This paper describes an experiment in an open sea environment and analyses the different configurations of the LoRa model (spreading factor, bandwidth, coding rate) for oversea signals. We determine a channel model and find the transceiver locations as a function of the received signal power. We show two locations can be differentiated with a 100 m error at 1.6 km maximal distance. We prove there is no significant differences between all possible configurations in terms of accuracy. Thus, we select the fastest configuration in order to increase the database measurement.