Atmospheric measurements and estimation of

Transcription

1 Atmospheric measurements and estimation of emissions of CH 4 Inverse modelling of North American methane emissions using GOSAT Proxy and Full-physics retrievals Ilya Stanevich Dylan Jones, Kimberly Strong, Feng Deng, John Lin, Kevin Wecht, Andre Butz, Robert Parker, Arlyn Andrews, Doug Worthy th 7 International GEOS-Chem Meeting, Harvard University May 4-7, 2015

2 Objectives Evaluate capability of GOSAT Proxy and Full-Physics retrievals to constrain North American fluxes Compare performance and results of different inversion approaches: GEOS-Chem model + GOSAT satellite retrievals STILT model + surface flask measurements Establish framework for inverse modelling of methane emissions at high spatial resolution resolution

3 From global to regional modelling GEOS-Chem for global modelling STILT for regional modelling Boundary conditions Global inversion creates optimized 3D field of methane. It is used to generate boundary conditions over North America. Zooming over North America we consider surface fluxes at higher spatial resolution. Global optimization: 1) 6 months spin up (January 2009 to July 2010) and 1 year monthly inversion 2) Consider last 2 months of inversion ( May-June 2010)

4 GOSAT coverage GOSAT Full-Physics May 2010 GOSAT Full-Physics May-June 2010 GOSAT Proxy May 2010

5 GEOS-Chem global inversion results: Scaling factors on emissions in May 2010 Proxy Full-Physics Proxy: use optimized CO2 fields from global GEOS-Chem inversion constrained by ACOS-GOSAT CO2 data as proxy for methane Full-Physics 2 month window: use 2 months observation time window to constrain monthly emissions Full-Physics 2 month window

6 GEOS-Chem global inversion results: Mean regional biases in May 2010 Proxy Full-Physics (FP) Negative bias in Full-Physics with 2 months window inversion Validation against North American hourly surface flask measurements shows no bias but poor correlation ( ); against global weekly measurements , with better performance of Proxy inversion in both cases Full-Physics 2 month window (FP-2months)

7 Regional inversion: Stochastic Time-Inverted Lagrangian Transport STILT model Measurement sites d[ X ] L P [X ] f P [X ] dt [ X ]0 Concentration at receptor point Influence function Concentration at initial point t1 C (x r, tr ) dt d 3 xi (x r, tr x, t ) S ( x, t ) d 3 xi (x r, tr x, t0 )C ( x, t0 ) t0 V Sources&Sinks Advection surface contribution V Advection initial tracer field

8 STILT model Input data and setup Input fields: WRF meteorological fields (30 km resolution) 3 types of boundary conditions from GEOS-Chem PROXY, FP and FP-2months optimizations GEOS-Chem surface fluxes, CH4 chemical loss rates Measurements: NOAA and Environment Canada surface flask measurements over North America (~11000 measurements) Setup: 10 day backward run for 100 particles Modelling period May-June 2010 Optimize 1 x 1 surface emissions using Optimal Estimation

9 Inversion results: Optimized scaling factors for May 2010 GEOS-Chem Proxy inversion STILT inversion PROXY boundary conditions Correlation of measurement versus model improves from 0.26 to 0.49 from GEOSChem to STILT optimization Strong sensitivity to Canadian emissions Surface measurements are not picking up signal from US emissions in the Gulf of Mexico compared GOSAT measurements

10 Comparison of STILT inversions Total Canadian emissions in May 2010 Mean bias between Proxy and Full Physics boundary conditions ~ 1.1 ppb Canadian methane emissions estimates for May 2010 equal ~ 0.98 Tg (global average monthly emissions ~ 45.0 Tg) Difference in Canadian emissions between Proxy and Full Physics inversions ~ 0.7%. Difference in scaling factors on a priori emissions PROXY FP-tgr PROXY FP FP FP-2months

11 Sensitivity to bias in BC (May 2010) Measurement sites Mean modelled inter-station surface contribution from North American emissions to methane concentration ~ 30 ppb or ~1.7% (values range from 0 to 500 ppb) Simple pseudo inversion test Sample the model at times and locations of measurements using a priori emissions and Proxy boundary conditions; use model output as pseudo observations Introduce bias in BC (uniformly increase/decrease BC by 2 ppb (~0.1%) and constrain a new state using pseudo observations Results Ideally, we get back a priori emissions In practice, total North American and total Canadian methane budgets change by ±2.5% and ±5%, respectively

12 Conclusions We obtain consistent estimates on total Canadian fluxes with the PROXY and FP boundary conditions; total Canadian emission in May 2010 are 17% lower than our a priori Although the total regional estimates are consistent, the PROXY and FP boundary conditions redistribute regional fluxes - this requires further investigation Using a longer assimilation window (2 months) in the global GEOS-Chem inversion generates lower emissions, which may indicate an issue with transport errors Total North American emissions are strongly sensitive to BC due to their small contribution to the total surface CH4 abundances and limited in situ measurements network coverage not capturing signal from strong CH4 emissions; sensitivity of high resolution fluxes is even stronger. Future plans Validate boundary conditions Bring in more observational constraints: 1) in STILT inversion (TCCON and NDACC FTIR sites) 2) in GEOS-Chem inversion (TES?) Apply optimized CO2 fields from a global inversion constrained by newer ACOS CO2 data as proxy for methane retrievals Acknowledgements This work was supported by the Canadian Space Agency through its funding of the Canadian FTIR Observing Network (CAFTON).