Walleye Pollock

This collection contains annual covariates related to walleye pollock spawning phenology and environmental conditions in the Gulf of Alaska (1983–2021). It is used to model how temperature and spawning timing influence the observed and model-predicted walleye pollock biomass. The data is used to study the validity of causal hypotheses in time series models through conditional-independence tests in Thorson et al. (2025). The second dataset of the simulation experiment in Thorson et al. (2025) about wolf and moose populations can be obtained as isle_royale through the R-pagacke dsem Thorson et al. (2024). Task: The dataset can be used to study causal discovery methods. Summary: Size of collection: 2 datasets with dimensions 40x4 and 53x5. Aggregated to 40x5. Task: Causal Discovery Data Type: Continuous Data Dataset Scope: Standalone Dataset Ground Truth: Unknown Graph Temporal Structure: Time Series Data License: CC BY-NC-SA 4.0 Missing Values: Existing Missing Values Features: Year: Calendar year of observation (1970 / 1983 - 2021) Fem30p_wt_logit: Logit-transformed proportion of females >30cm in a spawning or spent stage during the spawning-grounds survey (variable P in the paper) mismatchmed: Average number of days between the mean date of spawning (estimated from larval-derived hatch dates) and the mean date of the survey (variable A in the paper) MAR_SST: March sea surface temperature in the Gulf of Alaska (variable T in the paper) Full_data_pred: Model-predicted pollock biomass using all survey data. Present in the supplementary material but not used in the paper. No_indices_pred: Model-predicted pollock biomass excluding all surveys indices. Present in the supplementary material but not used in the paper. No_survey_pred: Model-predicted pollock biomass excluding the Shelikof spawning survey. Used as denominator to compute Q. Observed3p: Observed survey biomass from the Shelikof Strait spawning survey. Used as numerator to compute Q. The paper's variable 𝑄 is computed as the log-ratio between the surveyed biomass and predicted biomass where the latter is taken from one of the three population dynamics model fitted to the survey data without accounting for timing or temperature. Files: CatchabilityCovariates_2023-03-29.csv: Annual covariates related to walleye pollock spawning phenology and environmental conditions in the Gulf of Alaska expected_shelikof_indices_Jan2022.csv: This dataset contains observed and model-predicted survey biomass indices for walleye pollock in the Shelikof Strait, Gulf of Alaska. It provides the baseline population dynamics model output used to derive the survey availability metric Q. Graph: The estimated dynamic structural equation model is found to be T->A, A->P, A->Q ("Timing as mediator").