Subfolder contain the training and validation data that went into the final logistic regression models.
Input data is structured following the requirements of the LAND-SE R-language script written by Mauro Rossi and Paola Reichenbach:
Rossi, M., and Reichenbach, P., 2016, LAND-SE: a software for statistically based landslide susceptibility zonation, version 1.0: Geoscience Model Development 9, p. 3533-3543.
----------------
In each province folder there is:
modelParameters.txt
±finalModelParameters.txt
TrainingData subfolder
InternalValidation subfolder
ExternalValidation subfolder
-----------------
modelParameters.txt:
Contains the coefficients and coefficient statistics generated by the R glm() function using the family = binomial() specification. Tables include the coefficient estimate, the estimate standard deviation, and the Wald test statistics and Wald test p-value. This text file contains the direct output of the glm() function, which in the Mogollon-Datil and Southern Rocky Mountains areas is not the final parameterization used in the province susceptibility maps.
finalModelParameters.txt:
for the Mogollon-Datil and Southern Rocky Mountains areas, the slope angle parameter was treated as a categorical variable with a threshold minimum coefficient applied to steep slopes. This text file contains the parameterization used for generating the final province susceptibility maps.
TrainingData subfolder:
Contains shapefile of sample point locations keyed to a text table of sample point data. Output ROC plot from the LAND-SE script provided, as is the histogram of landslide-area model probabilities. Note that the training data was specifically stuctured to provide 50/50 landslide points and non-landslide points.
InternalValidation subfolder:
Contains shapefile of sample point locations keyed to a text table of sample point data. Output ROC plot from the LAND-SE script provided, as is the histogram of landslide-area model probabilities. "Internal" validation refers to the validation area being the original training area, but with randomly located sample points. Internal validation was determined from the final model parameters in the case of the Mogollon-Datil and Southern Rocky Mountains.
ExternalValidation subfolder:
Contains shapefile of sample point locations keyed to a text table of sample point data. Output ROC plot from the LAND-SE script provided, as is the histogram of landslide-area model probabilities. "External" validation refers to the validation area being a validation area separate from the training area, with randomly located sample points. External validation was determined from the final model parameters in the case of the Mogollon-Datil and Southern Rocky Mountains.
--------------------
Validation shapefiles:
MOD_PROB field - calculated model probability
MOD_MATCH field - numeric code for match between model probability prediction and actual landslide deposit occurrence. Threshold for "landslide predicted" is 0.5 for this comparison.
1 - No landslide predicted, no landslide deposit at location. (True negative)
2 - Landslide predicted, no landslide deposit at location. (False positive error)
3 - No landslide predicted, landslide deposit at location. (False negative error)
4 - Landslide predicted, landslide deposit at location. (True positive)