Workshop on Remote sensing in the O 2 A-band The O 2 A-band Spectrometer on the NASA Orbiting Carbon Observatory-2 (OCO-2) David Crisp for the OCO-2 Science Team Jet Propulsion Laboratory, California Institute of Technology July 7, 2016 Copyright 2016 California Institute of Technology. Government sponsorship acknowledged. 1
The OCO2 O 2 A-band Spectrometer The OCO-2 A-band spectrometer provides data for: Dry air mole fraction accuracy requirement 0.1%! Cloud screening (A-Band Preprocessor, ABP) Thin cloud & aerosol optical depth and vertical profile retrievals Surface pressure & Solar Induced Chlorophyll Fluorescence (SIF) Known issues with current, Version 7 data product O 2 channel sensitivity variations and related calibration artifacts Uncertainties in A-Band spectroscopy and solar spectrum Cosmic ray artifacts Updates Improvements in the O 2 A-band channel calibration Improvements in A-band spectroscopy Other algorithm and calibration updates 2
OCO-2 Sampling Approach 14.5 orbits per day 3 frames per second 12 seconds of data 3
The OCO Instrument Optimized for Sensitivity 0.765 m O 2 A-Band CO 2 1.61 m Band CO 2 2.06 m Band 0.7576 to 0.7726 m 1.5906 to 1.6218 m 2.0431 to 2.0834 m 17500 18500 19100 20500 19700 19900 4
OCO-2 X CO2 Products 5
Typhoon Dolphin Eye Crossing, May 16, 2015 The OCO-2 ground track passed through the eye of Typhoon Dolphin just before CloudSat CloudSat OCO-2 Cloud Top Pressure OCO-2 Max Retrieved Pressure: 932 hpa Official Estimated Central Pressure: 934 hpa Chris O Dell 6
Solar Induced Chlorophyll Fluorescence (SIF) OCO-2 Flies over Des Moines, Iowa. If not removed from the O 2 A-Band radiances, SIF will introduce biases in the dry air mole fraction and other A-band products SIF measurements also provide a constraint on the spatial distribution of CO 2 uptake by photosynthesis 7
The OCO-2 A-Band Cloud Screening Algorithm Taylor et al. 2016 85 % 10 % 5 % 8
SNR and Single Sounding Random Error SNR (4/2015) Single Sounding Random Error ABO2 10/2014 4/2015 WCO2 12/2014 6/2015 SCO2 High SNR 2/2015 8/2015 Low Random X CO2 Error 9
OCO2- GOSAT Spectra Comparison [Kataoka et al.] Comparisons between OCO-2 and GOSAT also indicate very good agreement Jul01,2015 [1] oco2_l1bsctg_05309a_150701_b7000r_150927171125.h5 GOSATTFTS2015070120440360242_1BSPOD201201.01 GOSAT Rad OCO2 average Rad within 5km of GOSAT cnt point X 2 ratio1 = OCO2/GOSAT OCO2: path137 (looking from East) GOSAT: path36 (looking from East) GOSAT footprint OCO2 obs point within 5 km of GOSAT center point 10
Differences between OCO-2 and ECMWF Surface Pressures Monthly average differences between surface pressure estimates from OCO-2 version 7 and ECMWF Differences are binned into 2 by 2 bins and global bias of 2.5 hpa has been removed. Spatially dependent biases of +/- 3 hpa remain. 11
A-Band Calibration Challenges 12
Calibration Challenges: Cosmic Rays Cosmic rays a particular problem, especially on orbits that pass through the South Atlantic Anomaly (i.e. just about every orbit over South America) O2 band meas-mod > 6sigma 0 15 30 45 60 75 90 105 120 135 150 165 180 195 The largest effects are seen in the O 2 A-band. An algorithm to screen the specific colors affected by cosmic rays has been implemented. OCO-2 A-band spectra from the South Atlantic Anomaly 13
Calibration Challenges: A-Band Channel Sensitivity Variations The sensitivity of the OCO-2 ABO2 channel has varied over time, while the WCO2 and SCO2 show much less variability The ABO2 sensitivity degradation has two components A fast degradation reversed by decontamination activities This component has been attributed to temporary degradation of the anti-reflection coating on the A-band focal plane array detector (FPA) due to the accumulation of a thin (< 100 nm) layer of ice on the FPA A monotonic slow degradation Lunar and Vicarious Calibration measurements indicate that this change is due to degradation of the solar diffuser rather than a throughput loss in the instrument 14
Sensitivity Variations Over Time 15
A Closer Look at the Fast Degradation The signal loss varies with footprint Signal loss varies with wavelength These variations are corrected in the V7 L1b product SIF observations show a zero level offset (or ILS change) that is correlated with the level of signal degradation [C. Frankenberg]. A correction for tis offset is currently under development for the next data product. 16
More Insight on the Slow Degradation The slow component of the degradation is mostly due to the calibrator, not the instrument. A reanalysis of the Lunar Calibration Data indicates that about 20% of the slow degradation is in the instrument optics 17
Evidence that the Slow Degradation is not a property of the Science optical path Lunar calibration data indicate that the slow degradation is a property of the solar calibrator, and not a the science optical path Upward Trend Unfortunately, the version 7 calibration correction did not discriminate between the slow and fast degradation. Comparisons with data acquired by MODIS over the Sahara indicate that the OCO-2 L1B calibration is introducing a progressively larger radiometric error 18
Gas Absorption Coefficient Updates 19
Improved O 2 A band Spectroscopy ABSCO Tables v4.2 (L2 v7) v5.0 13200cm -1 O 2 Line shape Voigt for main iso. Galatry for minor iso. Positions, intensities from Long [2010; 2011] Line mixing Tran & Hartmann [2008] Collision Induced Absorption (CIA) Tran & Hartmann [2008] Speed-dependent Voigt from self-consistent set of multispectrum fits, utilizing FTS and CRDS measurements (Drouin et al. 2016, JQSRT) From ground-based atmospheric measurements at Lamont (E. Mlawer, AER) and CRDS H 2 O-O 2 broadening Drouin et al. [2014] Drouin et al. [2014] ABSCO v5.0: Self-consistent set of parameters! Multispectrum fitting approach pioneered by Chris Benner and Malathy Devi Speed Dependent Voigt (SDV): Accounts for the fact that collisions between molecules take place with velocities spanning some distribution Line mixing: Accounts for collisional coupling (mixing) between spectral lines. Interactions described by a relaxation matrix. Collision Induced Absorption: Accounts for inelastic collisions between molecules 20
Preliminary Tests of the Revised O 2 Absorption Coefficients: ECMWF Version 7 (current O 2 A-band absorption coefficients) Revised O2 A- Band Absorption Coefficients The revised O 2 A-band Absorption coefficients substantially reduce the global bias with respect to the ECMWF surface pressure prior. The global mean land bias has been reduced from -2.4 hpa to less than 0.4 hpa. The largest improvements are seen for glint observations, where the bias essentially vanishes in these tests. 21
Maps of bias in Psurf Version 7, no EOFs (The V7 baseline (with EOFs) shows similar latitudinal structure) We assume that ECMWF surface pressure is OK at these latitudes Issues around coastlines (S. America sulfate aerosol) 22
Maps of bias in Psurf with O 2 ABSCO update, no EOFs The amplitude of the pressure biases has been reduced, but the large scale features are still present. 23
Other Improvements Being Tested 24
Revised Solar Line List and Continuum Recent improvements in the solar line list explain many of the most persistent spectral residuals Empirical fits to the Solar Doppler data further reduce spectral residuals further Impact on X CO2 retrievals under investigation Should reduce solar contributions to EOF s Empirical corrections to solar continuum based on 11 Solar Doppler Calibration tests (black lines) account for much of the structure seen in v7 EOF#1 (colored lines). Footprint-to-footprint variations, due to calibration errors, should now be the major contribution. 25
Absolute Difference Changes in the ABO2 Cloud Screening The version 7 A-Band Pre-processor (ABP) used for cloud screening does not screen out cosmic rays. Cosmic ray artifacts reduce the yield of cloud free scenes over the South Atlantic anomaly The ABP has been modified to screen out cosmic rays. Preliminary tests indicate that this change will increase the yield of cloud free scenes over the South Atlantic anomaly by up to 100%. Fractional (%) Diff 26
Improved BRDF Model Testing: Lamont Target, Orbit 1362 (2) Lambertian Lambertian Soil BRDF Soil BRDF Vegetation BRDF Vegetation BRDF Surface reflectance retrievals in the O2 A- band using a Lambertian BRDF (top) are compared to those using a simplified Soil (middle) and Vegetation (bottom) BRDF AOD retrievals in the ABO2 channel using a Lambertian BRDF (top) are compared to those using a simplified Soil (middle) and Vegetation (bottom) BRDF Simplified Soil and Vegetation BRDF functions reduce the scatter and systematic, observation-angle-dependence of the surface reflectance and aerosol optical depth compared to the Lambertian surface albedo. 27
Impact of Stratospheric Aerosols on Retrieved Sulfate Aerosol Distribution OMPS detected a significant enhancement in stratospheric H 2 SO 4 aerosols in mid 2015 28
Impact of Stratospheric Aerosols on Retrieved Liquid Water Distribution Adding a thin stratospheric aerosol layer does not change the retrieved optical depth of the water cloud, but changes the liquid water vertical distribution. 29
Future Prospects for A-Band Observations 30
LATER NEAR FUTURE PRESENT PAST The Evolving Greenhouse Gas Constellation EnviSat SCHIAMACHY Several greenhouse gas satellites are planned over the next decade 2002-2012 GOSAT OCO-2 Most have O2 A-band spectrometers. 2009 2014 Sentinel 5p TanSAT+ FengYun 3D GOSAT-2 OCO-3/ISS 2016 MicroCarb 2016 GOSAT-3 2018 2018 Sentinel # GEOCarbon 202X 2023 202X 202X 31
Summary OCO-2 was successfully launched on 2 July 2014, and has been returning about 100,000 full-column measurements of X CO2 each day since September 6, 2014 Over 18 months of data has been delivered to the GES-DISC http://disc.sci.gsfc.nasa.gov/oco-2 In addition to XCO2, OCO-2 is delivering a series of products based on the O 2 A-band measurements Cloud and aerosol products, surface pressure, SIF Improvements in the A-Band absorption cross sections and OCO-2 retrieval algorithm are expected to increase the overall accuracy of this product in the next data product release. 32