LUC: Land Use Change Dynamics Model
Last modified on:
Thursday, 24-Mar-11 14:04 CET
LUC is a web-based land use/land cover dynamics model that operates on
regular grids of a few square meters to hectare and square kilometer resolution,
and with time steps that can range from minutes to years.
Implemented as a diritributed client-server system, it can use high-performance
cluster computing for very detailed, long-term, and probabilistic scenarios.
Typical application domains include:
- urban and regional development, zoning, land development, mining, forestry
- integrated coastal zone management (ICZM), coastal erosion
- watershed management (degradation, deforestation, erosion control)
- environmental risk (forest fire, mud slides, floods, tsunamies)
LUC is originally designed to predict "land use/land cover dynamics"
over longer periods (10-50 years with annual or monthly time steps),
and subsequent socio-economic and environmental impacts
(the model interfaces with both
WaterWare water reources model
at the river basin scale, and
AirWare for urban and industrial air quality assessment and management,
where it can provide both source terms (emission, pollution, activity related water demand)
or population distribution for impact assessment, and can be a component model in
RiskWare for the representation of technological and environmental risks.
It is also an integrated components of EMIS
an Environmental Management Information Systems for industrial applications.
LUC is basically a stochastic model, where the transition probabilities
(modified by the RULES and physically based dynamic boundary conditions)
are "instantiated" by a Monte Carlo method including an iterative component
to maintain consistency of global constructs.
LUC can also be used to simulate much faster
(and largely stochastic) processes in real-time such as,
for example, forest fires, mud slides, floods that are mainly driven by physically based
numerical models, but use the cellular automata representation to integrate
approximate, symbolic reasoning representing the considerable uncertainty
in these processes.
The key feature is a unique combination of probabilistic Markov-chain
state transition modeling, physically based algorithmic parts and
embedded expert systems technology using first order production rules
to adapt transition probabilities.
These rules consider: current state and history of each cell;
any number of static, or externally driven attributes of this cell;
attributes of neighboring cells within a user defined radius;
global attributes, static, derived from the current state of the domain, or externally driven.
The Land use dynamic model is based on:
The model results are provided in a range of formats, including:
- a priori transition probabilities that describe
the likelyhood that any given landuse gets transformed into another
(or more general, that the sate of any one spatial units changes into another state);
- first order logic RULES that can modify these a priori
transition probabilities based on
- temporal and spatial "neighborhood" of the land parcel in question
including global properties (summed over all parcels of the model domain);
- properties of the land and again neighboring parcels; these may express
- physical attributes such as elevation, slope, soil;
- local or symbolic properties such as administrative entities;
- a land use master plan that defines a desired target class;
- connectivity based on networks of line features such as roads, railway, power grid, waterways.
- the dynamic boundary conditions that can be supplied by a
spatially distributed physically based model (for example,
long-term soil moisture, dynamic water budget, or wind field predictions).
- animation (time series) of landuse maps that can be viewed
step by step, or with continuous animation using a Java based interactive MPEG player;
- trajectories of individual classes, aggregated over the model domain or for selected locations;
- trajectories of derived features such as energy and water consumption,
employment, waste generation and emissions; these are computed from attributes
of land use class (area specific variables) that may be expressed as time series
to represent changing technologies or regulatory frameworks.
- the "fate" of any given spatial unit over time, or the frequency/probability
distribution of states over any or all units.
- the realtionships between states, correlation or co-existance.
When running the model in numerous scenario in parallel, the results
can be expressed as the probability distribution (PDF, or modeled frequency distribution)
of the possible sate (such as land use or land cover) at a certain location and time (intervall).