From Physics to Fish: Coupling 3-D Hydrodynamic-Biological Hypoxia

Models with Individual Based Fish Models

Dubravko Justic, Kenneth Rose, Lixia Wang,

Haosheng Huang

Department of Oceanography and Coastal Sciences School of the Coast and Environment

Louisiana State University

“When there were no computers, computing was not a problem. When we had a few weak computers, computing was a mild problem. Now when we have gigantic computers, computing is a gigantic problem.” E. Dijkstra

“Shift Happens”

1983 - “Sir Clive Sinclair with his latest innovative computer design” 16 KB RAM

“Shift Happens”

1983 - “Sir Clive Sinclair with his latest innovative computer design” 16 KB RAM

2011 16 GB RAM

Beowulf Cluster (Sterling and Becker, 1994)

September 14, 2009 LSU Launches New Supercomputing Cluster BATON ROUGE, La., Sept. 14 This week, LSU's High Performance Computing, or HPC, will debut the campus' newest supercomputing cluster, Philip….The University purchased Philip for $215,000 in May 2009....

User Friendly and Cheap

TR

FOt

NP

DO

From 2-Box Models to Mega Models

~106 boxes (Wang and Justic, 2009)

2 boxes (Justic et al., 1996)

“Conventional” Hypoxia Models 3-D Hydrodynamic model

Water quality model

Motivation Assess ecological effects of large scale hypoxia, i.e., direct and indirect impacts on commercially important fish and shrimp

Population Displacements

(Craig and Crowder, 2005)

Objective • Develop models of fish movement

within a 3-D coupled hydrodynamic-biological model framework and derive exposure histories of simulated fish to hypoxia

“From Physics to Fish”

Wang and Justic, 2009; Justic and Wang, 2009; Justic et al., in preparation

FVCOM-LaTex WASP5

IBM

Individual Processes Movement Growth Mortality Spawning

Conditions Dissolved O2 Temperature Salinity Food density Predator Density Individual Size

FVCOM-LaTex Computational Domain and Grid

Nodes: 14,740 Elements: 28,320 Resolution: 1-10 km 31 sigma levels

Wang and Justic, 2009

Model Calibration and Validation • Year 2002 – the largest hypoxic zone

on record (22,000 km2) • Simulation period: April 1 to September

30, 2002; Integration step = 5 s; 4 hours/month on Pelican (32 nodes)

• Initial conditions inferred from transect data and C6 and CSI03 moorings

• Model performance – evaluated using tidal gauges, ADCPs, shipboard measurements, and satellite imagery

Wang and Justic, 2009

Wang and Justic, 2009

WASP5

Coupled FVCOM-WASP5

Justic and Wang (in preparation)

Justic and Wang (in preparation)

Fish IBM Movement Smart Particles = “Smarticles”

Kinesis method (Humston et al., 2004)

• X, Y and Z velocities of each individual are computed based on kinesis behavior (response to DO, T and S)

• Kinesis = sum of random and inertial velocities • When conditions approach preferred state, fish

tend to maintain their previous heading; if environmental conditions are harsh, the directional persistence of fish movement is low, mimicking a random walk

Applications Deriving exposure histories of simulated fish to Deepwater Horizon oil spill (Justic and Wang, in preparation; Huang et al, in preparation)

NASA Aqua MODIS May 25, 2010

Challenges in “Physics to Fish” Modeling

• Movement algorithms • Data requirements • Number and locations of monitoring

stations • How many monitoring stations are required

to calibrate a complex “Physics to Fish” model that has 1 million elements?

“Shift Happens”

1983 - “Sir Clive Sinclair with his latest innovative computer design” 16 KB RAM

2011 16 GB RAM

Hypoxia models in 2039? Average computer 16 PB RAM (Petabyte) = 1015 Byte

~ 1012 boxes

< 1 m resolution La-Tex shelf

Microscope-Macroscope Shift "To see these patterns which are bigger than ourselves, let us take a special view through the macroscope.” H.T. Odum (1971)

< 1 m resolution La-Tex shelf

~ 1012 boxes

G. Müller-Niklas, 1997.

Acknowledgements:

NOAA NGI

NOAA/CSCOR

(NGOMEX)