Day One part 1

Day 1 Objectives

 * Understanding general concepts of spatio-temporal modeling and the fundamental structures of such models
 * Understanding SELES model structure and data requirements

Course Topics

 * Spatio-temporal modelling approaches and concepts
 * Getting comfortable in the SELES modelling environment
 * Structure and syntax of SELES simulation models and the spatio-temporal contexts of landscape events
 * Spreading events and analysis of landscape configuration
 * Building, debugging and verifying SELES models
 * Model design and application of SELES modelling for land-use planning

Outline

 * General modelling concepts
 * The SELES view of the world
 * The model building process
 * Demonstration: Building a basic SELES model
 * Overview of SELES Interface
 * Spatial data formats
 * Opening, exploring, manipulating and saving rasters
 * Static models
 * Tutorial 1 – Understanding Spatial Data Requirements for SELES
 * Introduction to structure of SELES models
 * Tutorial 2 – Manipulating Spatial Data Using the SELES User Interface

General Modelling Concepts-Types of Spatio-Temporal Models

 * Static vs. Dynamic Models
 * Stochastic vs. Deterministic
 * Optimization vs. Simulation
 * Event-based Simulation Models
 * Individual-based Population Models
 * Pattern Generation and Analysis

What is a model?

 * A model is an abstraction of a real system for the purpose of improving understanding, or for predicting system behavior
 * “All models are wrong. Some models are useful” Deming

Why model?

 * To clearly define a problem
 * storage/ organization device for current state of knowledge
 * To organize our thoughts and ideas
 * gain deeper understanding of system and relationships
 * To understand our data and observations
 * identify gaps in our data and knowledge

Why use a spatial model?

 * Ecological processes are spatially dependent
 * distribution of caribou depends on habitat patches, movement corridors
 * water flow depends on height of land and soil types
 * Can test how different management rules can influence location of developments and affect impacts
 * access roads and pipelines can avoid areas with high ecological and cultural values
 * Impacts on ecological/cultural values are often spatial in nature
 * e.g., fragmentation
 * Strategic vs. operation planning
 * a spatial model represents operational rules and projects a feasible pattern of development
 * usually, this is uncertain and can change from one simulation to the next
 * reporting is at a strategic level – averages over the study area
 * not to say that real-life developments will follow operation rules

Dynamic vs. Static Models

 * Static Models
 * e.g., landcover maps, digital elevation models
 * Dynamic Models
 * Temporally explicit
 * Time represented in either:
 * fixed time steps,
 * arbitrary, but discrete time steps,
 * or continuous (differential equations)
 * Temporal autocorrelation
 * Present and future events dependant on preceding events

Stochastic vs. Deterministic Models

 * Deterministic
 * no random variables
 * always produce the same results given the same initial conditions
 * Stochastic
 * produce a different output each time they are run
 * Model results may or may not differ from a deterministic counterpart
 * Analysis of stochastic models is complicated
 * If possible, run and understand deterministic behaviour prior to analyzing stochastic behaviour

Stochastic Models

 * Models variation, not just averages
 * Process error
 * e.g. unaccounted noise in regression
 * Use distributions as parameters
 * draw random values from distribution with specified mean and standard deviation
 * Model results should be presented as mean and variance for a large numbers of replicates

Optimization vs. Simulation - Optimization

 * Assumes a closed world with no external influences
 * Given a formal description of a problem, find the best solution.
 * Maximize the objective function subject to specified constraints
 * E.g., Find maximum timber volume available at each time period, while maintaining a minimum amount of forests in each seral stage
 * Techniques
 * Linear programming
 * Simulated Annealing (near optimization)

Optimization vs. Simulation - Stochastic Simulation

 * Each run of the model produces a different result
 * Sample from an unknown distribution
 * Does not need to assume closed world
 * Techniques:
 * Discrete-event simulation
 * Examples: Individual based models

Event-based Simulation Models

 * Sequence of events
 * An event is a discrete process
 * A process is a set of actions with a definite start and end
 * An event can influence other events
 * Example: natural disturbance models

Individual-based Population Models

 * Represent a population as a set of distinct individuals, each tracked separately
 * Persistence of identity (unlike event models)
 * Issues: data intensive, movement rules, mortality, spatial dependencies …
 * Example: Random walkers

Pattern Generation Models

 * Produce a pattern based on a set of rules
 * Neutral models: pattern in the absence of process (null hypothesis)
 * Habitat models (compositional analysis, logistic regression RSF’s)
 * Predictive static empirical models
 * insert pic slide20

Pattern Analysis

 * Landscape metrics
 * Composition: mean patch size, area/perim ratio
 * Configuration: contagion, nearest neighbour
 * Connectivity: spatial graphs
 * Spatial autocorrelation
 * e.g., Kriging

Spatial Modelling

 * Explicit representation of pattern and spatial relationships
 * Example: GIS overlays



Temporal Modelling
Explicit representation of process and change over time



Spatially Explicit Landscape Event Simulator

 * SELES is not a model.
 * SELES is a spatio-temporal modelling framework
 * A general tool for building models of landscape dynamics
 * A language for specifying models of landscape dynamics
 * A simulation engine for running these models

The SELES view of the world

 * Sets of georeferenced raster layers
 * Basic spatial unit is the cell
 * Sets of variables and constants
 * Together these define the state variables and initial state for the model
 * Behaviours of model is described by agents of change
 * Landscape Events or Landscape Agents

Landscape Events

 * Models of processes responsible for landscape change
 * human or natural
 * continuous or periodic
 * spreading or non-spreading