TH3.TO4.1.pptx

Improving the AMSR-E snow depth product: recent developments

Richard KellyUniversity of Waterloo, Canada.

Outline

• Current product: performance of the satellite only product

• Development issues– RFI– Lake behaviour– Forest attenuation

• Observations and Conclusions

Current snow depth product:

Approach (AMSR-E observation-based)

where

fd=50%ff=40%

r=12.5km

fd=10%ff=40%

IFOV radius=12.5 kmForest stand

Performance (2002-2003)

comments

• Retrievals are conducted at the IFOV level (not at the grid-averaged level)

• Retrievals are then aggregated to coarser grid (0.25x025 deg)

• Tb correction (forest only) is built into the retrievals but not perfectly formulated.

• RFI and water bodies not well accounted for.

RFI issues

RFI

Kidd, C.(2006) 'Radio frequency interference at passive microwave earth observation frequencies’, International Journal of Remote Sensing,27:18,3853 — 3865

MODIS-Daily

2 Feb. 2009

3 Feb. 2009

4 Feb. 2009

5 Feb. 2009

6 Feb. 2009

7 Feb. 2009

8 Feb. 2009

9 Feb. 2009

10 Feb. 2009

From sat.dundee.ac.uk

RFI impacts

UK AMSR-EDaily Snow 2.2.2009

Missing / bad

Water

320 – 540

160 - 320

80 - 160

40-80

1 - 40

0

SWE [mm]

RFI impacts

10V GHz 10V-36V (current)

5 February 2009

RFI impacts

Adjusted retrievals without the use of the 10 GHz channel in UK.

Lake issues

• Much of the Arctic and sub-Arctic are dominated by lake cover (fractions of 40% not uncommon).

• At 25 km grid resolution lakes will play an important role

• But what is that role and can lake signals be identified?

• Time series of data for 9 years.

Kang et al. (2010) GRSL 7(4)

A note on gridding and AMSR-E measurement samples

10 km

10 k

m

AMSR-E sampling grid

25 km

LOI

Depending on the proximity threshold (x), may or may not be daily, but it will be for a consistent location

x

Water Fraction

0-10%

11-20%

21-30%

31-50%

>50%

Forest Fraction

0%

1-10%

11-20%

21-30%

31-50%

>50%

0 400 km

MOD44W Water mask & MODIS VCF25 km diameter circle averaged on 1 km grid

25km

Snow

dep

th (c

m)

Date

AMSR snow depth (0.25x0.25 deg gridded)

Nearest WMO snow depth / air temperatureSn

ow d

epth

(cm

)

Date

Mean Tem

p (C)

0% FF & 37% WF

0% FF & 35% WF

Tb (K

)

Date

AMSR Tbs (36V & 18V): centre within 4 km of locations of interest

AMSR Tbs (36V & 18V)

Lake freeze-up

Snow

accumulation

on lake

Snow and

lake ice melt

Forest fraction

Water Fraction

0-10%

11-20%

21-30%

31-50%

>50%

Forest Fraction

0%

1-10%

11-20%

21-30%

31-50%

>50%

0 400 km

MOD44W Water mask & MODIS VCF25 km diameter circle averaged on 1 km grid

Snow

dep

th (c

m)

Date

AMSR snow depth (0.25x0.25 deg gridded)

AMSR Tbs (36V & 18V): centre within 4 km of locations of interestTb

(K)

Date

Forest & lakes combined

Water Fraction

0-10%

11-20%

21-30%

31-50%

>50%

Forest Fraction

0%

1-10%

11-20%

21-30%

31-50%

>50%

0 400 km

And when you mix forest and lakes….

MOD44W Water mask & MODIS VCF25 km diameter circle averaged on 1 km grid

Snow

dep

th (c

m)

Date

AMSR snow depth (0.25x0.25 deg gridded)

4% Water fraction and 55% forest fraction

AMSR Tbs (36V & 18V): centre within 4 km of locations of interestTb

(K)

Date

Areas of development• RFI can be accommodated in the current implementation by a of 10 GHz

observations with 18 GHz in the affected areas• Water fraction changes the nature of a Tb difference or ratio. 18 GHz is

sensitive to lake ice thickness so that DTb is not necessarily a function of DSD or SWE.

• Forest fraction/structure reduces sensitivity (known). But it is likely not a linear forest fraction attenuation effect

• Forest and water fraction together require correction before SWE inversion is applied. Recent work by Langlois et al. (2011) has demonstrated new approach to transmissivity correction and lake fraction can likely be corrected for (ongoing work)

• Operational implementation of forest and lake correction is yet to be completed.

• Is it not time we began looking at grids that are higher resolution – 25 km EASE grid was chosen because the SSM/I 37 GHz approximate this resolution?

• AMSR-E/GCOM-W are at 10x14 km at 36 GHz. Perhaps a higher spatial resolution (e.g. 12.5km grid)?

Thank you