Landslide state of activity maps by combining multi-temporal A-DInSAR (LAMBDA)

In this paper, a new methodology was developed to automatically update Landslide state of Activity Maps by
combining multi-temporal A-DInSAR data (LAMBDA). LAMBDA procedure was tested using ERS-1/2
(1992–2000), Radarsat-1/2 (2003–2009) and COSMO-SkyMed data (2011–2014) over an area of 2199 km2
located in Alps context of Piedmont region (north-western Italy). For the first time, a multidimensional landslide
activity matrix was implemented to update the landslide state of activity during the monitored time span. For the
definition of the state of activity, the representative velocity of each landslide was divided by the standard
deviation of the velocities along the slope of the whole dataset. Thus, a common stability threshold of±1 was
introduced for multi-sensors A-DInSAR data, allowing to distinguish a phenomenon with stable targets (PS-DS)
or unstable PS-DS. By combining activity classes estimated during different time spans allows to determine if a
phenomenon is active, reactivated, or dormant. Furthermore, an innovative confidence degree assessment was
carried out to verify the reliability of the procedure, by considering the measuring points distribution and the
variability of the movements for each landslide. The results were validated using the landslide inventory of the
study area and in-situ monitoring systems for representative case studies. Thanks to this approach an updated
state of activity until 2014 was assigned to 507 landslides out the 1657 which were previously mapped in the
study area.