Retrieval of Snow water equivalent SWE Eirik Malnes Contributions from: Norut, IMGI, HUT and IFAc

1
Retrieval of Snow water equivalent (SWE)Eirik
MalnesContributions from Norut, IMGI, HUT and
IFAc
2
Background

• Parameter Snow water equivalent (SWE)
• Definition Equivalent amount of water for a unit
  area of snow with a certain depth
• State of the art Presently only measured with
  passive microwave instruments (coarse resolution,
  10 km)
• Motivation
• SWE is THE MOST IMPORTANT snow parameter for
  hydrological use
• The returned SAR-signal contains information
  about SWE
• Problems
• Returned radar signal is a complex function,
  where SWE is one of several parameters
  contributing
• Dry snow Low sensitivity to SWE at C-band
  frequencies
• Wet snow C-band radar waves is attenuated in the
  surface layer of the snow pack. SWE can not be
  retrieved.

3
EnviSnow development

• Repeat pass interferometry
• Backscatter sensitivity to SWE at multiple
  frequencies
• Improved passive micro wave algorithms

4
Norut Repeat pass C-band interferometry

• Objective
• Measure phase change (i.e. increased T/R time)
  due to radar wave propagation in dry snow.
• Interferometric processing of repeated satellite
  pass (summer and winter) to measure phase change
• The phase-change is proportional to SWE
• Special Dk-processing to avoid ambiguities (phase
  wrapping)
• Demonstrations with ERS and Envisat

Summer Winter
5
Delta-K Interferometric processing
Corner reflectors for calibration
6
Results
The principle was demonstrated with Envisat ASAR
at Altevann, Norway in 2004. Verification using
field measurements
Altevann, Field measurements April 2005
SWE with SAR, March 2005
7
Correlation between field measurement and SAR
estimate
linear Pearson correlation coefficient 0.44
8
HUT Contribution

• SWE and snow depth in boreal forests
• Experiments using HUTSCAT (Helicopter borne
  scatterometer, CX-band, all-pol)
• Backscattering sensitivity to dry snow

9
HUTSCAT results X-band
Backscattering coefficient as a function of SWE.
X-Band, incidence angle 45 degrees off nadir.
Circle forest area, cross agricultural area.
Dashed line regression line including SWE 0 .
10
HUTConclusion

• SWE of dry or slightly moist snow has a rather
  small effect on the backscattering coefficient at
  C- and X-band for thin snow cover (SWE below 100
  mm).
• Retrieval of SWE from C- and X-band SAR data
  based on the backscattering coefficient seems not
  feasible.
• These data are more useful for determining
  snow-covered area (SCA).

11
IFAC contribution

• Retrieval of SWE of wet snow from SAR
  backscattering coefficient
• Retrieval of SWE from microwave space-borne
  radiometers

12
Retrieval of SWE in wide areas from space-borne
microwave radiometers

• Use of multifrequency data with an algorithm
  based on a multi-layer perceptrons ANN

• Retrieval from SSM/I data (19 37 GHz) over 12
  sites in Finland (By training only one network
  for all sites)  
• Different approaches for generating the training
  set
• simulated values of brightness temperature
• measured values

Retrieved versus measured SWE
13
Retrieval of SWE in wide areas from
space-borne microwave radiometers (cont.)

• Retrieval from AMSR-E data (10 -19 37 GHz H pol
  ) over a single site (Kautokeino - Norway, Lat.
  6901N Lon. 2304E )
• Ground measurements from meteo station
• AMSR-E Daily data collected in dry snow
  conditions From Dec. 03 to Feb 04.
• Training with experimental data (1 day for each
  week)

Retrieved versus measured Snow Depth
14
IMGI contribution

• Theoretical backscatter modelling
• Analysis of time series of backscattering
  coefficients and coincident ground-based snow
  measurements
• Analysis of temporal of backscatter changes in
  ASAR images from summer to winter and during the
  winter period.

15
ASAR backscatter sensitivity to dry snowLautash
valley (1100 m)

• Changes in the depth and SWE of a dry snow pack
  cause only small variations in C-Band ?,
• C-band backscattering signal is not suitable to
  retrieve SWE in the study region that is
  characteristic for large parts of the Alps

16
L-band repeat pass interferometry

• Work done in an ESA project revealed that L-band
  repeat pass interferometry may be used to
  retrieve SWE (i.e. same measurement principle as
  reported from Norut)
• L-band frequencies are favourable to C-band due
  to better coherence

17
Conclusions

• Not sufficient signal sensitivity to measure SWE
  using SAR backscattering at C- and X-band
  frequencies, particularly in forests
• Scientific progress to measure SWE with C-band
  repeat pass interferometric SAR. Principle
  demonstrated with ERS and ASAR
• Improvements in multi frequency passive mm wave
  algorithms applicable for wide areas
• Outlook
• Ku-band frequency and normal backscatter
  measurements are promising
• L-band repeat pass interferometry-promising
• C-band SAR Repeat pass interferometry should be
  pursued using improved sensors (Radarsat-2) since
  C-band SARs are the only operational sensors
  currently