ICLASS 1.1, a variational Inverse modelling framework for the Chemistry Land-surface Atmosphere Soil Slab model: description, validation, and application

Abstract

Abstract. This paper provides a description of ICLASS 1.1, a variational Inverse modelling framework for the Chemistry Land-surface Atmosphere Soil Slab model. This framework can be used to study the atmospheric boundary layer, surface layer, or the exchange of gases, moisture, heat, and momentum between the land surface and the lower atmosphere. The general aim of the framework is to allow the assimilation of various streams of observations (fluxes, mixing ratios at multiple heights, etc.) to estimate model parameters, thereby obtaining a physical model that is consistent with a diverse set of observations. The framework allows the retrieval of parameters in an objective manner and enables the estimation of information that is difficult to obtain directly by observations, for example, free tropospheric mixing ratios or stomatal conductances. Furthermore, it allows the estimation of possible biases in observations. Modelling the carbon cycle at the ecosystem level is one of the main intended fields of application. The physical model around which the framework is constructed is relatively simple yet contains the core physics to simulate the essentials of a well-mixed boundary layer and of the land–atmosphere exchange. The model includes an explicit description of the atmospheric surface layer, a region where scalars show relatively large gradients with height. An important challenge is the strong non-linearity of the model, which complicates the estimation of the best parameter values. The constructed adjoint of the tangent linear model can be used to mitigate this challenge. The adjoint allows for an analytical gradient of the objective cost function, which is used for minimisation of this function. An implemented Monte Carlo way of running ICLASS can further help to handle non-linearity and provides posterior statistics on the estimated parameters. The paper provides a technical description of the framework, includes a validation of the adjoint code, in addition to tests for the full inverse modelling framework, and a successful example application for a grassland in the Netherlands.