Spatial autocorrelation is common in data collected for ecological studies, and the use of statistical models for spatial autocorrelation has evolved. Initially, these models were used to improve linear model parameter estimation uncertainty, but more recently ecologists have considered spatial autocorrelation as a valuable tool for describing ecological patterns. The structure and water-driven continuity of stream-networks makes these landscapes unique, and has prompted development of new models for describing spatial autocorrelation within these networks. We evaluate the spatial autocorrelation detection and parameter estimation of four sampling protocols applied to complete censuses of coastal cutthroat trout (Oncorhynchus clarkii clarkii) habitat unit fish counts. We consider two cluster- and two non cluster-based sampling protocols. Spatially distributed clusters were the most apt to contain spatial autocorrelation. Spatial autocorrelation detection was also associated with headwater basin attributes. Differences among sampling protocols in regards to autocorrelation parameter estimation were less distinct.