FOSS4G korea 2017 발표자료 Himawari-8, VIIRS, MetopA/B, GOES, Sentinel, Aqua등의 기상위성(Weather satellite) 데이터를 처리하기 위한 Python 기반의 Pytroll pakages를 소개하고 활용법에 대해서 이야기 한다.

1. 기상위성 자료 처리를 위한
Python 기반의 Pytroll 활용하기
가이아쓰리디㈜ | 박진우 (swat018@gmail.com)
2017. 08. 31

2. Park, Jin woo
• Foss4g korea 2013, 스마트폰 기반의 TS 측량 지원시스템 개발
• Foss4g korea 2014, 무인항공기(UAV)를 이용한 모니터링을 위한 통신모듈 시스템 설계 및 개발
• Foss4g 2015 seoul, Development of Opensource-based Photogrammetric UAV System Using Smart Camera
• Foss4g korea 2016, Smart Camera UAV를 이용한 오픈소스 기반의 자동 촬영 및 UAV image 처리 시스템 개발
• Foss4g korea 2017, 기상위성 자료 처리를 위한 Python 기반의 Pytroll 활용하기

3. 기상 위성
▣ GEO and LEO

4. 기상 위성
▣ GEO (Geostationnary orbit : 35786km)
- COMS / MI
- Himawari-8 / AHI
- GK-2A / AMI

5. 기상 위성
▣ LEO (Sunsynchronous orbit : 약 833km)
- S-NPP / VIIRS
- MetopA,B / IASI
- Aqua / AIRS

6. 기상위성 자료처리
기상위성 자료 처리를 쉽게 할 수 없을까??
1. 많은 위성 이름과 센서
- COMS / MI
- Himawari-8 / AHI
- GK-2A / AMI
- S-NPP / VIIRS
- MetopA,B / IASI
- Aqua / AIRS
…
2. 정말 다양한 포맷들…
- RIT/LRIT
- HDF5 & NetCDF - NWCSAF
- HDF4 - MODIS L1B
- HDF5 - VIIRS (SDR, Compact, L1B)
- HDF5 - FY3
- AAPP
- NOAA GAC
- Metop PDF
- SAR
- NetCDF/CF
- GeoTIFF
- NinjoTIFF
- PNG, JPeG
…
3. 영상처리, 익숙치 않은 용어들
- RGB 합성
- GSICS
- 복사보정
- 기하보정
- VIS 채널
- IR 채널
- Calibration
- IOT …
4. 알고리즘을 만들어서 바로 볼 수 없을까?
- Fortran
- R
- IDL
- Python

7. 기상위성 자료처리
GEO DN
GEO Tb [K]
GEO rad. [WN] LEO rad. [WN]
LEO Tb [K]
LEO rad. [WL]
Current GSICS Correction
GEO rad. [WL]
IASI, AIRS, Cris
Meteosat?
[mW m-2 sr-1(cm-1)-1]
Himawari, GO
ES-R?,
MODIS, VIIRS
A)
C)

8. Pytroll
• Pytroll은 파이썬 기반의 기상위성데이터를 읽고, 처리하고, 쓰는 작업 등의 전반적인 프로세싱을 지원.
• mpop, satpy 라이브러리를 이용하여 VIIRS data을 읽을 수 있다.
• Pyresample : 위성자료의 resample, reprojection을 지원.
• Pyspectral : Planck radiation, spectral bands, solar irradiance and reflectance 등의 처리를 지원.

9. Pytroll
• Pyresample *
• Mpop/Satpy *
• Mipp
• Pyorbital *
• Pycoast *
• Geotiepoints
• Trollimage
• Posttroll
• Trollcast
Pytroll packages
• Pytroll-schedule
• Python-bufr
• Pyspectral *
• Trollduction
• Python-db
• Trollshft
• ….

10. Mpop/Satpy
- High level processing for satellite data
- Indexing by name or wavelength
- Many built-in composites
- Easy to extend and customize
- Both GEO and LEO

11. Mpop/Satpy
>>> from mpop.satellites import GeostationaryFactory
>>> import datetime
>>> time_slot = datetime.datetime(2009, 10, 8, 14, 30)
>>> global_data = GeostationaryFactory.create_scene("meteosat", "09",
"seviri", time_slot)
>>> from mpop.satellites import PolarFactory
>>> import datetime
>>> time_slot = datetime.datetime(2010, 2, 24, 11, 29)
>>> global_data = PolarFactory.create_scene("noaa", "19", "avhrr", time_slot,
orbit="05204")
mpop

12. Mpop/Satpy
>>> global_data.load([0.6, 0.8, 10.8])
>>> global_data.image.overview().show()
mpop example

13. Mpop/Satpy
>>> global_data.load([0.6, 0.8, 10.8])
>>> global_data.image.overview().show()
mpop example

14. Mpop/Satpy
>>> from glob import glob
>>> from satpy.scene import Scene
# Load data by filenames
>>> files = glob(“/data/viirs_data/*.h5”)
>>> scn = Scene(reader="viirs_sdr", filenames=files)
# Load data by sensor
>>> scn = Scene(sensor="seviri", base_dir=“/data/HRIT/”)
Satpy

15. Mpop/Satpy
>>> scn.load(["overview"])
>>> scn.show(“overview”)
Satpy example

16. Mpop/Satpy
>>> scn.load(["overview"])
>>> rs_scn.show(“overview”)
Satpy example

17. Mpop/Satpy
Satpy example
# Automatically load composites and
their dependencies
>>> scn.load(["true_color"])
# Resample multi-band data to a uniform
grid
>>> rs_scn = scn.resample("euron1")
# Save RGB geotiff
>>> rs_scn.save_dataset(“true_color”)

18. Mpop/Satpy
Satellite data formats
HRIT/LRIT
HDF5 & NetCDF - NWCSAF
HDF4 - MODIS L1B
HDF5 - VIIRS (SDR, Compact, L1B)
HDF5 - FY3
AAPP
NOAA GAC
Metop PDF
SAR
…
Output data formats
NetCDF/CF
GeoTIFF
NinjoTIFF
PNG, Jpeg

19. Pycoast
Adding shapes to images
• Adding coastlines
>>> cw = ContourWriterAGG('/home/esn/data/gshhs')
>>> cw.add_coastlines(img, (proj4_string, area_extent), resolution='l',
width=0.5)
• Adding rivers, borders
>>> cw.add_rivers(...)
>>> cw.add_borders(...)
• Adding graticules
>>> cw.add_grid(...)

20. Pycoast

21. Pyorbital
computing satellite orbital parameters and reading TLE’s
>>> from pyorbital.orbital import Orbital
>>> from datetime import datetime
>>> orb = Orbital("noaa 18")
>>> now = datetime.utcnow()
# Get normalized position of the satellite:
>>> orb.get_position(now)
([0.57529384846822862, 0.77384005228105424, 0.59301408257897559],
[0.031846489698768146, 0.021287993461926374, -0.05854106186659274])
# Get longitude, latitude and altitude of the satellite:
>>> orb.get_lonlatalt(now)
(-1.1625895579622014, 0.55402132517640568, 847.89381184656702)

22. Pyorbital
Uses SGP4 algorithm
(SGP4 algorithm: 궤도결정 알고리즘을 해석적 모델인 SGP4 모델)

23. Pyorbital
Use of sun zenith angles

24. Pyspectral
Because satellite remote sensing involves Planck radiation, spectral bands,
solar irradiance and reflectanc
- Reads sensor relative spectral responses (RSR)
- Reads solar irradiance spectra

25. Pyspectral
Triggered by the need to derive 3.x reflectances:
E.g. daytime microphysical RGBs

26. Pyspectral
>>> from pyspectral.rsr_read import RelativeSpectralResponse
>>> from pyspectral.solar import SolarIrradianceSpectrum
>>> from pyspectral.solar import TOTAL_IRRADIANCE_SPECTRUM_2000ASTM
>>> modis = RelativeSpectralResponse('eos', '2', 'modis')
>>> modis.load(channel='20', scale=0.001)
>>> solar_irr =
SolarIrradianceSpectrum(TOTAL_IRRADIANCE_SPECTRUM_2000ASTM,
dlambda=0.005)
>>> sflux = solar_irr.inband_solarflux(modis.rsr)
>>> print("Solar flux over Band: ", sflux)
('Solar flux over Band: ', 2.002927764514423)
example

27. Pyspectral
>>> from pyspectral.nir_reflectance import Calculator
>>> sunz = 80.
>>> tb3 = 290.0
>>> tb4 = 282.0
>>> refl37 = Calculator(modis.rsr, solar_flux=sflux)
>>> print refl37.reflectance_from_tbs(sunz, tb3, tb4)
0.251177702956
example

28. Pyspectral
Use MPOP to load SEVIRI data
Use Pyorbital to get sun zenith angles at SEVIRI pixels
 code
>>> r39 = refl39.reflectance_from_tbs(sunz, ch39.data, ch11.data,
lookuptable='/tmp/seviri_37_tb2rad_lut.npz')
MPOP, Pyorbital and Pyspectral

29. VIIRS의 영상처리 사례
- Loading
- Making RGB’s
- Map projection

30. VIIRS의 영상처리 사례
1. Loading Data (mpop)
>>> from mpop.satellites import PolarFactory
>>> from datetime import datetime
>>> time_slot = datetime(2013, 6, 25, 11, 15)
>>> orbit = "08599"
>>> global_data = PolarFactory.create_scene("Suomi-NPP", "", "viirs", time_slot, orbit)
>>> global_data.load([1.38])
>>> print global_data
-------> print(global_data)
'M01: (0.402,0.412,0.422)μm, resolution 742m, not loaded'
'M02: (0.436,0.445,0.454)μm, resolution 742m, not loaded'
'M03: (0.478,0.488,0.498)μm, resolution 742m, not loaded'
'M04: (0.545,0.555,0.565)μm, resolution 742m, not loaded'
'M05: (0.662,0.672,0.682)μm, resolution 742m, not loaded'
'M06: (0.739,0.746,0.754)μm, resolution 742m, not loaded'
'M07: (0.846,0.865,0.885)μm, resolution 742m, not loaded'
'M08: (1.230,1.240,1.250)μm, resolution 742m, not loaded'
'M09: (1.371,1.378,1.386)μm, shape (768, 3200), resolution 742m'
'M10: (1.580,1.610,1.640)μm, resolution 742m, not loaded'
'M11: (2.225,2.250,2.275)μm, resolution 742m, not loaded'
'M12: (3.610,3.700,3.790)μm, resolution 742m, not loaded'
'M13: (3.973,4.050,4.128)μm, resolution 742m, not loaded'
'M14: (8.400,8.550,8.700)μm, resolution 742m, not loaded'
'M15: (10.263,10.763,11.263)μm, resolution 742m, not loaded'
'M16: (11.538,12.013,12.489)μm, resolution 742m, not loaded'
'I01: (0.600,0.640,0.680)μm, resolution 371m, not loaded'
'I02: (0.845,0.865,0.884)μm, resolution 371m, not loaded'
'I03: (1.580,1.610,1.640)μm, resolution 371m, not loaded'
'I04: (3.580,3.740,3.900)μm, resolution 371m, not loaded'
'I05: (10.500,11.450,12.300)μm, resolution 371m, not loaded'
'DNB: (0.500,0.700,0.900)μm, resolution 742m, not loaded'

31. VIIRS의 영상처리 사례
>>> img = global_data.image.channel_image(1.38)
>>> img.enhance(stretch='histogram')
>>> img.show()

32. VIIRS의 영상처리 사례
2. Making RGB’s (mpop)
>>> global_data.load(global_data.image.truecolor.prerequisites)
>>> img = global_data.image.truecolor()
>>> img.save("./viirs_truecolor.png")

33. VIIRS의 영상처리 사례
3. Map projection
>>> global_data.load(global_data.image.truecolor.prerequisites)
>>> img = global_data.image.truecolor()
>>> img.save("./viirs_truecolor.png")
- Use Pycoast
>>> from PIL import Image
>>> from pycoast import ContourWriter
>>> from mpop.projector import get_area_def
>>> cw = ContourWriter('/local_disk/data/shapes')
>>> img = Image.open('./viirs_truecolor_proj.png')
>>> area_def = get_area_def("scan500m")
>>> cw.add_coastlines(img, area_def, resolution='i', level=3)
>>> img.save('./viirs_truecolor_proj_with_overlay.png')

34. VIIRS의 영상처리 사례
3. Map projection

35. Channel arithmetics
>>> ndvi = (local_data["2"] - local_data["1"]) / (local_data["2"] + local_data["1"])
>>> ndvi.show()

36. Pytroll
- 현재 지원하고 있는 기상 위성
•Meteosat series (tested with 7, 8, 9, 10)
•GOES series, in HRIT/LRIT format (tested with 11, 12, 13, 15)
•MTSAT series, in HRIT/LRIT format (tested with 1R, 2)
•Himawari 8, in HRIT/LRIT format
•Himawari 8, standard format (satpy only)
•Electro L, in HRIT/LRIT format (tested with N1)
•NOAA series, in AAPP, GAC and LAC format (tested with 15, 16, 17,
18, 19)
•Metop-A/B, in EPS 1a and 1b format
•Aqua and Terra, in hdf-eos format
•Suomi NPP, in SDR hdf5 format
•TerraSAR-X
•Radarsat-2 SAR
•COSMO-SkyMed SAR
•Sentinel-1 SAR
•Sentinel-2 MSI
•FY-3 viir

37. THANK YOU
swat018@gmail.com