Improvements in XCO2 accuracy from OCO-2 with the latest ACOS v10 product
<p>While initial plans for measuring carbon dioxide from space hoped for 1-2 ppm levels of accuracy (bias) and precision in the CO<sub>2</sub> column mean dry air mole fraction (XCO<sub>2</sub>), in the past few years it has become clear that accuracies better than 0.5 ppm are required for most current science applications.&#160; These include measuring continental (1000+ km) and regional scale (100s of km) surface fluxes of CO<sub>2</sub> at monthly-average timescales.&#160; Considering the 400+ ppm background, this translates to an accuracy of roughly 0.1%, an incredibly challenging target to hit.&#160;</p><p>Improvements in both instrument calibration and retrieval algorithms have led to significant improvements in satellite XCO<sub>2</sub> accuracies over the past decade.&#160; The Atmospheric Carbon Observations from Space (ACOS) retrieval algorithm, including post-retrieval filtering and bias correction, has demonstrated unprecedented accuracy with our latest algorithm version as applied to the Orbiting Carbon Observatory-2 (OCO-2) satellite sensor.&#160; &#160;This presentation will discuss the performance of the v10 XCO<sub>2</sub> product by comparisons to TCCON and models, and showcase its performance with some recent examples, from the potential to infer large-scale fluxes to its performance on individual power plants.&#160; The v10 product yields better agreement with TCCON over land and ocean, plus reduced biases over tropical oceans and desert areas as compared to a median of multiple global carbon inversion models, allowing better accuracy and faith in inferred regional-scale fluxes. &#160;More specifically, OCO-2 has single sounding precision of ~0.8 ppm over land and ~0.5 ppm over water, and RMS biases of 0.5-0.7 ppm over both land and water.&#160; Given the six-year and growing length of the OCO-2 data record, this also enables new studies on carbon interannual variability, while at the same time allowing identification of more subtle and temporally-dependent errors.&#160; Finally, we will discuss the prospects of future improvements in the next planned version (v11), and the long-term prospects of greenhouse gas retrievals in the coming years.&#160;</p><p>&#160;</p>