The impact of lakes in numerical weather prediction is investigated in a set of global simulations performed with the ECMWF Integrated Forecasting System (IFS). A Fresh shallow-water Lake model (FLake) is introduced allowing the coupling of both resolved and subgrid lakes (those that occupy less than 50% of a grid-box) to the IFS atmospheric model. Global fields for the lake ancillary conditions (namely lake cover and lake depth), as well as initial conditions for the lake physical state, have been derived to initialise the forecast experiments. The procedure for initialising the lake variables is described and verified with particular emphasis on the importance of surface water temperature and freezing conditions. The response of short-range near surface temperature to the representation of lakes is examined in a set of forecast experiments covering one full year. It is shown that the impact of subgrid lakes is beneficial, reducing forecast error over the Northern territories of Canada and over Scandinavia particularly in spring and summer seasons. This is mainly attributed to the lake thermal effect, which delays the temperature response to seasonal radiation forcing.Keywords: lakes impact in NWP; global lake simulations; lake-temperature; lake-ice(Published: 23 February 2012)Citation: Tellus A 2012, 64, 15829, DOI: 10.3402/tellusa.v64i0.15829