In the absence of eyewitness reports or clear sedimentary structures, it can be difficult to interpret tsunami deposits or reconstruct tsunami inundation patterns. The emplacement dynamics of two historical tsunami deposits were investigated at seven transects in Okains Bay, New Zealand, using a combined geospatial, geomagnetic and sedimentological approach. The tsunami deposits are present as layers of sand and silt intercalated between soils and become finer and thinner with distance inland. The deposits are attributed to the 1960 and possibly the 1868 tsunamis, based on radiometric dating and correlation with historical records. Measurements of Magnetic Fabric (MF: Anisotropy of Magnetic Susceptibility) and particle size were used to reconstruct the evolution of flow dynamics laterally and vertically. A combination of statistical methods, including spatial autocorrelation testing, Spearman's rank order correlation, Principal Component Analysis (PCA) and K-means cluster analysis, was applied to examine relationships between MF parameters and sediment texture, and infer depositional hydrodynamics. Flow patterns deduced from MF show the estuary channel acted as a conduit for inundation, with flow commonly aligned sub-perpendicular to the estuary bed. MF and sediment data suggest deposition occurred from settling during laminar flow. Evidence of both uprush and backwash deposition, as well as wave reflection from infrastructure, was found. Statistical analysis of data showed significant relationships between grain size parameters and MF parameters associated with flow speed and magnetic fabric type. PCA and cluster analysis differentiated samples into two primary hydrodynamic groups: (1) samples deposited from laminar flow; and (2) samples deposited close to the limit of inundation, which includes samples deposited further inland, those affected by flow convergence, and those in the upper part of tsunami deposits. This approach has potential as a tool for reconstructing hydrodynamic conditions for palaeotsunamis and by combining spatial and statistical analyses, large-scale investigations can be more easily performed. Copyright © 2016 John Wiley & Sons, Ltd.