Main content area

Slope instability analysis in South Patagonia applying multivariate and bivariate techniques on Landsat images during 2001–2015 period

Moragues, Silvana, Gabriela Lenzano, M., Moreiras, Stella, Vecchio, Andrés Lo, Lannutti, Esteban, Lenzano, Luis
Catena 2019 v.174 pp. 339-352
Landsat, glaciation, glaciers, lakes, landscapes, landslides, normal distribution, principal component analysis, remote sensing, ships, topographic slope, tourists, Argentina
We present results of the estimation of surface changes associated with slope instability processes on the Upsala Channel by remote sensing and statistic techniques. Hillslopes, involving lateral moraines, of Upsala Channel at the Argentino Lake have become potentially unstable due to the retreat of Upsala Glacier during the last century. The glaciation and deglaciation processes modify the tensions on the slope stability and may generate landslide processes. A landslide movement destroyed the western edge of the Upsala Channel in February 2013. In addition, the region represents a great tourist attraction usually ships navigate through the Upsala Channel seeing impressive landscapes, especially the Upsala Glacier, stand out. Due to this risky situation the necessity of increasing the knowledge in the area has arisen but has not been addressed yet. Therefore, the main goal of the present study is to investigate the multivariate statistical techniques by the principal component analysis (PCA) and discriminative analysis (DA) in four testing areas (TASn). Besides, we include Pearson's correlation coefficient (PCC) and normal distribution (ND) to determine whether the testing areas suffered slope movements or were temporally stable. The study is based on Landsat optical satellite images, acquired during the period from October 2001 to April 2015. The results show that TAS1,2,4 are less stable, whereas TAS5 is a more stable area as compared to another TAS. The PC1 and PC2 principal components explained the total variability of the 76% data. The total apparent error of DA reached 2.2% points. The PCC achieved a positive trend during the years when movement on the slope surface was not observed; while during the periods when slope instability was observed, the correlation showed a negative trend. The TASn that had shown a different behavior presented contrasted Gaussian bells; they are more flattened in those years with instability events.