Computational Science for Integrated Modeling of the Environment

1
Computational Sciencefor Integrated Modelingof
the Environment

• Deana Pennington

2
Presidents Information Technology Advisory
Committee Report on Computational Science
Engineering

• 3rd Pillar of ScienceWith theory and
  experimentation
• Computational experimentation
• in silico experiment
• When we cannot do a physical experiment
• Used to make inferences about the underlying
  system

3
Simulation vs Experimentation
Parameterizations Scenarios
Simulation Results
Experimental Design
Validity Reliability Error Uncertainty
Data Analysis
Inference
Theory
Empirical Results
Experimental Design
Treatments
4
Model Validity, Reliability, Error and
Uncertainty
Issues Unavailability Correct prediction ltgt
correct model
Validity gt Ground truthing
Reliability Repeatability
High, if correct procedures are followed
Coding mistakes, technical glitches Testing is
critical
Error
Can be huge depending on the number of elements
being modeled Forced articulation of the
details Unknowns gt guessing
Uncertainty
5
Medawar zone Grimm 2005
Payoff
Simplify the equation!
Model complexity
All Available Data
Single problem
Multiple Patterns Different Scales
6
Scenarios

• An approach to investigate systems
• Model validity (uncertain theory)
• Parameterization (uncertain data)

7
Example IPCC Climate Data

• Climate scenarios are plausible representations
  of the future that are consistent with
  assumptions about future emissions of greenhouse
  gases and other pollutants and with our
  understanding of the effect of increased
  atmospheric concentrations of these gases on
  global climate. It is important to emphasize that
  climate scenarios are not predictions, like
  weather forecasts are. Weather forecasts make use
  of enormous quantities of information on the
  observed state of the atmosphere and calculate,
  using the laws of physics, how this state will
  evolve during the next few days, producing a
  prediction of the future - a forecast. In
  contrast, a climate scenario is a plausible
  indication of what the future could be like over
  decades or centuries, given a specific set of
  assumptions. These assumptions include future
  trends in
• energy demand
• emissions of greenhouse gases
• land use change
• behaviour of the climate system over long time
  scales

IPCC Intergovernmental Panel on Climate Change
8
IPCC data

• 7 climate modeling centers
• 4 scenarios demographic, politico-societal,
  economic and technological storylines
• Production of greenhouse gases
• Aerosol precursor emissions
• 15 properties
• Precipitation, humidity, wind speed, maximum air
  temperature, etc.

9
Other issues
Analysis of voluminous modeled data Exploratory
spatial data analysis (Gahegan) Data/image mining
(NCSA) Web-based exploratory (Servilla Trends
project) Collaborative sensemaking
assessment Computer-supported collaborative work
(Pennington) Education Students -
eLearning Decision makers Public
outreach Participatory approaches Citizen science
10
SW Vegetation Potentially Relevant Patterns??

• Climate change increasing nighttime lows?
  Increasing precip variability?
• Climate variability seasonal, interannual ENSO,
  multi-decadal?
• Increasing wildfire
• Increasing invasive grass/shrubs
• Drought-related insect surge tree mortality
• Social increasing population, exurban
  migration, water usage

11
Important Components??
Vegetation Composition Structure
Invasive species
High uncertainty gt scenarios? social change,
exurban migration water usage invasive
management long-term climate change (IPCC)
12
Questions of interest???

• Carbon storage exchange
• Wildfire vulnerability places, structures,
  people
• Biota community composition, migration,
  biodiversity
• Zoonotic infectious disease rodents, mesquitos,
  bats

13
Strategic use of resources
NVAC
NCSA I2K GeoLearn
GeoVista
FIA
Habitat Growth
SEEK ENM
LTER CREATE
SEEK EcoGrid
DIREnet
SDSC CLEOS
SALVIAS
Competition Dispersal
Wildfire
Phenology
14
CI-Team project goal Pull this together ?

• Problem definition and planning
• CI-Seminar gt understand new IT approaches
• CI-Vision working meeting (early summer) gt build
  integrated scientific conceptual model
• CI-Strategy working meeting (late summer) gt
  integrated science and technology in proposals
• Choose the right team members
• Disciplinary excellence but not just this
• Communication and Availability
• Commitment to the problem (engagement)
• Hard to choose the right people until the problem
  is defined, but problem definition depends on who
  is involved
• Strong project leadership
• See the big picture cross-disciplines
• Earned trust and respect

Nicholson 2002, Interdisciplinary modeling
heuristics
15
Familiarity with resources
NVAC
NCSA I2K GeoLearn
GeoVista
FIA
Habitat Growth
SEEK ENM
LTER CREATE
SEEK EcoGrid
DIREnet
SDSC CLEOS
SALVIAS
Competition Dispersal
Wildfire
Phenology
16
Ecological Niche Modeling (ENM)
Known Species Locations
Environmental Characteristics from gridded GIS
layers
Temperature layer
Many other layers
17
Technical Challenges

• Problem Statement Technical challenges in
  ecological niche modeling include
• labor-intensive data preparation
• interoperation of analyses in multiple computing
  environments
• high-throughput analysis of many species with
  many model runs.
• The SEEK Solution
• Development of an analysis and modeling
  environment capable of
• Rapid incorporation of re-usable components with
  visual modeling
• Integration of multiple computing environments
  into a single environment
• Access to distributed resources
• Kepler Scientific Workflow System

18
ENM in Kepler Conceptual Workflow
TAXON approaches
Kepler Native
Filter out If n lt X, where n count of
occurrences X is user defined
IPCC future climate scenarios (S 21)
MaNIS Species Locations (L)
Append datasets
Store points as ASCII
1
IPCC present climate layers (C) n 7
Restructure
For each S
Convex Hull Mask
Convert layers to binary
Input Parameters
Rescale Projection Extent Grain
Restructure
Append layers
For each C
Rescale values
For each C, S T
Hydro1k topographic layers (T) n 4
Restructure
For each T
GIS GDAL/Java
Testing data sample set
For i 1 to n n of models
GARP model training prediction of present
distribution (P)
Select best models (m)
Calculate model error
Sample Data 2 sets
1
2
Combine prediction results gt probability map
For each S integrated with T
For each P F prediction from models (m) 22
Predict future distribution (F) from model
For each model in m
Dispersal analysis
2
GIS/R
19
ENM in Kepler Executable Workflow
Top level
20
Modular gt Reusable

• SEEKs prototype application mammals
• Every IPCC climate change model/scenario
• Modify for vegetation?
• Change occurrence point source data
• Change algorithm
• Demo