Print this pageSensitivity/Risk Assessment with Basin Models: Approach and Case Example from a Frontier Setting
LANDER, R. H, V. FELT A, and L. BONNELL
Most basin modeling studies are conducted using a set of 1most likely' input
parameters that are tuned to fit available calibration data. While such studies
can provide useful qualitative insights into controls on the hydrocarbon system
they cannot be applied rigorously to evaluation of risk. In this study we
investigated a number of key input parameter uncertainties for a frontier region
characterized by a single calibration well. The purpose of the study was to
determine the uncertainties in model predictions of the extent and timing of
hydrocarbon generation and oil to gas cracking. We reconstructed the history of
erosion/deposition, compaction, structural displacement, isostatic and tectonic
subsidence, heatflow, temperature, hydrocarbon generation, and oil to gas
cracking for a 
cross
section using BMT 3.3.
In all we constructed over 60 "optimized" models that agreed with the calibration well data to within measurement uncertainties. These models explore the effect of the following input uncertainties in model predictions: time to depth conversion factors; the number, magnitude, and timing of erosional events; the magnitude and timing of rifting events and associated heatflow pulses, crustal and subcrustal properties affecting subsidence, background heatflow values, thermal conductivities of lithologies, kerogen to hydrocarbon transformation kinetics, and vitrinite reflectance models.
Uncertainties in the timing of hydrocarbon generation vary greatly over the
modeled section
. In general predictions for the deep kitchen and shallow
platform areas are comparatively insensitive to input parameter uncertainties.
In intermediate locations, however, the extent of hydrocarbon generation and oil
to gas cracking varies greatly and is most sensitive to the history of erosion;
the heatflow history, and the oil to gas cracking kinetics. For the study area
large uncertainties in the following factors had little effect on the
predictions for the optimized models: time to depth conversion, crust/subcrust
properties, background heatflow, thermal conductivity, kerogen to hydrocarbon
transformation kinetics, and vitrinite reflectance algorithm.