Abstract We present a novel methodology for interpretation of high-angle and horizontal (HA/HZ) well data enabling comprehensive consideration of HA/HZ logs in large-scale reservoir characterization studies. The first step in the workflow is physics-based modeling and inversion of well logs, which yields reservoir structure (boundaries and faults, dips, cross bedding, etc.) and properties near the wellbore with high resolution. Subsequently, 3D geomodels are automatically updated with geometry and property information obtained in step one from log interpretation. We employ this in a giant carbonate field study to interpret hundreds of HA/HZ wells with the eventual objective to increase recovery by 25%, using our methodology integrated as a Web service into a geomodeling workflow. From the initial 3D geological model constructed using seismic and vertical well data, HA/HZ logs were inverted, and the results propagated back to the reservoir model. We perform refinements in well "curtain" cross-sections to match multiple log data by changing properties, dips, layer thicknesses, boundary and fault positions, and then automatically update the geometry and properties of the 3D model. Such automated 3D model update has never yet been attempted; in the case study, the logs were modeled at the rate of three wells per day, as compare to three days per well previously. Through this innovative approach, we (1) attain geomodels that honor high-resolution data through physics-based log modeling and automatic incorporation of inversion results into reservoir models; and (2) make feasible for reservoir engineers to directly refine geomodels while working on such tasks as formation evaluation (FE) and reservoir characterization. Introduction The structure of reservoir models is primarily based on low-resolution seismic data. High-resolution information about the formation structure and properties contained in the well logs—such as position and shape of bed boundaries, dip and azimuth of sub-seismic faults, or cross bedding— rarely becomes part of the geomodeling, and thus critical knowledge is not reflected in the full-field geological models. With the wealth of data from the new deep directional Electromagnetic (EM) tools on the market today [1], this is a considerable missed opportunity. Well logs are used as a data source for properties population, but usually environmental effects (such as invasion and borehole presence) are ignored. This assumption is invalid even in low-deviation, and especially in HA/HZ wells. Reservoir characterization applications are highly susceptible to model inaccuracies and greatly benefit from improved structure and properties distribution that come from formation evaluation; accurate geomodeling is vital in improving recovery and, ultimately, maximizing the profitability of an oilfield [2]. Building a structural and property earth model that would take well-log data into account requires that the results of log interpretation—updated geometry and physical properties—be integrated with the prior knowledge from seismic data in such a way that the resulting model agrees with both. This is a challenging task, for two main reasons. 1) Reliable interpretation of the modern logging tool responses in HA/HZ wells is only possible with proper, physics-based simulation and inversion; however, such log modeling capabilities are rarely available to geologists, reservoir engineers, and formation evaluation engineers: the simulation codes are very specialized and inherently computationally intensive, requiring high performance computing (HPC) resources. 2) Log interpretation yields high-resolution data, and integrating it with the low-resolution structural and property models is a daunting task.