Remote Sensing on land Surface Properties

advertisement
Remote Sensing on land
Surface Properties
Menglin Jin
Modified from Paolo Antonelli CIMSS, University of Wisconsin-Madison,
M. D. King UMCP lecture, and P. Mentzel
outline
•
•
•
•
Reflectance and albedo
Vegetation retrieval
Surface temperature retrieval
Theory of clouds and fire retrieval
MODIS Land Cover Classification
(M. A. Friedl, A. H. Strahler et al. – Boston University)
0 Water
1 Evergreen Needleleaf Forest
2 Evergreen Broadleaf Forest
3 Deciduous Needleleaf Forest
4 Deciduous Broadleaf Forest
5 Mixed Forests
6 Closed Shrublands
7 Open Shrublands
8 Woody Savannas
9 Savannas
12 Croplands
13 Urban and Built-Up
14 Cropland/Natural Veg. Mosaic
10 Grasslands
15 Snow and Ice
16 Barren or Sparsely Vegetated
11 Permanent Wetlands
17 Tundra
Reflectance
• The physical quantity is the Reflectance i.e.
the fraction of solar energy reflected by the
observed target
• To properly compare different reflective
channels we need to convert observed
radiance into a target physical property
• In the visible and near infrared this is done
through the ratio of the observed radiance
divided by the incoming energy at the top of
the atmosphere
Soil
Vegetation
Snow
Ocean
MODIS multi-channels
– Band 1 (0.65 m) – clouds and snow reflecting
– Band 2 (0.86 m) – contrast between vegetation and
clouds diminished
– Band 26 (1.38 m) – only high clouds and moisture
detected
– Band 20 (3.7 m) – thermal emission plus solar
reflection
– Band 31 (11 m) – clouds colder than rest of scene
-- Band 35 (13.9 m) – only upper atmospheric thermal
emission detected
MODIS BAND 1 (RED)
Low reflectance in
Vegetated areas
Higher reflectance in
Non-vegetated land areas
MODIS BAND 2 (NIR)
Higher reflectance in
Vegetated areas
Lower reflectance in
Non-vegetated land areas
RED
NIR
Dense Vegetation
Barren Soil
Vegetation: NDVI
The NDVI is calculated from these individual measurements as follows:
NDVI =
NIR-RED
NIR+RED
NDVI –Normalized Difference Vegetation Index
• Subsequent work has shown that the
NDVI is directly related to the
photosynthetic capacity and hence energy
absorption of plant canopies.
Satellite maps of vegetation show the density of plant growth over the entire globe.
The most common measurement is called the
Normalized Difference Vegetation Index (NDVI). Very low values of
NDVI (0.1 and below) correspond to barren areas of rock, sand, or snow.
Moderate values represent shrub and grassland (0.2 to 0.3), while high values
indicate temperate and tropical rainforests (0.6 to 0.8).
NDVI
• Vegetation appears very different at visible
and near-infrared wavelengths. In visible
light (top), vegetated areas are very dark,
almost black, while desert regions (like the
Sahara) are light. At near-infrared
wavelengths, the vegetation is brighter
and deserts are about the same. By
comparing visible and infrared light,
scientists measure the relative amount of
vegetation.
NDVI represents greenness
NDVI as an Indicator of Drought
August 1993
In most climates, vegetation growth is limited by water so the relative density of
vegetation is a good indicator of agricultural drought
Enhanced Vegetation Index (EVI)
• In December 1999, NASA launched the Terra spacecraft,
the flagship in the agency’s Earth Observing System
(EOS) program. Aboard Terra flies a sensor called the
Moderate-resolution Imaging Spectroradiometer, or
MODIS, that greatly improves scientists’ ability to
measure plant growth on a global scale.
• EVI is calculated similarly to NDVI, it corrects for some
distortions in the reflected light caused by the particles in
the air as well as the ground cover below the vegetation.
• does not become saturated as easily as the NDVI when
viewing rainforests and other areas of the Earth with
large amounts of chlorophyll
Electromagnetic spectrum
Radio waves
1000m
Microwave
1m
Longer waves
Infrared (IR)
1000 m
1m
Ultraviolet (UV)
Red Orange
Green
Yellow
(0.7m) (0.6m)
(0.5m)
Visible
Blue
X rays
0.001m
Shorter waves
1,000,000 m = 1m
Violet
(0.4m)
Gamma
Spectral Surface Albedo
(E. G. Moody, M. D. King, S. Platnick, C. B. Schaaf, F. Gao
– GSFC, BU)
• Spectral albedo needed for retrievals over land
surfaces
• Spatially complete surface albedo datasets have been generated
– Uses high-quality operational MODIS surface albedo dataset
(MOD43B3)
– Imposes phenological curve and ecosystem-dependent
variability
– White- and black-sky albedos produced for 7 spectral bands
and 3 broadbands
• See modis-atmos.gsfc.nasa.gov for data access and
further descriptions
Conditioned Spectral Albedo Maps
(C. B. Schaaf, F. Gao, A. H. Strahler
- Boston University)
MOD43B3
Indian Subcontinent during Monsoon
June 10-26, 2002
Spatially Complete Spectral Albedo Maps
(E. G. Moody, M. D. King, S. Platnick, C. B.
Schaaf, F. Gao – GSFC, BU)
Spectral Albedo of Snow
 Used near real-time ice and snow extent (NISE) dataset
– Distinguishes land snow and sea ice (away from coastal
regions)
– Identifies wet vs dry snow
» Projected onto an equal-area 1’ angle grid (~2 km)
 Aggregate snow albedo from MOD43B3 product
– Surface albedo flagged as snow
» Aggregate only snow pixels whose composite NISE snow
type is >90% is flagged as either wet or dry snow in any
16-day period
– Hemispherical multiyear statistics
» Separate spectral albedo by ecosystem (MOD12Q1)
Albedo by IGBP Ecosystem
Northern Hemisphere Multiyear Average (2000-2004)
???
urban
cropland
???
Surface Temperature: Skin
Temperature
• The term “skin temperature” has been
used for “radiometric surface temperature”
(Jin et al. 1997).
• can be measured by either a hand-held or
aircraft-mounted radiation thermometer, as
derived from upward longwave radiation
based on the Stefan-Boltzmann law
(Holmes 1969; Oke 1987)
Surface Temperature: Skin
Temperature
• The retrieval techniques for obtaining
Tskin from satellite measurements for land
applications have developed substantially
in the last two decades (Price 1984).
Tskinb = B-1( L)
Include emissivity effect:
Tskinb = B-1 [(L-(1-  )L )/  ]
Two unknowns!!
Surface Temperature: Skin
Temperature
• Split Window Algorith
•
Retrieving Tskin using the two channels (i.e., SWT)
was first proposed in the 1970s (Anding and Kauth
1970).
For example:
The NOAA Advanced Very High Resolution Radiometer
(AVHRR), which has spectral channels centered around
10.5 μm and 11.2 μm, has been widely used in this
regard for both land and sea surface temperature
estimation
Surface Temperature: Skin
Temperature
Split-window algorithms are usually written in
“classical" form, as suggested by Prabhakara
(1974)(after Stephens 1994):
Tskin ≈ Tb,1 + f(Tb,1 – Tb,2),
– where Tb,1 , Tb,2 are brightness measurements in
two thermal channels, and f is function of atmospheric
optical depth of the two channels.
– A more typical form of the split-window is
Tskin = aT1 + b(T1 –T2) – c
where a, b and c are functions of spectral emissivity
of the the two channels and relate radiative transfer
model simulations or field measurements of Tskin to
the remotely sensed observations
MODIS SST Algorithm
• Bands 31 (11 m) and 32 (12 m) of MODIS are
sensitive to changes in sea surface temperature,
because the atmosphere is almost (but not
completely) transparent at these wavelengths.
An estimate of the sea surface temperature
(SST) can be made from band 31, with a water
vapor correction derived from the difference
between the band 31 and band 32 brightness
temperatures:
• SST ≈ B31 + (B31 – B32) (just this simple!)
MODIS SST
Accuracy of Retrieved Tskin
• Accuracy of Tskin retrievals with SWT ranges from ≤ 1 to
≥ 5 K ( Prata 1993, Schmugge et al. 1998).
• SST is more accurate than LST (land skin temperature)
• Error sources:
split window equation;
Specifically, split window techniques rely on
assumptions of Lambertian surface properties, surface
spectral emissivity, view angle, and approximations of
surface temperature relative to temperatures in the lower
atmosphere (which vary more slowly). An assumption of
invariant emissivity, for example, can induce errors of 1-2
K per 1% variation in emissivity.
Lambertian surface properties
MODIS 2000-2007 averaged monthly Tskin
320
Land surface temperature
300
280
Jan
Apr
260
Jul
Oct
240
220
200
1
2
3
4
5
6
Modis land cover.
1. Evergreen Needleleaf Forest;
2,Evergreen Broadleaf Forest;
3,Deciduous Needleleaf Forest;
4,Deciduous Broadleaf Forest;
5,Mixed Forest;
6,Closed Shrubland;
7,Open Shrubland;
8,Woody Savannas;
9,Savannas; 10,Grassland;
11,Permanent Wetland; 12,Croplands;
13,Urban and Built-Up;
14,Cropland/Narural Vegetation Mosaic;
15,Snow and Ice; 16,Barren or Sparsely Vegetated
7
8
9
Land cover
10
11
12
13
14
15
16
Land Tskin vs Albedo
Land Tskin vs. Water Vapor
Transects of Reflectance
Band 4
(0.56 µm)
Band 1
Band 4
Band 3
Band 26
1.38 micron
Strong H20
1.38 μm
Only High Clouds
Are Visible
Band 26
1.38 µm
Visible
(Reflective Bands)
Infrared
(Emissive Bands)
Visible
(Reflective Bands)
Infrared
(Emissive Bands)
Emissive Bands
Used to observe terrestrial energy emitted by the Earth
system in the IR between 4 and 15 µm
• About 99% of the energy observed in this range is
emitted by the Earth
• Only 1% is observed below 4 µm
• At 4 µm the solar reflected energy can significantly
affect the observations of the Earth emitted energy
Spectral Characteristics of
Energy Sources and Sensing Systems
IR
4 µm
11 µm
NIR and VIS over Vegetation and Ocean
Vegetation
Ocean
NIR (0.86 µm)
Green (0.55 µm)
Red (0.67 µm)
RGB
NIR
Spectral Characteristics of
Energy Sources and Sensing Systems
NIR
IR
Radiation is governed by Planck’s Law
In wavelength:
B(,T) = c1 /{  5 [e c2 /T -1] } (mW/m2/ster/cm)
where
 = wavelength (cm)
T = temperature of emitting surface (K)
c1 = 1.191044 x 10-8 (W/m2/ster/cm-4)
c2 = 1.438769 (cm K)
In wavenumber:
B(,T) = c13 / [e c2/T -1]
(mW/m2/ster/cm-1)
where  = # wavelengths in one centimeter (cm-1)
T = temperature of emitting surface (K)
c1 = 1.191044 x 10-5 (mW/m2/ster/cm-4)
c2 = 1.438769 (cm K)
B(max,T)~T5
B(max,T)~T3
max ≠(1/ max)
Planck Radiances
180
160
B(,T) versus B(,T)
B(,T)
140
mW/m2/ster/cm (cm-1)
120
100
80
B(,T)
60
40
20
0
0
5
10
15
20
25
5
4
30
wavenumber (in hundreds)
100
20
10
6.
6
wavelength [µm]
3.
3
wavelength  : distance between peaks (µm)
Slide 4
wavenumber  : number of waves per unit distance
(cm)
=1/ 
d=-1/ 2 d 
Using wavenumbers
dB(max,T) / dT = 0 where (max) = 1.95T
Wien's Law
indicates peak of Planck function curve shifts to shorter wavelengths (greater
wavenumbers) with temperature increase. Note B(max,T) ~ T**3.
Stefan-Boltzmann Law
W/m2/deg4.

E =   B(,T) d = T4, where  = 5.67 x 10-8
0
states that irradiance of a black body (area under Planck curve) is proportional to T4 .
Brightness Temperature
c13
T = c2/[ln(______ + 1)] is determined by inverting Planck function
B
Brightness temperature is uniquely related to radiance for a given wavelength by the
Planck function
Planck Function and MODIS Bands
MODIS
MODIS BAND 20
Window Channel:
•little atmospheric absorption
•surface features clearly visible
Clouds are cold
MODIS BAND 31
Window Channel:
•little atmospheric absorption
•surface features clearly visible
Clouds are cold
Clouds at 11 µm look
bigger than at 4 µm
Temperature sensitivity
dB/B =  dT/T
The Temperature Sensitivity  is the
percentage change in radiance
corresponding to a percentage change in
temperature
Substituting the Planck Expression, the
equation can be solved in :
 = c2/T
Planck’s function (review lecture 1 )
First radiation constant
Wavelength of radiation
c1-5
B  (T) =
exp (c2 / T ) -1
Absolute temperature
Second radiation constant
Irridance:
Blackbody radiative flux
for a single wavelength at temperature T (W m-2)
Total amount of radiation emitted by a blackbody is a function of
its temperature
c1 = 3.74x10-16 W m-2
c2 = 1.44x10-2 m °K
Class Participation: Why this is different from the Planck Function on page 55?
∆B11> ∆B4
∆B11
∆B4
T=300 K
Tref=220 K
∆B4/B4= 4 ∆T/T
∆B11/B11 = 11 ∆T/T
∆B4/B4>∆B11/B11
 4
(values in plot are referred to wavelength)
> 11
(Approximation of) B as f
unction of  and T
∆B/B= ∆T/T
Integrating the Temperature
Sensitivity Equation
Between Tref and T (Bref and B):
B=Bref(T/Tref)
Where =c2/T (in wavenumber space)
B=Bref(T/Tref)

B=(Bref/ Tref) T 

B T 
The temperature sensitivity indicates the power to which the Planck
radiance depends on temperature, since B proportional to T
satisfies the equation. For infrared wavelengths,
 = c2/T = c2/T.
_____________________________________________________
____
Wavenumber
Typical Scene
Temperature
Temperature
Sensitivity
900
2500
300
300
4.32
11.99
Non-Homogeneous FOV
N
Tcold=220 K
1-N
Thot=300 K
B=NB(Thot)+(1-N)B(Tcold)
BT=NBThot+(1-N)BTcold
For NON-UNIFORM FOVs:
Bobs=NBcold+(1-N)Bhot
N
1-N
(1)
Bobs=N Bref(Tcold/Tref)+ (1-N) Bref(Thot/Tref) (2)
Bobs= Bref(1/Tref) (N Tcold + (1-N)Thot)
Tcold
Thot
(3)
For N=.5
Bobs= .5Bref(1/Tref) ( Tcold + Thot)
(4)
Bobs= .5Bref(1/TrefTcold) (1+ (Thot/ Tcold) ) (5)
The greater  the more predominant the hot term
At 4 µm (=12) the hot term more dominating than at 11 µm (=4)
Consequences: Cloud & Fire
Detection
• At 4 µm (=12) clouds look smaller than
at 11 µm (=4)
• In presence of fires the difference BT4BT11 is larger than the solar contribution
• The different response in these 2
windows allow for cloud detection and
for fire detection
MODIS clouds algorithm (As an example)
Band 29 (8.6 m)
Band 31 (11 m
The algorithm uses these thresholds to determine ice cloud:
Band 31 (11 m) Brightness Temperature < 238 K or
Band 29 – Band 31 difference > .5 K
The water cloud algorithm thresholds are
Band 31 (11 m) Brightness Temperature > 238 K and
Band 29 – Band 31 difference < -1.0 K
OR:
Or
Band 31 (11 m) Brightness Temperature > 285 K and
Band 29 – Band 31 difference < -0.5 K
Conclusions: Vegetation
Detection
• Vegetation: highly reflective in the Near Infrared and
highly absorptive in the visible red. The contrast
between these channels is a useful indicator of the
status of the vegetation;
• Planck Function: at any wavenumber/wavelength
relates the temperature of the observed target to its
radiance (for Blackbodies)
• Thermal Sensitivity: different emissive channels
respond differently to target temperature variations.
Thermal Sensitivity helps in explaining why, and
allows for cloud and fire detection.
Download