Assimilation of Infrared Radiance Observations
Will McCarty
Global Modeling and Assimilation OfficeNASA Goddard Space Flight Center
2015 JCSDA Summer Colloquium 28 July 2015
What is Infrared Radiation?
What is Infrared Radiation?
What is Infrared Radiation in Atmospheric Data Assimilation?
What is Infrared Radiation in Atmospheric Data Assimilation?
What is Infrared Radiation?In General
– Anything between the visible and microwave
What do we use in data assimilation– Primarily the parts of the IR that are emitted from the
earth– As our observations approach the middle-IR, there is
a region of overlap that consists of a mix of solar reflection and terrestrial emission (arrow)
Earth-Emitted RadiationFirst things first – Units
– In the infrared, particularly to the sounding community, units of wavenumber in cm-1 are traditionally used
– To the imaging community, units of wavelength in microns (micrometers) are typically used
– I have a bad habit of swapping back and forth on the fly
• If I start to just throw out numbers, call me out
• I will try to keep things somewhat generic by absorber
Earth-Emitted RadiationThe theoretical emission of the
earth is at about 280 K– Warmer than what is expected
value from the sun’s directly– greenhouse effect
In some regions, the actual emission is less than expected
– This is where absorption is occurring
– What is seen is the cool top of the greenhouse effect blanket Theoretical Emission
as a function of TMeasured Emission
Earth-Emitted RadiationMajor Absorbing
Constituents– Carbon Dioxide
Theoretical Emission as a function of T
Measured Emission
CO2
Earth-Emitted RadiationMajor Absorbing
Constituents– Carbon Dioxide
Theoretical Emission as a function of T
Measured Emission
CO2
Note: From a weather perspective, we consider CO2 constant and well mixed over short periods. Therefore, temperature is determined from CO2
Earth-Emitted RadiationMajor Absorbing
Constituents– Carbon Dioxide– Water Vapor
Theoretical Emission as a function of T
Measured Emission
CO2
H 2O v
H2Ov
Earth-Emitted RadiationMajor Absorbing
Constituents– Carbon Dioxide– Water Vapor– Ozone
Theoretical Emission as a function of T
Measured Emission
CO2
H 2O v
O3 H2Ov
Earth-Emitted RadiationMajor Absorbing
Constituents– Carbon Dioxide– Water Vapor– Ozone– Methane
Theoretical Emission as a function of T
Measured Emission
CO2 CH 4
H 2O v
O3 H2Ov
Earth-Emitted Radiation
Atmospheric Windows occur in regions of little absorption
– Surface-sensitivity
Theoretical Emission as a function of T
Measured Emission
CO2 CH 4
H 2O v
O3 H2Ov
Window
Window
Back to Simple Radiative Transfer
In the previous talk, I referred to radiation as being effected by reflection, absorption, and transmission
– Absorption is key in the infrared– The surface reflects, and will be discussed w/ surface
emissivity– The atmosphere in the IR generally only scatters with
certain-sized particles
So how does absorption work?– Quantum physics– Molecules absorb radiation, and re-emit radiation
Absorption at the Molecular LevelMolecules absorb in electronic, vibrational, and rotational modes
Absorption CoefficientThe absorption coefficient is
a complicated and highly non-linear function of molecule i and line j
Line Strengths, Sij, result from many molecular vibrational-rotational transitions of different molecular species and isotopes of those species(blue).
Where width of line, ij, is a function of the molecule structure (natural broadening), temperature (doppler broadening) and pressure (collisional broadening)
18
600 to 700 cm-1 700 to 800 cm-1
H2O
CO2
O3
N2O
HNO3
OCS
SO2
CH4
CO
Line Strengths @ 15 μm
16.6 to 14.3 μm 14.3 to 12.5 μm
19
900 to 1000 cm-1 1000 to 1100 cm-1
H2O
CO2
O3
N2O
HNO3
OCS
SO2
CH4
CO
Line Strengths @ 10 μm
11.1 to 10 μm 10 to 9.1 μm
20
1250 to 1350 cm-1 1350 to 1450 cm-1
H2O
CO2
O3
N2O
HNO3
OCS
SO2
CH4
CO
Line Strengths @ 6 μm
8.0 to 7.4 μm 7.4 to 6.9 μm
21
2100 to 2200 cm-1 2300 to 2400 cm-1
H2O
CO2
O3
N2O
HNO3
OCS
SO2
CH4
CO
Line Strengths @ 4 μm
4.8 to 4.5 μm 4.3 to 4.2 μm
Atmosphere Transmittance
The Optical Depth is the sum of absorption coefficients for all isotopes and species multiplied by the path-length, usually written in terms of pressure levels pi and pj and view angle
The transmittance of a layer is given by the exponential of the optical depth
The view angle can be included in the absorption coefficient and transmittance from a level in the atmosphere (at height z) to the top of the atmosphere can be written as
Slightly Less Simple Radiative Transfer
Let’s make some simple assumptions:– The atmosphere is now discrete isobaric layers– The atmosphere does not reflect
• So no scattering• Each layer either transmits or absorbs/emits
– The surface does not transmit• Either reflects or emits
– No clouds (will quickly address this later)
Slightly Less Simple Radiative Transfer
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
The Upward Surface Emission plus…
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
The Upward Surface Emission plus…
The Upward Atmospheric Emission plus…
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
The Upward Surface Emission plus…
The Upward Atmospheric Emission plus…
The Surface Reflection of Downward Atmospheric Emission
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
The Upward Surface Emission plus…
The Upward Atmospheric Emission plus…
For the sake of discussion, let’s assume the surface is a blackbody• The surface neither transmits or reflects• Absorptivity = Emissivity = 1.0
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
The Upward Surface Emission plus…
The Upward Atmospheric Emission plus…
For the sake of discussion, let’s assume the surface is a blackbody• The surface neither transmits or reflects• Absorptivity = Emissivity = 1.0
1
Slightly Less Simple Radiative TransferA Measured Radiance is equal to…
The Upward Surface Emission plus…
The Upward Atmospheric Emission plus…
For the sake of discussion, let’s assume the surface is a blackbody• The surface neither transmits or reflects• Absorptivity = Emissivity = 1.0
0
Slightly Less Simple Radiative TransferThe Upward Surface Emission is the…
Blackbody radiation emitted by the surface at a given surface skin temperature
Slightly Less Simple Radiative TransferThe Upward Surface Emission is the…
Blackbody radiation emitted by the surface at a given surface skin temperature Scaled by the transmissivity from the surface (ps) to the top of the atmosphere (TOA, 0 hPa)
Slightly Less Simple Radiative TransferThe Upward Surface Emission is the…
Blackbody radiation emitted by the surface at a given surface skin temperature Scaled by the transmissivity from the surface (ps) to the top of the atmosphere (TOA, 0 hPa)Scaled by the surface emissivity (assumed as one in this case)
Slightly Less Simple Radiative TransferThe Upward Atmospheric Emission is the…
The sum over the vertical of
Slightly Less Simple Radiative TransferThe Upward Atmospheric Emission is the…
The sum over the vertical ofThe blackbody radiation emitted by each atmospheric layer at that layer’s temperature
Slightly Less Simple Radiative TransferThe Upward Atmospheric Emission is the…
The sum over the vertical ofThe blackbody radiation emitted by each atmospheric layer at that layer’s temperature Scaled by the change in TOA transmittance in that layer
Slightly Less Simple Radiative Transfer
This term is known as the weighting function and illustrates the vertical sensitivity of a given channel to the atmosphere
The Path so far…
The atmosphere is made of varying molecules…These molecules interact with radiation via absorption and
emission…This radiation is ultimately emitted to space…Where it is observed by a satellite…
End observable – a radiance that is a result of the molecules over the path of the observation
Desired observable – the atmospheric distribution of those molecules
Thus, an inversion is needed
Methods to Assimilated Information from the Infrared
Two general methods to IR assimilation– Assimilation of retrieved atmospheric profiles– Direct assimilation of the measured radiances
History of both…– Satellite retrievals were the initial approach
• Vertical retrievals of temperature and moisture are inverted from the radiances and assimilated in as simple geophysical observations as such
• Source of satellite information in NCEP/NCAR Reanalysis (Kalnay et al. 1996)
– Advances in the 90s allowed for direct radiance assimilation• With the use of variational methods and fast radiative transfer models
(including their tangent linear and adjoint models), the inversion is done in line in the solution
• Derber and Wu 1998, McNally et al. 2000
Retrievals vs. RadiancesA beaten-to-death argument – Radiances are directly measured and uncorrelated
• Well, not really…especially in the infrared, but they’re far less correlated in spectral space than retrievals are in the vertical
– Retrieval errors are inherently correlated– Retrievals are performed off of some sort of a-priori.
• Are you assimilating the prior or the physical information• What is your prior, and would you even want to assimilate it?
– Why in the world would you assimilate a climatology-derived prior into a weather model?
Retrievals vs. RadiancesThere’s a way to do the direct comparison correctly, but I’ve never seen it
– Physical retrievals have estimates of the averaging kernel, which is analogous to a weighting function or vertical Jacobian
– Radiance assimilation should be compared against the proper assimilation of the averaging kernels, not by treating the retrievals as radiosondes
• Because they simply are not radiosondes
This could be a whole lecture, and radiances basically won this argument
IR Radiance Assimilation
Retrieved Cloud Height
Traditionally, infrared radiances are only assimilated in those scenes that are clear or for channels insensitive to clouds in cloudy scenes
IR Radiance AssimilationA key assumption to cloud screening is that an
accurate cloud height can be retrievedCloud Height Retrieval
Assumptions:– Single cloud– Flat, infinitesimally
thin cloud– Graybody cloud
(fractionally cloudy, black cloud)
IR Radiance AssimilationA key assumption to cloud screening is that an
accurate cloud height can be retrievedCloud Height Retrieval
Assumptions:– Single cloud– Flat, infinitesimally
thin cloud– Graybody cloud
(fractionally cloudy, black cloud)
These assumptions stinkG. Marseille, KNMI
CALIPSO Lidar Backscatter
IR Radiance Assimilation
Ultimately, clouds in the infrared are a sharp temperature signal
– Generally cold (in an atmosphere of positive lapse rate) Observed minus Forecast signal
– Cloudy IR assimilation attempts to include this signal in the solution
– What if the observation is clear, but the model erroneously warm at 200 hPa
– Cloud retrievals will determine this signal as cold
Infrared Instruments
In GMAO forward processing, infrared radiances are assimilated from IASI, AIRS, CrIS, GOES Sounder, SEVIRI and HIRS
– Heritage “multi”-spectral sounders like HIRS (~ 18 channels) and the GOES Sounder are being phased out
– The US HIRS instruments replaced by CrIS from NPP onward (hyperspectral – 1297 ch total, 399 for DA)
– The final HIRS launched on MetOp-B. MetOp-C will only fly IASI (hyperspectral – 8461 ch, 616 for DA)
– No Sounder in US GEO beginning w/ GOES-R– Hyperspectral sounding potentially in GEO in a number of
future longitudes
Infrared Instruments
Clearly more information from modern hyperspectral sounders (AIRS, IASI, CrIS) vs. older sounders (HIRS, GOES Sounder)
Clearly a lot of redundant information as well
Taken from a Tony McNally talk
Infrared Observation UsageIR observations make up ~65% of the current global observing system
But only a small portion of the total number of observations available are utilized:
• Spectral Thinning– AIRS: 281 of 2378 channels are available, 124 active (5.2% of total)– IASI: 616 of 8461, 137 active (1.6%)– CrIS: 399 of 1305, 81 active (6.2%)
•Spatial Thinning– 1 spectra per instrument for every 145x145 km thinning mesh (observation footprint
size is ~15km) (~1.5% of the previous percentages)•Quality Control
– Via traditional means, infrared observations sensitive to clouds are discarded via quality control (~25% of the previous 1.5%)
Infrared Observation ImpactSo based on real numbers:
– AIRS: 0.04% of all observations are assimilated– IASI: 0.04%– CrIS: 0.19%
So while the instruments provide as much bang as any other satellite instrument type out there, why can we only use such small percentages of the data?
Infrared Assimilation Research of the Present/FutureCloud-affected Assimilation
– My research– Implementation of a approach similar to that of McNally (2009) to exploit
cloud information– Some efforts are beginning to extract cloud microphysical information out of
IR (true all-sky)
Underutilized Parts of the IR Spectrum– The 4 μm CO2 band is typically avoided due to solar contribution– The use of a bidirectional reflectance function in the solution may
compensate for this
Assimilation of Principal Components– Europe is way ahead of the US on this– MetOffice is investigating reconstructed radiances from partial PCs (Fiona
Smith)– ECMWF is investigating direct assimilation of PC scores (Marco Matricardi)– When Geostationary Hyperspectral happens, PCs will likely need to be
considered
Famous Infrared Detectors
GOES Imager
The satellite images you see on the news
Famous Infrared Detectors
Atmospheric Infrared Sounder (AIRS)
First hyperspectral infrared instrument Carn et al. 2005
Famous Infrared Detectors
Predator
Fought two future US Governors and one
Carl WeathersGrossed $98.3M ($218M)