Naru Tsutsumida

presentation file for FOSS4F 2017 KYOTO.KANSAI

Transcript

1. ࣌ܥྻӴ੕ը૾σʔλॲཧπʔϧͷ঺հ Narumasa Tsutsumida అా੒੓ Assist. Prof. Graduate School of Global Environmental Studies, Kyoto University, ژ౎େֶ஍ٿ؀ڥֶಊɹॿڭ OSGeo Charter member 1
2. 2 Contents • Time series raster data processing by • Cloud based tools (Google Earth Engine) • Local tools (GDAL, R,…)
3. My interests Earth Observation GIScience Land Use / Land Cover studies
4. Earth Observation approach Tsutsumida et al (2013) Land, 2, 534-549 Tsutsumida et al (2016) Remote Sensing, 8(2), 143 Urban expansion monitoring Ulaanbaatar Jakarta
5. GIScience approach Tsutsumida et al (2016) Towards FUTURE EARTH: Challenges and Progress of Global Environmental Studies, Kaisei Publishing, pp.61-79 Spatio-temporal overall accuracy
6. 6 Massive remote sensing open data are now freely available! • Landsat • Sentinel • MODIS … https://landsat.usgs.gov/landsat-archive https://modis-land.gsfc.nasa.gov/MODLAND_grid.html
7. 7 Massive remote sensing open data Processing Download Mosaic Reprojection Analysis Output wget (shell) GDAL ModisTool … GRASS Python R C C++ … save time and resources to focus on your main task! Define ROI Your interest Data selection
8. If you are interested in using open RS data • No need to consider about RAM / Disk. • No need to download raw data. Google Earth Engine
9. Google Earth Engine
10. • Define a target area • Set a data product • Data DL→Format conversion ʢHDF→TiffʣˠMosaic→… RS data processing in Google Earth Engine
11. DEMO: Time series changes in MODIS NDVI average at global scale Is NDVI average stable over time? Not specific yet = global in 2001-2016 MODIS Download Mosaic Reprojection Analysis Output Define ROI Your interest Data selection GEE GEE GEE Whatever you want Whatever you want 460*24*16=176,640 hdf files…
12. Javascript-based Playground / Python API https://github.com/naru-T/2017FOSS4GHOKKAIDO https://code.earthengine.google.com/19cb8f156684edefc1c7f5ff005c65df
13. Local tools •gdal_calc.py •R (raster) •Python (rasterio) •ɾɾɾ Analysis (Parallel or loop) Aggregation Prepared large raster data Split large raster into patches Output rasterOptions(maxmemory = 1e+10) rasterOptions(tmpdir = 'tmpfolder') gdalbuildvrt gdal_translate qgis
14. NEW tools: geodocker/geonotebook/geopyspark https://github.com/OpenGeoscience/geonotebook https://github.com/geodocker/geodocker-jupyter- geopyspark
15. Thank you contact: naru@kais.kyoto-u.ac.jp, @naru-T 15