Just reprocessed a 2016 3D cube with pre-stack inversion in Petrel, and our main AVO Class III anomaly migrated about 400 m downdip, which completely reshuffles our lead ranking and charge model. How are you quantifying inversion uncertainty in prospect risking, and do you update STOOIP/GRV before revising petrophysical cutoffs, or the other way around?
I’d treat the “400 m downdip” shift as a structural update: recompute GRV/STOOIP on the new picks first, then run an ensemble inversion (perturb wavelet/LFM/noise) to get posterior Ip→porosity and N:G maps, only nudging cutoffs if blind-well calibration drifts. For risking, convert the inversion to a reliability curve (ROC from blind wells) and apply it as a Pg modifier on reservoir presence so you don’t double-count uncertainty — let the trap volume set the table and the cutoffs just pick the utensils.
After a similar reprocess, treating the inversion as a classifier and measuring its miss rate stabilized our risking. > so you don’t double-count uncertainty — let the trap volume set the table and the cutoffs just pick the utensils. We recompute GRV/STOOIP on the new structure, then use a blind‑well confusion matrix (sand vs predicted sand) to set the probability cutoff and scale Pg/N:G, rather than tuning cutoffs by eye. Do you have enough blind footage to build that error model?
On our last Petrel pre-stack inversion (also a 2016 cube), I first re-tied wells and rebuilt the rock-physics template, then did a cutoff sweep on the new inversion to generate P10/P50/P90 net pay maps and feed STOOIP — so the ‘400 m downdip’ move doesn’t force a new cutoff. Are your checkshots and phase stable after the reprocess?