Abstract: A relative grain size value for subterranean depth intervals is generated by executing a machine learning model to process the petrophysical well log data. The relative grain size value represents relative proportions of grain sizes of sedimentary rock materials at the un-cored depth intervals. The machine learning model is trained based on an observed grain size a petrophysical well log data of sedimentary rock materials at cored intervals. Based on the relative grain size value generated from the machine learning model, flow index values for the one or more un-cored depth intervals of the production well are determined. The flow index is a function of the effective porosity, the total clay content, and the relative grain size value. One or more target depth intervals for production of hydrocarbons are determined based at least in part on the flow index values for the un-cored depth intervals.