Logistic regression models utilize odds ratios for independent factors included in the model - factors that significantly impact the likelihood of gully erosion can be interpreted with their odds ratios.

What program are you using to run the regression?

A couple of our papers discussed some of the conditions we thought contributed to susceptibility, in my researchgate list. Certain soils are highly susceptible, uncolidated or highly weathered soils, such as saprolite, marine deposits, certain chemical imbalance soils. Soil piping and freeze-thaw expansion were other indicators. Soils may have highest importance. Areas with high intensity and magnitude rainfall, where water concentrates. Areas with soils exposed or weak soil cover, low density roots. Bottomland or stream channels where hydrologic modifications as channelization or dredging, or below agricultural terraces or mining ditches which have failed. Areas where concentration of flow and erosion has gone unchecked for many years. Areas adjusting to downstream base level or sediment load changes, such as inchannel mining, or clean water delivery effect below dams.