26
CIMSS Seminar December 8, 2006 A global infrared land surface emissivity database Suzanne Wetzel Seemann, Eva Borbas, Robert Knuteson, Elisabeth Weisz, Jun Li, and Hung Lung Huang

A global infrared land surface emissivity database

  • Upload
    london

  • View
    120

  • Download
    1

Embed Size (px)

DESCRIPTION

A global infrared land surface emissivity database Suzanne Wetzel Seemann, Eva Borbas, Robert Knuteson, Elisabeth Weisz, Jun Li, and Hung Lung Huang. Applications requiring a global land surface emissivity. - PowerPoint PPT Presentation

Citation preview

Page 1: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

A global infrared land surface emissivity database

Suzanne Wetzel Seemann, Eva Borbas, Robert Knuteson, Elisabeth Weisz, Jun Li, and Hung Lung Huang

Page 2: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Applications requiring a global land surface emissivity

Our goal is to produce a methodology for integrating the best information on land cover and surface emission on a high spatial (5km), high spectral (1 wavenumber), and high temporal (daily) that can be updated to support real-time operations and research for future instruments, including:

• GOES-R Proxy data set generation• GOES-R Surface characterization for TOA radiance calculations.3. GOES-R/NPOESS Training set (IR surface emissivity) for ABI retrieval4. GOES-R/NPOESS background field required for 1-D var data

assimilation

Synthetic (statistical) retrievals of atmospheric temperature, moisture, and ozone from MODIS MOD07 and UW IMAPP AIRS radiances

Selected Other Current Applications/Users:

• Assimilation of radiances over land (JCSDA) • Climate Monitoring SAF (EUMETSAT)• Cloud and Ozone retrieval from SEVIRI (EUMETSAT)• AIRS Retrieval of Dust Optical Depths (UMBC/ASL)• IASI-Metop Cal/Val (CNES, France)• Retrieval of hot spot data from AATSR (ESA/ESRIN)• Energy balance from ASTER over glacier (Univ of Milan)• AIRS trace gas retrieval for pollution monitoring • (Stellenbosch University, South-Africa)• Education (Seoul National Univ.; NTA, Konstantin)

Page 3: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Sensitivity of calculated BT to land surface emissivity

Difference between BT calculated using the prototype-CRTM model with emis = 1.0 minus emis = 0.95 for 3 Aqua MODIS bands. Each of the 8583 points represents a forward model calculation for one land SeeBor profile, and the colors correspond to the land surface type (IGBP ecosystem category)

Band 29 (8.6m) Band 31 (11m) Band 33 (13.4m)

Is emissivity important? In what spectral regions?

Page 4: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Average differences for 8583 land SeeBor profiles of BT for Aqua MODIS IR bands 25, and 27-36 (left) and all 2378 Aqua AIRS channels (right). Each symbol (MODIS) and dot (AIRS) represents the average BT difference over all profiles for one channel.

BT calculated with Emis = 1.0 minus that calculated with Emis = 0.95

BT calculated with Emis = 1.0 minus that calculated with Baseline Fit (BF) Emissivity

Page 5: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Sensitivity of retrieved atmospheric products to land surface emissivity: TPW

TPW % difference retrieved using a training data set with two different surface emissivities.

MOD07 retrievals used Terra radiances for 314 clear sky cases at the ARM SGP site between April 2001 and August 2005.

emis = 1.0 minus emis = 0.95

emis = 1.0 minus BF emis

Page 6: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Page 7: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Sensitivity of retrieved atmospheric products to land surface emissivity: T, q profiles

Profiles retrieved using the IMAPP AIRS algorithm (Elisabeth Weisz).

Temperature K

Land and coastlines (1402 profiles)

Barren/desert (242 profiles)

Mixing Ratio g/kg

Land and coastlines (1402 profiles)

Barren/desert (242 profiles)

Mean absolute differences between collocated radiosondes and retrievals with:• emis = 1.0 minus emis = 0.95 (dashed) • emis = 1.0 minus BF emis (solid)

Page 8: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Mean (solid) +/- 1 stdev (dashed) for emissivity assigned to the NOAA-88 training profiles in ATOVS and early MODIS retrieval algorithms

In the past, constant value or pseudo-random emissivity spectra have been assigned to the training data for retrieval of atmospheric temperature and moisture

Emissivity in regression retrievals of atmospheric properties

• The MODIS MOD07 synthetic regression retrieval algorithm uses 11 IR channels to retrieve atmospheric profiles of temperature and moisture, total precipitable water vapor (TPW), total ozone, lifted index, surface skin temperature.

• The algorithm uses clear-sky radiances measured by MODIS over land and ocean for both day and night. To compute the synthetic radiances from the profile training dataset to train the regression, surface emissivity values must be assigned to each profile.

Page 9: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

For Comparison:

Laboratory measurements of selected materials from UCSB (compiled by Dr. Zhengming Wan, MODIS Land Team):

Page 10: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

MODIS IR Channels

20 22 23 25 27 28 29 30 31 32 33 34 35 36

wavelength (m)

3.8 3.9 4.0 4.5 6.7 7.3 8.6 9.7 11 12 13.3 13.6 13.9 14.2

Channels in MOD07

X X X X X X X X X X X

Channels with Emissivity in MOD11

X X X X X X

One option for emissivity is the MODIS MOD/MYD11 operational land surface emissivity product but it is not at high enough spectral resolution

Page 11: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

To fill in the spectral gaps in MYD11 emissivity data, high spectral resolution laboratory measurements from the MODIS/USCB and ASTER emissivity libraries are used:

• High spectral resolution (wavenumber resolution between 2-4cm-1), • Not necessarily true representations of a global ecosystem as seen

from space.

The key to deriving a global emissivity database lies in the combination of the high spectral measurements made in the laboratory and moderate spectral resolution satellite observations of actual ecosystems.

There are a number of ways to combine the two. One approach, termed the “baseline fit” method is introduced here. Another effort is underway to generate a high spectral resolution emissivity dataset that uses principal component analysis to combine the laboratory data with MOD/MYD11 observations (Eva Borbas) .

Page 12: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Baseline Fit Methodology for Deriving Land Surface Emissivity

We use selected laboratory measurements of emissivity to derive a baseline conceptual model of emissivity and MODIS MYD11 measurements to adjust the emissivity.

First, we selected 10 inflection points, or hinge points that are important in characterizing the *shape* of a spectrum

Then, we developed a set of fitting rules to adjust the emissivity at these wavelengths based on the observed MOD/MYD11 values. The rules were developed based on careful inspection of and testing with 321 high spectral resolution laboratory-measured emissivity spectra.

Page 13: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Page 14: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

• Spectra typically slope up more steeply from 4.3 to 5m, then less steeply from 5 to 7.6m.

• In the 5-7m region, the spectra typically slopes more steeply from 5-5.8m, then levels off.

• Due to a lack of information from MOD11 in the 5-8m region, one value must be held constant in some cases. A value of 0.976 was used for the 7.6m emissivity based on an average over the laboratory spectra.

• Many, but not all, spectra have a broad reduction in emissivity centered around 8.6m.

• If MOD11 emissivity at 8.6m is greater than 0.97, these cases typically have relatively flat emissivity spectra, often with all emissivities higher than 0.97.

• The emissivity beyond 12m (the last wavelength for which MOD11 data is available) is assumed to have a constant slope for all spectra equal to a rise of 0.01 over 3.5 microns. This is based on inspection of the laboratory data.

Conceptual model of a land surface emissivity spectrum that was used to build the baseline fitting rules

Page 15: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Results of applying the baseline fit procedure to MYD11 emissivity at 4 locations.

The baseline fit spectra are shown by the solid, dotted, dashed, and dash-dot lines and the 6 input MYD11 emissivity values as the symbols.

Page 16: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Lab emis (black solid lines) was sampled at only the MYD11 wavelengths (vertical dotted lines) and input to the baseline fitting procedure.

The result is the baseline fit emissivity (blue dashed).

Sliced Santa Barbara Sandstone

Tropical Soil, Zimbabwe, Africa

Granodiorite

Page, Arizona

Laurel LeafAltered Volcanic Tuff

Evaluation of the Baseline Fitting Procedure

Page 17: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Average differences between 321 laboratory spectra and emissivity derived by the baseline fit method (blue). Differences are also shown for a constant emissivity of 1.0 (black) and those derived by linear interpolation between MYD11 wavelengths (red).

Page 18: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Impact of BF Emissivity on MODIS and AIRS Retrieved Temperature and Moisture

TPW (mm) at the ARM SGP site from Terra MODIS MOD07 (red), GOES-8 and -12 (blue), and radiosonde (black), with the ground-based ARM SGP MWR for 313 clear sky cases from 4/2001 to 8/2005.

MOD07 Statisticsbias = -0.04 mmrms = 2.49 mmn = 313

MOD07 Statisticsbias = 1.9 mmrms = 3.76 mmn = 313

Page 19: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Sahara Desert: Terra MODIS ascending (nighttime) passes on 1 August 2005

MOD07 TPW with emis = 0.95 MOD07 TPW with Baseline Fit emis

NCEP-GDAS TPW analysis

Page 20: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

MODIS MOD07 TPW for the 5 minute Terra granule beginning at 21:40 UTC on August 1, 2005.

A closer look at one of the Sahara desert granules

Emis = 0.95 Emis = 1.0 BF Emis

Page 21: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Some examples from the database…

Page 22: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

3.7 m 4.3 m 5 m

8.3 m 10.8 m 14.3 m

Global BF emissivity for 6 wavelengths, August 2002

Page 23: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

4.3 m 8.3 m 10.8 m

BF Emissivity in the Sahara Desert region for August 2003

Page 24: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

February May

Land Surface Emissivity Time Series

August November

Monitoring seasonal land changes in a region: 4.3m BF land surface emissivity

Page 25: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Monitoring changes over time at 10 point locations: 8.3m BF land surface emissivity

Page 26: A global infrared land surface emissivity database

CIMSS Seminar December 8, 2006

Future Work and LimitationsThe baseline fit emissivity database is tied to the accuracy of MYD11. Future work will apply a similar methodology to operational emissivity products from other platforms such as AIRS for comparison.

Monthly temporal resolution is not be sufficient for some applications. MYD11 L3 daily or 8-day global emissivity fields can be used to create BF emissivity with higher temporal resolution.

Spectral information between the inflection points is an approximation and will not be sufficient for some applications. Work is ongoing on a new version using a principal component analysis combining the MODIS MOD11 emissivity with laboratory measurements of emissivity. The dataset is based on a regression relationship between the observed MOD11 emissivity data and the principal components of selected laboratory spectra (Eva Borbas).

Work to compare the baseline fit database with ground-based measurements is planned. Work to comparison of this dataset with emissivity derived from other sources (AIRS, SEVIRI) is ongoing (Leslie Moy).

Dataset available at http://cimss.ssec.wisc.edu/iremisContact [email protected] Draft of paper submitted to JAMC September 2006 available upon request