SAFS hosts weekly lunch-time seminars where students and faculty share findings from their current research. Read through our past seminars to get an idea of topics covered and be sure to check out our events calendar to download upcoming seminars on your calendar.

Winter 2019  seminar

Fridays at 12:30-1:30pm in FSH 203 

Join the seminar remotely.


Arnaud Gruss (Acting Instructor, UW SAFS)

Title: Developing spatio-temporal models using multiple data types for supporting stock and habitat assessments  


Spatio-temporal models, which account for spatial and temporal autocorrelation, have become a key tool for evaluating population trends and habitat usage of marine populations. Fitting spatio-temporal models to a combination of encounter/non-encounter, count and biomass data collected by different monitoring programs (“combined data”) would allow for the generation of a diversity of products for assisting stock and habitat assessments. Previous research has combined count and encounter/non-encounter data, but our study is the first to combine these with biomass-sampling data. We predict the three data types using a computationally efficient approximation to a compound Poisson-gamma process. We analyze the predictions, precision, accuracy, error and confidence interval coverage of spatio-temporal models fitted to combined data for Gulf of Mexico (GOM) red snapper (Lutjanus campechanus) for the period 2006-2014. The spatio-temporal models fitted to combined data provided insights into the spatial distribution patterns of GOM red snapper, which we corroborated by comparing with past predictions generated using only encounter/non-encounter data. However, relying on biomass and count data in addition to encounter/non-encounter data also allowed us to reconstruct trends for GOM red snapper and to examine patterns of distribution shifts and range expansion/contraction for this fish population for the first time. We also found that combining multiple data types improved the precision of reconstructed population trends and some variables quantifying habitat usage. Moreover, scenarios and simulation experiments conditioned upon GOM red snapper data showed that the improvement in fitting to combined data (compared with just fitting to biomass data as conventionally done) is greater when biomass data for the study population are lacking for an entire subregion and, to a lesser extent, for an entire time period (e.g., in early years). Our framework will be useful to identify population trends and habitat usage for many populations and life stages, particularly those lacking biomass data for entire regions or times. Future studies should also consider additional data types, such as binned counts from citizen-science programs, and could use our modeling framework to assist climate-vulnerability and ecosystem assessments and monitoring program optimization.


Ingrid Spies (Research Fishery Biologist, Alaska Fisheries Science Center)




Sam Johnson (Graduate Student, Simon Fraser University REM)

Title: The effects of posterior sampling design on management procedure performance in management strategy evaluations, with examples from British Columbia pacific herring (Clupea pallasi) and sablefish (Anoplopoma fibria)


Fisheries management strategy evaluation ranks candidate management procedures according to their relative risk of failing to meet management outcomes, examining distributions of possible futures generated by monte-carlo simulations. The historical period, which sets up the jumping off point of these simulated futures, is often conditioned on the results of an assessment, usually by sampling the posterior distribution of the assessment in some way to integrate over uncertainty. Previous research on best practices suggests that sampling a Bayesian posterior is the best way, but makes no mention of how exactly the sampling should be designed. Furthermore, while a small number of posterior samples may be adequate to rank management procedures, objective risk measures may be biased under small sample sizes. We show two examples of sampling procedures that we used to improve representation of risk in previous management strategy evaluations for BC herring and sablefish, and discuss preliminary results from rigorous tests of alternative sampling designs.



Mira Sytsma (Graduate Student, UW SEFS)

Title: Responses of Wildlife to Tourism in Glacier Bay National Park, AK


Visitation to Glacier Bay National Park (GLBA) is very low compared to most national parks, but has nearly doubled in the past 20 years, leading to undocumented impacts of tourism on wildlife activity patterns and space use. We studied wildlife responses to tourism using 40 remote cameras installed at 10 study sites that were categorized as either areas where tour vessels drop off tourists on the shore (treatment), or areas where the vessels do not drop off tourists (control). The four mammals studied were Brown bear (Ursus arctos), Black bear (Ursus americanus), Moose (Alces alces), and Gray wolf (Canis lupus). Black bear and moose demonstrated behavior consistent with the human-shield effect, while wolves tended to avoid humans temporally. Brown bear were not impacted spatially or temporally by tourism. Detection for all species dropped to less than one photo/week when the number of minutes of human activity at a site/week reached 40, indicating that there is a threshold value for disturbance to wildlife in GLBA. This study demonstrates that wildlife responses to humans were detectable in a system with very low use, and that true baselines for estimating anthropogenic impacts may be difficult to obtain when using high-visitation national parks as study systems. The findings from this study will be used to assist park management in making yearly decisions regarding tour vessels and visitor access in GLBA to ensure significant resource degradation does not occur in popular tourist destinations.


Chelsea Wood (Assistant Professor, UW SAFS) & Grant Adams (Graduate Student, UW SAFS)

Title: Habitat area integrates over spatial and temporal variability in snail abundance to predict human urinary schistosomiasis burden


Schistosomiasis is the world’s second-most common parasitic disease of humans, affecting over 250 million people globally and causing the loss of over 3 million years of healthy life annually. Recently, the World Health Organization recognized that efforts to control schistosomiasis transmission exclusively through mass drug administration have been ineffective; their new recommended strategy for global schistosomiasis control emphasizes destroying the aquatic snails that host larval stages of the parasite. But it has been surprisingly difficult to predict the spatial location of human infections from information on snail distributions, frustrating efforts to leverage snail control for durable schistosomiasis transmission reduction. We thought that previous studies might have failed to link snails to human infections because snails occur in ephemeral and aggregated patches, which means that intensive spatial and temporal sampling are required to obtain stable estimates of their abundance. We sought to identify the optimal sampling design for predicting infection risk to humans by sampling snails at finer spatial and temporal resolutions than in previous studies. We worked at the site of the world’s largest recorded schistosomiasis epidemic: the Lower Senegal River Basin in Senegal, above the Diama Dam. Our intensive sampling strategy did reveal positive relationships between the presence of snails (in our case, total snail abundance, infected snail abundance, total snail density, and snail prevalence) and human infection burdens (in our case, village-level prevalence and intensity of infection) across sites. However, it also showed that snail abundance was extremely variable in space and time – a rapidly moving target for snail control efforts – and obtaining data on snail distributions required a massive input of labor and resources. For this reason, we asked whether there might exist better alternatives to counting snails: stable environmental proxies that are more predictive and less difficult to estimate than snail counts. Indeed, we found that site area was a strong predictor of human urinary schistosomiasis re-infection (after praziquantel treatment), probably because increasing area increases the available space for non-emergent vegetation, which serves as a substrate and resource for schistsome-competent snails. Indeed, the abundance of suitable snail habitat (i.e., floating, non-emergent vegetation) also predicted human urinary schistosomiasis re-infection. Our results suggest that the greater the available habitat for snails near human water-access sites, the greater the schistosomiasis transmission risk. Unlike snail counts, which require hundreds of person-hours to accurately estimate, site area and habitat type can be observed by drone or satellite, making possible large-scale, fine-grained estimation of human urinary schistosomiasis risk. Once identified, simply removing plant species that support snails could reduce rates of urinary schistosomiasis to people, complementing mass drug administration in campaigns to eliminate schistosomiasis.


Lukas DeFillipo (Graduate Student, UW SAFS)

Title: Recruitment variation disrupts the stability of alternative life histories in an exploited salmon population


Males of many fish species exhibit alternative reproductive tactics, which can influence the maturation schedules, fishery productivity, and resilience to harvest of exploited populations. While alternative mating phenotypes can persist in stable equilibria through frequency‐dependent selection, shifts in tactic frequencies have been observed and can have substantial consequences for fisheries. Here, we examine the dynamics of precocious sneaker males called “jacks” in a population of sockeye salmon (Oncorhynchus nerka) from Frazer Lake, Alaska. Jacks, which are of little commercial value due to their small body sizes, have recently been observed at unusually high levels in this stock, degrading the value of regional fisheries. To inform future strategies for managing the prevalence of jacks, we used long‐term monitoring data and a Bayesian state-space approach to identify what regulates the frequencies of alternative male phenotypes in the population over time. Expression of the jack life history could not be explained by environmental factors expected to influence juvenile body condition and maturation probability. Instead, we found a strong positive association between the proportion of individuals maturing as jacks within a cohort and the prevalence of jacks among the males that sired that cohort. Moreover, due to differences in age‐at‐maturity between male phenotypes, and large interannual variability in recruitment strength, jacks from strong year‐classes often spawn among older males from the weaker recruitments of earlier cohorts. Through such “cohort mismatches,” which are amplified by size‐selective harvest on older males, jacks frequently achieve substantial representation in the breeding population, and likely high total fertilizations. The repeated occurrence of these cohort mismatches appears to disrupt the stabilizing influence of frequency‐dependent selection, allowing the prevalence of jacks to exceed what might be expected under equilibrium conditions. These results emphasize that the dynamics of alternative life histories can profoundly influence fishery performance and should be explicitly considered in the management of exploited populations.


Jim Thorson (Habitat and Ecosystem Process Research Program Leader, Alaska Fisheries Science Center)

Title: New uses for vector autoregressive spatio-temporal (VAST) models:  MICE-in-space, oceanographic indices, and expanding composition data from multiple surveys

Authors: James Thorson, Grant Adams, Lorenzo Cianelli, Kirstin Holsman, Jim Ianelli, Stan Kotwicki, and Mike Litzow

Spatio-temporal models are undergoing rapid development and are being used in stock assessments and ecosystem-status reports worldwide.  However, the many studies regarding development and testing make it hard to keep up with how these new models can answer old questions.  I therefore highlight three new uses for spatio-temporal models in addressing long-running fisheries questions, while focusing on features in the R package VAST.  First, I discuss efforts to estimate a matrix of species interactions within multivariate models.  New research shows that these spatial models of intermediate complexity for ecosystems (“MICE in space”) can estimate fishing and biomass-based biological reference points given external estimates of fishing mortality rates, as demonstrated using data for four species in the Gulf of Alaska.  I next demonstrate that spatial models can generalize the “empirical orthogonal functions” (EOF) that are widely used by oceanographers to define ocean indices.  To do so, I replicate the EOF calculation of the Pacific Decadal Oscillation, and then use a multivariate extension to show that sea-ice extent arises as an EOF based only on biological sampling in the Eastern Bering Sea.  Finally, I show that VAST can assimilate data from multiple surveys while expanding compositional data for use in stock assessment, and demonstrate this using length-composition data from the Eastern and Northern Bering Sea for Alaska pollock.  Overall, I highlight that multivariate spatio-temporal models remain a fertile ground for research and development, with potential to address age-old questions in stock assessment and fisheries oceanography.


Kiva Oken (Post Doc, UW SAFS)

Title: Quantifying and explaining nearshore species assemblages’ response the Deepwater Horizon oil spill


The Deepwater Horizon oil spill released approximately 210 million gallons of oil into the Gulf of Mexico. There was a clear signal of oil exposure in individuals across taxa, experiments have shown oil to be a stressor that leads to physiological responses, and there were changes observed in lower trophic level communities. Combined, this indicates a strong potential for population declines of commercially and recreationally valuable fishes and aquatic invertebrates. First I will briefly present work using spatiotemporal models to analyze years of monthly monitoring data around the Mississippi Delta to quantify potential impacts of the oil spill at the population scale. Both our study and others have shown little evidence of population declines. Several hypotheses have been proposed to explain this apparent paradox. Two possibilities include a fishing moratorium following the spill and changes in predation pressure following predator die offs. Using food web models, we quantified how much species would be expected to increase given only a fishing moratorium or predator die off without any oil mortality. Predicted increases of a magnitude much greater than what was observed indicate evidence for some other source of population mortality; we hypothesize this source is likely oil exposure. We emphasize that oil spills are one influence within a large socio-ecological system, and understanding oil spill impacts requires consideration of this complexity.


Jessica Badgeley (Graduate Student, UW ESS):

Title: Arctic climate through the Holocene: A data assimilation approach


Investigations of Holocene climate commonly rely on paleoclimate proxy compilations, which are spatially incomplete, or climate model simulations, which are quite sensitive to initial conditions. We show that a data assimilation approach can be used to optimally and objectively combine proxy records and climate models by taking into account the uncertainty in each. Using the ensemble Kalman filter we produce skillful reconstructions of Holocene climate fields. Advantages of this method include, 1) spatially- and temporally-complete reconstructions, 2) allowance for non-stationary teleconnections through time, 3) unlimited incorporation of proxy records, 4) easily-calculated uncertainty estimates, 5) physically-consistent reconstruction of multiple climate variables, and 6) minimal computational cost.


Jameal Samhouri (Ecosystem Science Program Manager, Northwest Fisheries Science Center)





Archived seminar schedules

Year Coordinator links to archives
2014–2015 Hilborn Lab fall/winter/spring
2013–2014 Branch Lab fall/winter/spring
2012–2013 Punt Lab fall/winter/spring
2011–2012 Anderson Lab fall/winter/spring
2010–2011 Kotaro Ono fall/winter/spring
2009–2010 Chantel Wetzel fall/winter/spring
2008–2009 Dawn Dougherty fall/winter/spring
2007–2008 Essington Lab fall/winter/spring<
2006–2007 Ian Taylor fall/winter/spring
2005–2006 Eric Ward all quarters
2004–2005 Jason Cope fall/winter/spring
2003–2004 Lucy Flynn fall/winter/spring
2002–2003 Gavin Fay fall/winter/spring
2001–2002 Carolina Minte-Vera fall/winter/spring
2000–2001 Juan Valero fall/winter/spring
1999–2000 Arni Magnusson fall/winter/spring
1998–1999 Ivonne Ortiz fall/winter /spring
1997–1998 Carlos Alvarez-Flores fall/winter/spring
1996–1997 Billy Ernst fall/winter/spring

Spring 2015 Seminars

Back to Top