Red mullet in northern Spain (GSA 6) and the Adriatic (GSAs 17-18)
A spatio-temporal species distribution model: biomass, habitat suitability, trawling and marine heatwaves
21 October 2026
Goodness of fit
- Both models converge (all
sanity()checks passed) - Residuals on the normal line: Tweedie fits zeros and large catches
- Observed vs fitted (log): r = 0.90 (GSA 6), 0.88 (Adriatic)
- In-sample fit; cross-validation would test new hauls
Bottom temperature: two areas compared
+17% biomass per +1 °C in the Adriatic (95% CI +14% to +21%)
+18% per +1 °C in GSA 6 (95% CI −4% to +45%): not significant
Discussion point: GSA 6 hauls span only 12.9-17.6 °C (Adriatic 11.2-28.2 °C), so the temperature effect is poorly constrained there and depth dominates.
Predicted biomass: GSA 6
Highest on the inner shelf along the whole coast, dropping quickly offshore.
Predicted biomass: Adriatic
Highest in the shallow northern basin and along the Croatian coast; a gentler depth gradient over a much wider shelf.
Spatial and spatio-temporal random effects
ω (first column): persistent hotspots not explained by depth and temperature
ε by year: small in GSA 6, large in the Adriatic (up to ±2.9, a factor of ~18 in biomass)
Biomass index
Sum of predicted biomass × cell area, per year
−3% GSA 6, 2015 → 2021 (Tweedie GAM of Section 2b: −12% in density)
77% to 146% The Adriatic index fluctuates more, between 77% and 146% of its 2014-2021 mean
Discussion point: each area peaks in its least-sampled year. Index standard errors (get_index()) would show how uncertain these years are.
Habitat suitability index (HSI)
Fixed effects (depth + bottom temperature) rescaled to 0-1 across all years in each area; reproduces the group’s maps. Highest on the shallow inner shelf; in the Adriatic, the northern basin scores high almost everywhere.
Change in habitat suitability, 2014 to 2021
- GSA 6: median +0.1%, almost no change
- Adriatic: median −1.8%; 63% of cells lose suitability, mainly the northern basin
Discussion point: the % change of a 0-1 index explodes where HSI ≈ 0 (up to +217% in one GSA 6 cell): use the difference in HSI and a 2014-2021 trend rather than two years.
Overlap between red mullet and trawling
Local correlation between red mullet and GFW trawling effort, by year
- Red: trawling concentrated where red mullet is abundant
- GSA 6: mostly negative, positive patches on the outer shelf
- Adriatic: a positive band through the north-western and central basin every year; coastal edge negative
Discussion point: in small windows, correlations near ±1 are common by chance: add a significance test or a whole-area overlap measure.
Marine heatwaves in the two areas
Summer heatwaves (Hobday et al. 2016) computed on the prediction grid of each area
- GSA 6 strongest summers: 2017, 2018, 2020
- Adriatic strongest summers: 2015, 2018, 2019
Discussion point: the two areas do not share their heatwave years (2020: strong in GSA 6, almost none in the Adriatic), so comparisons need area-specific years.
Heatwaves, biomass trends and trawling trends: GSA 6
Trends 2014-2021 per cell, as % of the cell’s mean per year; grey: not significant (p ≥ 0.05). Significant biomass trends: 3% of cells; none after correcting for multiple tests.
Heatwaves, biomass trends and trawling trends: Adriatic
Grey in A: no SST data north of 45°N. Significant biomass trends: 12% of cells; none after correcting for multiple tests.
Trawling in strong and weak heatwave summers
- Daily GFW trawler effort, June-August, 3 strong vs 3 weak summers, from 2015
- Total hours: +1.9% in GSA 6, -1.3% in the Adriatic, within year-to-year noise
Do the fleets fish the same grounds?
Core grounds (cells with 80% of hours): blue-grey in both types of summer, red strong only, dark blue weak only.
| GSA 6 | Adriatic | |
|---|---|---|
| Overlap D, strong vs weak | 0.72 | 0.72 |
| Overlap D, same type | 0.71, 0.7 | 0.71, 0.72 |
| Centre shift | 1.4 km | 3.9 km |
| Effort on red mullet hotspots | 13% / 14% | 23% / 20% |
Same grounds in strong and weak heatwave summers: overlap as high as between two summers of the same type (D: 1 = identical effort distribution).
Heatwaves and red mullet biomass across years
Adriatic: biomass drops in the two strongest heatwave summers (2015, 2018), recovers the year after; overall ρ = -0.14
GSA 6: the opposite, its strongest heatwave summers are among its higher-biomass years (ρ = 0.53)
Discussion point: eight years per area, two areas in opposite directions: a hypothesis to test, not a result.
Heatwaves and red mullet biomass within years
Year-specific deviation of predicted biomass (ε) in each cell, against heatwaves and trawling, with depth, latitude and year in the model
- GSA 6: small effects
- Adriatic: not monotonic, a dip at moderate heat, higher biomass in the hottest cells
- No trawling effect in either area
- Does not reproduce the annual dips