Results in Transition Potential Modelling

[juda-garcia]

how interpret the results in the transition potential modelling? how to know the confidence intervals? in the technical information methods and algorithms does no t appear

[simgislab]

may be this will be of help http://wiki.gis-lab.info/w/Landscape_change_analysis_with_MOLUSCE_-_methods_and_algorithms#Logistic_regression

[KolesovDmitry]

Short answer

Unfortunately MOLUSCE doesn't provide confidence intervals. The results on your screenshot are taken by classic multinomial logistic regression. The MOLUSCE implementation of it is here: https://github.com/nextgis/molusce/blob/master/algorithms/models/lr/multinomial_logistic_regression.py

Long Answer

A pdf user-guide is provider with the plugin (see 'About' menu). More techical details are provided here: http://wiki.gis-lab.info/w/Landscape_change_analysis_with_MOLUSCE_-_methods_and_algorithms

MOLUSCE implements a lot of methods (they are different for different stages of analysis). Bellow are provided some references. The references are examples, you can prefer other authors ;)

Change Detection Stage

Methods: post-classification change detection - Singh 1989 Output:

• land use and forest cover change matrix
• land use and forest cover change maps

Modelling transition potential or areas at risk of deforestation

Input data:

• land use and forest cover change maps (e.g., 2000-2010);
• biophysical datasets e.g., distance to roads, slope etc.
• socio-economic data e.g., population density, crop production

Methods:

• artificial neural networks (ANN) - Pijanowski et al. 2002 and 2005.
• weights of evidence (WoE) - Bonham-Carter et al. 1988; Almeida et al., 2005.
• multi-criteria evaluation (MCE) - Wu and Webster 1998; Eastman et al 2005.
• logistic regression - Eastman et al. 2005; Mertens and Lambin 1997.

Output: deforestation risk maps

Simulating future land use and forest cover changes

Input data:

• base land use and forest cover map (e.g. 2000)
• transition potential or deforestation risk maps

Methods:

• cellular automata model - Wolfram 1984; Wu and Webster 1998.
• validation (quantity) using Kappa statistics - Pontius et al 2004.

Output: simulated future land use and forest cover changes

[juda-garcia]

@simgislab @KolesovDmitry Thank you. I'm going to check that links and bibliography. Can i translate your quick_help.pdf to spanish for Latin America users?

[simgislab]

sure, send the translation our way too when it's done