LANDSLIDE SUSCEPTIBILITY MODELING WITHIN GLACIOMARINE PRESUMPSCOT FORMATION, SOUTHERN MAINE

A landslide susceptibility map of southern Maine was created using probability based statistical analysis and the ArcGIS ModelBuilder software. Suspected landslide locations were identified using aerial photo interpretation and compiled on 7.5' landslide inventory maps of portions of southern Maine. A Geographic Information System database was compiled comprised of raster and vector data including: a 10 meter Digital Elevation Model, surface water features mapped at a 1:24,000 scale and the suspected landslide locations converted to digital data points.

The data were used to calculate the relationships among the suspected landslide locations and factors that contributed to slope instability. These factors included: distance to the nearest dynamic water body, slope, slope aspect, and the presence of the glaciomarine Presumpscot Fm. These factors were given equal weighting and the percent probability of a landslide occurring near them was determined. Using these calculations, an automated model to calculate slope instability was created in ModelBuilder and extrapolated to the rest of southern Maine. Data created by the model were then compiled into landslide susceptibility maps identifying zones as: “not probable,” “lower probability” or “higher probability”.

The output indicates that zones rated as “higher probability” are much more likely to occur in areas that are: underlain by the Presumpscot Fm., affected by marine processes, or lie in close proximity to streams and tributaries. Zones identified as “lower probability” are defined by the presence of Presumpscot Fm. and are restricted to those locations that are underlain with it. Zones identified as “not probable” occur infrequently within Presumpscot Fm. and account for the entirety of areas located outside of it.

The maps do not take into consideration developed areas, depth to bedrock data or temporal data that would help make distinctions based on seasonality. Such data would help refine the model and allow for more accurate interpretation of the area which could be incorporated into the model.