It was not easy to me to understand completely your question. About ROC curve, AUC is a number ranging from 0.5 (random classification) to 1 (perfect classification) that measure the area under the ROC curve and it is considered as the probability that the classification is good.

I use XLSTAT software to apply logistic for landslide susceptibility mapping. The input data consists in values of 1 for landslide areas and values of 0 for non-landslide area, along with the values of predictors (such as slope, land use etc.). The logistic regression module computes an equation which can be used to map landslidesusceptibility in GIS environment. Also, for each landslide (1) and non-landslide (0) points the equation is used to compute predicted landslide susceptibility values. Generraly, if predicted value is greater than 0.5, then the respective point is classified as landslide point. The number and percentages of correctly classified points are given in a classification table, also known as confusion matrix. The ROC curve and area under the curve (AUC) are also generated by logistic module. Generally, if AUC is greater than 0.7, the model is considered good. For more information, you can check out some of my publications of this topic on my research gate page.

Thanks for Cristian Patriche’s explanation. Currently I know how to get AUC.

By the way, as you know, in GIS, like slope, aspect is the raster type. I am wondering how to change into the digital number to use in some statistical software, such as SPSS, SAS? Maybe ASCII value? It sounds that.

So far I am reading your paper.

If I understand correctly, you want to extract values from raster layers and export them into a statistical software. My approach is to generate random points inside and outside landslide area. For this purpose I use Hawths Tools which can run under ArcGIS. In order to extract the values, I also use an option from Hawths Tools. So finally you will get a table (dbf) containing the predictors' values for the point samples which can be opened in a statistical software.

Dr Cristian Patriche, very happy to receive your reply. Your understanding is quite right. Do you know how use Arcgis to extract into the values?

In ArcGIS follow the path: Spatial Analyst Tools - Extraction - Extract Values to Points. The module will create a new shapefile with a new column into the attribute table containing the extracted values. However, the module allows only one extraction at a time. So if you want to extract values from say 3 layers, you must run the module 3 times. An alternative is to use Hawths Tools which allows you to extract from multiple layers at once. It is a free tool which can be downloaded from http://www.spatialecology.com/htools/ at it runs under ArcGIS. The path to follow if using this tool is : Hawths Tools - Analysis Tools - Intersect Point Tool

Dr Cristian,

Very appreciated your help. Later I will try.

thanks so much.