A stratified random survey design proposed to increase the efficiency of estimating numbers of waterfowl wintering on Mississippi catfish ponds was evaluated. The optimally allocated sample generally produced estimates with coefficients of variation <50%, similar to those obtained from a completely random design used previously. Coefficients of variation were not associated with survey date. Stratified random sampling reduced the number of catfish pond clusters surveyed and flight time, compared to completely random sampling. We recommend using the stratified random design to estimate numbers of waterfowl on aquacultural impoundments when sampling of impoundments must be limited.