Description: The Regional Freshwater Index Layers dataset is composed of five single-band raster layers in GeoTIFF format. Each layer corresponds to a marine region, which generally coincide with the following layers from the Species Distribution Modelling Boundaries dataset: Nearshore_HG, Nearshore_NCC, Nearshore_QCS, Nearshore_WCVI, and Shelf_SalishSea. The main purpose of the dataset is to supplement existing layers that are used for species distribution modelling in the Pacific nearshore marine environment. Each regional freshwater index layer has the same spatial resolution and extent as other predictor layers for the corresponding region. While salinity layers exist from oceanographic models, they may not capture local difference from smaller scale rivers and streams entering the marine environment. Therefore, these layers are meant to complement salinity layers and are not suitable as a replacement for salinity data in species modelling. Methods: The cell values represent an estimate of freshwater influence on a 0-1 scale, where a higher value represents a greater level of freshwater influence. Details on how these values are determined is described in the supplemental information section of the metadata. The main data source for these derived products is the B.C. Freshwater Atlas, including the stream network and river polygons layers. Uncertainties: The values in the rasters are not a measure of salinity. The units are an index representing the level of freshwater influence weighted by the stream order and rescaled across regions on a 0-1 scale where only the region with the greatest value has a range of values 0-1 and the other regions are scaled relatively. This is done to ensure that values in one region can be compared to values in another region. As a result, some regions have very small values because the Salish Sea with the Fraser River is dominant, even after applying a rescale factor to the data.

Description: In the coming decades, warming and deoxygenation of marine waters are anticipated to result in shifts in the distribution and abundance of fishes, with consequences for the diversity and composition of fish communities. Here, we combine fisheries-independent trawl survey data spanning the west coast of the USA and Canada with high-resolution regional ocean models to make projections of how 34 groundfish species will be impacted by changes in temperature and oxygen in British Columbia (BC) and Washington. In this region, species that are projected to decrease in occurrence are roughly balanced by those that are projected to increase, resulting in considerable compositional turnover. Many, but not all, species are projected to shift to deeper depths as conditions warm, but low oxygen will limit how deep they can go. Thus, biodiversity will likely decrease in the shallowest waters (less than 100 m), where warming will be greatest, increase at mid-depths (100–600 m) as shallow species shift deeper, and decrease at depths where oxygen is limited (greater than 600 m). These results highlight the critical importance of accounting for the joint role of temperature, oxygen and depth when projecting the impacts of climate change on marine biodiversity. The rasters available in this dataset project the occurrence of each of the 34 groundfish species in a 3 km^2 grid cell for the historical baseline, as well as for two emissions scenarios, from each of the two regional ocean models (BCCM and NEP36). Each projection layer is provided as the mean projected occurrence as well as the lower and upper 95% confidence interval of projected occurrence. Methods: Estimated species response curves: We estimated how the observed distribution of groundfish species is determined by temperature, dissolved oxygen and seafloor depth using data from fisheries-independent scientific research trawls spanning the entire American and Canadian west coast. We included data from 4 surveys (NOAA West Coast, NOAA Alaska, NOAA Bering or DFO Pacific) from 2000 to 2019. For each species, we modelled occurrences in the coastwide trawl dataset using a generalized linear model (GLM) using the sdmTMB package in R v. 4.0.2. The predictors were temperature, log dissolved oxygen, log depth and survey. We included quadratic terms for temperature and log depth to allow species occurrences to peak at intermediate values. We fitted a breakpoint function for log dissolved oxygen to reflect the fact oxygen is a limiting factor. We assessed the forecasting accuracy of the SDM by comparing how well a model fitted to only data from 2000 to 2010 could forecast species’ occurrences in trawls within our focal region for the period of 2011–2019. We assessed all 77 groundfish species that were present in the overall trawl dataset, however the final analysis included only the 34 species for which the models had adequate forecasting ability. Projecting groundfish biodiversity changes: We based our groundfish biodiversity change projections on two regional models that downscale climate projections: the British Columbia Continental Margin model (BCCM) and the North-Eastern Pacific Canadian Ocean Ecosystem model (NEP36-CanOE). We used a historical baseline of 1986–2005 and future projected values for 2046–2065 based on RCP 4.5 and 8.5 emissions scenarios. Using the models that we validated in our forecasting accuracy assessment, we projected the occurrence of each species in each 3 km^2 grid cell for the historical baseline, as well as for two emissions scenarios, from each of the two regional ocean models. Uncertainties: Source survey data was collected by consistent methods with survey-grade GPS for all years included. Data quality is expected to be high. Modeled data are at 3 km resolution. Outputs are as accurate as source input models and are deemed to be of high quality and accurate based upon the precision of model inputs. Projecting biodiversity responses to climate change involves considerable uncertainty and our approach allows us to quantify some aspects of this. Of the uncertainty that we could quantify, roughly half was due to uncertainty in our SDMs and the remainder was due to regional ocean model uncertainty or scenario uncertainty. This amount of uncertainty in the SDMs is typical, stemming from the fact that contemporary species distributions are also influenced by other factors that we have not included in our model. In addition, although oxygen demand is understood to vary with temperature, limitations in the implementation of breakpoint models prevented us from estimating a temperature-dependent oxygen breakpoint. However, although somewhat unrealistic, this limitation is unlikely to have greatly increased the uncertainty in our SDMs because low oxygen concentrations occurred almost exclusively at depths where temperature variation and projected change was small. To reduce uncertainty due to year-to-year variation in climate, our model projections are based on 20-year climatologies with a future period that is far enough ahead to ensure that changes are unambiguously due to greenhouse gases. We have made projections based on two different emissions scenarios, and two different regional ocean models that are both downscaled from the same global model, the second generation Canadian Earth System Model (CanESM2), using different downscaling techniques. While the BCCM model was run inter-annually and then averaged to produce the climatologies, the NEP36 model used atmospheric climatologies with augmented winds to force the ocean model and produce representative climatologies. Comparing these regional projections provides an estimate of the uncertainty across different regional downscaling models and methods. We find that the projected impacts of climate change on the groundfish community are more sensitive to the differences in the regional ocean models than they are to the emissions scenarios used. However, these differences are in magnitude (changes tend to be larger based on NEP36 compared with the BCCM) rather than in direction, with both models resulting in similar overall patterns of biodiversity change and turnover for the groundfish community. Over the 60-year time period (1986–2005 versus 2046–2065) used in our study, our projections suggest that groundfish community changes are similar regardless of the scenario used.

Description: This dataset contains layers of predicted occurrence for 65 groundfish species as well as overall species richness (i.e., the total number of species present) in Canadian Pacific waters, and the median standard error per grid cell across all species. They cover all seafloor habitat depths between 10 and 1400 m that have a mean summer salinity above 28 PSU. Two layers are provided for each species: 1) predicted species occurrence (prob_occur) and 2) the probability that a grid cell is an occurrence hotspot for that species (hotspot_prob; defined as being in the lower of: 1) 0.8, or 2) the 80th percentile of the predicted probability of occurrence values across all grid cells that had a probability of occurrence greater than 0.05.). The first measure provides an overall prediction of the distribution of the species while the second metric identifies areas where that species is most likely to be found, accounting for uncertainty within our model. All layers are provided at a 1 km resolution. Methods: These layers were developed using a species distribution model described in Thompson et al. 2023. This model integrates data from three fisheries-independent surveys: the Fisheries and Oceans Canada (DFO) Groundfish Synoptic Bottom Trawl Surveys (Sinclair et al. 2003; Anderson et al. 2019), the DFO Groundfish Hard Bottom Longline Surveys (Lochead and Yamanaka 2006, 2007; Doherty et al. 2019), and the International Pacific Halibut Commission Fisheries Independent Setline Survey (IPHC 2021). Further details on the methods are found in the metadata PDF available with the dataset. Abstract from Thompson et al. 2023: Predictions of the distribution of groundfish species are needed to support ongoing marine spatial planning initiatives in Canadian Pacific waters. Data to inform species distribution models are available from several fisheries-independent surveys. However, no single survey covers the entire region and different gear types are required to survey the range of habitats that are occupied by groundfish. Bottom trawl gear is used to sample soft bottom habitat, predominantly on the continental shelf and slope, whereas longline gear often focuses on nearshore and hardbottom habitats where trawling is not possible. Because data from these two gear types are not directly comparable, previous species distribution models in this region have been limited to using data from one survey at a time, restricting their spatial extent and usefulness at a regional scale. Here we demonstrate a method for integrating presence-absence data across surveys and gear types that allows us to predict the coastwide distributions of 66 groundfish species in British Columbia. Our model leverages the use of available data from multiple surveys to estimate how species respond to environmental gradients while accounting for differences in catchability by the different surveys. Overall, we find that this integrated method has two main benefits: 1) it increases the accuracy of predictions in data-limited surveys and regions while having negligible impacts on the accuracy when data are already sufficient to make predictions, 2) it reduces uncertainty, resulting in tighter confidence intervals on predicted species occurrences. These benefits are particularly relevant in areas of our coast where our understanding of habitat suitability is limited due to a lack of spatially comprehensive long-term groundfish research surveys. Data Sources: Research data was provided by Pacific Science’s Groundfish Data Unit for research surveys from the GFBio database between 2003 and 2020 for all species which had at least 150 observations, across all gear type and survey datasets available. Uncertainties: These are modeled results based on species observations at sea and their related environmental covariate predictions that may not always accurately reflect real-world groundfish distributions though methods that integrate different data types/sources have been demonstrated to improve model inference by increasing the accuracy of the predictions and reducing uncertainty.