This thesis investigates wireless sensor network design for industrial petroleum refineries. A wireless channel measurement campaign is conducted in a Shell Canada gas refinery west of Calgary. The propagation measurements are the first to characterize the large-scale channel statistics for peer-to-peer transmission in an outdoor refinery environment. It is shown that the propagation is well characterized by the familiar standard pathloss and log-normal shadowing model. This model is then used in building a machine-to-machine network performance simulation for process control in a petroleum refinery. The simulation requires the use of cross-layer optimizations to determine the routes which maximize the network lifetime. Linear Programming and Mixed Integer Linear Programming problems are formulated and evaluated to solve the cross-layer design problem. Carrier Sense Multiple Access (CSMA) and Time Division Multiple Access (TDMA) schemes are both used, but it is shown that TDMA has superior performance.