Efficiency and Sensitivity Analysis of Observation Networks for Atmospheric Inverse Modelling with Emissions

TL;DR

This paper introduces a theoretical framework combining Kalman filtering, ensemble methods, and singular value decomposition to evaluate and optimize observation networks for atmospheric inverse modeling, focusing on emissions and initial conditions.

Contribution

It presents a novel quantitative assessment method for measurement configurations, enhancing the optimization of initial values and emission rates in atmospheric models.

Findings

Identifies optimal observation network configurations for improved model control.

Determines the sensitivity of observations to model parameters.

Provides a method to evaluate the efficiency of different measurement setups.

Abstract

The controllability of advection-diffusion systems, subject to uncertain initial values and emission rates, is estimated, given sparse and error affected observations of prognostic state variables. In predictive geophysical model systems, like atmospheric chemistry simulations, different parameter families influence the temporal evolution of the system.This renders initial-value-only optimisation by traditional data assimilation methods as insufficient. In this paper, a quantitative assessment method on validation of measurement configurations to optimize initial values and emission rates, and how to balance them, is introduced. In this theoretical approach, Kalman filter and smoother and their ensemble based versions are combined with a singular value decomposition, to evaluate the potential improvement associated with specific observational network configurations. Further, with the same singular vector analysis for the efficiency of observations, their sensitivity to model control can be identified by determining the direction and strength of maximum perturbation in a finite-time interval.