Two methods applying natural heat as a tracer to quantify surface water–groundwater exchange were evaluated using field data. Arrays capable of monitoring and recording the streambed response to diurnal temperature variations in the surface water were deployed for a 2-month period in three locations in perennial pools at Maules Creek, New South Wales, Australia. Multi-level array design, field deployment and parameter estimation are discussed. The applicability of analytical solutions derived from the heat transport equation to the streambed environments was analysed using the recorded temperature time series. The stream was found to lose water to the aquifer, which was supported by simultaneously recorded hydraulic gradients. However, the one-dimensional (1D) analytical solutions did not adequately describe the observed streambed thermal response at two locations. The resulting artefacts in the estimated flow velocities are discussed. It was hypothesised that the artefacts originate from model limitation due to streambed heterogeneity and application of 1D solutions to multi-dimensional and dynamic streambed flow. This consequently imposes limitations on the field applicability of the methods. Nevertheless, in combination with time series of surface water and streambed water levels, the use of heat as a tracer provided a powerful tool for better understanding the shallow hydrogeological system.