Surface Reflectivity from OMI using MODIS to Eliminate Clouds: Effects of Snow on UV-Vis Trace Gas Retrievals

Gray O’Byrne,1 Randall V. Martin,1,2 Aaron van Donkelaar,1 Joanna Joiner3 and Edward A. Celarier4

[1] Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada

[2] Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts, USA

[3] National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, Maryland, USA

[4] SGT, Inc., Greenbelt, Maryland, USA

Selecting Cloud- and Aerosol-Filtered Scenes

Grid MODIS Cloud Mask Check OMI Footprint

~12 min

transport

Clouds in Red

Cloud- and Aerosol-Filtered SceneAnalysis repeated for scenes with AOD>0.2

Use LER from OMRRCLD as Surface LER for filtered scenes

Separate Snow-Free and Snow According to NISE Dry Snow flagReject Additional scenes:

-According to Sun Glint flag-If OMRRCLD cloud (or scene) pressure is 100hPa away from Surface Pressure-If LER > 0.3 (snow-free case only)

Snow-free surface LER at

354 nm (unitless)

Snow-covered surface LER at 354 nm (unitless)0 0.2 0.4 0.6 0.8 1

OMI LER[Kleipool et al., 2008]

GOME MinLER[Koelemeijer et al.,2003]

TOMS MinLER[Herman & Celarier, 1997]

Mean Diff. = 0.0002Std (σ) = 0.011

Mean Diff. = 0.012Std (σ) = 0.026

Mean Diff. = -0.008Std (σ) = 0.022

Snow-Covered LER Difference (Previous Climatology – Snow-Covered Surface LER)-0.8 -0.6 -0.4 -0.2 0 0.2

OMI LER

GOME MinLER

TOMS MinLER

Snow Weakly Represented in Previous Climatologies

Unrealistic Relation in OMI NO2 versus Cloud & Snow(Inconsistent with in situ data)

OMI Reported Cloud Fraction

≥ 5cm of snow

0 > snow < 5cm

no snow

Win

ter

Mea

n T

rop

. N

O2

(mo

lec/

cm2 )

Winter OMI NO2 over Calgary & Edmonton

OMI NO2 for Snow-Covered Scenes

corrected

correctedoriginal Bias NO Relative 2

With CloudFractionThreshold (f < 0.3)

-0.5 0 1.0

To correct NO2 retrieval for snow• Use snow-covered surface reflectivity• Use MODIS-determined cloud-free scenes to correct clouds

NO2 bias for MODIS-determined cloud-free scenes•Positive (negative) bias from underestimated (overestimated) surface LER•OMI reports clouds when surface LER is underestimated

Moving Forward

• Separate LER databases for snow-free and snow-covered scenes

• BRDF representation of surface

• MODIS for snow detection

• Future instruments with discrete bands at longer wavelengths (for cloud and snow discrimination)

Removed Slides

Corrected NO2

Over Snow

NO2 BiasOver Snow

MODIS Filtered

OMI Scenes

Snow-Covered Surface LER

OMI Clouds

SurfaceReflectivity

OMI NO2

Previous “Statistical” Climatologies

Kleipool et. al [2008]

Is Minimum Best?

Previous Reflectivity Climatologies

Mean DifferenceStandard Deviation

OMI 0.0002 0.011

OMI Mininum -0.002 0.033

GOME Mininum 0.012 0.026

TOMS Mininum -0.008 0.022

NISE Classification

No Snow(0 cm)

Thin Snow(0 < snow depth ≤ 5 cm)

Thick Snow(snow depth > 5 cm)

Snow-free Land3872 observations

0.31 0.49 0.20

Dry Snow4301 observations

0.06 0.18 0.76

Table 2. Comparison of the NISE classification in the OMI snow flag to collocated ground based measurements of snow depth. For the Snow-free and Dry Snow classifications a breakdown is given of the fraction of measurements that fall into 3 different snow depth categories. The data are from November, December, January, February and March of 2005 and 2006 over Edmonton and Calgary, Canada.

Vegetation Type

95%354 nm

Max Vegetation354 nm

95%360 nm[Tanskanen and Manninen, 2007]

Max Vegetation470 nm[Moody et al., 2007]

Water (Lakes) 0.82 0.82 - -

Evergreen Needleleaf Forest 0.22 0.38 0.28 0.36

Deciduous Needleleaf Forest 0.32 0.39 0.30 0.43

Deciduous Broadleaf Forest - 0.17 - 0.43

Mixed Forest 0.21 0.32 - 0.39

Open Shrubland 0.80 0.75 0.83 0.73

Woody Savannas - 0.50 - 0.47

Grasslands 0.76 0.75 0.72 0.72

Permanent Wetlands- 0.70 - 0.69

Croplands 0.71 0.66 0.38 0.76

Cropland/Natural Vegetation Mosaic - 0.66 - 0.65

Table 3. OMI derived surface LER of various snow-covered land types. The IGBP percentage land types are taken from the MODIS land cover product. The first method (95%) uses only grid squares containing at least 95% of a single land type to infer the mean LER. The second method (Max Vegetation) uses the maximum land cover type for each grid square. Results from two other sources are presented for comparison.

Figure 4. Monthly mean LER of seasonal snow-covered lands at 354 nm. Only locations with clear-sky observations of non-climatological snow cover for all six months (Nov-Apr) are used in computing the mean LER. Mountainous regions are masked. Error bars represent the standard deviation of the spatial mean.

Figure 6. Random AMF error versus surface reflectivity for tropospheric NO 2 over Edmonton, Canada.