Assimilating Compact Phase Space Retrievals (CPSRs): Comparison with Independent Observations (MOZAIC CO in situ and IASI CO Retrievals) and Extension to Assimilation of Retrieval Partial Profiles

Abstract for oral presentation at the Fifth AMS Symposium on the Joint Center for Satellite Data Assimilation (JCSDA) at the 97th AMS Annual Meeting, January 22-26, 2017, Seattle, WA 

Assimilating Compact Phase Space Retrievals (CPSRs): Comparison with Independent Observations (MOZAIC CO in situ and IASI CO retrievals) and Extension to Assimilation of Retrieval Partial Profiles 

Arthur P. Mizzi^*, David P. Edwards^*,

and Jeffery Anderson^#

^*Atmospheric Chemistry Division

National Center for Atmospheric Research

Boulder, CO 80307

^#Institute for Mathematics Applied to Geosciences

National Center for Atmospheric Research

Boulder, CO 80307 

mizzi@ucar.edu

303-497-8987

Assimilation of atmospheric composition retrievals presents computational challenges due to their high data volume and low information density.  Assimilation of compact phase space retrievals (CPSRs) meets those challenges and offers an innovative new approach for assimilating satellite observations for air quality analysis and prediction.  Comparison of CPSR assimilation results with assimilated observations (MOPITT CO) and with independent observations (MOZAIC in situ CO and IAIS CO retrievals) shows that they perform as well or better than assimilation of raw or quasi-optimal retrievals (reductions of ~60% for root-mean square error (RMSE) and bias when compared to assimilation of raw retrievals) at reduced computational cost (reductions of ~35% for MOPITT CO and ~50% for IASI CO).  CPSRs offers great promise because they can be applied to retrieval profile obtained from any optimal estimation method. 

One result from our independent observation comparisons is that assimilation of MOPITT CO CPSRs degrades the fit/skill in the upper troposphere (p < 300 hPa) when compared with not assimilating MOPITT CO and verified against IASI CO.  That result is due to the assimilation of biased MOPITT CO retrievals (MOPITT CO retrievals have a known positive bias of ~14% in the upper troposphere).  In this talk, we propose discarding/not assimilating the biased retrievals and extending CPSRs to assimilation of retrieval partial profiles (profiles with the biased retrievals discarded).  Our results show that not assimilating the biased retrievals resolves the fit/skill degradation in the upper troposphere but introduces fit/skill degradation in the middle and lower troposphere.  We show that the middle/lower tropospheric degradation is due to a: (i) ~17% reduction in the information content of the assimilated observations, and (ii) reductions in the amplitude of the transformed averaging kernels (reductions of ~70% for Mode 1, ~18% for Mode 2, and ~6% for Mode 3).  We are continuing to investigate methods to ameliorate the impacts of the biased retrievals.