Predicting abundance indices in areas without coverage with a latent spatio-temporal Gaussian model

Breivik, Olav Nikolai; Aanes, Fredrik Lohne; Søvik, Guldborg; Aglen, Asgeir; Mehl, Sigbjørn; Johnsen, Espen

doi:10.1093/icesjms/fsab073

Cited by 20 publications

(9 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Evaluating the fit of predictive process SDMs and validating assumptions about structure involves considerations that are shared among many statistical models, but also includes others that are specific to—or particularly important for—spatial or spatiotemporal modeling. Examples of diagnostics for SDMs include the analysis of temporal or spatial autocorrelation in residuals ( Cressie & Wikle, 2015 ; Ward et al, 2018 ), randomized quantile residuals ( Dunn & Smyth, 1996 ; Hartig, 2021 ), one-step ahead residuals ( Thygesen et al, 2017 ; Breivik et al, 2021 ), residuals from MCMC draws from the posterior ( e.g ., Rufener et al, 2021 ), and examining evidence of non-stationarity. The choice of additional metrics of fit often follows from previous decisions about model structure.…”

Section: Methodsmentioning

confidence: 99%

“…There may be a number of applications where making predictions beyond the range of observations may be a goal—examples include using SDMs to predict the spread of invasive species, making predictions into areas that are protected or otherwise physically inaccessible, quantifying range shifts related to warming environments, or identifying effects of habitat restoration for species that are locally extirpated ( e.g ., making predictions based on habitat suitability). Predictions beyond the edges of the sampling domain are expected to be imprecise, however there are cases where such predictions may still be useful ( e.g ., Pearson et al, 2007 ; Breivik et al, 2021 ). As an alternative example, there may be scenarios where the prediction grid should be smaller than the sampling domain: geographic features may constrain some species (freshwater lakes for terrestrial species, or islands in marine applications), or protected areas where sampling is not permitted.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

The shadow model: how and why small choices in spatially explicit species distribution models affect predictions

Commander

Barnett

Ward

et al. 2022

PeerJ

View full text Add to dashboard Cite

The use of species distribution models (SDMs) has rapidly increased over the last decade, driven largely by increasing observational evidence of distributional shifts of terrestrial and aquatic populations. These models permit, for example, the quantification of range shifts, the estimation of species co-occurrence, and the association of habitat to species distribution and abundance. The increasing complexity of contemporary SDMs presents new challenges—as the choices among modeling options increase, it is essential to understand how these choices affect model outcomes. Using a combination of original analysis and literature review, we synthesize the effects of three common model choices in semi-parametric predictive process species distribution modeling: model structure, spatial extent of the data, and spatial scale of predictions. To illustrate the effects of these choices, we develop a case study centered around sablefish (Anoplopoma fimbria) distribution on the west coast of the USA. The three modeling choices represent decisions necessary in virtually all ecological applications of these methods, and are important because the consequences of these choices impact derived quantities of interest (e.g., estimates of population size and their management implications). Truncating the spatial extent of data near the observed range edge, or using a model that is misspecified in terms of covariates and spatial and spatiotemporal fields, led to bias in population biomass trends and mean distribution compared to estimates from models using the full dataset and appropriate model structure. In some cases, these suboptimal modeling decisions may be unavoidable, but understanding the tradeoffs of these choices and impacts on predictions is critical. We illustrate how seemingly small model choices, often made out of necessity or simplicity, can affect scientific advice informing management decisions—potentially leading to erroneous conclusions about changes in abundance or distribution and the precision of such estimates. For example, we show how incorrect decisions could cause overestimation of abundance, which could result in management advice resulting in overfishing. Based on these findings and literature gaps, we outline important frontiers in SDM development.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

The shadow model: how and why small choices in spatially explicit species distribution models affect predictions

Commander

Barnett

Ward

et al. 2022

PeerJ

View full text Add to dashboard Cite

show abstract

“…These priors may be useful in cases where estimation is difficult because of identifiability issues or relatively flat likelihood surfaces, or to impart prior information or achieve regularization. Following other recent SPDE implementations in TMB (Osgood-Zimmerman & Wakefield 2021; Breivik et al . 2021), penalized complexity (PC) priors (Simpson et al .…”

Section: Model Descriptionmentioning

confidence: 99%

sdmTMB: An R Package for Fast, Flexible, and User-Friendly Generalized Linear Mixed Effects Models with Spatial and Spatiotemporal Random Fields

Anderson

Ward

English

et al. 2022

Preprint

View full text Add to dashboard Cite

Geostatistical data---spatially referenced observations related to some continuous spatial phenomenon---are ubiquitous in ecology and can reveal ecological processes and inform management decisions. However, appropriate models to analyze these data, such as generalized linear mixed effects models (GLMMs) with Gaussian random fields, are often computationally intensive and challenging to implement, interpret, and evaluate. Here, we introduce the R package sdmTMB, which implements predictive-process SPDE- (stochastic partial differential equation) based spatial and spatiotemporal models. Estimation is conducted via maximum marginal likelihood with Template Model Builder (TMB) but can be extended to penalized likelihood or Bayesian inference. We describe the statistical model, illustrate the package's use through two case studies, and compare the functionality, speed, and interface to related software. We highlight advantages of using sdmTMB for this class of models: (1) sdmTMB provides a flexible interface familiar to users of glm(), lme4, glmmTMB, or mgcv; (2) estimation is often faster than alternatives; (3) sdmTMB provides simple out-of-sample cross validation; (4) non-stationary processes (time-varying and spatially varying coefficients) are easily constructed with a formula interface; and (5) sdmTMB includes features not available as a combination in related packages (e.g., penalized smoothers and break-point effects, anisotropy, abundance index standardization). We hope that sdmTMB's user-friendly interface will open this useful class of models to a wider audience within species distribution modelling and beyond.

show abstract

“…Other studies have applied statistical approaches to infer stock structure (i.e., stock abundance-at-length) from incomplete survey data. In Breivik et al (2021), a model predicts the number of fishes per year and length class in the unsurveyed hauls. It uses a linear combination of multi-variate Gaussian functions dependent on time, location, and length class.…”

Section: Introductionmentioning

confidence: 99%

Filling Gaps in Trawl Surveys at Sea through Spatiotemporal and Environmental Modelling

et al. 2022

View full text Add to dashboard Cite

International scientific fishery survey programmes systematically collect samples of target stocks’ biomass and abundance and use them as the basis to estimate stock status in the framework of stock assessment models. The research surveys can also inform decision makers about Essential Fish Habitat conservation and help define harvest control rules based on direct observation of biomass at the sea. However, missed survey locations over the survey years are common in long-term programme data. Currently, modelling approaches to filling gaps in spatiotemporal survey data range from quickly applicable solutions to complex modelling. Most models require setting prior statistical assumptions on spatial distributions, assuming short-term temporal dependency between the data, and scarcely considering the environmental aspects that might have influenced stock presence in the missed locations. This paper proposes a statistical and machine learning based model to fill spatiotemporal gaps in survey data and produce robust estimates for stock assessment experts, decision makers, and regional fisheries management organizations. We apply our model to the SoleMon survey data in North-Central Adriatic Sea (Mediterranean Sea) for 4 stocks: Sepia officinalis, Solea solea, Squilla mantis, and Pecten jacobaeus. We reconstruct the biomass-index (i.e., biomass over the swept area) of 10 locations missed in 2020 (out of the 67 planned) because of several factors, including COVID-19 pandemic related restrictions. We evaluate model performance on 2019 data with respect to an alternative index that assumes biomass proportion consistency over time. Our model’s novelty is that it combines three complementary components. A spatial component estimates stock biomass-index in the missed locations in one year, given the surveyed location’s biomass-index distribution in the same year. A temporal component forecasts, for each missed survey location, biomass-index given the data history of that haul. An environmental component estimates a biomass-index weighting factor based on the environmental suitability of the haul area to species presence. Combining these components allows understanding the interplay between environmental-change drivers, stock presence, and fisheries. Our model formulation is general enough to be applied to other survey data with lower spatial homogeneity and more temporal gaps than the SoleMon dataset.

show abstract

Predicting abundance indices in areas without coverage with a latent spatio-temporal Gaussian model

Cited by 20 publications

References 30 publications

The shadow model: how and why small choices in spatially explicit species distribution models affect predictions

The shadow model: how and why small choices in spatially explicit species distribution models affect predictions

sdmTMB: An R Package for Fast, Flexible, and User-Friendly Generalized Linear Mixed Effects Models with Spatial and Spatiotemporal Random Fields

Filling Gaps in Trawl Surveys at Sea through Spatiotemporal and Environmental Modelling

Contact Info

Product

Resources

About