Evaluation of TrueCell program for estimating point capillary pressure - saturation parameters for Flint sand

•Means of inversely estimated point parameters and neutron values were similar.•Individual TrueCell predictions were variable compared to neutron estimates.•Few TrueCell predictions fell within 95% confidence intervals of neutron estimates.•Deviations were related to variations in column packing.

The purpose of this paper is to compare inverse estimates of point capillary pressure-effective saturation, Se(ψ), parameters and functions with independently measured Se(ψ) parameters and functions. Average capillary pressure-effective saturation functions, < Se >(ψ), were measured for Flint sand using a hanging water column setup and 9 packed columns, with heights varying between 4.3 cm and 55.0 cm. TrueCell, a Fortran program, was employed to inversely estimate point Se(ψ) parameters and functions from the < Se >(ψ) data sets. The TrueCell estimates were then compared with point Se(ψ) parameters and functions determined by neutron radiographic imaging of a single hanging water column setup. One sample t-tests indicated that there were no significant differences (at p < 0.05) between the mean values of the inversely estimated point parameters and the corresponding neutron imaging values. However, the individual TrueCell predictions produced variable results compared to the measured point parameter set. Relatively few parameter estimates fell within the 95% confidence intervals of the neutron imaging estimates, and some deviations were quite large. These deviations were related to subtle variations in column packing, rather than differences in column height. Although this study has produced support for inverse modeling with TrueCell, it is important to note that individual predictions of Se(ψ) parameters and functions were often at odds with the independent measurements. Thus, TrueCell should be used with caution. In future studies of a similar nature, it would be desirable to independently determine point parameters and functions on multiple columns, rather than on a single column as was done here. Additionally, future research might want to examine the relationship between variations among individual parameters within a parameter set and the resulting predicted function.