Presentation on theme: "WHERE IS F3 IN MODELING LARVAL DISPERSAL? Satoshi Mitarai, David Siegel University of California, Santa Barbara, CA Kraig Winters Scripps Institution of."— Presentation transcript:
WHERE IS F3 IN MODELING LARVAL DISPERSAL? Satoshi Mitarai, David Siegel University of California, Santa Barbara, CA Kraig Winters Scripps Institution of Oceanography, La Jolla, CA Flow, Fish & Fishing A Biocomplexity Project
GOAL OF THIS WORK Assess fundamental mechanism of larval dispersal & dispersal kernel in California Current system – Using “idealized” simulations Develop modeling to establish dispersal kernels from available data sets
MATHEMATICALLY,... ❶ # of larvae released at a source location y ❷ Fraction of larvae transported from source y to destination x ❸ Fraction of larvae that recruit to adult Dispersal kernel (or connectivity matrix)
Self settlement DIFFUSION MODELS Do not account for regional differences Valid for long term dispersal
MARKOV CHAIN MODELING (SIEGEL ET AL, 2003) F3 requires seasonal dispersal kernels – Larval releases ~ 90 days – Decorrelation of larval dispersal ~ 3 days – 30 independent larval release Long term kernel (or diffusion model) Short term kernel (or Markov chain model)
IDEALIZED SIMULATIONS Based on ROMS (regional ocean model system) – Solves fundamental fluid dynamics equations, given initial & boundary conditions Initial & boundary conditions are specified using observation data – For strong and weak upwelling cases “Idealized” = statistically stationary & homogeneous in alongshore direction
SIMULATION FIELDS Strong upwelling case (summer northern California)
VALIDATION OF SIMULATION: MEAN TEMPERATURE Simulation field (mean over 180 days) CalCoFI data (July, Line #70) Shows good agreement with CalCOFI seasonal mean
VALIDATION OF SIMULATION: LAGRANGIAN STATISTICS Time scale Length scale Diffusivity (zonal/merid) (zonal/merid) (zonal/merid) 2.7/2.9 days 29/31 km 4.0/4.3 x10 7 cm 2 /s 2.9/3.5 days 32/38 km 4.3/4.5 x10 7 cm 2 /s Surface drifter data (Swenson & Niiler) Simulation data Show good agreement with surface drifter data
LARVAL DISPERSAL IN SIMULATIONS Modeled after typical rocky reef fish Nearshore habitat = waters shallower than 150 m Larvae are released daily for 90 days, uniformly distributed in habitat (1280 each day) Tracked as Lagrangian particles Settle when they are in habitat within competency time window of 20 to 40 days
LARVAL DISPERSAL Larval dispersal Sea level (stream line) Larvae are released every day for 90 days, uniformly distributed in habitat (1280 each day) Red dots: settling larvae
ONLY THE LARVAE THAT SETTLE Larval transportSea level (stream line) Let us observe where settlers to this subpopulation come from
DISPERSAL KERNEL Simulation Diffusion Model Dispersal kernel is heterogeneous Large spatial structures of “hot spots” exist Self settlement Mean 130 km
THE NEXT GENERATION OF SETTLERS Year t Year t+1 Self settlement Results suggest that dispersal kernel is stochastic
CONCLUSION Simulated results suggest... – Larval settlement is episodic – Dispersal kernels are stochastic & heterogeneous even in homogeneous environment – Large spatial structures of hot spots exist This results will have important consequences for predicting coastal fish stock variations
NEXT STEPS FOR IDEALIZED SIMULATIONS Investigate weak upwelling case Assess the role of topography – Coastline may create consistent “hot spots” Assess the role of larval behavior – Vertical migration may be important – Will behavior change kernel spatial structures? – Or, just change mortality?
NEXT STEP FOR MODELING DISPERSAL KERNEL Modify Markov chain model – To account for spatial structures of hot spots – How to obtain necessary information for Markov chain model from available data sets? – Should we simulate Channel Islands?
MARKOV CHAIN MODELING: APPLICATION FOR COMPLEX COAST LINE Dave’s Catalina Island 5000 independent larval release from each cite Dispersal Kernel Reasonable?