1	"Bias Determination and Inverse Source..." Bias Determination and Inverse Source Identification with MOPITT and MOZART
2	Role of CO in the tropopshere Oxidation capacity of the atmosphere main sink for OH, impacts t HC, GHG... impact on ozone Precursor of tropospheric ozone air quality climate Indoor and Urban pollutant people’s health
3	Global Budget of CO in the troposphere The nature of CO sources and sinks are fairly well established: 1. CO is a byproduct of the incomplete combustion of fossil fuel and biomass. 2. Significant amount of CO also comes from the oxidation of hydrocarbons, mainly methane. 3. CO averaged global lifetime is 2 months. 4. 90 % of CO sink is due to its reaction with the hydroxyl radicals OH (the rest being due to dry deposition). 5. CO sources and sink are quite variable in time and space. è CO global budget still highly uncertain
4	"For the first time," For the first time, scientists get a broader and finer picture of CO distribution thanks to the MOPITT instrument onboard NASA Terra satellite MOPITT which measures tropospheric content of CO covering the global surface of the Earth in a few days.
5	March 2000 Total column of CO : MOZART2 (top) and MOPITT (bottom)
6	July 2000 Total column of CO : MOZART2 (top) and MOPITT (bottom)
7	Hypothesis Transport in the model is perfect (for the CMDL/IMAGES inversion) Statistics of the observations known (mean and cov matrix) Statistics of the a priori sources known All Errors are gaussian and independent Chemistry weakly non linear. CO sink not optimized. a posteriori sources close ENOUGH to a priori è no big change in [OH] è Use linearized version of the model
8	The solution of the inverse problem minimizes a cost function
9
10	Optimal Interpolation Analytical solution x_aexists for linear problem (or weakly non-linear)
11	Principle The discrepancies between the observed and the modeled CO distributions are used to optimize poorly known parameters in the model – here, CO surface emissions.
12	"In this talk we present..." In this talk we present two results pertinent to assimilation and inverse modeling of trace gases in the troposphere. Both are related to MOPITT CO measurements. Persistent model biases can significantly affect the quality and success of the trace gas assimilation. A bias correction scheme proposed by Dee and Da Silva for humidity assimilation was adopted to MOPITT CO assimilation in a global 3-D CTM MOZART 2 and used to identify the biases and improve the assimilation. To our knowledge this is a first application of this technique to trace gas analysis. Uncertainties in the sources of tropospheric trace gases such as CO can be rather large. Errors in the source strength can thus severely influence model simulations and lead to incorrect results. We present final results of a 3-year project on numerical identification of global surface sources of CO obtained via inverse modeling with MOPITT and MOZART.
13	MOZART 2 MODEL 3-D chemistry-transport model Full tropospheric chemistry, with analyzed H₂O fields (NCEP and ECMWF analyses) ~ 60 species and ~ 200 reactions. 2.8^ox 2.8^ohorizontal resolution on the NCEP 28 sigma-vertical levels with ~ 7 levels below 850 mb 20 min time step
14	CO emitted by factories in Colorado on December 7 2000
15	Time evolution of CO plume
16	Modeling and Data Assimilation Physics-based numerical models can be used to simulate spread of air-borne agents. These models rely on a set of input parameters (winds, temperatures, surface fluxes, etc) that contain errors. Additionally, model representations of the actual physical processes also contain errors. In order to accurately monitor and predict evolution of air-borne pollutant one needs to reduce these errors using data in the process called Data Assimilation.
17
18	MOPITT AVERAGING KERNELS
19	MOPITT CO RETRIEVALS
20	Major mechanisms that control CO distribution
21
22	TRACER ASSIMILATION: VARIANCE EQUATION and MODEL (CTM) ERRORS
23	Optimization of the monthly CO emissions using Phase-I MOPITT data based on approach of Petrone et al., JGR,2002
24	Online bias-estimator in the CDA with the MOPITT CO retrievals
25	Two-stage bias correction algorithm
26	Bias-state evaluation algorithm for the MOPITT CO retrievals
27	Bias forecast equation
28	CO April 2001 at 700 mb
29	Fig 3
30
31
32
33	Seasonal Variations of the Surface CO
34	Seasonal Variations of the Surface CO
35
36	Looking at MOPITT minus FORECAST CO fields
37	Monthly mean OmF (ppbv) at 700 mb, March 2001
38	ASSIMILATION and CTM RESULTS
39	ASSIMILATION and CTM RESULTS
40	Seasonal Variations of the Surface CO
41
42	Tracer data analysis: Conclusions and Challenges
43	Mathematical Basis
44	Source Identification (Inverse Modeling)
45	Wireless Sensor Nets – Good Things Wireless sensor networks show a great promise in advancing our ability to provide data for an assimilation system. Since the sensors are inexpensive, data coverage can be increased dramatically. Wireless communication ability facilitates data collection and can eliminate operational delays.
46	Wireless Sensor Nets – Bad Things Low sensor cost often results in low S/N and low sensitivities and selectivities. Low production costs for the integrated board can lead to large (and often changing) biases, higher fault frequencies, and large errors. Unless a GPS chip and a GPS antenna are included on the board, position determination is a big issue.
47	What one can expect Lifetime: 5yr+ Calibration time: 5-10 minutes per sensor Time between calibrations: months to years Cost: $20 to $50 (tens), $10 to $15 (10,000's) Power: 50 to 500mW continuous during operation (adding GPS on board will reduce this)
48	Sensor Response Examples
49	Wireless Sensor Nets (from Kevin L. Moore, Director Center for Self-Organizing and Intelligent Systems Utah State University)
50	Wireless Sensor Nets (from Kevin L. Moore, Director Center for Self-Organizing and Intelligent Systems Utah State University)
51	Wireless Sensor Nets (from Kevin L. Moore, Director Center for Self-Organizing and Intelligent Systems Utah State University)
52	Wireless Sensor Nets (from Kevin L. Moore, Director Center for Self-Organizing and Intelligent Systems Utah State University)
53	Conclusions Modern design schemes for wireless sensor nets tend to include a software “integrator” and control module. Numerical atmospheric models and data assimilation methods readily provide such back end module for airborne agent tracking. Bayesian approach inherent in all data assimilation schemes provides a rigorous framework for dealing with low-cost sensor errors and biases. Wireless sensor net implementation issues can be rather complex. Practical problems related to coordinated communication, queries and power consumption will be reviewed tomorrow by Dr M. Murphy
54	Emissions Validation
55
56
57
58
59
60
61
62
63
64
65
66
67	Time Evolution of 150 ppb CO iso-surface
68
69	Jan, Feb, Mar, Apr
70	May, Jun, Jul, Aug
71	Sep, Oct, Nov, dec
72
73	TRACE-P
74	TOPSE, 6 flights
75	CMDL
76	CMDL Stations
77
78	TRACE-P
79	Introduction Carbon monoxide, CO, is a byproduct of fossil fuel and biomass incomplete combustion. The incomplete oxidation of hydrocarbons also produces substantial amounts of CO. CO is the principal sink of hydroxyl radicals OH in the free troposphere. Its global mean lifetime is 2 months. CO controls indirectly the lifetime of many other species, such as methane CH4. In presence of nitrogen oxides, NOx (>10-15 pptv), and sunlight, CO is a precursor of tropospheric ozone O3. The uncertainties on CO sources are as high as a factor of 2 to 4. The objective of the study is to combine the information contained in a numerical and in observations to improve CO surface emissions.