Double-observer Sightability Model Update for Mount Rainier and Olympic National Parks, 2014

Lubow, Bruce; Jenkins, Kurt; Happe, Patti; Griffin, Paul; Beirne, Katherine

It is well established that not all animals are detected during aerial surveys, and that estimation of detection bias (proportion of animals not seen during a survey) is key to estimating abundance accurately. Previously, we developed and published an analysis methodology for double-observer sightability (DO-S) models used to adjust helicopter-based elk survey results for aerial detection bias in Mount Rainier National Park (MORA). We applied this same methodology to summarize trends in the abundance of elk using high elevation summer ranges within MORA from 2008-2011. The DO-S models were based on detection patterns of two independent observer pairs in a helicopter and on telemetry-based detections of collared elk groups. The models included effects of specified covariates on the unconditional probability that two pairs of observers in a helicopter (front-seat pair or back-seat pair) detected a given elk group. The models were fit to data from two sources: DO-S trials in which two independent observer-pairs in a survey helicopter attempted to detect known groups of elk containing one or more radio-marked animals; and double observer (DO) trials of elk groups in which the same observer pairs attempted to locate groups containing no radio-marked animals. The DO trials provided estimates of conditional detection probabilities because the estimated probabilities were conditional upon one or both observers seeing the elk groups. The DO-S trials provided independent data to estimate the implicit residual heterogeneity bias that would not be quantifiable from DO observations alone, thus producing unconditional estimates of detection probabilities. We had planned to develop DO-S sightability models based on DO-S and DO trials obtained in both MORA and Olympic (OLYM) National Parks so that they could be applied to adjust survey results for detection bias in both parks. Because of large-scale GPS-collar failure (24 of 43) in OLYM, there were not enough DO-S trials completed by 2011 to develop the full model for both parks. Over the last several years, elk were radiocollared in OLYM so that the sightability model could be updated based on a combined set of DO and DO-S trials in MORA and OLYM. In this report, we describe the model development process and the resulting combined models. We apply these models to update abundance estimates of elk within selected trend count areas in both parks. We structured 16 candidate models based on 12 variables and 9 2-way interactions between variables expected to influence detection probabilities. Based on the a priori expectation that sightability differs between parks, we structured all models with a baseline intercept and heterogeneity effect plus additive intercept and heterogeneity parameters for OLYM to distinguish the two parks. We made inferences about population size using weighted-averages across all 16 models in the final set. In OLYM from 2008-2013, percent of elk seen based on model correction averaged 93.9 (range 82.3-99.5%). In MORA from 2008-2013, percent of elk seen based on model correction averaged 82.3 (range 60.9-93.5%). The model will be updated again following the completion of additional surveys and detectability trials in both parks during 2014 and 2015 elk population surveys. The updated models will be used to compute elk abundance values before examining elk population trends in the high elevation summer ranges of both parks.

Double-observer Sightability Model Update for Mount Rainier and Olympic National Parks, 2014

Series

Collections