We explore the problem of energy-efficient, time-constrained path planning of a solar-powered robot embedded in a terrestrial environment. Because of the effects of changing weather conditions, as well as sensing concerns in complex environments, a new method for solar power prediction is desirable. We present a method that uses Gaussian Process regression to build a solar map in a data-driven fashion. Using this map and an empirical model for energy consumption, we perform dynamic programming to find energy-minimal paths. We validate our map construction and path-planning algorithms with outdoor experiments, and we perform simulations on our solar maps to further determine the limits of our approach. Our results show that we can effectively construct a solar map using only a simple current measurement circuit and basic GPS localization, and this solar map can be used for energy-efficient navigation. This establishes informed solar harvesting as a viable option for extending system lifetime even in complex environments with low-cost commercial solar panels.