Thin reservoir sands less than 40m thickness at an approximate depth of 2800m were successfully delineated using both P-wave (PP) and converted-wave (PS) shear data. Reflectivity associated with the reservoir sands is more clearly seen on the PS data than on the PP data. 58 wells with SP logs were used to train a neural network using both PP and PS data as input. Including the PS data was required to successfully predict the distribution of the reservoir sands. One well has been drilled on these results, with two others planned in early 2006. Training the neural network to predict the reservoir sands using P-wave data (migrated data and AVO attribute volumes) could not accurately delineate the reservoir sands without the PS data.