Hello ,
I am doing my M.Sc. thesis on landslide susceptibility mapping using GWR method (ArcGIS 10). My slope unit polygon map has total number of 2943 polygons in which landslides only occupied in 472 polygons. I have calculated the landslide proportion (as %) in each polygon and added a new field with those data to represent dependent variable. Independent variables consist of Geology, landuse, slope, aspect, curvature etc. (please refer the slopeunit polygon shape file) , which were selected based on the dominated class of variable present in each slope unit polygon and assigned respective values of Frequency ratios calculated for each class of the variable using bivariate analysis- please also refer the table) .
My problem is when running GWR using all independent variables, message of "severe model design error" appears continuously, (I have run OLS prior to GWR but couldn't identify any variable with VIF >7) .However when omitting some of the independent variables it is possible to run GWR, but in every case( with different combinations of independent variable options) I got only R2 which is <40%. I cant find any missing variables either in my data set .
Please help me.. can anybody tell me how to improve R2 of the model. I just need to clarify is there any thing wrong with my data set (e.g. values I have used etc. ) I have prepared?
Thank you very much ,
Narmada
I am doing my M.Sc. thesis on landslide susceptibility mapping using GWR method (ArcGIS 10). My slope unit polygon map has total number of 2943 polygons in which landslides only occupied in 472 polygons. I have calculated the landslide proportion (as %) in each polygon and added a new field with those data to represent dependent variable. Independent variables consist of Geology, landuse, slope, aspect, curvature etc. (please refer the slopeunit polygon shape file) , which were selected based on the dominated class of variable present in each slope unit polygon and assigned respective values of Frequency ratios calculated for each class of the variable using bivariate analysis- please also refer the table) .
My problem is when running GWR using all independent variables, message of "severe model design error" appears continuously, (I have run OLS prior to GWR but couldn't identify any variable with VIF >7) .However when omitting some of the independent variables it is possible to run GWR, but in every case( with different combinations of independent variable options) I got only R2 which is <40%. I cant find any missing variables either in my data set .
Please help me.. can anybody tell me how to improve R2 of the model. I just need to clarify is there any thing wrong with my data set (e.g. values I have used etc. ) I have prepared?
Thank you very much ,
Narmada