ESA PyOSSE: Package for Observation System Simulation Experiments
ESA DA Edinburgh University ESA PyOSSE: Package for Observation System Simulation Experiments ESA-DA Logo - Link to ESA-DA Home Page
Reading University
Section Contents
[*] Home
[*] Python
[*] Documentation
[*] Publications
[*] Reference
[*] Installation
[*] Visualization
[*] 2014 CO2 (3-hourly)
[*] Full Chem
[*] GC CO2 2009-2015 (v2.1)
[*] New 2015 CH4 (daily)
[*] 2009.01-2017.03 CO2 (daily)
[*] UoE L4 flux (map)
[*] prior flux 2009-2015
[*] modified 2012 prior flux
[*] revised posterior flux for modified 2012 apriori
[*] obspack sampling
[*] 2015.01-2017.03 CO2 (3-hourly)
[*] 2009-2015 CH4 (revised)
[*] monthly natural flux (2009-2016)
[*] NCEO presentation
[*] CO2 flux from UoL v7.2
[*] 3-hourly GC CO2 (2014-2017)
[*] 3-hourly GC CO2 (2009-2013)
[*] 3-hourly CO2 flux (In-situ)
[*] 3-hourly CO2 flux (OCO2)
[*] 3-hourly CO2 (2009-2018)
[*] enkf_tansat
[*] 3-hourly CO2 (2020)
[*] FF comparison
[*] Insitu comparison
[*] OCO2 v10r inversion
[*] 2014 CO2 (3-hourly)
[*] 2015 CO2 (3-hourly)
[*] 2016 CO2 (3-hourly)
[*] 2017 CO2 (3-hourly)
[*] 2018 CO2 (3-hourly)
[*] 2019 CO2 (3-hourly)
[*] 2020 CO2 (3-hourly)
[*] 2021 CO2 (3-hourly)
[*] CASA NEE
 

Introduction

   Developing a new space-based observation system represents a substantial financial investment.   Observation System Simulation Experiments (OSSE) are a cost-effective numerical approach to realistically describe space-borne measurements and to evaluate their impact on current knowledge as part of preparing a science case for a particular space-borne mission; in our example experiment we quantify the impact of atmospheric measurements of a trace gas on improving our current prior understanding of surface fluxes (emission minus uptake) of that gas. The example OSSE, relevant to atmospheric composition, includes methods to: 1) simulate the 4-D distribution of atmospheric constituents and sample it as it is observed by the space-borne sensor; and 2) estimate the magnitude and distribution of surface fluxes by fitting prior model surface fluxes, which describe the physics, chemistry and biology underpinning the surface flux processes, to the pseudo observations using a data assimilation algorithm, accounting for model and observation errors. Despite the obvious limitations of OSSEs (Lahoz et al., 2006), self-consistent numerical experiments have been useful to evaluate newly proposed space-based instruments.  For example, the impact of data from the NASA Orbiting Carbon Observatory (OCO) was examined by several OSSE systems based on different transport models and different data assimilation techniques (e.g., Baker et al., 2006; Chevelliar et al, 2007; Feng et al., 2009).

 

  It is time-consuming to develop a robust, realistic OSSE for any individual space-borne sensor, reflecting the challenges associated with combining an atmosphere transport model, a surface (process or empirical) model, and an instrument observation model with modern data assimilation techniques.  Such a tool should also be flexible, considering all possible configurations of a proposed instrument and its potential applications, which can evolve during the design phase. We have developed the software package PyOSSE to help researchers to build their own OSSE systems with minimum overhead. PyOSSE consists of modules written in FORTRAN and Python, with emphasis on flexibility and simplicity, so that the user can easily understand, and apply these modules to their research.

  We have also provided a complete example OSSE system for an instrument based loosely on the specification of the NASA Orbiting Carbon Observatory that measured column-integrated dry-air CO2 mole fraction XCO2. The default specifications of the OCO instrument such as the scene-dependent averaging kernels, observation errors as well as aerosol and cloud climatology are taken from Boesch et al (2011). They can be extended or modified to describe other similar instruments. We have used the GEOS-Chem CTM as the default CTM to simulate the temporal and spatial distributions of atmospheric constituents, but the user is able to replace it with their own CTM without extensive changes  

  We have used an Ensemble Kalman Filter to digest simulated observations. The below plots show the (1) simulated OCO XCO2 observations for 2009.01; 2). prior model XCO2; 3). posterior model XCO2.


 

Simulated XCO2
To simulate XCO2, we sample CO2 concentrations simulated by GEOS-Chem along aqua orbits.
Prior Model XCO2
In the prior simulation, we have artificially enlarged the surface flux by 50%.

 

Posterior Model XCO2
In the inversion, we have used EnKF to optimally fit model simulations with observations.

 

Contact us
Email: lfeng@staffmail.ed.ac.uk