This thesis presents a sampling policy which offers the experimenter some statistical grounds upon which to base a decision as to when subsequent stream samples should be taken so as to detect a polluted condition. For any proposed sampling policy, the experimenter, without actually sampling, can evaluate the frequency of required sampling, can obtain a "feel" for how quickly changes in stream conditions can be detected, and can make comparisons with capabilities offered by other proposed sampling policies. Regression analysis of field data was conducted to explore possible relationships between stream dissolved oxygen content and physical factors. The resulting models were incorporated into the above mentioned sampling policy. / Master of Science