Mount St. Helens Elk Population Assessment: 2009-2013

In 2009, we initiated a study of the Mount St. Helens elk population to better quantify elk abundance, develop a practical and defensible population monitoring approach, and document recent trends in elk condition, productivity, and survival. During 2009-2012, we captured and radiomarked 150 unique elk aged . 1-yr-old (110 F: 40 M) by helicopter darting in a 5-Game Management Unit (GMU) study area (GMUs 520, 522, 524, 550, and 556) in the core of the Mount St. Helens elk herd area. Among the issues motivating our work were episodic high overwinter elk mortality, recent evidence of sub-par condition among elk translocated to the North Cascades in 2003 and 2005, and apparent elk herbivory impacts on plant communities in the vicinity of Mount St. Helens. In response to these issues and concurrent with the initiation of our work, antlerless elk harvesting was liberalized across several GMUs to reduce local elk densities.

Using ultrasound examination and body condition scoring we estimated mean ingesta free body fat (IFBF) for elk we live captured in February, 2009-2012, was 5.64% (95% CI = 5.08-6.21) for non-lactaters and 3.26% (95% CI = 2.34-4.18) for lactaters. These levels suggest food limitation. We found that GMU, lactation status, and pregnancy status affected IFBF, but year did not. Overall, 73 of 109 cow elk (67%) we examined for pregnancy via ultrasound were pregnant. Pregnant elk had higher IFBF than did non-pregnant elk. We also used organ samples from 364 hunter-harvested cow elk to estimate fall (Nov) IFBF for elk in the Mount St. Helens herd, 2009-2011. We detected effects of geographic subarea and lactation status on IFBF, but not effects attributable to year or cow age. IFBF was higher for cow elk harvested in GMU 560 and Columbia Gorge GMUs than from the managed forest portion of our 5-GMU study area. We estimated mean IFBF during the fall at 12.51% for non-lactaters and 10.84% for lactaters, controlling for other factors.

We collected data during intensive late winter helicopter surveys (2 complete survey replicates yearly 2009-2012, 1 survey in 2013) over the 5-GMU study area. We used data from Mar-Apr flights, 2006-2007 to fit logistic regression models to predict the sightability of elk groups based on group and environmental covariates. Several covariates influenced sightability in univariate logistic regression models. We then used multi-model inference and an information-theoretic criterion (AIC_c) to compare several alternative multivariate models of varying complexity; our results indicated the best multivariate model predicted sightability of elk groups based on: 1) transformed (log₂) group size, and 2) forest canopy cover (%). Predicted sightability increased with increasing group size and with decreasing cover.

We also used the logit-normal mixed effects (LNME) mark-resight model to generate estimates (2009-2012) of total elk population size and the sizes of the cow and branch-antlered bull subpopulations at a variety of spatial scales. We explored 11 LNME models to estimate total population size, 10 models to estimate total subpopulation sizes for cow elk and branch-antlered bulls, and 15 models to estimate GMU-specific estimates of cow elk abundance. We also used the Lincoln-Petersen model to generate mark-resight estimates for total population size and total cow elk subpopulation size for 2013 using data from the single survey conducted that year. We again used multi-model inference and AIC_c to evaluate the evidence in our data for the various models in our LNME model sets.

Sightability model estimates appeared to underestimate true abundance, relative to LNME estimates. This result is common and relates to how the 2 types of models account for undetected elk. Mark-resight models are virtually always more effective at accounting for such animals. However, trend estimates from the 2 modeling approaches were relatively congruent and time-specific estimates from both approaches were highly correlated, suggesting that sightability model estimates, although biased low, provided a useful and consistent abundance index. The application of a sightability modeling approach is a much more practical strategy, relative to mark-resight, for large-geographic-scale monitoring such as is needed for elk at Mount St. Helens.

Sightability model and LNME mark-resight estimates, 2009-2013, suggested a decline in overall elk abundance and cow elk abundance; bull abundance estimates indicated a relatively stable bull population. We found evidence of strong spatial variation in the decline in overall elk abundance and cow elk abundance. Estimates indicated substantial a reduction in elk abundance in GMUs 520, 524, and 550. We did not detect any decline in GMU 522 elk abundance, nor in GMU 556 abundance; however, estimated elk abundance in GMU 556 during the last survey year that we report on, spring 2013, was the lowest we recorded across the 5 years of data from GMU 556. Across our individual counting units, the units the furthest west showed the most consistent and dramatic declines in raw elk counts; units further east in the same GMUs produced more stable counts.

For virtually every geographic scale of abundance estimates for total elk and total cow elk, the 2013 point estimate was the lowest estimate obtained 2009-2013, except for GMU 522 estimates. For total elk and total cow elk across the 4-GMU landscape (excluding GMU 522), 2013 estimated abundance was on the order of 30-35% lower than the 2009 estimates. GMU-specific sightability model estimates of total elk and total cow elk abundance were on the order of 60-70% lower in 2013 than in 2009 for GMUs 520 and 550, were ~40-60% lower for GMU 524, and were ~20-25% lower for GMU 556.

We also used radiomarked elk to estimate survival rates and explore possible sources of variation in survival. We explored 15 survival models with known-fate modeling using AICc and model weights to draw conclusions about Mount St. Helens elk survival during 2009-2013 (4 survival years). The best model had a common cow survival parameter for GMUs 520, 522, 524, and 556 that was constant during 2009-2011, a common cow survival parameter for all GMUs during the last survival year (2012-2013), a unique survival parameter for GMU 550 cows during 2009-2011, and constant bull survival across years. Bull elk survival was estimated to be 0.56 (95% CI = 0.43-0.68). Annual cow survival was estimated to be 0.85 (95% CI = 0.78-0.91) during 2009-2011 in GMUs 520, 522, 524, and 556. During the same years, cow survival was estimated at 0.64 (95% CI = 0.48-0.78) in GMU 550. Cow survival in the final survival year (2012-2013) was estimated to be 0.52 (95% CI = 0.38-0.65) across all 5 GMUs. Low survival of radiomarked elk, 2012-2013, corresponded to a fairly high number of unmarked, winter-killed elk (n= 71) tallied during the annual mortality survey on the mudflow. During the previous 3 years, the annual winter mortality survey yielded tallies ranging 2-46 elk.

Spring calf recruitment varied considerably during 2009-2013. Calf:cow ratios exceeded 35:100 during 2010 and 2011. Calf recruitment was lower in the spring of 2009 and much lower in 2012, 2013. Overall, observed estimates were in the 25-30:100 range for the study area and in the 25-35:100 range for most GMU-specific estimates. After attempting to correct the observed ratios for fall removals of antlerless elk via hunter harvest, calf recruitment was indexed mostly in the high teens to 100 cows range for 2012, 2013 and in the 20-30-ish calves per 100 cows in 2009. Indexed recruitment in spring 2013 was the lowest.compared to other study years.for almost all GMUs. Depressed calf recruitment in the spring of 2013 corresponded to high mortality among radiomarked elk that same year, high observed overwinter mortality of unmarked elk, and elk abundance estimates that were also low.

Spring calf recruitment, 2009-2013, was strongly related to late summer-fall precipitation metrics (r² = 0.91-0.96); calf recruitment was higher in years with significant late summer-fall moisture, presumably because of enhanced forage production/quality during the time when calf elk are becoming increasingly dependent on foraging. Overwinter elk mortality, as indexed by the annual mortality survey on the mudflow, was strongly related (r² = 0.90) to a metric reflecting daily snowpack during mid-to-late winter; in years with substantial late winter snowpack, overwinter mortality was higher than in years with milder winter conditions.

Collectively, our estimates of elk condition, productivity, and survival indicated fairly strong food limitation in this population that may have been a function of elk density. Attempts to reduce the elk population via liberalized hunter harvest beginning in 2007 were apparently successful, based on our estimates of elk abundance. However, links between weather covariates and recruitment and survival, coupled with a substantive overwinter mortality event, 2012-2013, suggest that reducing the elk density has not eliminated the risks of overwinter mortality, at least in the short-term. It is likely that plant community responses to lower elk herbivory are still evolving and benefits likely will take some time to be fully realized. We discuss the implications of both density-dependent and density-independent influences on elk demography and management in the Mount St. Helens elk herd. Our work did not address issues surrounding elk hoof disease, as these issues were beyond our research scope. The role of hoof disease in elk population processes at Mount St. Helens remains unclear, as does the degree that the conditionfs presence will complicate meeting management objectives.