Hydraulic conductivity estimation by NMR data in unconsolidated geological materials: insights from SDR model calibrations

• NMR-based K prediction using the SDR model for unconsolidated geological materials. • Covers both organic & inorganic geological units from multiple sites. • Assesses K estimation using both field-based and laboratory-based datasets. • Tests general, site-specific, and geology-specific SDR calibration approaches. • Reducing K estimation uncertainty in sediments by excluding porosity from SDR model. Accurate hydraulic conductivity ( K ) estimation is essential for effective groundwater resource management, yet conventional methods struggle with spatial heterogeneity and scale disparities. This study evaluated nuclear magnetic resonance (NMR)-based K prediction using the Schlumberger-Doll Research (SDR) model across diverse unconsolidated environments. We analyzed 92 laboratory samples from nine sites and 198 borehole screens at eight sites in Denmark and Germany, encompassing sediments (sand, clay, till) and organic deposits (peat, gyttja). Bootstrapping and statistical analysis revealed that general calibration of the SDR model was successful for inorganic sediments but failed when organic geological units were present in the dataset. No correlation was found between the mean logarithm of the NMR transverse relaxation time distribution ( T 2ML ) and K in peat soils, making reliable K estimation impossible in these materials. However, gyttja – despite being an organic soil – allowed for reliable K estimation using geology-specific calibration, although with different SDR constant coefficients than those for inorganic sediments. Generally, geology- or site-specific calibrations significantly improved K accuracy for inorganic geological units when sufficient data points were available for the curve fitting, and T 2ML and K values spanned a sufficiently broad range (e.g., at least 1 to 1.5 orders of magnitude). We demonstrated that using a simplified SDR model – excluding porosity ( K = b T 2 M L n ) – for unconsolidated inorganic geological materials reduces uncertainty in the free-fitting coefficient parameters ( b and n ). Moreover, it avoids significant errors arising from magnetic susceptibility-induced porosity underestimation, which reached up to 70 % in our samples.