satellite sensor radiometric calibrationimage.sciencenet.cn/olddata/kexue.com.cn/upload/blog/... ·...
TRANSCRIPT
Satellite Sensor Radiometric Calibration
Du Yongming, Cao Biao, Zhong Bo
Outline
The Fundamental Theory of the calibration Laboratory calibration (Pre-launch calibration)In-flight calibration (Post-launch calibration)
Onboard calibrationIn situ calibrationCrossing calibration
Relative calibration, by Cao BiaoBreakA new sensor calibration algorithm.Some remote sensors’ calibration plans
CEOS’s Definition
Calibration is the process of quantitatively defining the system response to known, controlled signal inputs.
Calibration
observed scene
Sensor onboard
instrument calibrationdata calibration
geophysical calibration
validation instruments
raw data
geophysical variableEstimate/ Retrieval
validation=
comparison
feed-back to calibration(if needed)
quality statement
data release touser community
One simple calibration process
L=DN*gain+offset
Out put ( DN )
In put L ( illuminance intensity)
Low Response region
Linear Response region
Saturation Response region
Calibration can be classified as:
Pre launch calibration;In flight calibration;
Onboard calibration;In situ calibration;Cross calibration;
The relations between calibration methods
Pre launch calibration is to initial the calibration system;Onboard calibrators are to do operational work;In situ calibration and cross calibration are to verify and correct the onboard calibrators
The Fundamental Theory of the calibration
ARF is the Area Response Function, which describes the sensitivity distribution over the detector element.GRF is the Geometric Response Function, which describes the sensitivity of the sensor in dependence of the direction under which it enters the aperture.SRF is the Spectral Response Function, which describes the sensitivity of the sensor for radiance at different wavelength.TRF is the Temporal Response Function, which describes the sensitivity of the instrument during the time interval of the measurement.RRF is the Radiometric Response Function, which gives the relation between the raw digital counts and the radiance in physical units.
GRF: Geometric Response FunctionTo calibrate a sensor, we should select the most uniform scene as the calibration field!In lab, integrating sphere is used to ensure the FOV is homogenous for VIR bands, black body is used for TIR bandsWhat in a pixel in nature?When you split a pixel into some sub_pixel, The GRF should be considered!
GRF measured in lab
FOV
Parallel Pipe
SRF: Spectral Response Function
If the spectral response range is wide, the calibrated index gotten from a set of input spectral illuminance maybe bring errors when used for another set of spectral illuminaceSo the natural objects’ spectral distribution should be examined, and this study results should be adopted as the calibration input standard. 500 550 600 650 700 750 800 850 900 950 1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1System Transmittance
Wavelength (nm)
Tran
smitt
ance
RedNIR
SRF measured in lab
LIGHT SOURCES
100W TungstenHalogenbulb
MONOCHROMETER
AgCam
M3
M2
M4M1
Spectral Calibration Configuration
CHOPPERFILTER WHEEL
Collimator
Each measure only represents the sensor’s response to a narrow band spectral intensity
Measurements repeated when the monochrometer changed over all the spectral range, the SRF can be gotten.
0
64
128
192
256
550 600 650 700 750
array low endarray centerarray low midarray high endarray high mid
0
64
128
192
256
750 800 850 900 950
Array centerArray low endArray low midarray high end
TRF: Temporal Response FunctionM
ean
pixe
l va
lue
Image number
Image number
Mea
n pi
xel
valu
e
Variation in exposure time
“Memory effect”
RRF: Radiometric Response Function
The measured range should cover all the work range.Each detector’s RRF should be measured one by one.The measure should be a table include input and output. Especially in the non_linear region.
L
DN
RRF measured in lab (VIR bands)
RRF measured in lab ( Thermal bands )
Blank body is used for thermal bands
MTF: Modulation Transfer FunctionMTF: There are several factors affect MTF, such as optical system, detector, and read out electrocircuit.PSF: Point spread function. Can be converted into MTF by Fourier transfer method.
The right is a target to measure MTF in lab;
The left is the calibration field to measure MTF
Other factors affect the calibration
Scatter lights inner the sensor;Thermal radiance inner the sensor;Thermal effects (Especially for TIR band);Electromagnetism effectsElectric parts crosstalk;Dark current;……
Scan patterns
Example: Removal the memory effects
It is a whiskbroom scan image ;It is a multi_detector scaning image;The “Memory effect” is very serious, we can see the TRF as the following;The scan direction is forward and backward;
Onboard calibration
The basic concept is as following:
Two points method
L=DN*gain+offset
Out put ( DN )
In put L ( illuminance intensity)
Low Response region
Linear Response region
Saturation Response region
MODIS Onboard calibrators
Solar diffuser (SD) Blank body (BB)Solar DiffuserStability Monitor (SDSM)Space look window
Solar Diffuser degradation
The Solar Diffuser was made of Spectralon materials.Its Bi-directional Reflectance Factor (BRF) from 400 to 1700 nm was characterized pre-launch using a scattering goniometer and a known BRF standard reference sample.Figure shows the SD degradation at three different SDSM detectors wavelengths, corresponding to MODIS B8 (412 nm), B3 (469 nm), and B11 (531nm). The degradation for other longer wavelength is much smaller
A Better Solar diffuser need to be developed
The SD degradation is caused by the cosmic ray ( X ray, β ray) destroy the reflective surface of SD, which is exposed in the space.A new Solar Diffuser has been reported to be developed to substitute this kind of SD. It is designed as a board with many very small holes in it. The diameter of the hole is near with the wavelength of the bands to calibrate.This new SD are stable to resist the cosmic ray destroy.
Other bands calibration indexes change
The calibration indexes for RSF change continuously. This is partly caused by the detector’s degradation, partly caused by the mirror’s degradation.The calibration indexes for TEB change seriously in the beginning 200 days, and ware stable later.
In situ calibration Work steps:
Measure field reflectance/radiance when the sensor overpass;Measure atmospheric characters’profiles when the sensor overpass ;Use a radiative transfer code to calculate the “at sensor” radianceUse the SRF to integrate the spectral intensity into the band intensity;Use “two points method” to calibrate the sensor.
Measure items: Field reflectance/radiance;Field temperature;Field BRDF;Atmospheric characters;Sample strategy;
地 面 测量 数 据
红外传感器定标需求分析
定标场的选择
探 空数 据
定标实验设计
通 道 响应 函 数
入 瞳 处辐 亮 度
传 感 器输 出 值
定 标系 数
Modtran4.0
已 定 标 传感 器 数 据
待 定 标 传感 器 数 据
图像几何配准
定标地点的选择
空间分辨率转换
通道响应的转换
定 标系 数
已 定 标 传感 器 数 据
待 定 标 传感 器 数 据
误差分析和精度评价
定标报告
CE313
FR ASD
Reflectance
Atmospheric observation at Atmospheric observation at DunhuangDunhuang SiteSite
OlOl--754 radiometer754 radiometerCECE--318 sun318 sun--photometer photometer
敦煌场地相对反射率因子随观测角的变化
Field BRF changed with view angle
Sample strategy
Calibration fields in the world White sands;Railroad Valley;La Crau;Lunar Lake;Dun Huang, China;……
1.1. White Sands, US White Sands, US
The White Sands Missile Range test site in New Mexico has been in use for vicarious calibration since the mid 1980s. It is located in the desert southwest of the United States in a region of low aerosol loading and an elevation of 1.2 km. The coordinates of the test site are 32.919°N latitude and 106.351°W longitude. The site is relatively devoid of vegetation. The level of reflectance varies with season with the lowest reflectance values occurring during the winter months when portions of the missile range are either underwater or wet from the higher water table. Highest reflectance values are typically seen in late fall after the surface has dried after summer-season rains. The size of the White Sands area is the largest of the test sites with an overall size of about 50 km
2. Railroad Valley 2. Railroad Valley
The Railroad Valley playa is situated in central Nevada at 38°280N and 115°410Wand at an elevation of 1435 m above sea level (ASL) (Table 1). It is very homogeneous and consists of compacted clay-rich lacustrine depositsforming a relatively smooth surface compared to most land covers.
Imaged by SPOT-1 HRV on June 18,1998 (north is to the top).
3. La 3. La CrauCrau
The La Crau test site is located in southeastern France at 4.87°E longitude and 43.50°N latitude about 50 km northwest of Marseilles. It is a flat area of about 60 km2. The ground is uniformly composed of pebblesand sparsely covered by low and dry vegetation. The climate in this region is dry and sunny and the optical properties of the ground vary little during the year.
Enlarged MOMS-2P-scene (2 July 1998) with 400 ×400 m2 calibration area
4. Lunar Lake 4. Lunar Lake
Lunar Lake playa in Nevada locates in a high-altitude desert region approximately 1.7 km above sea level. The site receives little precipitation and has a high percentage of cloud-free days. The playa is actually a dry lakebedconsisting of hard-packed clay that is uniform over a large area. Based on previous measurements, the surface of the playa is approximately Lambertian at viewing angles up to 30°off nadir.
IKONOS image of Lunar Lake Playa acquired on June 7, 2000
5. 5. 敦煌校正场及地面测量敦煌校正场及地面测量
敦煌辐射校正场位于敦煌市西北方向大约20km处的党河再生冲击扇上,东西长约60km,南北宽约40km。其均匀面积约40×30km2,地
表基本无植被覆盖。场地属典型的大陆性气候,具有干旱少雨、太阳辐射强、降水量少、能见度高等特点。地表层主要成分是含细砾的中细沙、沙砾石和含砾粗沙。砾石分布较均匀,以细砾为主,中砾次之,结构较松散,颜色主要是黑色、灰色和白色,其中黑色约占40%,灰色、白色约占58%。 敦煌场区资源一号卫星(CBERS-1)影像图
SSC Remote Sensing Radial Target
and Edge Response Targets
6 NASA6 NASA’’s Calibration fields Calibration field
Cross Calibration
Main work steps:
Select a well-calibrated sensor as the “Source”; which should be similar with the “target” sensor ( To be calibrated sensor). The two sensor should have the similar band response function, similar space resolution, similar overpass time;Select a field as the calibration field, which is the overlap area of the two sensor’s FOV, is homogenous and large enough to find “pure pixel”;Pick out the reference radiance value of the cross calibration field from the source sensor image, and pick out the DN from the target sensor image;Do the band convert, view angle correction, time correction, to get the reference radiance;Using “two points method” to calibrate the target image;
Relative calibration, by Cao Biao
Pseudo-invariant Sites Method to Calibrate Landsat data(Dennis Helder, EROS Data Center, 2007 )
Work done or in progress by several groups:FranceAustraliaU.S. – GSFC & SDSU
Concept:Many locations on the Earth exhibit ‘constant’ surface reflectance and BRDF over short and long periods of timeLocations are homogeneous spatially and temporallyAtmospheric effects are minimal, fairly constant, and can be accounted for in a reasonably simple mannerCurrent sites are primarily located in deserts
Additional Calibration Sources
Desert Sites used in this studySaharan location: Path 181 Row 40Collaboration with CNESProcessing steps:
Center 3000 x 3000 pixels usedLevel 0R dataCheck for saturated pixels (Band 5)Sun angle > 48.5o
Earth-sun distance correctionOutgassing correction for cold focal bands
Potential Geo-locations for Pseudo-Invariant Site Selection
Number of potential sites = 51
Potential Arabian & African Sites
Additional Saharan Middle East Desert Sites
Arabia 1 Libya 1
Egypt 1 Libya 4
EROS Archive of Potential Arabian & African Sites
32451899.6921.67Niger 4
41451907.9621.57Niger 3*
494518810.5921.37Niger 2*
42461889.8119.67Niger 1*
374617727.8920.23Sudan 4
244817628.7717.34Sudan 3
354817825.6917.34Sudan 2
374517728.2221.74Sudan 1*
474516349.8621.67UAE 1
364616152.6120.23Arabia 9
454616349.5220.23Arabia 8
474716349.1918.79Arabia 7
514716546.918.79Arabia 6
554016745.4428.87Arabia 5
684017040.8128.87Arabia 4
764016843.7328.92Arabia 3*
334616250.9620.13Arabia 2*
534716446.7618.88Arabia 1*
No of L5 TM Scenes**RowPathLongitude, ELatitude, NSite Name
* Sites presented by Cosnefroy et al. ** EROS Archive through Dec 2006
EROS Archive of US Sites
475302997.143.2EROS
475292996.844.3Brookings
4473838114.531.7Yuma
4393340115.5938.23Lunar Lake Playa, NV
4503539115.36735.651Roach Lake Playa, NV
4503539115.38835.55Ivanpah, NV
4393340115.69238.504RailRoadValley
4333733106.35132.919White Sands, NM
No of L5 TM Scenes*RowPathLongitude, WLatitude, NSite Name
* EROS Archive through Dec 2006
Consistent calibration of the Landsat archive through use of pseudo-invariant sites
Major ObjectivesDevelop a worldwide set of sitesDevelop techniques to deal with ‘small sites’
Geometric & Radiometric processingAutomate the processDevelop lifetime radiometric gain trending for Landsat archiveUse as an additional tool for absolute gain and cross-calibration of Landsat archive
Techniques for relative gain calibration/correction of large linear arrays
Relative Gain—whiskbroom to pushbroom scanner issue:Landsat 4/5 TM—16 detectors/refl. band + 4 thermal det. 100 det.Landsat 7 ETM+ -- add the pan band & 30m thermal band 136 detectorsAdvanced Land Imager 320 multispectral detectors/sca x 4 sca’s/band x 9 bands + 960 pan detectors/sca x 4sca’s/band = 15,360 detectorsLDCM ≥ 57,000 detectors!
Relative gain estimation is a critical element for LDCM!Methods to estimate Relative Gain
Image uniform fields90o yaw maneuversStatistical based methods
Statistics-based Relative Gain Algorithm Overview
Cumulative Histogram statistics extracted from the databases to be used as an input to the algorithmFirst Pass : Calculate scene statistics from bulk trending database- Screen data for scenes that likely exercise detectors over significant linear rangeSecond Pass: Calculate detector statistics and SCA-level statistics(µ and σ) for all valid scenesCalculate relative gains from detector and SCA-level statistics
Relative gain for ith detector gi,relative = (σi/ σ)Correct relative gains within an SCACalculate SCA discontinuities using 10 overlaps detectors from adjacent SCA’sCorrect SCA discontinuities
EO12005070130654_SGS_01
1. Post-Image Bias removal2. SCA based RG correction
1. Relative SCA-to-SCA Correction based on the ten detector overlap
EO12005070130654_SGS_01
SCA 4 SCA 3 SCA 3 SCA 2
SCA 3 SCA 2 SCA 2 SCA 1
Linear 2 % Stretch
Techniques for relative gain calibration/correction of large linear arrays
Summary/Objectives:Develop theoretical basis for scene statistics approach‘Perfect’ the technique using ALI data as a precursorDevelop methodology for near 90o yaw maneuversExplore usage of near uniform Earth fields and internal lamps for relative gain estimation
Vicarious calibration of LDCM and Landsat TM/ETM+ instruments
History of vicarious calibration of Landsat 5 TM and Landsat 7 ETM+Continuation of this activityVicarious calibration of LDCM instrument…(?)
Landsat 7 Lifetime Gainwith 5% error bars (uncert of mean)
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
0 365 730 1095 1460 1825 2190 2555
Days Since Launch
Gai
n(W
/m^2
μm
sr/D
N)
Band 1
Band 2
Band 3
Band 4
Band 5
Band 720
00
2001
2002
2003
2004
2005
2006
Landsat 7 Lifetime Gainwith 5% error bars (uncert of mean)
y = -8.040E-04x + 2.179E+01
y = -1.082E-04x + 7.327E+00
1.0
6.0
11.0
16.0
21.0
0 365 730 1095 1460 1825 2190 2555
Days Since Launch
Gai
n(D
N/W
/m^2
μm
sr)
Band 1
Band 2
Band 3
Band 4
Band 5
Band 7
2000
2001
2002
2003
2004
2005
2006
Calibration Status for Current Instruments and Plans for Future Instruments: AVHRR, Landsat, MODIS, VIIRS, APS, ABI
B. Guenther1 and Brian Markham2
1 University of Maryland, Baltimore County Goddard Earth Science and Technology Center [[email protected]]2 NASA’s Goddard Space Flight Center, Mail Code 614.4, Greenbelt, MD 20771
AVHRR
Picture of Sensor
Climate ApplicationsVegetation IndicesISCCP (cloud) products
Primary Calibration Strategy
PrelaunchVicarious on-orbit
Ascending Node 1340 (11), 1400/1420/1430 (7, 9, 14, 16, 18), 1930 (6, 8, 10, 15), 2200 (17)
Wavelength in micrometers
1.58 - 1.643
0.725 - 1.100.725 - 1.100.725 - 1.102
0.58 - 0.680.58 - 0.680.58 - 0.681
NOAA 15, 16, 17, 18
NOAA 7, 9, 11, 12, 14
NOAA 6, 8, 10
Band No.
Landsat MSSLandsat-1 1972-1978; Landsat-2 1975-1982; Landsat-3 1978-1983;Landsat-4 1982-1992; Landsat-5 1984-1992
4 VNIR Bands - 80 meters
185 km swath
18 day repeat cycle (Landsat 1-3)
16 day repeat cycle (Landsat 4-5)
0850-0945 equatorial crossing
Climate Applications
• Regional land cover land use change & carbon consequences
• Regional biomass change & carbon consequences
Primary Calibration Strategy
Internal lamps on scene-by-scene basis
Landsat TM/ETM+ Landsat-4 TM (1982-3; 1987-1993), Landsat-5 TM (1984->), Landsat-7 ETM+ (1999->)*
Climate Applications
• Regional land cover land use change & carbon consequences
• Regional biomass change & carbon consequences
Primary Calibration Strategy • L4/L5 TM
Initial: Internal lamps on scene-by-scene basisCurrent L5 TM : Reconstructed history based on
lamps, vicarious, outgassing model and ETM+ Xcal 5/2003 ->
• L7 ETM+Pre-launch calibration; monitored by on-board lamps,
diffuser and vicarious calibration
4 VNIR; 2 SWIR bands -- 30 m 1 Panchromatic - (ETM+) -- 15 m
{1 TIR -- 60 m (ETM+); 120 m (TM) }
0945-1000 equatorial crossing
185 km swath; 16-day repeat L7- seasonal global “cloud-free”coverage
* SLC failure May 2003
LDCM Sensor(s)*
No pictureavailable
Climate Applications
• Regional land cover land use change & carbon consequences
• Regional biomass change & carbon consequences
Primary Calibration Strategy
• Internal Lamps• Lunar views• Solar diffuser
5 VNIR bands - 30 m
3 SWIR bands including cirrus detection band - 30 m
1 Panchromatic band - 15 m
{Possible thermal band option}
* Based on draft documents on LDCM website 5/06, subject to change
Landsat Climate Change ApplicationsLandsat Ecosystem Disturbance Adaptive Processing System (LEDAPS)
• North Americans forest disturbance and surface reflectance mapping using Landsat data for North American Carbon Program
• http://ledaps.nascom.nasa.gov/ledaps/ledaps_NorthAmerica.html
Land Cover Land Use Change (LCLUC) studies
• Forest Biomass and Land-Use Change in Central Africa: Reducing Regional Carbon Cycle Uncertainty
• Comparative Studies on Carbon Dynamics in Disturbed Forest Ecosystems: Eastern Russia and Northeastern China
• http://lcluc.umd.edu
Statewide Landcover and Trees Study (SLATS):
• Monitoring land cover change and greenhouse gas emissions in Queensland, AU
• http://www.nrm.qld.gov.au/slats/index.html
National Carbon Accounting System (NCAS)
• comprehensive picture of land cover change over the Australian continent for the past 30 years using Landsat data.
• http://www.greenhouse.gov.au/ncas/activities/landcover.html
Landsat-7 ETM+ VIS/NIR Calibration Strategy
Radiometric Calibration Hardware•Internal lamps --every scan
•Primary - 99.9% of scenes•Secondary - 0.1% of scenes
•Solar diffuser - Full Aperture Solar Calibrator - monthly•Partial Aperture Solar Calibrator-daily
Calibration Working Group•NASA•USGS•2 University Vicarious Calibration Teams•{2 thermal band vicarious calibration teams}•Twice Yearly meetings
MODIS
Picture of Sensor S
20 Bands from 0.4 to 2.4 μm, with NadirGIFOV of 250, 500 or 1000 mWall-to-wall coverage with wisk-broomscanner13:30 and 22:30 constant local timeorbits
Climate ApplicationsAdvances CZCS and AVHRR
nLw, Chlor_a, aerosols, clouds, H2O vapor, vegetation indices, snow/ice, land cover/dynamics
Primary Calibration StrategyFull aperture Solar Diffuser with SDSMSpectro-radiometric calibration assembly for spatial and spectral partial aperture characterizationLunarVicarious techniques
VIIRS for NPOESS
Picture of Sensor S
- 11 Moderate and 3 ImagingBands, near constantresolution of 750m (375m) sq- 1330 or 2130 synchronous
Climate ApplicationsBuild on MODIS data set heritage
Primary Calibration Strategy
Spectralon Solar Diffuser (100% Earthshine rejection)Solar Diffuser Stability Monitor (8 bands)Lunar observations
Picture of Sensor
Aerosol Polarimeter SensorShown is the RSP
Climate Applications Primary Calibration StrategyOn-orbit calibration approachesData HeritageResearch Scanning Polarimeter, POLDER, MODIS, Sun Photometer network [EOSP]
APS Science RequirementsAerosol refractive index, single
scattering albedo, and shapeAerosol optical thicknessAerosol particle size
CoveragePolarimetry and radiometry between 400 and 2300 nmAlong track scanning, narrow swath
A-train position
Advanced Baseline Imager for GOES-R
Picture of Sensor
Climate ApplicationsAerosols, vegetation, clouds (& thin cirrus), fire/burn scar detection, snow/ice
Primary Calibration Strategy
On-orbit, sun-activated targetProposed special target, Swales development (Bremmer)
6 Spectral Bands, resolution of 0.5 to 2 kmFull disk or CONUS in 5 min, 30 s mesoscaleimages