The impact of piscivores in aquatic systems is often estimated by assuming that predation rate and predator density can be characterized as means throughout large, homogeneous areas. Predation losses of juvenile Pacific salmonids Oncorhynchus spp. migrating through Columbia River reservoirs were previously estimated with the assumption that each reservoir consisted of one or two homogeneous areas. Data from the John Day Reservoir and throughout the river system showed that predation rate and predator density vary greatly between near-dam and midreservoir areas, suggesting that reservoirs in the Columbia River should be divided into at least three or four areas for estimating salmonid losses. For example, the estimated number of salmonids annually eaten by northern squawfish Ptychocheilus oregonensis in John Day Reservoir decreased from 2.9 million when all samples in the reservoir were pooled into one area, to 1.4 million when samples were partitioned among four areas. Variance about the estimates also decreased steadily with finer partitioning. Mortality of juvenile salmon from predation was substantial with any type of partitioning; however, spatial variation in predation rates and other density-dependent processes may be especially important in river models of migrating juvenile salmon that repeatedly apply predation rates in a series of reservoirs or river reaches.
