Data Products for "Spectral Filtering of Multitemporal SAR Imagery for Multiscale Landslide Detection, Mapping, and Characterization"

This dataset contains SAR backscatter intensity ratio (IR) maps, landslide-detection outputs, spectral-analysis products, and reference data supporting the study of earthquake- and rainfall-triggered landslides in Hokkaido and Hiroshima, Japan. All products are derived from the Copernicus Sentinel-1 C-band SAR imagery processed in Google Earth Engine and analyzed using Fourier spectral filtering methods described in the associated publication. 

Study areas and events:

·       Hokkaido, Japan: 2018 Eastern Iburi earthquake on 6 September 2018. Study area: ~512 km².

·       Hiroshima, Japan: 2018 extreme rainfall event, 28 June–8 July 2018. Study area: ~655 km².

Dataset Contents

1. SAR-Derived Products

Sentinel-1 VV and VV+VH backscatter intensity ratio (IR) maps for four post-event stacking scenarios: 1-day/1-week (low-stacking), 1-month, 6-month, and 1-year (high-stacking). Both original (unfiltered) and spectrally filtered (lowpass) IR maps are provided for averaged orbit directions. Products are available for both the Hokkaido and Hiroshima study areas.

Format: GeoTIFF | Resolution: 10 m | CRS: WGS84 (EPSG:4326)

Units: Backscatter intensity ratio in decibels (dB)

Example: VH IR map for 1-year post-Hokkaido

• GEE JavaScript script: Hokkaido_VH_1year_IR.js
• GEE shareable link: https://code.earthengine.google.com/f7fcb4505517c62effe4cd3a4fd97d01
• Full instructions: README_GEE_Hokkaido.txt

2. Landslide Inventories

Modeled inventories: Binary landslide classification maps generated by applying uniform (90th percentile) and LULC-adaptive percentile thresholds to filtered IR maps in high-stacking scenarios. 

Published reference inventories used for validation are also included:

• Z19 inventory: Zhang et al. (2019), ~4,897 landslides, Hokkaido
• AJG18 inventory: Association of Japanese Geographers (2018), ~6,724 landslides, Hiroshima

Format: Shapefile (SHP) and KMZ | CRS: WGS84 (EPSG:4326)

3. Spectral Analysis

Computation of 1D and 2D Fourier power spectra from representative landslide and non-landslide subareas for high-stacking (1-year) and low-stacking (1-week or 1-day) IR maps in both study areas. Filter transfer functions (boxcar and hybrid boxcar + half-Gaussian) used for lowpass spectral filtering are also provided. 

MATLAB script (Figure 6): SpectralAnalysis_Fig6.m. Requires input GeoTIFFs from the Data folder and six custom MATLAB functions included in this repository.

Full instructions: README_SpectralAnalysis.txt

4. Reference Data

Ancillary geospatial layers used for analysis and validation:

• Land Use and Land Cover (LULC) maps clipped to each study area, from the ALOS High-Resolution LULC Map Products.
• Slope angle and slope aspect grids derived from the Copernicus Global Digital Elevation Model (30 m resolution).

Format: GeoTIFF | Resolution: 30 m | CRS: WGS84 (EPSG:4326)

Data Sources

• Copernicus Sentinel-1 SAR GRD Imagery by the European Space Agency (ESA), available via Google Earth Engine (https://earthengine.google.com).
• NASA NASADEM HGT v001, available via Google Earth Engine.
• ALOS High-Resolution LULC Map Products (JAXA EORC)(https://www.eorc.jaxa.jp/ALOS/en/dataset/lulc_e.htm).
• Copernicus Global DEM, available via OpenTopography (https://portal.opentopography.org/).
• Z19 inventory: Zhang S, Li R, Wang F, Iio A (2019), Characteristics of landslides triggered by the 2018 Hokkaido Eastern Iburi earthquake, Northern Japan. Landslides 16:1691-1708. https://doi.org/10.1007/s10346-019-01207-6
• AJG18 inventory: Association of Japanese Geographers Disaster Response Committee (2018), Heavy rains in July 2018 (Western Japan Heavy Rains). https://ajg-disaster.blogspot.com/2018/07/3077.html. Accessed 6/14/2025.