Ground robots are often tasked to autonomously explore and characterise unknown and sensitive environments. While performing these searching tasks, for applications such as interplanetary exploration, it is important that these robots also minimise how much they disturb the area that they are exploring to decrease contamination of the original environment. This work develops a Minimising Disturbance Informative Path Planner (MDIPP) which models an unknown environment using discrete observations whilst prioritising finding high concentrations of samples of interest. Disturbance to the environment is also incorporated into the path planner to minimise how much of the environment is being disturbed. MDIPP uses Gaussian processes (GPs) to model the unknown environment and Bayesian optimisation (BO) to determine the next best position to move to. MDIPP reduces the disturbance to the environment by at least 48.6% compared to a baseline informative path planner (BIPP), without sacrificing accuracy in the GP prediction generated from its online observations. The results have been validated in hardware on a tracked ground vehicle at CSIRO’s lunar testbed in Queensland, Australia to mimic planetary exploration.