Permeability evaluation in deeply buried overpressured shaly sandstone gas reservoirs remains challenging because permeability is controlled not only by pore volume but also by pore-throat architecture, connectivity, and heterogeneity. In addition, differences between laboratory nuclear magnetic resonance (NMR) measurements and downhole NMR logging introduce echo-spacing dependence into transverse relaxation time ( T 2 ) distributions and derived descriptors, limiting the direct transfer of laboratory-calibrated permeability models to logging applications. To address this issue, we propose a pore-type-aware and echo-spacing-consistent NMR workflow for permeability evaluation in overpressured shaly sandstone gas reservoirs. Samples were first classified into three pore-system types using independent descriptors reflecting pore-structure characteristics, and the echo-spacing dependence of key NMR parameters was then examined under multiple acquisition conditions. The results show that, relative to the 0.2 ms reference condition, increasing echo spacing ( T E ) up to 1.2 ms progressively attenuates short- T 2 components, leading to underestimation of NMR porosity and systematic shifts in the geometric mean transverse relaxation time ( T 2GM ) and the fractal dimension ( D f ). Both the T 2GM and the D f derived from the T 2 distribution decrease monotonically with increasing echo spacing; this decrease is weakest in large-pore-dominated samples but strongest in micropore- and clay-rich samples. The corrected T 2GM and D f were then incorporated into a pore-type-specific modified Schlumberger-Doll Research (SDR) permeability model. Compared with the conventional SDR model, the proposed model improves internal predictive consistency and practical laboratory-to-log transferability within the present reservoir-specific calibration data set, with the most pronounced improvement observed in tight and structurally complex pore types. A field NMR log example further illustrates improved consistency between laboratory calibration and logging interpretation.
He et al. (Wed,) studied this question.