LTER/NASA COLLABORATION ON ATMOSPHERIC CORRECTION OF
REMOTELY SENSED DATA
WORKSHOP REPORT
AUGUST 16-18, 1996
John R. Vande Castle - University of Washington
Eric F. Vermote - University of Maryland
With LTER and NASA collaborators: Phylis Adams, David Bolgrien, Mathew Clark, Warren Cohen, Gayle Dana, Dwight Goodin, Joshua Greenberg, Brent Holben, Robert Kennedy, Mary Martin, Barbara Nolen, John Porter, Hassan Quaidrari, Steve Running, Gregory Shore, David Verbyla, Sally Westmoreland, Randy Wynne
This workshop was funded in part by NASA grant#NAGW-4565 and
held at the Long-Term Ecological Research Network Office
College of Forest Resources, University of Washington
Background and status of the Project
Atmospheric Correction Software
North Temperate Lakes/Madison LTER site (NTL)
Virginia Coast Reserve LTER (VCR)
General use of Hyperspectral data for LTER sites
Earth Observing System (EOS) overview
Example of atmospheric correction
ATMCOR Workshop II Participants
Background and status of the Project (John Vande Castle)
The LTER/NASA atmospheric correction project is proceeding in its second year. The primary goal is develop an operational method to correct satellite data for artifacts introduced by the presence of aerosols and water vapor in the atmosphere. This is the second workshop for the project, held August 16-18 at the LTER Network Office at the University of Washington. The workshop included discussions of the status and future plans for the project by the LTER and NASA researchers at this meeting.
The goals outlined for the LTER/NASA atmospheric correction project have followed the objectives, which are defined within the original proposal to NASA, with modifications and augmentation from discussions during the initial workshop. The primary goals of this project are to develop an operational method to apply atmospheric corrections to remote sensing data and to make these methods and corrected data available to ecological researchers for verification purposes.
The report of the first workshop has been distributed, and is available online through the main LTER home page (http://lternet.edu), through links in online documentation describing LTER/NASA collaborations (http://lternet.edu/nasa) as well as the direct link to the ATMCOR project home page (http://lternet.edu/nasa/atmcor). In the first year of the project, computer resources were established including accounts for project scientists on the LTER servers, addition of disk storage for correction software and data storage, and system upgrades to speed processing of the corrected data. The atmospheric correction programs were ported to the LTER servers, and 18 full Landsat-TM scenes were placed on-line for use in the project.
The primary LTER server dedicated to this project is known as "time" (time.lternet.washington.edu). All on-line archive data for the project can be accessed in the directory called "/archive" which links to online LTER datasets. In addition direct links for the project of "/d12/atmcor" are used to more easily separate data used in the projects from other remote sensing data in the archive. A more complete description of the data and examples of running the atmospheric correction software are given at the end of this report.
Data, as well as the processing software "demonstrat" from the NASA sunphotometer network (AERONET) has been updated in the last year at NASA/Goddard (GSFC). This includes all the data from the LTER sunphotometer network used in this project, which is now part of the larger AERONET network. All of the new data has be mirrored on the LTER servers, as well as an updated version of the "demonstrat" software. This software is described in the ATMCOR meeting report of 1995. The software is now installed and running on servers at GSFC as well as at the LTER Network Office.
Atmospheric Correction Software (Hassan Ouaidrari)
The atmospheric correction software used in this project is based on part of the "6S" code. Documentation on the specific software correction process can be found on the ATMCOR web page (http://lternet.edu/nasa/atmcor/tm_atc.htm), and a paper of the "6S" user manual and code documentation also exists (Vermote, E. F., Tanre, D., Deuze J. L., Herman, M. and Morcrette, J. J. (1997), Second Simulation of the Satellite Signal in the Solar Spectrum: an overview, accepted to IEEE Trans. Geosci. Remote Sens.
Vermote, E. F., Tanre, D., Deuze, J. L., Herman, M. and Morcrette, J. J., 1996, Second Simulation of the Satellite Signal in the Solar Spectrum: user manual, Version 1.0 University of Maryland, Dept. of Geography / Laboratoire d'Optique Atmospherique.)
The atmospheric correction software relies on data inputs obtained in the header documentation file of Landsat-TM data, converting the TM digital number data to radiances. This is based on water vapor information from climatology or sunphotometers, aerosols, from dark targets in the TM data or sunphotometers, surface reflectances in the TM data, and ozone from climatology or TOMS data. The latter has relatively small effects and generally low spatial variability.
The software correction can be applied with options of top of atmosphere (file extension .TAR), atmospheric corrections only (.ACP), or atmospheric and adjacency effects corrections (.ACT). The TM data are processed in grids of 16 cells to reduce processing time. Processing for .ACT data takes the longest, .TAR data takes the least. In addition, a "mask" data file is generated which contains an 8 bit quality control flag with the following information:
Based on the current data, a regression of LTER site aerosol optical density (AOD) from the TM data vs. AOD derived from sunphotometer data shows a good correlation. The regressions are best for lower wavelength data (i.e. blue) since these data are most stronly affected by atmospheric effects. For TM band 1 (blue), the correlation is strong with r2 = 0.89 and slope close to one. Points from SEV data do not fit as well due to a lack of good dark targets in the scene. At the VCR LTER (Hog Island) two scenes were evaluated to compare data from "hazy" and "clear" day. Changes in the surface reflectance information in the data of 12 and 28 July 1993 were assumed to change very little. In raw, uncorrected data, a comparison of the histograms from TM band 1 data were quite different, but more similar from in histograms from the corrected data. These result are, however preliminary and point to needs of future work including:
Sun Photometer – (Brent Holben)
Sunphotometers are used in the ATMCOR work for validation of the atmospheric correction. The original sunphotometer network that incorporated the LTER sites has grown substantially. It is now known as AERONET (formerly SPAM) - AErosol RObotic NETwork. It is funded through the Earth Observing System (EOS) for aerosol validation and characterization for the EOS platform. The network has land sites in many continents as well as oceanic sites. These sites can be permanent, seasonal or for short time observations.
The data processing system for the sunphotometer network has been modified for simplified access. The original "demonstrat" processing program still exists for access to raw data, but an automated Web-network access has also been installed at an address of http://spamer/gsfc/nasa/gov/. The Web access represents a "second generation of the demonstrat processing software, with a purpose of simplifying access to the sunphotometer data archive. In this version the data will be filtered through a number of automated processes, so the data can be more easily used by the remote sensing community.
MODLERS Project – (Robert Kennedy)
The "MODLERS" project grew out of specific plans for collaboration between the NASA MODIS land science team (MODLAND) and scientists associated with the LTER program. The MODLERS project was convened after a number of planning meetings including one held by Warren Cohen at the H.J. Andrews Experimental Forest LTER site to finalize plans for the project. The project, in its present form will, among other research, produce products of LAI, IGBP Landcover classes and Net Primary Productivity over a range of scales from 30m to a 10x10km area for 14 LTER sites.
NDVI - John Vande Castle
John Vande Castle presented information on initial evaluation applying atmospheric correction to Landsat TM data and its effect on full-scene NDVI. A comparison of the corrected (ATC) data compared to more conventional top of the atmosphere (TOA) corrections show that, for example data of the Madison (NTL-LTER) data, the ATC data produce NDVI results 25% greater than the TOA adjustments. For less forested/agricultural regions such as the Sevilleta and Jornada LTER sites, ATC data lower full scene NDVI estimates and increase the NDVI variance within the scene.
North Temperate Lakes/Madison LTER site (NTL) - Randy Wynne
Randy presented results from analysis of ATC data for estimating water quality data. Since the ATC processing often reduces the dynamic range of Landsat TM data at the lowest reflectance values, the standard processing of ATC data will need applications of scaling factors so the signal at low reflectances (i.e. clearer water) is not diminished.
Harvard Forest LTER (HFR) – Mary Martin
Mary presented data comparing work at HFR using the PnET Model using input parameters derived from AVIRIS data, with comparisons to species composition, foliar nitrogen, and foliar biomass data. There was a good correlation with ground nitrogen data, and also highly correlated with lignin content.
Bonanza Creek LTER – (BNZ) - Phyllis Adams & Dave Verbyla
David and Phyllis described the vegetation and land cover of the site and the continued operation of the sunphotometer station. They mentioned that compared to past years, the site required more maintenance and cloud cover was more of a problem. Coordination for future sunphotometer data acquisition as well as processing two 1994 Landsat TM datasets in the LTER Network Office archive were also discusses.
McMurdo LTER (MCM) - Gayle Dana
Gayle described the Antarctic LTER dry valley area as a "polar" desert with no visible land vegetation. The deployment of the MCM sunphotometer know as "Mr. Sunny" was again successful, with plans to continue data acquisition through at least 1998. Since the site is not in view of any geostationery satellites for data transmission, the instrument requires Daily download of data, which is later sent to GSFC by computer ftp transmission. Primary plans for use of these data is for comparison and correction of AVHRR data.
Sevilleta LTER (SEV) - Greg Shore
Greg covered background on the use of Landsat TM data for vegetation mapping on the LTER site. The mapping project was started in 1992 with interagency partners to produce a multi-temporal analysis of vegetation change. Some problems in comparison of different scenes were noticed, which may be due to data calibration problems. Using atmospherically corrected data might solve this problem, but all data would need to be reprocessed.
Virginia Coast Reserve LTER (VCR) - John Porter
John described the dynamic character of the barrier islands, with dramatic changes from storm events. Landsat TM data has been useful for determining water depth and vegetation changes of the island. A sunphotometer has been installed for short term use in the past and there are plans for future deployment because of its proximity to Wallops Island.
General use of Hyperspectral data for LTER sites- John Vande Castle
A short overview was presented to introduce the possibility of incorporating atmospheric correction techniques using hyperspectral data, to take advantage of AVIRIS data collected over a number of LTER sites. Collaborations developing between the LTER Network Office and facilities such as the San Diego Super Computer Center and the UNM-Maui High Performance Computing Center could be used in these efforts. Plans to acquire future hyperspectral data including that from the LEWIS HSI sensor could be a part of this effort.
Hyperspectral Tools - Sally Westmoreland
A demo package of hyperspectral tools developed by Photon Research Incorporated, showed potentials for initial use of atmospheric corrections to hyperspectral data. The routines are similar for both broad-band as well as hyperspectral data using radiative transfer code of MOZART with some elements for MODTRAN. The toolbox currently runs on the SGI computer platforms with potential plans to port the code to Sun workstations.
Earth Observing System (EOS) overview- Steve Running
Steve discussed validation-planning efforts for future EOS efforts. An important link for this can be found at the EOS project science web page at http://spso2.gsfc.nasa.gov/spso_homepage.html. Included in this effort will be global validation efforts for the Global Climate Observing System (GCOS) and Global Terrestrial Observing System (GTOS). Linkages between GTOS and the LTER and International LTER ("http://ilternet.edu) programs might be starting points for future collaborative EOS validation efforts.
Final discussion focused around what action items need to be worked on from here. These were developed as a list of action items for a continuation of the ATMCOR efforts:
Example of atmospheric correction:
The atmospheric correction of Landsat data is accomplished in a number of steps, with options for different levels of correction. Listed below is an example run for a Landsat-TM scene.
Step 1 – Generate a parameter file using the Landsat TM header. This is done by running the program "mkparmsfile". The first step in the atmospheric correction process is to generate a parameter file as input to the atmospheric correction program. This parameter file is constructed from information in the Landsat-TM Fast Format header file, or a header file created from scratch with pertinent information included. The command line to create the parameter file includes the program name, the input header file name, the output parameter file name, the location and root name of the input data files, and the path to the ancillary data used in the corrections. The input data for correction should be in standard "FAST FORMAT" with a common base name and extensions of "b1, b2, b3, etc. for the individual TM bands. On the LTER server machines, the following command is used to generate the parameter file:
"/homes/hassan/LTER/bin/mkparamsfile 940723.hdr 940723.prm /d12/atmcor/raw/bnz/tm/940723 /homes/hassan/LTER/ancillary"
Note that "/homes/hassan" may be something like "/home/nasa/hassan". The parameter file that is generated is a simple text file that can be viewed and edited. For example, to control the location of the output files, a line can be added to the parameter file to do this:
DATA PATH OUT = /My-Directory
An example listing of a parameter file looks like this:
# this params file was generated by mkparamsfile on 10/16/1996
# using TM header : 940724.hdr
#
DATA PATH = /d12/atmcor/raw/and/tm/940724
ANCILLARY DATA PATH = /homes/hassan/LTER/ancillary
DATE = 940724
NUMBER OF LINES = 8510
NUMBER OF PIXELS = 8975
PIXEL SIZE = 30.000000
CAL1 = 0.916214,-2.272727
CAL2 = 1.438396,-3.414634
CAL3 = 1.207517,-1.791045
CAL4 = 0.638823,-1.171875
CAL5 = 0.049999,-0.170507
CAL6 = 0.005656,0.123800
CAL7 = 0.022699,-0.059524
UOZ = -1.000000
UWV = -1.000000
ECMWFSP = -1.000000
ECMWFST = -1.000000
CENTER = 4255,4488,44.617404,-121.410333,34.701828,125.987320,0.389232,130.499542
GP1 = 1063,1121,45.342638,-122.464195,35.749310,125.419945,9.296780,101.860191
GP2 = 1063,3364,45.337039,-121.748598,35.345119,126.317726,4.423517,103.691658
GP3 = 1063,5607,45.326960,-121.033296,34.939770,127.234604,0.535031,261.136658
GP4 = 1063,7850,45.312406,-120.318455,34.536560,128.161392,5.416768,280.274323
GP5 = 3190,1121,44.864006,-122.468653,35.461040,124.899261,8.226439,102.083633
GP6 = 3190,3364,44.858499,-121.759009,35.051151,125.797211,3.346770,104.497940
GP7 = 3190,5607,44.848585,-121.049649,34.643269,126.705109,1.593697,274.979095
GP8 = 3190,7850,44.834271,-120.340738,34.237514,127.623001,6.488310,280.606049
GP9 = 5317,1121,44.385333,-122.473001,35.172771,124.378578,7.156098,102.307076
GP10 = 5317,3364,44.379917,-121.769162,34.760368,125.267319,2.264368,106.063942
GP11 = 5317,5607,44.370166,-121.065600,34.349953,126.166084,2.674932,277.740143
GP12 = 5317,7850,44.356087,-120.362472,33.941647,127.074974,7.560765,280.845673
GP13 = 7444,1121,43.906619,-122.477243,34.887829,123.848381,6.076991,102.601379
GP14 = 7444,3364,43.901292,-121.779067,34.472885,124.727829,1.180555,110.493080
GP15 = 7444,5607,43.891702,-121.081159,34.059929,125.617348,3.761808,278.919189
GP16 = 7444,7850,43.877855,-120.383673,33.645779,126.526947,8.633221,281.085297
Step two:
Once the parameter file is generated, the atmospheric correction program is run. It has options to generate data corrected to the top of the atmosphere, (the output files will have a ".TAR" extension), partial correction, i.e. corrected without adjacency correction (output with ".ACP" extension) or complete correction (".ACT" extension). Processing times vary from an hour to several hours on the current LTER servers (SUN SPARC). The "ACT corrections are the slowest.
The correction program uses the base file name of the input data, the input parameter file name, the correction level (-noatmos for ".TAR", -noadja for ".ACP" and –standard for ".ACT". Again input data for correction should be in standard "FAST FORMAT" with a common base name and extensions of "b1, b2, b3, etc. for the individual TM bands. The output data will consist of the log file that contains calibration coefficients to convert the digital number to reflectance and other information used in processing, the corrected data, a cloud and a quality control mask data file. The thermal channel is not corrected, but coefficients are given in the log file to convert the data to uncorrected degree Kelvin. An example for producing both top of atmosphere and complete correction are:
"/homes/hassan/LTER/bin/process_TM-V1 940723 940723.prm -standard 940723.log"
"/homes/hassan/LTER/bin/process_TM-V1 940723 940723.prm -noatmos 40723.TAR_log"
The correction program reads data in blocks to speed processing. The program does not always write out an incomplete block which might remove 1 or more lines of data. If this is a problem, the incomplete data can be blank filled using a program on the server called "fixsize".
Landsat-TM Data used for validation:
At the time of the workshop, 18 scenes have been assembled for use in the atmospheric correction verification. All of the data were acquired with concurrent sunphotometer data acquisitions. The 18 scenes listed by LTER site acronym and date are as follows:
SEV: 950317, 950418, 950520, 950605, 950707
NTL: 950606, 950622, 950926, 951012.
VCR: 930712, 930728.
AND: 940724, 940809, 940910, 940926.
BNZ: 940723, 940909
HFR: 930705
LTER/NASA ATMCOR Workshop Participants:
Phyllis Adams Site: BNZ
Institute of Northern Forestry
308 Tanana Drive
Fairbanks, AK 99775-5500
LTERnet address: PAdams@LTERnet.edu
E-Mail: pAdams@spruce.lter.alaska.edu
Phone: (907) 474-3310
FAX: (907) 474-3350
Gayle Dana Site: MCM
Desert Research Institute
University and Community College System of Nevada
7010 Dandini Boulevard (89512), PO Box 60220
Reno, NV 89506-0220
LTERnet address: Gdana@LTERnet.edu
E-Mail: gdana@maxey.unr.edu
Phone: (702) 674-7538: (702) 674-7539
FAX: (702) 673-7397
Organization(s): LTER
Matthew Clark Site: NET
College of Forest Resources Box 352100
University of Washington
Seattle, WA 98195
LTERnet address: mateo@LTERnet.edu
E-Mail: mateo@lternet.edu
Phone: (206) 543-4512
Organization(s): LTER
Brent N. Holben
NASA Goddard Space Flight Center
Mailstop 923.0
Greenbelt, MD 20771
LTERnet address: BHolben@LTERnet.edu
E-Mail: brent@kratmos.gsfc.nasa.gov
Phone: (301) 286-2975
FAX: (301) 286-1757
Organization(s): NASA
Robert Kennedy
Site: AND
Department of Forest Science
Oregon State University
020 Forestry Sciences Lab
Corvallis, OR 97331-7501
LTERnet address: rkennedy@LTERnet.edu
E-Mail: kennedyr@fls.orst..edu
Phone: (503) 750-7498
FAX: (603) 862-0188
Mary Martin Site: HFR
University of New Hampshire
Complex Systems Research Center
Morse Hall
Durham, NH 03824
LTERnet address: MMartin@LTERnet.edu
E-Mail: mem@.unh.edu
Phone: (603) 862-4508
FAX: (603) 862-0188
Barbara Nolen Site: JRN
New Mexico State University
Department of Biology
Box 30001, Dept. 3AF
Las Cruces, NM 88003
LTERnet address: BNolen@LTERnet.edu
E-Mail: bNolen@nmsu.edu
Phone: (505) 646-4465
FAX: (505) 646-5665
Hassan Quaidrari
NASA Goddard Space Flight Center
Code 923
Room N8, Building 16W
Greenbelt, MD 20771
LTERnet address: hquaidrari@LTERnet.edu
E-Mail: hassan@kratmos.gsfc.nasa.gov
Phone: (301) 286-6232
FAX: (301) 286-3221
Organization(s): NASA
Steven W. Running
University of Montana
School of Forestry
Missoula, MT 59812
LTERnet address: SRunning@LTERnet.edu
E-Mail: swr@hps1.ntsg.umt.edu
Phone: (406) 243-6311
FAX: (406) 243-4510
Gregory A. Shore Site: SEV
University of New Mexico
Department of Biology
Biology Annex
Albuquerque, NM 87131
LTERnet address: GShore@LTERnet.edu
E-Mail: gShore@sevilleta.unm.edu
Phone: (505) 277-2109
FAX: (505) 277-0304
John R. Vande Castle Site: NET
Associate Director
LTER Network Office
University of New Mexico
Department of Biology
Albuquerque, NM 87131-1091
LTERnet address: JVandeCastle@LTERnet.edu
E-Mail: jvc@lternet.edu
Phone: (505) 272-7315
FAX: (505) 272-7080
Dave Verbyla Site: BNZ
University of Alaska, Fairbanks
Department of Forest Sciences
Fairbanks, AK 99775
LTERnet address: DVerbyla@LTERnet.edu
E-Mail: dVerbyla@merlin.salrm.alaska.edu
Phone: (907) 474-5553
Organization(s): LTER
Eric Vermote
NASA Goddard Space Flight Center
Code 923
Room N8, Building 16W
Greenbelt, MD 20771
LTERnet address: EVermote@LTERnet.edu
E-Mail: eric@kratmos.gsfc.nasa.gov
Phone: (301) 286-6232
FAX: (301) 286-3221
Organization(s): NASA
Sally Westmoreland
Department of Geography
San Diego State University
San Diego, CA 92182-4493
LTERnet address: SWestmoreland@LTERnet.edu
E-Mail: sjw@photon.com
Phone:
FAX:
Randolph H. Wynne Site: NTL
UW-Madison Environmental Remote Sensing Center
Atmospheric, Oceanic, and Space Sciences, Rm 1249
1225 West Dayton Street
Madison, WI 53706-1695
LTERnet address: RWynne@LTERnet.edu
E-Mail: rhw@albert.ersc.wisc.edu
Phone: (608) 263-6584
FAX: (608) 262-5964
This works was made possible by the interest and support of NASA Mission to Planet Earth Headquarters, in particular that of Diane Wickland. The collaborations with Brent Holben’s NASA "AERONET" group is also deeply appreciated. This project was accomplished thanks to the interest and (unpaid) support of all NASA and LTER "ATMCOR" participants. The project and workshops would not have been possible without the usual expert logistic support provided by Adrienne Whitener and Lynne Hendrix of the University of Washington LTER Network. Office.