Masters Thesis

Environmental stochasticity and the reliability of redd count data: a simulation study of redd construction, redd loss, and survey frequency in a small coastal stream in northern California

Redd counts – total number of distinct redds observed during a spawning season - are often used as an indicator of the abundance of salmonids spawning in streams and have recently been adopted for statewide assessment of status and trends in abundance of Pacific salmon in California's Coastal Salmonid Monitoring Plan (CMP). Methods currently used to generate an estimate of spawner abundance from redd counts make no attempt to account for redd loss between successive survey occasions. Stochastic periods of high stream discharge between survey occasions may scour or obscure newly constructed redds, so that the total number of distinct observable redds at the time of surveys may be substantially less than the total number of redds actually constructed. Therefore, even if field observers have perfect detection probability for those new redds that are present at the time of surveys, redd counts may have substantial negative bias as an estimator of total redd construction. I developed a Monte Carlo simulation model that allows assessment of the potential influence of stochastic variations in stream discharge on loss of constructed redds over the course of a spawning season. Submodels used in simulations were based on empirical redd survey data collected in Freshwater Creek, a small northern California stream, following CMP survey protocols. A six-year time series of daily discharge gauged in Freshwater Creek, and a longer 55-year surrogate time series of discharge gauged in a nearby stream (Little River), were used to simulate the interannual variation in redd loss that might result from interannual variation in pattern of stream discharge. In each simulated survey year, a realistic temporal pattern of recruitment of constructed redds was superimposed upon a randomly selected hydrograph. Time between successive redd count surveys was set at a minimum of 10 days, but simulated inter-survey interval exceeded 10 days whenever simulated depth of visibility was predicted insufficient to allow successful observation of redds. Loss of newly constructed redds was a function of the temporal discharge regime from date of redd construction to the first survey occasion. Based on the selected hydrograph, a simulated temporal pattern of redd construction, and a simulated survey schedule, the expected number of distinct redds that would be observable at the time of surveys was compared to the actual total number of redds constructed, for low, moderate and high abundance (numbers of redds constructed) scenarios and for dry, moderate and wet water year types. Simulation results suggest that the proportion of observable redds varies by abundance and by water year type. Degree of negative bias and interannual variability in the proportion of observable redds was greatest in wet water years and at low abundance. Simulation results suggest that redd counts unadjusted for redd loss will have substantial negative bias and low reliability, especially in wet water years, even if perfect detection is assumed.

Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.