Aerosols Observations from MODIS

Pawan GuptaNASA ARSET- AQ – California Training

September 10-12, 2013

ARSET - AQApplied Remote Sensing Education and Training – Air Quality

A project of NASA Applied Sciences

Topic covered…

• EOS satellites• MODIS on Terra and Aqua• Data Products • Atmospheric Data Products• Aerosols Data …

Five Instruments:• ASTER• CERES• MISR• MODIS• MOPITT

Terra Spacecraft

Aqua Spacecraft

Six Instruments:• MODIS• CERES• AIRS• AMSU - A• AMSR - E • HSB

Satellite Remote Sensing Instrument - MODIS

Newer Earth-observing satellite remote sensing instruments typically make observations at many discrete wavelengths or wavelength bands.

36 wavelength bands covering the wavelength range

405 nm (blue) to 14.385 μm (infrared)

MODIS – MODerate resolution Imaging Spectroradiometer

MODIS Reflected Solar Bands

250 M

500 M

MODIS Thermal Bands

MODIS products are divided into three main categories

Due to the size and variety of products there is no single sourcefor information about the MODIS products.

Atmosphere

Land

Ocean

Level 1 Products - Raw data with and without applied calibration.

Level 2 Products - Geophysical Products (sometimes gridded)

Level 3 Products - Globally gridded geophysical products

This structure is common to many satellite products

Data Levels

Level 1 Products – Raw or calibrated radiances

Level 2 Products – Geophysical products

Less Processing

More Processing

Granules

Level 3 Products – Global composites

Global compositesof level 2 granules

MODIS Product Hierarchy

Granule - A 5 minute piece of an orbitUsed to produce

Granules

Used to produce

(MOD for Terra/MYD for Aqua)

Understanding a MODIS File Name

MOD04_L2.A2001079.0255.005.2006289012028.hdf

Product Name Date - year, Julian day

Time Collection

File processing information

Aerosol Product Filename

MOD - Terra product MYD - Aqua product

Few more things to know about MODIS DATA

• All MODIS products come in HDF format • In HDF format each file contains both data and

metadata

• SDS - Each parameter within a MODIS HDF file is referred to as an SDS (Scientific Data Set)

• An SDS must be referenced precisely according to name when analyzing the data with your own computer code.

• Calibration accuracy.

• Quality Assurance – product creators estimate• of the quality of the data.

• Data formats.

• Product Resolutions.

• How level 3 products are created from level 2• temporally and spatially.

• Current release of the data and data history.

And therefore you need to learn!

Things that change with each instrument

Start with aerosol detection …

Aerosol Retrieval

MODIS-Terra, May 08, 2007

Satellite Observations & Pollution

Land

Water

Smoke

Clouds

MODIS-Terra, May 14, 2007

Complex Image: Smoke & Clouds

Radiance -to- Aerosol Products

MODIS-Terra, May 2, 2007

High

Low

NoRetrievals

Fundamental Inputs• Separate Algorithms for Ocean and Land• Calibrated and Geolocated Reflectances (MOD02, MOD03, Errors <2%)• Surface Information (land, water, desert)• Meteorological Data from NCEP• Other Ancillary data sets

Remer et al., 2005

Land Algorithm

MODIS Aerosol Products

Land Ocean

Dark Target Deep Blue – Used over bright land surfaces

Three Separate Algorithms

Currently the dark target and deep blue products are separate.When both are available the user must select which one to use

In collection 6 there will be a joint product that uses an automatedprocedure to select the appropriate product.

MODIS Aerosol Products

Land Ocean

Dark Target Deep Blue

Three Separate Algorithms

MODIS Aerosol ProductsAll three algorithms create a 10 Km product.Land and Ocean 400 half kilometer pixels. Deep Blue 100 one kilometer pixelsOcean and Land (dark target) algorithms assume that aerosols brighten the scene.

Retrievals can only occur where there are asufficient number of spectrally dark pixels.

MODIS Aerosol Products

Important Points to Remember for MODIS Aerosol Products- Ocean and Land Products are produced using totally separateand distinct algorithms. All current Level 2 aerosol products are at 10 Km (10 x 10) resolution.

- The most important products for air quality are Aerosol Optical Depth. These exist for both Ocean and Land. - The most current version of Aerosol Product is collection 5 and collection 6 should release soon.

MODIS Aerosol SDS

The individual Land or Ocean SDS is generally preferred because - it contains more wavelengths - gives more information about quality - at level 3 it gives a quality weighted product that screens out anomalies

Land_And_Ocean Is useful if you need both together but may not give the same results as Land or Ocean

The Ocean and Land algorithms each produce their own Scientific Data Sets or SDS’s. When both algorithms retrieve the same parameter they may be combined into a joint Land_and_Ocean SDS.

MOD04 Aerosol Products MYD04Main Products - Ocean

Effective_Optical_Depth_Average_OceanRetrieved AOD at .47, .55, .66, .86, 1.24, 1.63, 2.13

Optical_Depth_Ratio_Small_Ocean_0.55*Fraction of Fine Mode AOD at 0.55

Optical_Depth_Small_Average_OceanAOD * Fine Mode Ratio

AOD

FineFraction

FineAOD

Main Products - Land

Corrected_Optical_Depth_LandRetrieved AOD at .47, .55, and .66

Optical_Depth_Ratio_Small_LandFraction of Fine Mode AerosolNot reported for AOD < 0.2

Optical_Depth_Small_LandAOD * Fine Mode Ratio(may be a threshold)

A Qualitative Product Only!

Aerosol Products

Main Products - Land

Corrected_Optical_Depth_LandRetrieved AOD at .47, .55, and .66

Optical_Depth_Ratio_Small_LandFraction of Fine Mode AerosolNot reported for AOD < 0.2

Optical_Depth_Small_LandAOD * Fine Mode Ratio(may be a threshold)

Aerosol Products

AOD

FineFraction

FineAOD

Quality Assurance is Extremely Important!!

Quality_Assurance_OceanScale is 0 - 3

Recommend Ocean QA above 0

Factors:Number of pixelsError fittingHow close to glint

QA indicates the confidence in the quality of the retrieval.

Quality_Assurance_LandScale is 0 - 3

Recommend Land QA of 3

Factors:Number of pixelsError fittingSurface reflectance

Optical_Depth_Land_and_OceanImage_Optical_Depth_Land_and_Ocean

For Visualization purposes when youwant an image of as much of the sceneas possible.This SDS includes poor quality(QA level 0) data

Corrected_Optical_Depth_Land

Daily L3 aerosol optical thickness over land and ocean

The large gaps over ocean are due to Sunglint The gaps over land are due to bright surfaces and and clouds Regularly spaced gaps near the equator are due to lack of coverage between orbits.

MODIS 10 KM and3 KM Products

Fires: Zambia, day 205, 2005. 1 km resolution of AOT allows tracing smoke plumes

-0.50-0.250.000.250.500.751.001.251.50

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50

0.0 0.2 0.4 0.6 0.8

MAIAC

MOD04

Aerosol Retrievals: Biomass Burning

Backup Slides

Land Algorithm…• First Step: Organize reflectance in three channels (0.47, 0.67, and 2.13) into 10X10 Km box (20X20 =400 Pixels) at 500 m resolution

20 Pixels (10 Km)

20 P

ixel

s (1

0 K

m)

Remer et al., 2005

Dark Target Algorithm Summary(Land)

1) Screen out unwanted features

- clouds - water - snow/ice pixels

Land Algorithm…Pixel Selection• The 400 pixels in the box are evaluated pixel by pixel to identify whether the

pixel is cloudy, snow/ice, or water

• The land algorithm will retrieve aerosol for coastal boxes that contain one or more pixels identified as ocean, but will decrease the quality of that land retrieval

• An ocean retrieval requires all 400 pixels in the box to be identified as water

• Cloud mask based on spatial variability to identify low clouds • The reflectance in the 1.38-m channel to identify high clouds

• The algorithm is sensitive to small subpixel patches of snow/ice, now all eight pixels contiguous to a pixel identified as “snow/ice” by MOD35 will also be labeled as “snow/ice.”

• The pixels are further screened for subpixel water by determining NDVI for each pixel. NDVI values less than 0.10 are identified as containing subpixel water and are excluded along with cloudy and snowy pixels from the remainder of the algorithm

Remer et al., 2005