Boracéia Biological Station

Understanding patterns of habitat use across trophic levels and the physical drivers of multispecies aggregations is essential to inform ecosystem-based management. To achieve this, we quantified the spatial distribution and co-occurrence of hotspots (defined using the Getis-Ord statistic) for euphausiids and nine of their commercially important fish and whale predators on the west coast of Canada during summer. We first developed fine-scale spatiotemporal distribution models of euphausiids and Pacific hake using high-resolution acoustic data from coast-wide surveys conducted between 2007 and 2018. We found that the spatiotemporal distribution of hotspots of euphausiids and hake was variable between years with low direct overlap (apart from 2017). The summer of 2015, during the 2014-2016 marine heatwave event, was a particularly anomalous year, as euphausiids and hake showed spatial mismatch in their biomass hotspot distributions. For the other eight predator species, predictions from published species distribution models were used to identify spatial hotspots as an average across years. Co-occurrence patterns were associated with the depth gradient across the shelf and slope and along the canyon and sea valley systems that characterize the Pacific coast of Canada. One assemblage was associated with the deeper parts (200-1000 m+) of the continental slope (euphausiids, hake, redbanded rockfish, sablefish, Pacific ocean perch, and humpback and fin whales) and a different assemblage (redstripe and yellowtail rockfish, and dogfish) was associated with the shallower shelf regions. Important ecological areas with co-occurring multispecies hotspots occurred along the west coast of Vancouver Island, the sea valleys of Queen Charlotte Sound, and the northwest coast of Haida Gwaii. Our results identify areas where multiple species aggregate, which can inform better management and hopefully protection of these regions that support complex food webs, commercial species, and large predators, and are therefore essential for overall ecosystem health.

Geostatistical spatial or spatiotemp oral data are common across scientific fields. However, appropriate models to analyze these data, such as generalized linear mixed effects models (GLMMs) with Gaussian Markov random fields (GMRFs), are computationally intensive and challenging for many users to implement. Here, we introduce the R package sdmTMB, which extends the flexible interface familiar to users of lme4, glmmTMB, and mgcv to include spatial and spatiotemp oral latent GMRFs using the stochastic partial differential equation (SPDE) approach. SPDE matrices are constructed with fmesher, and estimation is conducted via maximum marginal likelihood with TMB or via Bayesian inference with tmbstan and rstan. We describe the model and explore case studies that illustrate sdmTMB's flexibility in implementing penalized smoothers, non-stationary processes (time-varying and spatially varying coefficients), hurdle models, cross-validation, and anisotropy (directionally dependent spatial correlation). Finally, we compare the functionality, speed, and interfaces of related software, demonstrating that sdmTMB can be an order of magnitude faster than R-INLA. We hope sdmTMB will help open this useful class of models to more geostatistical analysts.

Kelp forests offer substantial carbon fixation, with the potential to contribute to natural climate solutions (NCS). However, to be included in national NCS inventories, governments must first quantify the kelp-derived carbon stocks and fluxes leading to carbon sequestration. Here, we present a blueprint for assessing the national blue carbon capacity of kelp forests in which data synthesis and Bayesian hierarchical modeling enable estimates of kelp carbon production, storage, and export capacity from limited data. Applying this blueprint to Canada's extensive coastline, we estimate kelps hold 0.6 to 2.8 Tg C in short-term biomass, producing 1.1 to 6.2 Tg C yr-1, of which 0.04 to 0.4 Tg C yr-1 could be exported to the deep ocean. While modest compared to terrestrial sinks, our findings suggest kelps have comparable carbon sequestration to marine and freshwater wetlands, warranting further consideration in Canada's NCS inventories. Our transparent, reproducible blueprint represents an important step towards accurate carbon accounting for kelp forests.

Multivariate spatio-temporal models are widely applicable, but specifying their structure is complicated and may inhibit wider use. We introduce the R package tinyVAST from two viewpoints: the software user and the statistician. From the user viewpoint, tinyVAST adapts a widely used formula interface to specify generalized additive models, and combines this with arguments to specify spatial and spatio-temporal interactions among variables. These interactions are specified using arrow notation (from structural equation models), or an extended arrow-and-lag notation that allows simultaneous, lagged, and recursive dependencies among variables over time. The user also specifies a spatial domain for areal (gridded), continuous (point-count), or stream-network data. From the statistician viewpoint, tinyVAST constructs sparse precision matrices representing multivariate spatio-temporal variation, and parameters are estimated by specifying a generalized linear mixed model (GLMM). This expressive interface encompasses vector autoregressive, empirical orthogonal functions, spatial factor analysis, and ARIMA models. To demonstrate, we fit to data from two survey platforms sampling corals, sponges, rockfishes, and flatfishes in the Gulf of Alaska and Aleutian Islands. We then compare eight alternative model structures using different assumptions about habitat drivers and survey detectability. Model selection suggests that towed-camera and bottom trawl gears have spatial variation in detectability but sample the same underlying density of flatfishes and rockfishes, and that rockfishes are positively associated with sponges while flatfishes are negatively associated with corals. We conclude that tinyVAST can be used to test complicated dependencies representing alternative structural assumptions for research and real-world policy evaluation.

In 2017, a new regulatory management system, the traffic light system (TLS), was implemented to estimate the effects of salmon lice from Norwegian salmon aquaculture on marine survival of wild Atlantic salmon and forms the basis of aquaculture capacity regulation. The TLS relies on observational data and a set of models to estimate the risk for negative impact on wild salmon populations. This review of the literature that forms the basis for the TLS as well as other relevant studies is presented in the context of the currently practiced TLS and suggestions are made for immediate and long-term improvements. The main findings of this review are that: (1) assumed timing and duration of smolt migration contribute to unreliable observational and modelled data and overestimates of infection pressure; (2) production of lice larvae from farmed salmon is overestimated; (3) TLS model systems rely on or are calibrated by the same potentially flawed data; (4) lice-associated mortality in wild salmon smolts may be overestimated; and (5) lice infection levels on farms are not associated with measurable effects on wild salmon. Recommendations to improve the accuracy and reliability of the TLS, and hence its environmental efficiency include the more complete use of available biological and physical environmental variables, adjusting the time period that observational data are registered and modelled data are integrated, adjusting the interpretation of data including recognition of uncertainty in model outcomes, and use of more realistic assumptions concerning lice-induced mortality thresholds.