The exploration of natural H2 is beginning in several countries. One of the most widely used methods for detecting promising areas is to measure the H2 percentage of the air contained in soils. All data show temporal and spatial variabilities. The gradient versus depth is not usually measured since the standard procedure is to drill and quickly install a tube in the soil to pump out the air. Drill bits used do not exceed one meter in length. These limitations have been overcome thanks to the development of a new tool that enables percussion drilling and gas measurements to be carried out with the same tool until 21 feet deep. This article shows the results obtained with this method in the foreland of the Colombian Andes. The variation of the gradient as a function of depth provides a better understanding of H2 leakage in soils. Contrary to widespread belief, this gradient is also highly variable, and, therefore, often negative. The signal is compatible with random and discontinuous H2 bubbles rising, but not with a permanent diffusive flow. Near-surface bacterial consumption should result in a H2 increase with depth; it may be true for the first tens of centimeters, but it is not observed between 1 and 5 m. The results show that, at least in this basin, it is not necessary to measure the H2 content at depths greater than the conventional one-meter depth to obtain a higher signal. However, the distance between the measured H2 peaks versus depth may provide information about the H2 leakage characteristics and, therefore, help quantify the near-surface H2 flow.