A Language to Support Spatial Dynamic Modeling Bianca Pedrosa, Gilberto C

1
A Language to Support Spatial Dynamic Modeling
Bianca Pedrosa, Gilberto Câmara, Frederico
Fonseca, Tiago Carneiro, Ricardo Cartaxo
Brazils National Institute for Space
ResearchPennsylvania State University
TerraML
2
TerraML Purpose

• Support spatial dynamic modeling
• Discrete and continuous behavior
• Inhomogeneous space
• Extensible framework

3
Outline

• Requirements of a dynamic modeling environment
• The TerraML computational environment
• The TerraML theoretical foundations
• The TerraML structure and syntax
• Future work

4
Spatial dynamic modeling
Demands Requirements

• discretization of space in cells
• generalization of CA
• discrete and continous processes
• Extensibility to include user-defined models
• Flexible neighborhood definitions

• Locations change due to external forces
• Realistic representation of landscape
• Elements of dynamic models
• Different types of models
• Geographical space is inhomogeneous

5
Inhomogeneous Space
Spaces of fixed location and spaces of fluxes in
Amazonia
6
TerraMLComputational Environment
7
Spatial Information Engineering

• Technological change
• Current generation of GIS
• Built on proprietary architectures
• Interfacefunctiondatabase monolythic system
• Geometric data structures archived outside of
  the DBMS
• New generation of object-relational DBMS
• All data will be handled by DBMS
• Standardized access methods (e.g. OpenGIS)
• Users can develop customized applications

8
TerraLib the support for TerraML

• Open source library for GIS
• Data management
• object-relational DBMS
• raster vector geometries
• ORACLE, Postgres, mySQL, Access
• Environment for customized GIS applications
• Web-based cooperative development
• http//www.terralib.org

9
TerraLib and TerraML

• TerraML is integrated with TerraLib
• access to typical GIS analytical tools

10
Computational Model
BUILDER
Parser
TerraLib Component Library
DOM/XERCES
TerraML XML based
TerraLib Code Generator
11
Theoretical Foundations for TerraML
12
TerraML Cellular Model
Cellular Space
13
Cell-space x Cellular Automata

• CA
• Homogeneous, isotropic space
• Local action
• One attribute per cell (discrete values)
• Finite space state
• Cell-space
• Non-homogeneous space
• Action-at-a-distance
• Many attributes per cell
• Infinite space state

14
Hybrid Automata

• Formalism developed by Tom Henzinger (UC
  Berkeley)
• Applied to embedded systems, robotics, process
  control, and biological systems
• Hybrid automaton
• Combines discrete transition graphs with
  continous dynamical systems
• Infinite-state transition system

15
Hybrid Automata

• Variables
• Control graph
• Flow and Jump conditions
• Events

Event
Event
Jump condition
Control Mode A Flow Condition
Control Mode B Flow Condition
16
Neighborhood Definition

• Traditional CA
• Isotropic space
• Local neighborhood definition (e.g. Moore)
• Real-world
• Anisotropic space
• Action-at-a-distance
• TerraML
• Generalized calculation of proximity matrix

17
Supporting Different Models

• Cells Potential for Change is Function of
• Global Demand
• e.g. 2 of forest area will be deforested per
  year
• Neighborhood Influence
• e.g., 80 of deforestation occurs near existing
  roads
• Local Attributes
• e.g., cells wìth more than 2800 mm of rain/year
  will not be feasible for agriculture

18
TerraML Structure
19
An Example in Hydrology

• A water balance Automata

input
input
WET Surplussoilwater-infilcp Soilwaterinfilcp
DRY soilwatersoilwaterpre-evap
soilwatergtinfilcp
discharge
input
TRANSPORTING MOVE(LDD, surplus, infilcp)
Surplusgt0
Control Mode Flow Condition Jump Condition Event Transition
DRY Solwatsolwatpre-evap Solwatgtinfcap WET
WET Surplussoilwater-infilcap Surplusgt0 discharge TRANSP
TRANSP MOVE(LDD,surplus, infilcap) Surplusgt0 input DRY
20
TerraML Example
ltcellprocessor author"bianca"
date"04/03/02" model"simulation of runoff"
case" timesteps of 6 hours gt modelling time
one week"gt ltinputgt ltlayer
name"infilcap.map attributeinfil"gt InfilCap
/gt ltlayer name"soil.map
attributeclass"gt SoilType /gt ltlayer
nameLDDmap attributeldd"gt LDD
/gt lttemporal name"rain.tss"gt
RainTimeSeries /gt lt/inputgt
ltoutputgt ltTemporal name"rainfall
attributesolwat"gt SoilWater /gt ltlayer
name"runoff gt Surplus
/gt lt/outputgt
21
TerraML Example
lttransitiongt ltmode
controlmode DRY
flowconditionsoilwaterpreevap
jumpcondition soilwatergtinfl_cap
towet /gt ltmode
controlmode nameWET
flowcondition Surplussoilwater-infilcp
Soilwaterinfilcp
jumpconditionsurplusgt0
toTRANSP eventdischarge
/gt ltmode
controlmode nameTRANSP
flowcondition MOVE(ldd,surplus,infilcap)
jumpconditionsurplusgt0
toDRY
eventinput /gt lt/transitiongt
22
TerraML Example
ltsimulationgt ltcellularspace neighborhoodLDD
resultsoilwater /gt lttimer init"1" end"28"
step"1" timeUnit"6 hours"gt
ltTransitgt lt/timergt lt/simulationgt
23
Future Work

• Formalization of model types
• Constructions of real-life applications
• Hydrology
• Deforestation
• Web availability

24
Acknowledgments

• ESRI
• Methodist University of Piracicaba, Brazil