Transcript Agent-based simulations of biocomplexity: Effects of

Agent-based simulations of
biocomplexity: Effects of
adsorption to natural organic
mobility through soils
Leilani Arthurs and Patricia Maurice
Civil Engineering and Geological
Sciences
Gregory Madey, Xiaorong Xiang,
Yingping Huang, and Ryan Kennedy
Computer Science and Engineering
University of Notre Dame
Natural Organic Matter (NOM)





Ubiquitous in aqueous and terrestrial
environments
Breakdown product of decaying plant
material
Controls many biogeochemical processes
Polydisperse mixture
Molecular weight controls NOM reactivity
Figure from Cabaniss et al. (2000)
Development of NOM Simulator




Complex interactions of NOM through porous
media results in emergent behaviors amenable
to a “biocomplexity” approach.
Design and use an agent-based stochastic model
for NOM interactions.
We focus specifically on how NOM molecular
weight affects adsorption to mineral surfaces
and mobility through soil.
Additional research by Cabaniss et al. focuses on
higher order biogeochemical reactions.
The NOM Simulator Design
• Java language, J2EE architecture
• Swarm and Repast software
• WEB interface
•Can be used interactively as part of a
collaboratory
•Allows for data mining
• Low surface coverages: adsorbed fraction mimics initial
• Higher surface coverages: preferential adsorption of
intermediate to high molecular weight components
• Kinetic data show that smaller molecules adsorb fast,
gradually replaced by larger molecules
Zhou et al.
(2001)
Adsorption & Desorption Probabilities to Fit
Batch Data
• High MW adsorbs slowly and desorbs slowly.
• Low MW adsorbs fast and desorbs fast.
P
desorb
 0.89  e
x
2000
 0.01
x  MolecularWeight(MW )
NOM Input Distribution
1. Example of WEB interface:
2. Initial Molecular Distribution:
Initial NOM Input Distribution
50
Percentage of Total Distribution
45
40
35
30
25
20
15
10
5
0
1
10
100
1000
10000
100000
1000000 10000000
Molecular Weight (log scale)
f
i

1
 2
 (   logM i) / 2 2
e
2
(Equation Cabaniss et al. 2000)
• Zhou et al. showed that average MW in solution
decreased over time, indicating replacement of
fast adsorbing small molecules by larger
molecules.
• The NOM Simulator captures this behavior for
batch adsorption example.
Batch experiment data at pH 5.5 by Zhou (1999)
Total # Adsorbed Molecules vs. Molecular Weight
# Adsorbed Molecules
2300
MW in Solution
2280
2260
2240
2220
2200
2180
2160
0
1
2
3
4
Time (hours)
sample
control
5
6
1600
1400
1200
1000
800
600
400
200
0
1
10
100
1000
10000 100000 1E+06 1E+07
Molecular Weight (log scale)
Time Step 1
Time Step 2
Time Step 4
Time Step 500
Probability equations optimized from batch
experiments applied to flow model (column
experiment).
Total # Adsorbed Molecules vs. Molelcular Weight
Total # Adsorbed Molecules vs. Time Step
# Adsorbed Molecules
# Adsorbed Molecules
25000
20000
15000
10000
5000
0
0
1000
2000
3000
Time Step
4000
5000
6000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
1
Time Step 10
10
100
1000 10000 100000 100000 1E+07
0
Molecular Weight(log scale)
Time Step 50
Time Step 100
Time Step 200
Flow simulation will be compared to
future laboratory column experiments.
Time Step 300
Visualization of Simulation
Settings
Legend
Visualization Capture 1
Visualization Capture 2
Visualization Capture 3
Visualization Capture 4
Results and Conclusions




Developed an agent-based stochastic
model for NOM adsorption.
The simulator is accessible through the
WEB.
Promotes the use of a “collaboratory” for
geographically separated interdisciplinary
scientists.
Allows users to set/refine parameters and
equations.
Acknowledgements




Dr. Steve Cabaniss
(University of New
Mexico)
Center for Environmental
Science and Technology
and Environmental
Molecular Science
Institute at the University
of Notre Dame
National Science
Foundation (EMSI, ITR)
PPG Industries