High data rate and energy-efficient design are of paramount importance for 5G systems. Massive MIMO can ensure both of these by extending the number of antennas at base station (BS) to a few hundreds and operating in time division duplex (TDD) mode to serve several tens of terminals. This study evaluates the impact of user location distribution (ULD) variation on the energy efficiency of a load-adaptive massive MIMO system. Then, it suggests a dynamic resource allocation strategy that exploits the advantage of ULD variations to attain a more energy-efficient design. Daily ULD variation is modeled by splitting the cell into a certain number of coverage areas and assigning different user densities to each one for each hour. This modeling yields different ULD variations, such as boundary focused (BF), middle focused (MF), uniform, and center focused (CF) ULD variations. For clarity, all cells are assumed identical in terms of BS configurations, cell loading and ULD variation. The simulation performed in this study benchmarks the proposed dynamic strategy with a fixed strategy dimensioned at maximum cell load and BF ULD model. The results show that the optimal number of antennas depends primarily on ULD model and secondarily on cell loading. Up to one third of the active antennas can be turned off on a daily basis, and this in turn conserves up to 36-50% of the consumed energy. (C) 2016 Elsevier GmbH. All rights reserved.