in three remote observation wells, resulting in 10GB of raw data. These arrival times were then inverted for the event locations, from which the hydrofracture geometry was inferred. A five-dimensional conjugate-gradient algorithm with a depth-dependent, but otherwise constant shear wave velocity model (CVM) was developed for the inversions. To validate CVM, we coated a layered shear wave velocity model of the formation and used it to calculate synthetic arrival times from known locations chosen at various depths along the estimated fracture plane. These arrival times were then inverted with CVM and the calculated locations compared with the known ones, quantifying the systematic error associated with the assumption of constant shear wave velocity. We also performed Monte Carlo sensitivity analyses on the synthetic arrival times to account for all other random errors that exist in field data. After determining the limitations of the inversion algorithm, we hand-picked the shear wave arrival times for both hydrofractures and inverted them with CVM. Finally, to correct for the areal inhomogeneity of the rock, we calculated the distortion of conical waves that were generated by air gun blasts in a remote observation well. This novel technique improved significantly the accuracy of the event locations in the shallow hydrofracture. The azimuth of both hydrofractures was N21 degrees +/- 4 degrees E. In each treatment well, there were two separate hydrofractures at two different depths that correspond to the diatomite layers with higher permeabilities. Both shallow hydrofractures were asymmetrical. Initially, the upper NE wing was 230 ft long, whereas the lower SW wing was only 30 ft long. The deep hydrofracture was symmetrical and the wings of its two parts were initially 130 and 10 ft long, respectively. These conclusions agree well with temperature surveys in the surrounding observation wells during steam injection.