Model-Assisted Estimation of Steelhead Redd Abundance

Steelhead Oncorhynchus mykiss redd abundance is frequently estimated using design-based redd surveys.  Redd abundance estimates may be biased low because redds may not be distinguishable prior to or in between sampling events.  In this study steelhead redd survival time was evaluated in Northeast Oregon and a model-assisted estimation technique to correct for removal bias was described.  Accelerated failure time (AFT) models were fit to redd data to estimate redd survival probabilities and the model-assisted estimation approach was evaluated via simulation.  Model selection suggested that spatial and temporal heterogeneity existed in redd survival times.  The simulation showed negative bias in raw redd counts depending on the length of the sampling interval and the longevity of redds.  Bias was as high as 32% when redd longevity was relatively short at a 25 d sampling interval.  Bias was reduced to 2% using the model-assisted approach.  Benefits of the model-assisted approach decreased as redd longevity increased and there was a slight tendency for the model-assisted approach to overestimate redd abundance at longer sampling intervals.  Nonetheless, the model-assisted estimation methods evaluated in this study improved the accuracy of redd-abundance estimates with little to no additional effort beyond what is currently expended in most redd surveys.