**4. Well-based 3D modeling using geostatistics and artificial intelligence: propagating the surface drilling derived logs across the reservoir volume**

 In Sections 2 and 3, it is shown how the rock mechanical properties, pore pressure and stresses are extracted along the lateral section of the wellbore and are used as reference properties to engineer the completion to improve the overall efficiency. In this section, these properties are propagated in 3D space to accurately characterize the reservoir, so that these inputs can be fed into the hydraulic fracturing design and reservoir simulation workflows explained in the subsequent sections. The large number of wells drilled in unconventional assets combined with the estimation of critical logs at all the wells from surface drilling data provides the opportunity to propagate the well information into a 3D reservoir model. Since many companies do not have seismic on their acreage or for budgetary reasons do not plan to license the existing seismic, these multiple logs derived at all of the wells allow the construction of reliable 3D reservoir models. These 3D models could be estimated in a stratigraphic framework over a large area that encompasses many wells. In such cases, geostatistics could be used to estimate the distribution of gamma ray, porosity, Young's Modulus, Poisson's ratio and shear modulus. However, the pore pressure, minimum stresses and natural fracture are more complex continuous properties that need to be estimated with neural networks [8, 9] and other artificial intelligence tools able to capture the complex geology that control their variability.

 One major reason for propagating these rock properties in 3D is to provide that information to a 3D planar hydraulic fracturing simulator as well as to geomechanical software. To achieve this goal, all the wells are used together in a large reservoir grid to create the 3D models from which smaller well grids (**Figure 4A**) will be extracted around a well or a pad. With this approach, all the available well data will be used to improve the 3D distribution of the key properties needed for the 3D

#### **Figure 4.**

*(A) A large stratigraphic 3D geocellular grid is built from all the available wells to propagate 3D reservoir models. A smaller, higher resolution grid is extracted around the well that provides to the 3D planar hydraulic fracturing simulator: (B) Young's Modulus, (C) Poisson's ratio, and (D) unconfined compressive strength (UCS). (E) A cross section in the well grid of the minimum stress and (F) Poisson's ratio honoring the lateral and vertical variability captured by the 3D models."* 

planar hydraulic fracturing simulator. The other benefit of these derived 3D models is the estimation of the stress gradients resulting from the interaction between regional stresses and the three sources of stress perturbation created by the local geology: (1) variable geomechanical properties, (2) pore pressure and (3) natural fractures all available from the extrapolation in 3D of the logs derived from surface drilling data. Because of their importance in estimating these stress gradients, the modeling of the natural fractures requires some particular attention.
