Estuarine residence and growth of juvenile chum salmon (Oncorhynchus keta) from Netarts Bay, Oregon were estimated from daily-formed growth increments of sagittal otoliths which are distinguishable from accretion patterns formed during freshwater residence. Estuarine residence time was inversely related to the average size at which juvenile chum salmon entered Netarts Bay. Among fin-clipped, hatchery fish, large individuals disappeared from Netarts Bay more rapidly than smaller members of the same release group during the first 9 days of estuarine residence. This selective removal was not evident among creek-reared chum which were smaller than the fin-clipped fish. Size-related emigration explains this variation in residence time. Growth rates were estimated from back-calculations of fork length from otoliths. The growth rates of juvenile chum salmon in Netarts Bay (0.6 mm FL/d; 3.5% bw/d) were lower than growth rates in other estuaries where harpacticoid copepods were a major prey item. Juvenile chum salmon that entered the estuary early in the spring grew faster than those which entered later in the year, possibly because of lower water temperature and lower density of small (50 mm FL) chum during the early period. High water temperature reduced apparent growth rates by slowing growth and causing emigration of larger, faster-growing individuals. / Graduation date: 1990

In response to the increasing need for ecosystem services throughout the Southeast Alaska region, decision makers are tasked with balancing the need for natural resources with salmon conservation. However, accurate historical and current information on salmonid population abundance, freshwater distribution, and habitat quality are sparse with limited resolution for large portions of this remote and rugged landscape. Here, I created Intrinsic Potential (IP) models for chum and pink salmon to predict the potential for portions of coastal rivers to provide high-quality spawning habitat. I developed IP models for both species from field redd surveys and synthetic habitat variables derived from 1-m resolution digital elevation models. The surveys were performed at 49 study reaches in five coastal drainage basins on the north end of Chichagof Island, Southeast Alaska. I used a spatially balanced random sampling design that included field surveys for redds during two field seasons with contrasting precipitation patterns and disparate adult salmon escapements. The IP models predict probable spawning habitat for both species based on persistent landform characteristics and hydrologic processes that control the formation and distribution of spawning habitat across the landscape. Selection of persistent reach variables for both species IP models was informed by principal component analysis (PCA), resource selection ratios, random forest modeling, and regression models of field and synthetic variable comparisons. I observed primarily one spawning strategy by chum salmon associated with mainstem channels, and two distinct spawning strategies for pink salmon related to small moderate-gradient channels and tributaries, and lower drainage basin mainstem channels. The relationships suggest that chum and pink salmon primarily selected for unconstrained channel types in large-and small-size channels, with chum salmon being more selective toward the larger mainstem channels, and pink salmon selecting for smaller channels and tributaries. The prediction of chum salmon redd presence within a specific reach for both high and low streamflow regimes was explained by channel gradient, floodplain width, and mean annual flow in order of importance. In general, chum salmon redds were observed in larger unconstrained low-gradient floodplain reaches where accumulation of deposited gravels and adequate flow produce habitat heterogeneity suitable for spawning. Pink salmon redd presence for both survey years was explained by channel gradient, reach elevation, and mean annual flow, in order of importance. Specifically, when flows allowed upstream access, spawning pink salmon utilized smaller moderate-gradient channels where substrate size and flows were better suited to their smaller body size. Remotely sensed persistent fish habitat data is valuable information for helping understand fish population distributions across the landscape. These synthetic metrics enabled the identification and evaluation of persistent landscape features as probable predictors of IP. Validation of LiDAR-derived channel characteristics indicated channel lengths measured from the DEM were 12% longer than field measured channel length, primarily for channels wider than 10 meters. Thus, understanding the limitations of the data is important so that decision makers do not unintentionally set unrealistic objectives. This research highlights the utility of using IP models with high resolution remote sensing to expand known distributions and quality of spawning habitat for these two species in Southeast Alaska coastal streams.