Determining the optimal soil temperature scheme for atmospheric modelling applications

The treatment of the land surface can have a significant impact on the performance of atmospheric models, influencing the surface energy balance and near surface atmospheric variables. In numerical weather prediction models it is especially important to reproduce the observed diurnal cycle in screen-level temperature, which requires an accurate representation of the surface temperature, and therefore an accurate and computationally efficient representation of soil heat storage and transfer. We present a technique for analysing the accuracy of numerical soil temperature schemes, and a methodology for choosing the optimal layer thicknesses for schemes with a given number of layers. Furthermore, the analysis suggests that first generation land surface schemes, which typically used a layer-average surface temperature, may be more accurate in this respect than the latest land surface schemes, which tend to use a 'skin' surface temperature boundary condition.