a review on different methodologies employed in current swe products from spaceborne passive...
TRANSCRIPT
A review on different methodologies employed in current SWE products from spaceborne passive
microwave observations
Nastaran Saberi, Richard KellyInterdisciplinary Center on Climate Change (IC3) and Department of Geography and
Environmental Management, University of Waterloo, ON, Canada
Outline
• Introduction– Snow physical properties retrieval using
passive microwave observations• Emission Modeling (HUT, MEMLS, DMRT)
• SWE Products & Validation Process
• Research questions and summary
Why measuring snowpack physical
properties?
How?
Which properties?
Snow Properties Retrieval Using Passive Microwave RS
• Passive microwave observationsBrightness temperature: TB
– Appropriate frequency channels:10 GHz-19GHz-37GHz
X , Ku, Ka (IEEE)
o Goal: Modeling microwave-medium
interactions to retrieve snow physical propertiesAMSR2 instrumenthttp://www.drroyspencer.com/2012/05/
Snow properties retrieval
Empirical approaches
SD & TB differences
Emission modeling
Physical
DMRT-ML
Semi-empirica
l
MEML
S
HUT
(Matzler et al., 1982)
(Wiesmann and Matzler, 1999)
Snow Properties Retrieval Using Passive Microwave RS
Emission Modeling
http://www.iup.uni-bremen.de/iuppage/psa/members/MariaHoerhold.php
HUT snowpack emission model Pulliainen et al. (1999)
Semi-empirical model, adapted for remote sensing observations
1
2
3
d
4
MEMLS Snowpack emission modelA. Wiesmann and C. Matzler (1999)
A semi-empirical six-flux radiative transfer model
For a multi layered snowpack
MEMLS inputs (for each layer of snowpack) o Grain size (correlation length)o LWCo Temperatureo Deptho Density
For substratum, reflectivity and temperature is needed
DMRT-MLLeung Tsang et al. (2000) , Picard et al. (2012)
QCA-CP
DMRT (ks, ke)
DISORT (RT)o Mono disperse, stickinesso Poly disperse, Rayleigh
DMRT-ML inputs (for each layer of snowpack) o Grain size (Optical)o Temperatureo Deptho Densityo Substratum model
Emission Models Summary
• Differences HUT, MEMLS, DMRT-ML:
– Radiative transfer solution
– Wave propagation– Substratum – Representation of a
snow grain
Empirical Physical
Complexity
Sensitivity Analysis
Simplicity
Calibration
Parsimonious modeling: physics-based | semi-empirical | empirical
Snow Depth & Snow Water Equivalent Products
Snow Depth and Snow Water Equivalent Products
SWE estimation using in-situ data =>sparse observing networks
• Data assimilation • Reanalysis snow cover using land surface or
snow modelsProblem: dependency on precipitation data
• Satellite passive microwave derived SWE datasetsChallenge: complex topography
Glo
bSno
w -
SW
E
AMSR2 Snow Depth and Snow Water Equivalent
Product by R. Kelly
V1 & V2 :Predicated on the AMSR-E algorithm (2003/4)
V1: Regression based, came in response to deficiencies in static algorithm (Kelly ,2009).
V2: Physical modeling based (kelly, 2003)
V2Grain size & Density are dynamic
Forest correction is model-based
Atmospheric correction
Lake ice addressed
RFI determination (10 Ghz)AMSR
2 - S
WE
V2
Snow Detection based on history of snow
Snow Depth/SWE Retrieval using DMRT-ML
Methods that Use Machine Learning Techniques
Neural network
Emission model: HUT/MEMLS/DMRT
Training ProcessTB<->SD/SWE
Passive microwave observation (TB)
Inversion by NNSD/SWE
Training dataset
TB 19V
TB 37V
SWE/SD
MicroWave Radiation Imager (MWRI) on Feng-Yun 3
Grass Land Bare Soil
Farm Land Forest
SD=fgrass×SDgrass+ fbarren×Sdbaren+fforest×Sdforest + ffarmland×SdfarmlandSDfarmland=4.235+0.432×d18h36h+1.074×d89v89h;
Validation Process
Validation ProcessGeneral overview of SWE dataset assessment by SnowPEx
Statistical Assessment Tools
Summary & Research Questions!
Challenges in gathering In-situ data
• High spatial variability in snow physical properties and limitation in accessibility
=> Quantifying errors
In electromagnetic modeling
• Adaptation of physical model to metamorphic processes and a layered snowpack structure, also adapting to spaceborne scale => Key emission controllers of seasonal snow evolution
Thank you
Aknowledgements: Karem Chokmani