1. Karthaus, September 2005 Wouter Greuell IMAU, Utrecht, NL -Why? -Cloud masking -Retrieval method -An application: estimate surface mass balance from satellite data REMOTE SENSING - RETRIEVAL OF THE SURFACE ALBEDO

2. WHY ? (Surface) albedo =fraction of short-wave radiation reflected by the surface Important for surface mass balance: 1)Net short-wave radiation dominates energy balance 2)Large variations 3)Positive feedback between albedo and mass balance Spatial variations cannot be measured at surface Satellite data But: conversion of measured signals into albedo can cause considerable errors

3. Spectral response: no block shape ! AVHRR band nr Wavelength range (µm) 10.58 - 0.68 20.73 - 1.10 33.55 - 3.93 410.3 - 11.3 511.5 - 12.5 SATELLITE NARROWBANDS

4. DATA PROCESSING -geolocation -cloud mask -conversion of signal into albedo

5. Difficult over snow and ice Reason: albedo of clouds is similar to that of the surface temperature of low clouds is similar to that of the surface Conclusion: sensor with band around 1.64 µm most suitable But: no such sensor available for AVHRR and MISR Method: use (multiple) thresholds CLOUD MASK

6. Why? Because of post-launch degradation of sensors. Two methods 1)On-board: by measurement of solar radiation reflected from a panel with known reflectance 2)Ground target by measurement of radiance reflected by a ground target with constant optical properties (e.g. desert, dry-snow on polar ice sheets) CALIBRATION OF SATELLITE SENSORS

7. UNCERTAINTY IN CALIBRATION

8. Relevant processes: Absorption: mainly by ozone and water vapour; little by aerosols Scattering: gas molecules (Rayleigh scattering), aerosols ABSORPTION AND SCATTERING - ATMOSPHERE

9. is a function of: aerosols, H 2 O, O 3 solar zenith angle (  s ) view zenith angle (  v ) elevation (z) Aim:calculate the surface albedo (  s ) as a function of the planetary albedo (  p ) Usual approach: calculate  p with a radiative transfer model for a (large) number of  s, aerosol loads, H 2 O amounts, etc. Store results in a Look Up Table (LUT), which is used during the retrieval. ATMOSPHERIC CORRECTION

10. REFLECTION VARIES WITH VIEW ANGLE

11. isotropic reflection: intensity of reflected radiation does not vary with the direction of reflection Lambertian surface reflects radiation isotropically Snow and ice reflect radiation anisotropically Why important ? Satellite measures reflected radiation from a single direction (radiance), but we are interested in all the reflected radiation (flux) Correction needed! Directional distribution of the reflected radiation is described by functions called Bi-directional Reflectance Distribution Functions (BRDFs) ANISOTROPIC REFLECTION AT THE SURFACE

12. CO-ORDINATE SYSTEM BRDF  s  solar  zenith angle      zenith angle     azimuth angle

13. Nadir = Straight down Zenith view angle of 75˚ (90˚ = horizontal)  = 0˚ Azimuth direction of sun = backward scattering  = 180˚ forward scattering TM2 (green light); solar zenith angle =50˚; albedo = 0.52 EXAMPLE OF A MEASURED BRDF

14. SPECTRAL ALBEDOS AND NARROWBANDS

15. narrowband: spectral band in which the satellite sensor is sensitive to radiation broadband: entire solar spectrum would not be necessary if the albedo was constant with wavelength would have been easy if the narrowbands covered the entire broadband:  : broadband albedo  i : the narrowband albedo in band i S i : incoming solar radiative flux in band i. NARROWBAND TO BROADBAND (NTB) CONVERSION

16. BAND ALBEDO MEASUREMENTS

17.  bb = 0.539  TM2 + 0.166  TM4 ( 1 +  TM4 ) Solution: develop equations from simultaneous and coincident measurements of narrowband and broadband albedos EQUATION FOR NTB CONVERSION

18. converts satellite counts into surface albedo RETRIEVAL METHOD

19. Weakest processing steps retrieval method: 1)insufficient knowledge BRDFs of snow and ice 2)uncertainty in calibration coefficients More accurate processing steps of the retrieval method: 1)NTB conversion 2)atmospheric correction in relatively dry and clean polar and high-mountains atmosphere ACCURACY RETRIEVAL METHOD

20. VALIDATION OF SATELLITE-DERIVED ALBEDO

21. SensorTMAVHRRMODISMISRASTER full name Thematic Mapper Advance Very High Resolution Radiometer MODerate resolution Imaging Spectroradio meter Multi-angle Imaging SpectroRadio meter Advanced Spaceborne Thermal Emission and Reflection radiometer resolution at nadir 30 m1.1 km 250 to 1000 m 275 to 1100 m15 to 90 m coverage (equator) once per 16 days once per day every 1 to 2 days once per 9 days sensors are pointable data since 1982 (MSS since 1972) 19782000 number of channels 7536414 remarks measurements from 9 angles VARIOUS SATELLITE SENSORS

22. Surface albedo derived from AVHRR data western margin of the Greenland ice sheet 16 July 1995 SURFACE ALBEDO GREENLAND

23. ESTIMATE SURFACE MASS BALANCE FROM SATELLITE-DERIVED ALBEDO

24. BASIC IDEA OF THE METHOD B:mean annual surface mass balance :potential absorbed short-wave radiative flux I 0 :extraterrestrial incoming short-wave radiative flux  :satellite-derived surface albedo (glacier mean) day 1 - day 2:beginning and end of ablation season

25. WHY COULD METHOD BE SUCCESSFUL? lower albedo  more melt; absorbed short-wave radiation largest contribution to melt more melt  lower albedo less accumulation  lower albedo

26. CONVERSION INTO ENERGY BALANCE EQUATION AND THEN INTO MASS Add transmission of the atmosphere (  atm ): Add long-wave and turbulent fluxes (Q 0 ): Convert energy into mass (L f is latent heat of fusion):

27. ABSORBED RADIATION 1995

28. RELATION SATELLITE-DERIVED WITH MEASURED MASS BALANCE Mean 13 years: Measured: -1202 mm w.e. Satellite: -1142 mm w.e. Interannual variability (standard deviation): Measured: 366 mm w.e. Satellite: 395 mm w.e.

29. SUM UP -Why -Cloud masking -Retrieval method -An application: estimate surface mass balance from satellite data

30. CLOUD MASK (2) Scheme developed for AVHRR data Antarctica: multiple tests; if cloudy according to one test, then pixel is cloudy Band(s)Day / night Type of cloudsPhysics 3 (3.75 µ)daylowwater droplets have higher reflectivity than ice / snow 5 (12.0 µ)bothcold, highclouds have lower temperature than ice / snow 4 (10.8 µ)bothuniform clouds over heterogeneous surface and vice versa 4 - 5 and 3 - 5 bothcirruscloud transmissivity differs from band 4 to 5 4 - 3nightfog / low emissivity of water droplets differs with ; not for ice / snow 3 - 5nightpartially- covered pixels ?

31. CLOUD MASK (3) Comments: -a single test is not enough -each test has physical basis -threshold values are variable in space and are subjectively chosen -combination of tests is subjectively chosen -low clouds over snow and ice remain difficult to detect

32. ALBEDO OF A STABLE TARGET FROM AVHRR NOAA 11 NOAA 14 NOAA 16 Dry-snow area at top of Greenland ice sheet

33. EXPERIMENTAL SET UP TO MEASURE BRDFs pyrheliometers measure radiance pyranometers measure flux (irradiance)