The Atacama Desert in northern Chile is the driest place in the world. Nevertheless, it hasexperienced some rare, extreme rainfall events that have caused landscape changes. Rainevents of March 2015 and July 1991 caused extensive damage in some locations. Attempts atspatially comprehensive detection of areas that change due to rainfall include use of InSARafter the March 2015 rain event in Antofagasta Province, Chile (Scott et al., 2017). In thiswork, we employ a new method to delimit the areas of important change in the physicalcharacteristics of the surface using LANDSAT imagery; one of the cases allowscomparison to the InSAR approach. The process is an easy, semiautomatic, andreproducible methodology that has the potential to identify the types of surfacechange. The method is based on a succession of filters that constrain the area ofchange due to extreme rain, while deleting areas of human activities and of shadowchange.The filter sequence starts with the identification of areas of human activity using two differentinformation sources: a map of historical mine disturbances based on visual inspection ofoptical satellite images, and the extreme differences between two before-rain LANDSATscenes. The latter are identified using the extreme Chi-square values produced using IrMAD(Nielsen and Morton, 2011), an unsupervised change-detection procedure which is also ableto characterize the degree of change. The two products are merged into one Human Activitypolygon (HA). Areas with high degrees of change in shadows are identified using simplehillshades (Shadow Changes polygon, SC). Then, we combine HA and SC to form a non-raineffect area (Non-Rain polygon, NR). Finally, with the NR polygon masked, weextract the high-value IrMAD results from a set of images for dates before andafter a rain event (Second Chi-Square polygon, SCS) to define the area of true raininfluence.Using the area of true change as a mask of the MAD (multivariate alteration determination)rasters, which corresponds to the variance in every spot in different wavelengthof the six bands (1,2,3,4,5,7 for the 1991 rain event and 2,3,4,5,6,7 for the 2015event), we performed an unsupervised classification in ArcGIS software, selecting 10classes.The unsupervised classification shows great potential for delimiting categories of change –erosion, deposition or moisture content - in the playa-lakes and channels. One class correlatestrongly to alluvial fans and colluvium-covered hillslopes. However, it is ambiguous whetherthe hillslope changes result from non-filtered shadow changes or real soil surface changes.Additionally, there are some classes that mainly show cloud change and LANDSATmisdetection areas.Nielsen, A. A, and Morton, C. J., 2011, A Method for Unsupervised Change Detection andAutomatic Radiometric Normalization in Multispectral Data: 34th International Symposiumon Remote Sensing of Environment (Sydney, Australia: ISPRS, 2011).Scott, C., Lohman, R., and Jordan, T.E., 2017, InSAR constraints on soil moisture evolutionafter the March 2015 extreme precipitation event in Chile: Scientific Reports, 7(4903),10.1038/s41598-017-05123-4.2. [ABSTRACT FROM AUTHOR]