Liu et al. Science. 347,(2015) Torres, J. et al. Science. 357, 970–971 (2017). Liu et al. Ecology & Society. 18 (2013).
Quantifying cascading effects in trade networks
A case study on salmon
Telecoupling and teleconnections
“human actions in one place may create unintended consequences elsewhere”
Climate teleconnections: distant interactions between climate systems
Urban-land teleconnections: land changes linked to underlying urbanisation dynamics
Economic globalization: trade and distant interactions between productive systems
- By 2016 salmon trade accounted for 11.6B US$:
- 93 countries
- 406 bilateral relationships
- In 2016 toxic algae bloom “red tide” killed 20% of Chile’s salmon:
- 27M salmon = 70M pounds
- >300 whales
- 40k tons sardines
- >800M dollars lost
Theory: what’s an attractor or manifold?
Sugihara et al. Detecting Causality in Complex Ecosystems. Science (2012)
Theory: Taken’s theorem
Sugihara et al. Detecting Causality in Complex Ecosystems. Science (2012)
Method: convergent cross-mapping
If x causes y, then the behaviour of the causal variable (x) leaves a signature on the affected variable (y)
Criterion for CCM to be a rigorous test of causality:
- The cross-map prediction skill ρ with the full time-series is statistically significant
- Cross-map prediction converges ρ increases as more time-series is used
Sugihara et al. Detecting Causality in Complex Ecosystems. Science (2012)
Method: convergent cross-mapping
If x causes y, then the behaviour of the causal variable (x) leaves a signature on the affected variable (y)
Criterion for CCM to be a rigorous test of causality:
- The cross-map prediction skill ρ with the full time-series is statistically significant
- Cross-map prediction converges ρ increases as more time-series is used
Sugihara et al. Detecting Causality in Complex Ecosystems. Science (2012)
Each link correspond to the relationship A -> B <- C, where the forecasting skill allow us to predict C dynamics given the time series of A with a ρ > 0.1 (t-test, p < 0.05)
Each link correspond to the relationship A -> B <- C, where the forecasting skill allow us to predict C dynamics given the time series of A with a ρ > correlation coefficient (t-test, p < 0.05)
41 out of 239 trading countries are causally related on the salmon trade network (~45%)
Prediction skill ρ is weakly correlated to network structure but not necessarily with trade volume
Lessons
- It is possible to detect interdependencies between natural resource exploitation in far way places = teleconnections
- Non-linear causality detection techniques help to identify relevant links on the network; relevant for management
- Governance of natural resources transcends national borders, but it’s important to know with whom to team up to better manage ecosystems
Tack | Gracias
Questions?
email: juan.rocha@su.se
twitter: @juanrocha
slides: juanrocha.se/presentations/trade_networks
preprint: coming soon!
Stockholm Resilience Centre
Subscribe to our newsletter at www.stockholmresilience.org/subscribe
1
Quantifying cascading effects in trade networks
A case study on salmon
Juan C. Rocha