Soil moisture has an obvious dependence on atmospheric variability, and also feeds back on the atmosphere on a range of timescales. If soil moisture variability and the coupling to the atmosphere could be well represented in models, this could provide increased forecast skill, particularly at seasonal timescales. The Global Land-Atmosphere Coupling Experiment (GLACE; Koster et al., 2006, J. Hydromet.) provided a method for quantifying the influence of a model's soil moisture on its precipitation. In the GLACE model intercomparison project, the land-atmosphere coupling in the Met Office Hadley Centre's model, HadAM3, was found to be one of the weakest of the twelve models assessed.

HadGEM3-A is the atmospheric component of the latest version of the Met Office Hadley Centre's model, currently under development. Compared to HadAM3, it has increased horizontal and vertical resolution and incorporates updated representation of the land surface, boundary layer and convection as well as a new dynamical core. We present an analysis of HadGEM3-A using the GLACE framework, and demonstrate a strengthening of the coupling diagnostic compared to HadAM3. The existence of this stronger signal enables an investigation into the factors contributing to the coupling strength. In particular, a correction was recently made to the way in which soil hydraulic parameters are specified in the model. This produced an increase in moisture stress for surface evaporation over most of the globe. We discuss how this simple, universal change impacts on the GLACE coupling signal.