On the Use of Linearized Measurements from Satellite Observing Systems for Global Air Quality Forecasting

As satellites are at the forefront in numerical weather prediction (NWP), future observations of tropospheric constituents from geostationary satellites will play a major role in regional-to-global air quality forecasting (AQF) systems. These observations provide a wealth of information that fully complements current AQF capability. However, the potential for higher spatio-temporal and spectral resolution of these observations entails a more efficient and practical AQF system that can handle and effectively assimilate the large volume of data in operational mode. This includes a reasonably accurate radiative transfer model (RTM) within the AQF system to calculate the observation operator across multiple spectral measurements. In addition, some practical forms of reducing the data without significant loss of information must be implemented for computational expediency. Here, we explore the applicability of using a reduced form of linearized measurements as input to the AQF system. In particular, we revisit the concept of using the Jacobians, which are calculated nonetheless during the retrieval of these constituents, in lieu of a full RTM calculation within the AQF system. A singular-value-decomposition of the Jacobian can then be carried out for each measurement to reduce the volume of data to be assimilated. This concept serves as a practical alternative to conventional approaches like full radiance or retrieval data assimilation. We demonstrate the applicability of this concept using the radiance measurements from the NASA/Terra Measurement of Pollution In The Troposphere instrument (MOPITT) and the corresponding Jacobians from the MOPITTv4 retrieval algorithm. Assimilation experiments are carried out under an ensemble-based chemical data assimilation framework that mimics an advance AQF system. We show results of the assimilation and comparisons with results from a retrieval assimilation of the same data.