Introduction in to SAMT

1. Concept
2. Simulation Control

State of the art of landscape modeling

Landscape models operate on maps. That means: to read information from maps and produce new maps as modeling results. Therefore GIS (geographic information system) like ARCGIS or GRASS are the "working horses" in this part of science. Landscape modeling is characterized by uncertainty of the underlying information and modeling knowledge. So techniques like fuzzy modeling or neural networks are common. The traditional GIS is specialized to perform spatial analysis, to produce maps in print-quality and last but not the least to organize the data handling of the maps. But there is a lack of modeling capacity in most of the GIS software. We have the following types of models:

1. mathematical expressions (like regression equations)
2. some modern techniques like SVD, Fourier, cluster algorithms
3. neural networks (as an alternative kind of statistical models)
4. fuzzy models (as a method to include expert knowledge)
5. dynamic models, using differential equations
6. dynamic models as cellular automata

The Concept

A commercial GIS is very expensive while a free GIS (GRASS) is often not easy to use. We need a modeling system with includes few GIS-features, like spatial analysis, some spatial techniques and a simple in and export function to a GIS. The main focus is to have an open system to include different models (especially fuzzy-models). Here are the main features of the Spatial Analysis and Modeling System (SAMT)

1. It is a grid based system, that means all maps are matrices and it can read and write grids in ASCII in ARCGIS format and all operations are performed on grids.
2. There is a connection to a mysql database. The models can get information from this database or can store information on it. Additionally grids can be used as index maps to the database, that means spatial information can be coded using mysql.
3. A HDF system is used as spatial database
4. There are some simple grid functions (SGRID), like set a grid to a constant value, construct a random grid, multiply a grid with constant and so on.
5. Additional advanced grid functions (AGRID), like scaling, interpolation, distances, focal functions are integrated.
6. There are analysis functions, like histogram, statistic, etc. Integrated in this part are additional the sensitivity analyses and some useful techniques like semi variogram or power spectra.
7. To work with sample data (point theme) is essential in spatial modeling. Therefore some techniques like ordinary kriging or spatial interpolation using Shepard method where integrated.
8. Additionally there are some 3D-views and analysis techniques like Splatter technique and Elevation view.
9. Different models are included: a simple cluster algorithm, a Kohonen feature map (Model). This models can be used as a basis for a radial basis function network (RBF). This type of neural network is useful for spatial modeling and can be a comprehensive alternative to simple regression approaches.
10. A set of special function are integrated. This includes edge detection algorithms (Sobel, Laplace); rotation of grid with an angle 90 deg, 180 deg, 270 deg; some point algorithms (edit point theme, transformation of a point theme to a grid) and last but not least special functions: flood fill algorithm, convolution etc.
11. It has a prototype to communicate with an external dynamical model via sockets (linked model)
12. Everything is controlled by an interactive graphical user interface
13. An new model component, the dynamic simulation system DESIRE was included. This system can be used stand alone like Matlab or it can be used as part of SAMT via an interface.

Simulation control

The simulation can be controlled using the menu bar and the tool bar. In the menu bar all operations are listed. To point to a entry a short help appears in the status line in the button of SAMT. The three inputs P1,P2,P3 are read line windows. They can be used to input parameter of a simulation or change options. Often has one operation different options. For example the histogram uses as default 250 bins. This can be changed by input the number of pins in P1. This technique was often used. Please keep care of the status line. In the status line is always printed a short message how to use the command. This short help line is for the experienced user.