Energy-efficency is a key concern when designing protocols for wireless sensor networks (WSN). This is of particular importance in commercial applications where demonstrable return on investment is a crucial factor. One such commercial application that motivated this work is telemetry and control for freight railroad trains. Since a railroad train has a global linear structure by nature, we consider in this paper linear WSNs as sensor networks having, roughly, a linear topology. Aiming at such networks, we introduce two routing schemes that efficiently utilize energy: Minimum Energy Relay Routing (MERR) and Adaptive MERR (AMERR). We derive a theoretical lower bound on the optimal power consumption of routing in a linear WSN, where we assume a Poisson model for the distribution of nodes along a linear path. We evaluate the efficiency of our protocols with respect to the theoretical optimal lower bound and with respect to other well-known protocols. AMERR achieves optimal performance for practical deployment settings, while MERR rapidly approaches optimal performance as sensors are more densely deployed. Compared to other protocols, we show that MERR and AMERR are less complex and have better scalability. We also postulate how both protocols might be generalized to a two-dimensional WSN.