AMERICAN DIPPERS NESTING NEAR JUNEAU, ALASKA 
in alpine zones and are fed largely by snow melt in spring and summer. Three 
streams originate at glaciers, nine originate in mid-elevation bogs, and two 
originate in natural nonglacial lakes. Seven of these streams were accessible 
only by boat, and we visited them only once; we include these streams only in 
the characterization of occupied and unoccupied watersheds. The remaining 
33 streams were accessible within a 3- or 4-hour round-trip hike from local 
roads. This set of streams we surveyed regularly includes all streams in the 
Juneau area except the three largest glacial rivers, mostly at a low gradient, 
and tiny intermittent streams. Initial surveys quickly showed that such very 
small streams did not support nesting dippers (see also Results). 
METHODS 
We assessed factors that have been thought to limit the dipper’s distribu- 
tion and abundance (e.g., Kingery 1996). To this end, we estimated stream 
flow, as an index of potential foraging space, measured characteristics of 
territories around known nest sites, sampled densities of macroinvertebrate 
prey in known foraging areas, and characterized nest sites. In addition, we 
monitored annual variation in apparent annual survival of banded birds and 
territory occupancy, using this information to assess the relative role of nest 
sites and food in limiting the local population of dippers. 
Estimating Stream Flow 
Because our sites were not equipped with stream gauges, we characterized 
streams by size (flow) as estimated by an equation based on watershed area, 
elevation, and precipitation (Wiley and Curran 2003). We delineated water- 
sheds by using a digital elevation model from the Shuttle Radar Topography 
Mission (SRTM; Werner 2001) in combination with digital hydrography 
interpreted by the U.S. Forest Service (2002) from aerial photography. The 
SRTM’s digital elevation model was the source for preliminary watershed 
boundaries drawn on the basis of predicted surface flow in the direction of 
maximum slope (Tarboton et al. 1991). These were compared with observed 
streams in the Forest Service’s database. Where we found discrepancies, 
we adjusted the digital elevation model to fit the flow pattern observed 
in the Forest Service’s database and recalculated watershed boundaries (Wer- 
ner 2001). The accuracy of the final delineation of watershed boundaries was 
verified with the Forest Service streams as well as with USGS topographic 
maps at a scale of 1:63,360. Later, we estimated the area of two additional 
watersheds visually by extrapolating boundaries with georeferenced aerial 
photos and topographic maps in an ArcGIS database; these estimates have a 
wider margin of error than those calculated by the first method. We estimated 
the average elevation of each watershed by placing a grid over the watershed 
on a topographic map (scale 1:25,000) and sampling grid squares at each 
elevation increment according to a random-number table (sample size was 
proportional to watershed area). 
Precipitation data were obtained from maps in Jones and Fahl (1994). It 
is important to emphasize that each of these variables is only an estimate, 
so the estimate of stream flow is rough. Precipitation is the variable with 
the least detail and the greatest potential source of error because there are 
193 
