Recent years have seen an increase in surface water temperatures in several streams and fisheries, which has a detrimental effect on cold-water species such as trout. One possibility to resolve this issue is to create localized refugia of colder water generated through active cooling. The current work focuses on a prototype thermal refugia design and field testing. Various configurations of the prototype thermal refugia were tested in a stream, which could benefit from additional refugia regions. The prototypes featured a staggered, tube-bundle heat exchanger placed inside an enclosure with an aperture. The results demonstrate that in remote locations, man-made refugia can be provided. While the base enclosure (91.5 cm × 91.5 cm × 45.8 cm) allowed for excess mixing with the warmer free stream and resulted in low performance (dimensionless temperature difference of θ¯avg= 0.07), additional modifications improved performance. By utilizing a panel or an extension, the dimensionless temperature difference quadrupled (θ¯avg= 0.26) while the average heat transfer per dimensionless temperature difference was reduced to approximately one-fifth (1.92 kWth/θ¯avg) of the base enclosure. But these configurations increased the standard deviation of the temperature differences inside the refugia due to localized cooling. The combination of the panel and the extension did not further increase the standard deviation but resulted in an even higher dimensionless temperature difference (θ¯avg= 0.55) and a lower heat transfer per dimensionless temperature difference (0.81 kW/θ¯avg). This suggests that the enclosure design can be used to achieve a desirable temperature differential while maintaining a reasonable spatial fluctuation in that temperature difference and power requirement.