How can observations be used to best evidence the influence on climate of human activities, among other forcings? Statistical methods of Detection and Attribution (D&A)were designed to answer this question. Conventional D&A methods are based on linear regression of spatial or temporal patterns extracted from one or several climate models ("optimal fingerprinting").
On the other hand, how can observations be used to best constrain a numerical model’s state variables and parameters? Methods of Data Assimilation (DA) meet this general purpose. Could D&A methods take advantage of recent progresses in DA? We argue that one may hope so. Indeed, under an inverse problem formulation of D&A, observations can be seen as a complex function of the forcings consisting of the full climate model itself. Under this perspective, D&A consists "merely" in reconstructing forcings from available observations by inverting the full climate model itself - a challenge for which recent DA-based parameter estimation schemes might be an answer.
We will discuss this general idea and illustrate it qualitatively by applying a simple data assimilation algorithm (AEKF) to a 1D radiative column model. The model simulates the vertical temperature structure of the atmosphere resulting from prescribed vertical profiles of key optical properties, as well as its dynamics when these optical properties are affected by changes in atmospheric constituents. In particular, it allows to represent the characteristic pattern of cooling of the stratosphere and warming of the troposphere under greenhouse gas forcing, which is a historical fingerprint used in D&A. We show that it is possible to reconstruct these forcings from the observation of the evolution of the vertical temperature profile based on data assimilation and thereby to evidence the influence of anthropogenic forcing on the evolution of climate.