Soil moisture monitoring is a fundamental process to enhance agricultural\noutcomes and to protect the environment. The traditional methods for measuring\nmoisture content in soil are laborious and expensive, and therefore there is a\ngrowing interest in developing sensors and technologies which can reduce the\neffort and costs. In this work, we propose to use an autonomous mobile robot\nequipped with a state-of-the-art non-contact soil moisture sensor that builds\nmoisture maps on the fly and automatically selects the most optimal sampling\nlocations. The robot is guided by an autonomous exploration strategy driven by\nthe quality of the soil moisture model which indicates areas of the field where\nthe information is less precise. The sensor model follows the Poisson\ndistribution and we demonstrate how to integrate such measurements into the\nkriging framework. We also investigate a range of different exploration\nstrategies and assess their usefulness through a set of evaluation experiments\nbased on real soil moisture data collected from two different fields. We\ndemonstrate the benefits of using the adaptive measurement interval and\nadaptive sampling strategies for building better quality soil moisture models.\nThe presented method is general and can be applied to other scenarios where the\nmeasured phenomena directly affects the acquisition time and needs to be\nspatially mapped.\n