Estimates of Escapement and an Evaluation of Abundance Methods for North Fork Lewis River Fall-run Chinook Salmon, 2013 – 2017

The Washington Department of Fish and Wildlife (WDFW) has estimated the abundance of fall-run Chinook salmon spawners in the North Fork (NF) Lewis River for more than five decades. Over this time period, the methods used to collect spawner data and generate estimates of abundance have varied. Specifically, estimates from 1964 - 1999 were calculated using a peak count expansion (PCE) factor of 5.27 that was derived using the relationship between the peak count (797) and total abundance (4,199) in a single year (1976). In the early-2000s, WDFW re-evaluated the PCE estimator and from this work developed a new expansion estimator known as the Bright-eye method (BEM). The BEM estimates annual abundance for NF Lewis River Chinook salmon using weekly carcass counts and average age-specific recovery rates observed during two years of data collection (2001 and 2002). Although the BEM was thought to be an improved estimator relative to the historical PCE factor, the main assumption of the BEM (i.e., constant within and among year age-specific carcass recovery rates) has never been evaluated and thus it is unknown if derived BEM estimates are unbiased. Additionally, the BEM does not generate an estimate of uncertainty around the point estimate. Therefore, the current BEM estimator does not meet the monitoring recommendations for ESA-listed salmon and steelhead populations that have been outlined by NOAA Fisheries and local watershed management plans.

In an effort to evaluate the BEM and gather additional years of the PCE ratios, WDFW conducted mark-recapture spawning ground surveys for five years (2013 - 2017). The objectives of the mark-recapture carcass tagging surveys were to (1) generate independent and unbiased estimates of spawner abundance and composition with estimates of uncertainty, (2) evaluate whether or not the BEM and PCE can generate unbiased estimates of abundance, and (3) provide recommendations for future surveys and analyses based on the results. Using a mark-recapture Jolly-Seber (JS) model, we generated estimates of abundance for NF Lewis River Chinook by stock (tule, bright), origin (hatchery, wild), sex (jack, female, male), and total age (2 - 6). Across the five years of surveys, estimates of total fall-run Chinook salmon abundance (i.e., tules and brights combined) generated with the JS estimator ranged from approximately 10,000 to 27,000 spawners per year (CV of 2 - 13%) of which approximately 66 - 85% were late-run ("bright" stock) Chinook salmon.

Using the JS estimates, which were assumed to be unbiased, we evaluated the accuracy and precision of abundance estimates derived with the BEM and three different PCE estimators. Among years, the absolute percent error for estimates of abundance derived with the BEM estimator ranged from 3 - 55% (mean absolute percent error: 13 - 24%) while the absolute percent error for estimates of abundance derived with the PCE estimators ranged from <1 - 65% (mean absolute percent error: 7 - 28%). Therefore, in general, the accuracy of abundance estimates derived with the BEM and PCE estimators were quite similar. However, one advantage of the PCE estimators is that abundance estimates are derived with uncertainty, albeit relatively imprecise (CV 28 - 49%). Overall, the main assumptions of the BEM and PCE estimators were not consistently met, which led to inaccurate estimates of abundance in some years. Based on these results, we recommend the continuation of annual mark-recapture surveys (JS method) to estimate the abundance of NF Lewis Chinook salmon until a more cost-effective, alternative method has been developed that can generate abundance estimates by stock, origin, sex, and age with comparable uncertainty and robustness to model assumptions.