The impact of physical temperature on brightness temperature observations over snow for NASA

1
The impact of physical temperature on brightness
temperature observations over snow for NASAs
AMSR-E

• Richard Kelly
• Department of Geography University of Waterloo
• Ontario, Canada
• Marco TedescoCity College of New York - CUNYNew
  York, USA
• Thorsten Markus James FosterNASA/GSFC, USA

2
Observation There are high temporal frequency
variations in the brightness temperatures (and
therefore retrievals) at 36, 18 and 10
GHz. Question What controls/causes high
frequency (day to a few days) changes?
57.07N, 86.22E
3
Outline

• Simple theory
• Met station measurements
• AMSR-E observations
• Summary further work

4
Simple theoretical standpoint

• What controls the brightness temperature (Tb)
  variation from a snow-covered scene as observed
  by a spaceborne microwave radiometer?
• (1)
• Tbs is snow Brightness Temperature
• Tbv is vegetation (tree canopy) Brightness
  Temperature
• Gt is a atmospheric transmissivity
• Tbatm atmospheric brightness temp (up down)
  (assume negligible in this case)
• NB Tb responses are frequency dependent.

5

• What controls the brightness temperature (Tb)
  variation from a snow-covered scene as observed
  by a spaceborne microwave radiometer?
• Deconstructing previous expression
• (2)
• Ts is snow physical temperature
• Air temperature is the driver here and changes
  through time the snowpack thermal gradient is
  constantly adjusting.
• Sub-nivean temperature probably stable
• es is snow emissivity and related to bulk snow
  properties
• grain size, snow crystal packing, number of
  scatters in the path length SWE, water content
  free or bounding
• probably (?) buried vegetation effects too
• Tv is vegetation physical temperature
• ev is vegetation emissivity

6

• What is the role of Tv or Ts ?
• In the models, Tv and Ts are often equated or
  combined as the effective temperature, T0, where
• (3)
• T0 is also computed through (e.g.)
• (4)
• where Tair is the air temperature and Ts is the
  snow temperature.
• But, are there overlooked implications to these
  assumptions ?

7

• What do physical temperature measurements suggest?

8
CLPX Experiment Data

• Colorado 19-24 Feb. 2003
• 3 MSAs (25x25km)
• Each MSA had 3 ISAs (1x1km)
• Fraser ISA moderate snow accumulations denser
  forest fraction
• Rabbit Ears deep snow accumulations less dense
  forest fraction

9
Fraser Experimental Catchment MSA

• St. Louis Creek ISAs (forest and moderate snow)
  and LSOS site.
• SWEmean 189mm
• SWEs 55 mm
• Depthmean 80 cm
• Depths 20 cm

10
Rabbit Ears MSA

• Walton Creek ISA (moderate forest and deep snow)
• SWEmean 580 mm
• SWEs 115 mm
• Depthmean 189 cm
• Depths 55 cm

11
In situ measurements dense pine at CLPX LSOS
12
In situ measurements Rabbit Earsless dense
forest
13

• Summary of in situ measurements
• Scene Tbs are sensitive to (constituent surface)
  physical temperature.
• (Tv) Vegetation canopy temperature is likely
  affected by air temperature
• overall large fluctuations
• (Ts) Snow temperature at the near air-snow
  interface varies more than at near basal snow
  temperature.
• overall small fluctuations

14

• How might Tphys affect PM SWE retrievals?
• AMSR-E Observations

15

• Retrieval approaches based on R-T theory (Chang
  et al., 1987 1996)
• where a is a calibration coefficient and ff the
  forest fraction. If this is deconstructed
  further
• where es18 and es36 are snow emissivities at 18
  and 36 GHz respectively.
• Is SWE a function of To / Tv / Ts ?

16
AMSR-E Tbs
17
Tbs at adjacent CLPX ISA sites (separated by 8km)
18
Fraser St. Louis data

• Variations of surface temperature (Tair) and Tbs
  at 18V 36V

19
Variations of surface temperature (Tair)
Tb18V-Tb36V K
But which of these channels contributes most to
the variations?
20

• Variations of Tair match Tb variations (somewhat)
  at low frequencies but less at 36 GHz
• Fraser (St. Louis Creek), Colorado - dense tree
  cover.

10V GHz 18V GHz 36V GHz
21

• Again, variations of Tair match well variations
  at low frequencies and to some extent the 36 GHz
  
• Rabbit Ears, (Walton Creek), Colorado - dense
  tree cover.

10V GHz 18V GHz 36V GHz
22

• Summary
• What causes apparent fluctuations in the SWE
  estimates or Tb18-Tb36?
• Contribution of Tair to Tbs at lower frequencies
  is greater than higher frequencies
• Surface temperature-related effects (driven by
  air temps) are a likely cause of Tb fluctuations
• Vegetation temperatures are likely to change with
  air temperature
• Vegetation emissivity changes are small
  (excepting snow in the canopy)
• Snowpack temperature variations Ts are not a
  likely cause
• Ground temperature/emissivity variations are not
  a likely cause
• Snow emissivity changes in response to punctuated
  snowfall events and seasonal snowpack evolution
  but not at the time scale under consideration.

23

• Conclusions Further Work
• We are looking at correcting for Ts Tv in the
  retrievals.
• Can we estimate Tair from AMSR-E? (synergy w/
  John Kimball). If achievable, Tair could be used
  to help drive a snowpack stratigraphy model
  (information needed in retrieval
  parameterization).
• Other sites under test (Canada tundra and Boreal
  forest Russia).
• A simple fix could be to ratio Tb18/Tb36 rather
  than subtract Tb18-Tb36
• Validation of current version is in progress for
  Sept 2008 - refinement activity will follow.

24
(No Transcript)
25

• Rationale
• Retrieval approach is often snapshot in scope
• Algorithms generate coarse-resolution SWE
  estimates at 25 x 25 km
• Uncertainties in the estimates are related to
  algorithms and spatial resolution

Monthly average
26
In situ measurements open pine at CLPX LSOS