Abstract
The problem of detection of an increase in runoff due to precipitation management is considered. Classical methods of detection are reviewed and their shortcomings are discussed. The concept of grouping of observations is introduced. The paper answers two fundamental questions:
Given a region consisting of B basins in which changes are suspected and given that (economic, or financial, or other) constraints limit to b the number of basins where measurements can be obtained, which basins should he selected?
How should the measurements in individual basins be combined?
The answers are obtained by a constrained optimization procedure. When applied to the Colorado River Basin Pilot Project area the power of the test, expressed in years needed for detection, is increased by a factor of 2.
