Habitat and Distribution Modeling Across Terrains and Disciplines: Addressing Common Challenges in Fisheries and Wildlife: Part I

Fisheries and wildlife management and conservation are increasingly interdisciplinary processes, and this is especially true for approaches that use habit and distribution models to identify the ecological niches essential for feeding, rearing, and reproduction. Multiple analytical methods have been developed to understand relationships between environmental factors and species distributions, and how these distributions may change under different climate- and land-use scenarios. In addition, scientists develop habitat suitability models that rely on combinations of physical, biological, and chemical models to determine how conditions differ between geographic locations where species are present and absent. Such approaches have been successfully used to identify areas to prioritize for conservation and protection of threatened species, and to identify species that might be at risk due to loss of habitat associated with changing climate and land-use patterns. In this symposium, we welcome talks that combine species distribution or movement data with habitat models to address questions related to fish and wildlife conservation and management.

8:00AM Relationships Among Side-Scan Sonar Classified Habitat and Fish Densities at Multiple Spatial Scales
  Jerrod Parker, Stephen Pescitelli, John M. Epifanio, Yong Cao
Knowledge of habitat structure is needed to make informed fish management decisions. Underwater habitats in nonwadeable lotic systems have previously been difficult to quantify at large spatial scales. We used consumer grade side-scan sonar to map the substrate of two nonwadable rivers in Illinois, USA, and conducted three years of standardized fish surveys at 40 fixed sites. We characterized substrate composition at four spatial scales (0.5km, 1km, 2km, and 5km) around each of the 40 sites and used linear regressions to explain site variation in the abundance or biomass of four fish species. We hypothesized that larger spatial scales would better explain densities of larger predatory species, and that biomass would be better explained than abundance. We found that substrate explained varying portions of the fish density variation among sites (0.05 ≤ adjR2 ≤ 0.76, mean adjR2 = 0.41), but the relationship between spatial scale and fish species were not as conclusive as hypothesized. However, we did find strong evidence that biomass is more closely linked to variations in substrate than abundance. We conclude that side-scan sonar is a useful management tool, multiple spatial scales should be investigated, and biomass may be more intrinsically linked to ecosystem processes than abundance.
8:20AM Planning for the Future: Utilizing Habitat and Climate Modeling to Prioritize Restoration Actions for Salmonids
  Laura McMullen, Jon Walker, Chip McConnaha
Restoration planning can be enhanced by integrating future predicted conditions, including climate change, into the process of evaluating and ranking opportunities. Portland (Oregon) has sought Salmon Safe Certification of its land management practices, including restoration planning for several salmonids (Chinook and Coho Salmon and steelhead). We supported the certification process as independent evaluators by incorporating predictions of climate change and future restored habitat characteristics into baseline conditions of a spatially-explicit ecosystem model, allowing prioritization of restoration opportunities through the lens of the salmonids’ habitat requirements. We organized a multi-party process to incorporate empirical data, spatial analysis, expert knowledge, and mathematical modeling into a holistic habitat modeling evaluation of city-wide aquatic habitat restoration actions and opportunities. Our results highlight locations in Portland’s urban streams that are most important to protect or restore for these salmonids, as well as indicating where the largest current and future impacts may be. In addition to evaluating future restoration priorities, Portland is incorporating the results into its Stormwater System Plan that evaluates stormwater infrastructure needs to align built and natural green infrastructure investments to also benefit salmonids and watersheds.
8:40AM A New Parametric Mixture-Cure Survival Model to Predict Mortality Risk in the Context of Time-Varying Covariates
  Nicholas A. Som, PhD, Russell Perry, Julie Alexander
Standard parametric survival models (e.g., accelerated failure time) assume that all individuals will eventually experience mortality associated with the risk under study. These models also generally assume that mortality risk is a function of fixed covariate values, where fixed is defined as a single value of each covariate for each individual. These assumptions are problematic for modeling populations that contain a proportion of individuals that survive (often referred to as a cure fraction), and for populations that experience varying mortality risk associated with covariate values that change over time. These traits can be present in survival data collected for fish and wildlife populations that are exposed to disease pathogens. Research developments have improved the flexibility of survival models to account for either cure fractions or time-varying covariates, but a model that addresses both aspects has not been developed. In this talk, we present a new survival model that accounts for a cure fraction, allows for fixed and time-varying covariates, and also allows for censored event times. In addition to model details, we present an application of the model to Klamath River Chinook Salmon, and demonstrate the benefits of our model over fixed-covariate models in both model-fit and simulation scenarios.
9:00AM Fish Predation on a Landscape Scale
  Cyril J. Michel, Christopher Loomis, Mark Henderson, Joseph Smith, Nicholas Demetras, Ilysa Iglesias, Brendan Lehman, David D. Huff
California’s Central Valley salmonid populations are in decline, and it is believed that one of the major contributors to these declines is low survival during residence in the Sacramento-San Joaquin River Delta. The mechanism of their mortality is unclear, but it is believed that a significant contributor is predation by the large populations of predators present there. However, it is currently not clear what proportion of juvenile salmonid mortality can be directly attributed to fish predation, largely because empirical data on predation has only been collected at limited spatial scales. In 2017, we quantified predation mortality rates, predator abundance, and relevant environmental covariates in 21 randomly selected in the Delta, using a GRTS selection protocol. Predation mortality rates were quantified using Predation Event Recorders (standardized predation monitoring devices), and predator densities were quantified using Dual-Identification Sonar cameras. This site selection protocol allowed for the inference of relationships between the environment and predation across a broader spatial scale than previous studies. Using these relationships, we then developed the capability to produce high-resolution spatially and temporally-explicit predation risk estimates. Finally, predation risk estimates were then compared with survival estimates of salmonid populations to discern what proportion of mortality is attributable to predation.
9:20AM Fishes in the San Francisco Estuary: Some like It Hot, Some like It Cold, Some High Salinity, Some Low DO
  Alexander Scott, Levi Lewis, Malte Willmes, Christina Parker, Micah Bisson, Arthur Barros, Christian Denney, James Hobbs
The San Francisco Estuary (SFE) historically provided vast tidal salt marsh habitat, but more than 90% of these wetlands have been developed for agricultural, industrial, and urban use. Remaining marshes face multiple anthropogenic perturbations, including modification of freshwater outflow, invasive species, pollution, and climate change. Understanding relationships between water quality and habitat use by fishes and invertebrates is central to effective management of aquatic ecosystems. In order to explore these relationships, bottom trawls were conducted from 2016-2019 in creeks and sloughs adjacent to restored marshes at the southern and northern ends of the SFE, with water quality measurements taken simultaneously. Generalized additive models (GAMs) were used to analyze the relationships between water quality parameters and catch per unit effort (CPUE) of selected native and invasive fishes and invertebrates. We observed large variation in the responses of different species to water temperature, salinity, and dissolved oxygen. For example, Longfin Smelt catch was highest in colder water with intermediate salinity and higher dissolved oxygen, whereas Threespine Stickleback catch was highest in warmer water with low salinity and dissolved oxygen concentrations. These results suggest that species-specific responses to water quality can and should be incorporated into the management of San Francisco’s tidal marshes.
09:40AM Break
1:10PM Spatial and Temporal Patterns of Land Use, Land Cover, and Land Change in Great Lakes Watersheds
  Mark Nelson, William Severud, Trevor Host, Joseph Knight, Jody Vogeler, Charles Perry
Presence and abundance of cold water fish are directly affected by riverine conditions, which vary with land use, land cover, and land change characteristics. To support a Great Lakes Restoration Initiative project, we quantified the spatial and temporal distribution of landscape characteristics across all United States watersheds within Great Lakes basins. Land cover attributes were obtained from the National Land Cover Database (NLCD), a tree canopy cover dataset, and Landsat time series-based forest canopy disturbance data. Attributes of land use were obtained from permanent remeasured field plots of the Forest Inventory and Analysis (FIA) database. We analyzed landscape attributers at two spatial scales and across two decades to capture spatiotemporal variation. We also assessed the impact of definitional differences between cover and use on area estimates. Excluding Great Lakes water bodies, the basins span 69 million acres, comprised by land uses of agriculture/range/undeveloped (34.5%), developed (14.8%), forest (47.9%), and open water (2.8); forest area is slightly higher than during the previous decade (46.6%). Composition varied among five Great Lakes basins: 8-52% agricultural, 5-26% developed, 20-84% forested, and 1-4% water; and among 97 8-digit watersheds: 0-84% agricultural, 0-92% developed, 2-100% forested, and 0-37% water. We present additional results and maps.
1:30PM Lake Physics and Fish Habitat: Basin Characteristics, Sedimentation Zones and Habitat Partitioning Among Benthic Feeding Fish
  Mark Ridgway, Allan Bell, Trevor Middel
Seiche events in stratified lakes generate two categories of waves that affect lake substrate. One kind is surface waves we all observe. These waves scour the nearshore zone with depth of scouring a function of lake surface area. A second kind is hidden from view and occur along the thermocline density gradient of the metalimnion. Thermocline waves scour the lake bottom defining an area of substrate known as the wash zone. Recent physical limnology and acoustic bottom typing showed the sharpest gradient in the acoustic properties of lake substrate occurred in the wash zone. Wave action of these types aligns with sedimentation zones of lakes (erosive, transport and focusing), affecting substrate characteristics and potentially habitat for benthic feeding fish. We estimated habitat selection for three benthivores, White Sucker (Catostomus commersoni), Longnose Sucker (C. catostomus) and Lake Whitefish (Coregonus clupeaformis), and their use of these areas of lakes. Habitat selection models were based on depth stratified random netting surveys of large lakes to estimate occupancy via multiple survey passes or hurdle models. Among many insights, the catostomids showed the sharpest habitat partitioning with White Sucker selecting the wash zone and Longnose Sucker selecting the focusing zone.
1:50PM Fish Community Response to Rapidly Changing Conditions in the Nearshore Beaufort Sea
  Sarah M. Laske, Vanessa R. von Biela, Randy Brown, Kenneth Dunton
On the Alaskan Beaufort Sea inner shelf, climate change effects are reflected in a lengthening of the ice-free season, increased water temperatures and turbidity, and reduced salinity. These responses are amplified in shallow, nearshore lagoons that also serve as important fish habitat for both marine and freshwater fish species. To determine how fish populations may have changed, we revisited fish collection sites in August 2017 and 2018 that were first sampled annually from 1988–1991 in the eastern Beaufort Sea. Preliminary analyses showed within-season fish catch rates in 2017 and 2018 varied by an order of magnitude per day for some species, likely resulting from wind-driven recruitment. We noted declines in abundance of Arctic-specialists and increased abundance of Arctic-Boreal Pacific species since the 1988-1991 studies, which are in agreement with expectations of a warmer and fresher Beaufort Sea. Our examination of species distribution patterns using an occupancy modeling approach should provide estimates of detection and occurrence probabilities across the range of habitat conditions that fish experience. Repeated sampling will also help clarify the relative importance of habitats to species-specific occurrence, including use and suitability for both resident fishes and new arrivals that have migrated northward into the nearshore Beaufort Sea.
2:10PM Haddock without Cod: A Habitat Model to Help Recreational Anglers Catch One Species and Avoid Another
  Micah Dean, William Hoffman, Gregory DeCelles, Matthew H. Ayer, Douglas Zemeckis, Emily Keiley, John Mandelman
Due to overlapping habitat preferences, cod and haddock are frequently caught together in the Gulf of Maine recreational fishery. While the haddock stock is currently at record abundance, cod remain near an all-time low. Despite a prohibition on keeping cod, concerns over discard mortality have led managers to impose limits on harvesting haddock. To counteract this issue, we sought to offer guidance to the recreational fishery by identifying areas where the catch rate for haddock is high, yet low for cod. We developed a seasonally-resolved regression-kriging model, fit to bottom trawl survey data, that accounts for the non-linear relationship between abundance and habitat (depth, temperature, seafloor rugosity). After translating the predicted density into an expected catch rate for baited hooks, we classified the resulting maps into areas to target (high haddock; low cod) and to avoid (high cod; low haddock). These guidance maps will be validated through standardized charter fishing trips during the summer of 2019. Once validated, guidance maps will be distributed freely as waterproof booklets, and via a location-aware smartphone app. Our goal is reduce cod bycatch, thereby preventing further measures to restrict recreational anglers’ access to the abundant haddock resource.
2:30PM The Challenges of Modeling the Distribution of Reef Fishes: Lessons from a Massive External Validation
  Zach Siders, Nicholas Ducharme-Barth, William F. Patterson, Robert Ahrens
Species distribution models increasingly rely on remote sensing data products to acquire environmental covariates. What happens when these products are unavailable, at coarse resolutions, or are proxy covariates for the true driver of a species’ distribution? In reef fish, all three of these data product problems frustrate the modeling process. However, the worldwide bio-socio-economic importance of such species necessitates generating species distribution models. We recently undertook such an endeavor with Gulf of Mexico Red Snapper (Lutjanus campechanus) under the auspices of an external abundance estimate from the federal stock assessment. This provides the backdrop for perhaps the largest external validation of a species distribution model ever. We review the various modeling approaches used to model Red Snapper before, potential additional methods, and, ultimately, the motivation behind choosing Random Forests, a decision tree based machine-learning algorithm. The utility of Random Forests lie in facilitating data integration of multiple surveys, often a necessity in species distribution models and, especially with wide-ranging aquatic organisms, as well as the ability of the model to incorporate non-linear interacting covariate responses. We discuss the challenges in model inference across the spectrum of species distribution models and offer considerations for others wishing to undertake the task.
2:50PM Refreshment Break
3:20PM Predicting Hydrologic Disturbance of Streams Using Species Occurrence Data
  J. Tyler Fox, Daniel Magoulick
Rapid landscape change coupled with a growing human demand for surface and groundwater have altered natural flow regimes of many rivers and streams on a global scale. Using a machine learning approach and long-term, georeferenced species occurrence data compiled by the USGS Aquatic Gap and state agencies, we modeled and mapped spatial patterns of hydrologic disturbance for streams in Arkansas, Missouri, and eastern Oklahoma. Fish presence/absence data had a similar overall model prediction accuracy of 77% (95% CI: 0.74, 0.80) as flow variables 76% (CI: 0.73, 0.80). Including topographic variables in the fish model increased the RF prediction accuracy to 90% (CI: 0.88, 0.92) compared to 86% (CI: 0.84, 0.89) for flow metrics. Correlation analysis of HDI by flow regime showed groundwater stable streams had the lowest disturbance frequency, with over 50% of stream reaches with low HDI located in forested land cover. HDI was highest for big rivers and intermittent runoff streams and streams associated with agricultural land use. Our results show that long-term georeferenced biological data can provide a valuable resource for predictive modeling of hydrologic disturbance for ungaged rivers and streams.
3:40PM Quantifying Spatio-Temporal Variability in Thermal Patterns Using a Fiber-Optic Distributed Temperature Sensing System
  Shannon Brewer, Skylar Wolf, Evan Tanner, John Polo, Bruce Noden, Sam Fuhlendorf
The thermal environment can be limiting during seasonally-harsh periods. Organism stress due to physicochemical conditions is common among both terrestrial and aquatic environments. Although organisms may perceive temperature differently, terrestrial and aquatic organisms seek thermal refugia when temperature conditions are especially harsh (i.e., hot or cold). A fiber-optic distributed temperature sensing (FO-DTS) system is a useful tool to measure temperature at both fine spatial and temporal scales with high accuracy. FO-DTS has become more prominent in aquatic and terrestrial applications over the last decade. Our objective is to demonstrate the relationships between thermal patchiness across riverine and terrestrial landscapes and temperature use as a habitat element by a variety of organisms. Specifically, we show microhabitat selection patterns by stream fishes and potential for interactions among native and non-native fishes during seasons with more heterogeneous thermal environments when thermal conditions can become limiting for species. Additionally, we demonstrate that encroachment of grasslands by Red Cedar decreases spatial variance of thermal conditions and diurnal temperature ranges, likely resulting in increased tick abundance. Collectively, we illustrate how FO-DTS systems may be implemented in ecological research to better understand spatio-temporal variability of thermal conditions across both aquatic and terrestrial landscapes.
4:00PM Fish Habitat Modeling: A Comparative Study
  Matthew Spetka, Barnali Dixon, Meagan Schrandt
Through this project, comparisons will be made between two methods of habitat modeling in different software: ArcGIS (HSM) and MaxEnt (ENFA). Using data collected by fisheries independent monitoring (FIM), a group out of the Florida Fish and Wildlife Research Institute (FWRI), long-term habitat suitability models will be created, individually validated, and then compared. A habitat suitability model (HSM) will use calculated habitat suitability indices (HSI) of juvenile gag grouper (Mycteroperca microlepis) to map, using ArcGIS, spatial suitability areas in estuaries on the west Florida shelf. Ecological niche factor analysis (ENFA) will then be computed and mapped in the software MaxEnt for the same species and study area. Both models will use water quality, seagrass cover, and depth to calculate and map suitability of gag in the following estuaries in Florida: Tampa Bay, Charlotte Harbor, and Apalachicola Bay. Model maps will be evaluated using the continuous Boyce index in order to determine predictive accuracy since both maps display suitability on a continuous scale. The results from this comparison of two habitat associated modeling techniques will be presented.
4:20PM Quantifying Habitat Suitability for Forage Fishes in Chesapeake Bay: A Coupled Modeling Approach Using Fishery Surveys and a Hydrodynamic Model
  Mary C. Fabrizio, Troy D. Tuckey, Aaron J. Bever, Michael L. MacWilliams
The sustained production of sufficient forage is critical to advancing ecosystem-based management in Chesapeake Bay. Yet, factors that affect local abundances and habitat conditions necessary to support forage production remain largely unexplored. Our study aims to quantify suitable habitat for forage fishes by combining information from 17 years of monthly fisheries surveys with a 3-D hydrodynamic model of the Bay. This coupled modeling approach was used to develop habitat suitability models and quantify suitable habitat for four key forage fishes – bay anchovy, spot, weakfish, and spotted hake – in the Bay and its tributaries. The modeled salinity, temperature, and velocity results were subsampled at the times and locations of the fisheries surveys to provide dynamic habitat metrics that are not generally observed at the time of fish sampling (e.g., velocity, stratification). Sediment composition and dissolved oxygen metrics were also considered for inclusion in fish habitat models. Boosted regression trees were used to identify influential habitat metrics that were then used to construct habitat suitability models for each species. Our approach demonstrates that numerical model hindcasts can be combined with fisheries observations to explore environmental conditions at multiple spatial and temporal scales to improve understanding of observed biological data.

Organizers: Mary C. Fabrizio, Thomas Edwards, Jennifer Wilkening
Supported by: AFS Marine Fisheries Section; AFS Fish Habitat Section

Location: Reno-Sparks CC Date: September 30, 2019 Time: 8:00 am - 5:00 pm