Table of Contents
Predicting the Impact of Global Climatic Change on Land Use in Europe
Paper presented at International Conference on Geographic Modelling and Environmental Systems with GIS, Hong Kong 23-25 June 1998
Contents
Why?
Some Background
Results reported today reflect a 5 person-year sub-project concerned with adding a socio-economic systems modelling component to the physical environmental models
The research challenge!
The hardness of this challenge should not be under-estimated!
The Model Design Brief
The mainproblem is ..
itsIMPOSSIBLEat a micro level of detail!
Yet.. there is an increasing urgency to know what is happening to our world!
Previous Research has been very deficient
The CLUE modelling framework
CLUE Model
Mapping Environmentally Sensitive Areas is a non-model alternative
Modelling Design Checklist
Building a Synoptic Prediction System (SPS)
SPS Modelling
PPT Slide
SPS is limited by the following:
Other Problems
but
but
but
but
there is little that can be done about any of this!
The question is HOW to OPERATIONALISE this schematic model in the best way that is possible right now?
In essence it is a kind of non-linear regression model
PPT Slide
SPS is based on a neurocomputing approach
Do not PANIC!
Do not PANIC!
Do not PANIC!
Its just an artificial neural network!
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
Some Key Assumptions
Artificial Neural Networks are used for all the modelling in this presentation
Building a SPS
Building a SPS (Part 2)
Building a SPS (Part 3)
Step1: Assemble data for a common EU wide geography
Data required for Land use Predictor Variables
Why these variables?
1 Decimal Minute EU database
EVERY data set caused problems and required its own set of GIS operations in order to create the 1 DM grid data base
Estimating and Interpolating population data
PPT Slide
Impossible?
Maybe BUT that is the degree of difficulty associated with this environmental-socioeconomic modelling task!
Methodology
PPT Slide
PPT Slide
PPT Slide
RIVM method seemed best
PPT Slide
PPT Slide
The Errors are too LARGE
Maybe it is possible to do better using a neural net to perform the interpolation
PPT Slide
Population Predictor Variables
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
The Errors are still LARGEbut far smaller
but
Population interpolation maps look good!
PPT Slide
Lets have a closer look at the UK and Italy
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
(continue..) Step1: Assemble data for a common EU wide geography
PPT Slide
PPT Slide
Other Data
PPT Slide
PPT Slide
Other Data (continued)
PPT Slide
PPT Slide
PPT Slide
Basically EU data and all aspects relating to it are in
Step 2. Obtain or make forecasts for these data for 25 and 50 years time
Step 3. Construct Neural Nets to model the relationships between climate-soil-biomass-elevation-population in order to predict present day land use
PPT Slide
SPS Neural Net
Results
PPT Slide
PPT Slide
PPT Slide
Deficiencies!
(yet more grave) Deficiencies!
Good Points?
Good Points (more??)
Step 5. Create maps of changesStep 6. Consider modifying the predictions and forecasts to reflect knowledge expressed as fuzzy rules
Fuzzy Interpretation of Impacts
PPT Slide
16 Fuzzy Rules
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
Step 7. Repeat everything to test different change scenariosStep 8. Make estimates of uncertainties using Monte Carlo simulation
Conclusions
The results presented today are:
BUT
No need to be too pessimistic!
BUT
BUT
BUT
its very important to the future of the world that we try!
|