PowerPoint-Pr

1
The Orbiting Carbon Observatory (OCO) Mission
Retrieval Characterisation and Error Analysis H.
Bösch1, B. Connor2, B. Sen1, G. C. Toon1 1 Jet
Propulsion Laboratory, California Institute of
Technology, Pasadena, CA, USA, 2National
Institute of Water and Atmospheric Research,
Lauder, New Zealand Contact Hartmut Bösch,
Phone [phone], Email hartmut.boesch_at_jpl.nas
a.gov

• Error Analysis
• We show results from a linear error analysis for
  nadir observations at Park Falls, WI (46 N) in
  July (conifer) and Lauder, NZ (45 S) in July
  (frost), which represents well the expected range
  of our geophysical scenarios.
• The total error budget for XCO2 has several
  components which can be random and/or
  systematic
• Measurement noise (random)
• Smoothing error (random and systematic)
  retrieved XCO2 depends on an a priori CO2
  profile and its covariance. Fine structures in
  true CO2 profiles are smoothed by the
  retrieval which causes a random error. An a
  priori CO2 profile which is systematically too
  large/small results in a bias.
• Interference Error (random and systematic)
  Errors in XCO2 due to the interference of
  non-CO2 retrieval parameters with CO2
• Model Parameter Errors (systematic) Errors due
  to uncertainties in forward model inputs (e.g.
  instrument parameters). We assume that these
  parameters will have systematically varying
  uncertainties. Spectroscopic errors are not
  included here, since they are predictable and
  can be largely reduced by validation.
• Forward Model Error (systematic) Errors due to
  inadequacies in the forward model (not included
  in the presented error budget). Aerosol optical
  properties, cirrus clouds or polarization can
  cause errors in the range of several ppm and our
  current forward model has to be improved to
  reduce these errors to less than 1 ppm.

• Retrieval Algorithm
• XCO2 (dry air, column averaged, mole fraction of
  CO2) will be retrieved from a simultaneous fit
  of O2 and CO2 bands using Optimal Estimation
• The forward model is based on Radiant, a
  multi-layer, spectral resolving, multiple
  scattering radiative transfer model Mick
  Christi, CSU. Polarization is corrected with a
  2 orders of scattering approach.
• The algorithm retrieves profiles of CO2, H2O,
  temperature and aerosol optical depth as well as
  surface pressure, surface albedo and spectral
  dispersion
• XCO2 is computed from the retrieved state after
  the iterative retrieval has converged

• Motivation
• OCO will measure CO2 with very high precision
  and accuracy (0.30.5 ) which puts
  unprecedented demands on both instrument and
  analysis
• Here, we present an error analysis and retrieval
  characterization for OCO nadir observations
  which allows for verification and quantification
  of the precision and accuracy of our retrieval
  algorithm. Furthermore, a good understanding of
  the sensitivity and the errors of the space-based
  measurements is critical for inverse modeling of
  carbon sources and sinks.

• The OCO Mission
• OCO is a space-based mission solely dedicated to
  CO2 measurements with precision, accuracy and
  resolution needed to quantify CO2 sources and
  sinks
• OCO will target a regional, monthly averaged
  precision of 1-2 ppm without significant
  geographically coherent biases
• The payload consists of three bore-sighted, high
  resolution grating spectrometers (CO2 bands at
  1.61 ? m and 2.06 ? m and O2 A-band at 0.76 ?m)
• OCO will switch from nadir observations (small
  footprint size of 3 km2) to glint observations
  (high signal over oceans) every 16 days

Overview of the OCO retrieval algorithm
2.06 ?m CO2 Band
1.61 ?m CO2 Band
O2 A-Band
(CO2)
(CO2)
Simulated radiance spectra for the 3 OCO
spectrometers
(H2O)
(H2O)

• OCO Averaging Kernel
• The sensitivity of a space-based CO2 measurement
  varies with height due to the physics of
  spectroscopy and radiative transfer and the
  instrument characteristics
• The averaging kernel describes this sensitivity
  as a function of height  
• ak(z) ?XCO2/?x(z)
• The OCO averaging kernel depends on the solar
  zenith angle and surface albedo or type (and
  aerosol optical depth not shown here)

Error reduction for Park Falls due to the
information in the measurement
Cross talk for Park Falls. Shown is the averaging
kernel scaled by the retrieved and the a priori
uncertainty
OCO observation strategy
OCO ground track for nadir observations
The Earth Observing System Afternoon
Constellation, or A-Train
Single Sounding error budget for Park Falls and
Lauder for an aerosol optical depth of 0.3. The
total errors are 1.01 ppm (Park Falls) and 2.97
ppm (Lauder). For regional, monthly averages,
random errors reduce by a factor 100 and the
total errors are dominated by systematic
contributions.
OCO averaging kernels for nadir observations as a
function of solar zenith angle for 4 different
surface types