Do fishers follow fish displaced by climate warming?

Climate change profoundly affects oceanographic conditions and, consequently, the expected distribution of many fish species. Because harvested fish is integral to the world food system and supports millions of livelihoods, it is essential to assess how fishers respond to “species on the move.” Prior studies often assume that fishers follow fish displaced by warming, yet fishers may face technological, economic, regulatory, and behavioral constraints that limit spatial adaptation, even in industrialized fisheries. Surprisingly little is known about actual fisher responses to shifting species distributions. This study matches twenty years of individually observed fisher behavior with model-based predictions of fish distribution at fine spatio-temporal resolution to assess to what extent fishers follow fish. Focusing on the Norwegian Northeast Arctic (NEA) cod fishery—high value, diverse gears, and few constraints on mobility—the study asks whether fishers relocate effort in line with shifting spawning-suitable habitats predicted from oceanographic conditions.

Global studies using the mean temperature of the catch (MTC) document warming signatures in fisheries, but this metric cannot disentangle fish behavior from fisher behavior. Case studies from large marine ecosystems and lakes show warming effects on fisheries, while a limited number of studies matching survey with catch/landing data (e.g., Alaska, US coasts) indicate sluggish fisher reactions to geographic shifts in target species. Prior work highlights the importance of economic, regulatory, and behavioral drivers of spatial effort (e.g., location choice models and habit persistence). For NEA cod, earlier research documents climate-driven changes in spawning timing and locations and the emergence of potential new spawning sites under warming. This study contributes by using highly resolved Norwegian landing data to capture within-season behavior, linking effort and catch to exogenous, model-derived spawning suitability at fine spatial and temporal scales.

Study system and data: The NEA cod fishery along Northern Norway peaks mid-February to early May during coastal spawning aggregations. Daily landing ticket data from the Norwegian Directorate of Fisheries (2001–2021; per-trip observations with species-specific weights, catch locations, vessel, gear, and regulatory characteristics; anonymized IDs) were aggregated to the weekly level by statistical catch areas to reduce noise and account for multi-day trips. The count of vessels operating in each area-week is the primary effort measure; mean catch per trip in each area-week measures catch potential. Oceanographic model and suitability index: Spawning suitability was estimated using the Regional Ocean Modeling System (ROMS) hindcast for the Northern Atlantic (4×4 km resolution), daily from February–May for 2001–2021. A grid point is suitable if (1) local depth < 180 m; (2) temperature within an appropriate spawning range at some depth between 50–150 m (described as 4–6 °C in Results; Methods list temperature criteria and 34.0–34.9 salinity); (3) salinity 34.0–34.9 at 50–150 m. Suitability for each catch area-week is the average daily fraction of grid points within the area meeting criteria. An area-based index is used (cod segregate by depth; females higher in the water column). Empirical strategy: Three-step approach: (i) document temporal changes and spatial shifts in spawning-suitable habitat; (ii) validate that suitability predicts catch potential via regressions of area-week mean catch per trip on suitability with extensive fixed effects (year, week, area, region, gear category, vessel length group); (iii) test whether fishers follow fish by regressing the number of vessels in an area-week on suitability and on lagged effort (previous week and previous year), with progressively richer fixed effects (including area fixed effects) to absorb time-invariant spatial characteristics (e.g., bathymetry, distance to ports, accessibility). Standard errors are clustered by area and year.

- Oceanographic changes: Spawning-suitable habitat has shifted over time, occurring earlier in the year and further north, consistent with prior findings. - Suitability predicts catch per trip: In regressions of area-week mean catch per trip on the suitability index (with year, week, area, region, gear, vessel length group fixed effects), suitability is a positive, statistically significant predictor. Table 1 (All areas): Observations = 105,515; R² = 0.3551; Within R² = 0.0002. Results are robust when restricting to coastal areas north of 62°N (not shown in the excerpted table but described in text) and controlling for vessel attributes, notably length group. - Suitability does not predict where vessels fish once area characteristics are controlled: In models explaining the number of vessels per area-week, suitability is significant when not controlling for area fixed effects (e.g., coefficient ≈ 0.359–0.366, p < 0.01). However, with area fixed effects included, suitability loses the expected positive association and is not a positive predictor (coefficients around −0.289 with p < 0.05). Lagged effort strongly predicts current effort: number of vessels in the previous week (≈ 0.58, p < 0.001) and previous year (≈ 0.32, p < 0.001) remain robust predictors across specifications. Table 2: Observations = 1,183,200; R² up to ≈ 0.739. - Behavioral pattern: Evidence points to site fidelity and habit persistence rather than dynamic relocation tracking spawning-suitable habitat within season and across years, despite few apparent regulatory or technological mobility constraints in this fishery.

The analysis directly addresses whether fishers follow fish displaced by climate warming by comparing observed fishing locations with model-predicted spawning-suitable areas at high spatio-temporal resolution. Spawning suitability robustly predicts catch potential, validating the suitability index as a proxy for fish availability. Yet, once time-invariant spatial characteristics are controlled via area fixed effects, suitability does not drive vessel presence; instead, past presence (week- and year-lagged) strongly predicts current effort. This indicates that fishers rely on established knowledge, experience, and possibly local economic conditions (e.g., distance to ports, safety, accessibility), exhibiting site fidelity rather than tracking shifting habitats in real time. These findings imply that projections of climate change impacts on fisheries must incorporate human behavioral dynamics—simple assumptions that fishers will follow fish can misrepresent realized effort distributions, economic outcomes, and ecological pressures. The results also suggest opportunities: if real-time suitability information were disseminated effectively, it might alter fisher expectations and behavior, potentially improving alignment between effort and dynamic habitat, though adoption and trade-offs (e.g., diversification vs specialization) need evaluation. The geopolitical implication for NEA cod is that limited fisher movement in response to shifting distributions could temper transboundary competition relative to expectations based solely on fish movement.

This study couples high-resolution oceanographic modeling with two decades of detailed Norwegian landing data to test the common assumption that fishers follow fish as climate change alters species distributions. While spawning suitability predicts catch per trip, it does not predict vessel presence once area characteristics are controlled; lagged effort patterns dominate, indicating site fidelity. The main contributions are: (1) validating a fine-scale suitability index against realized catch, (2) demonstrating that fisher behavior may not track shifting habitats even in a technologically advanced, relatively unconstrained fishery, and (3) highlighting the necessity of integrating human behavioral dynamics into climate impact assessments for fisheries. Future research should develop structural models of fisher location choice under changing environments, assess the role and uptake of real-time ecological information, evaluate diversification-specialization trade-offs under anticipated change, and extend analyses to other fisheries and regions to test generality.

- The oceanographic suitability index is a model-based proxy and an imperfect representation of true, variable ecological conditions (particularly salinity performance in some regions). - The study uses a reduced-form approach and does not structurally model fisher decision-making; mechanisms (e.g., costs, risk, information constraints) are not identified causally. - Area fixed effects absorb important economic and geographic factors (e.g., bathymetry, port distance, exposure), limiting direct attribution of drivers. - Suitability criteria and some parameter ranges in the text contain minor inconsistencies (e.g., temperature range notation), which could affect interpretation though core results are robust. - Results pertain to NEA cod during February–May over 2001–2021 and may not generalize to other species, seasons, or regions without caution. - Landing data are aggregated and anonymized; catch locations are self-reported areas rather than precise GPS tracks, potentially introducing spatial measurement error mitigated by weekly aggregation.