The schemes proposed in part I are tested using single‐column data sets from tropical field experiments (GATE, BOMEX, ATEX) and an arctic air‐mass transformation. Both the deep and shallow schemes perform well. The sensitivity of the schemes to adjustable parameters is also studied. Preliminary global forecasts show significant improvements in the global surface fluxes and mean tropical temperature tendency over the operational Kuo convection scheme.

This review discusses the land‐surface‐atmosphere interaction using observations from two North American field experiments (First International Satellite Land Surface Climatology Project Field Experiment (FIFE) and Boreal Ecosystem Atmosphere Study (BOREAS)) and the application of research data to the improvement of land surface and boundary layer parameterizations in the European Centre for Medium‐Range Weather Forecast (ECMWF) global forecast model. Using field data, we discuss some of the diurnal and seasonal feedback loops controlling the net surface radiation and its partition into the surface sensible and latent heat fluxes and the ground heat flux. We consider the impact on the boundary layer evolution and show the changes in the diurnal cycle with soil moisture in midsummer. We contrast the surface energy budget over the tropical oceans with that over both dry and wet land surfaces in summer. Results from a new ECMWF model with four predicted soil layers illustrate the interaction between the soil moisture reservoir, evaporation and precipitation on different timescales and space scales. An analysis of an ensemble of 30‐day integrations for July 1993 (the month of the Mississippi flood) showed a large sensitivity of the monthly precipitation pattern (and amount) to different initial soil moisture conditions. Short‐range forecasts with old and new land surface and boundary layer schemes showed that the new scheme produced much better precipitation forecasts for the central United States because of a more realistic thermodynamic structure, which in turn resulted from improved evaporation in an area that is about 1‐day upstream. The results suggest that some predictability exists in the extended range as a result of the memory of the soil moisture reservoir. We also discuss briefly the problem of soil moisture initialization in a global forecast model and summarize recent experience with nudging of soil moisture at ECMWF and improvements in the surface energy budget coming from the better prediction of clouds.

A scheme for the representation of subgrid-scale orography (SSO) in numerical weather prediction and climate models is presented. The new scheme arose in part from a desire to represent nonlinear low-level mountain drag effects not currently parametrized. An important feature of the scheme is that it deals explicitly with a low-level flow which is 'blocked', when the effective height of the subgrid-scale orography is sufficiently high. In this new scheme, it is assumed that, for this 'blocked' flow, separation occurs at the mountain flanks, resulting in a form drag. This drag is parametrized on model levels which are intersected by the SSO, and provides a dynamically based replacement for envelope orography. The upper part of the low-level flow goes over the orography and generates gravity waves. At the model resolutions considered (T106 and T213) it is assumed that the length scales characteristic of the SSO are sufficiently small for the Coriolis force to be neglected. The various parameters of the scheme are adjusted using an off-line procedure in which the scheme is used to estimate the mountain drag and the momentum profiles above the Pyrenees; and these estimates are validated with the PYREX data. Forecasts using T106 and T213 resolutions with this new scheme, and with mean orography, show that the forecast mountain drag consistently reproduces the drag measured during PYREX whenever the flow component normal to the ridge is large. Isentropic flow diagnostics, further, show that the new scheme has a realistic impact on the flow dynamics, reinforcing the low-level wake observed in mesoscale analyses of the flow. With this new scheme and a mean orography, the ECMWF model outperformed, in forecast skill, a version of the model which had an envelope orography and the old gravity-wave-drag scheme, while no longer suffering any disadvantages of envelope orography. The proposed low-level drag parametrization should also be relevant at model horizontal resolutions much higher than T213.

The sensitivity to the horizontal resolution of the climate, anthropogenic climate change, and seasonal predictive skill of the ECMWF model has been studied as part of Project Athena—an international collaboration formed to test the hypothesis that substantial progress in simulating and predicting climate can be achieved if mesoscale and subsynoptic atmospheric phenomena are more realistically represented in climate models. In this study the experiments carried out with the ECMWF model (atmosphere only) are described in detail. Here, the focus is on the tropics and the Northern Hemisphere extratropics during boreal winter. The resolutions considered in Project Athena for the ECMWF model are T159 (126 km), T511 (39 km), T1279 (16 km), and T2047 (10 km). It was found that increasing horizontal resolution improves the tropical precipitation, the tropical atmospheric circulation, the frequency of occurrence of Euro-Atlantic blocking, and the representation of extratropical cyclones in large parts of the Northern Hemisphere extratropics. All of these improvements come from the increase in resolution from T159 to T511 with relatively small changes for further resolution increases to T1279 and T2047, although it should be noted that results from this very highest resolution are from a previously untested model version. Problems in simulating the Madden–Julian oscillation remain unchanged for all resolutions tested. There is some evidence that increasing horizontal resolution to T1279 leads to moderate increases in seasonal forecast skill during boreal winter in the tropics and Northern Hemisphere extratropics. Sensitivity experiments are discussed, which helps to foster a better understanding of some of the resolution dependence found for the ECMWF model in Project Athena.

SUMMARYIn the context of the European Cloud Systems project, the problem of the simulation of the diurnal cycle of convective precipitation over land is addressed with the aid of cloud-resolving (CRM) and single-column (SCM) model simulations of an idealized midlatitude case for which observations of large-scale and surface forcing are available. The CRM results are compared to different versions of the European Centre for Medium-Range Weather Forecasts (ECMWF) convection schemes using different convective trigger procedures and convective closures. In the CRM, maximum rainfall intensity occurs at 15 h (local time). In this idealized midlatitude case, most schemes do not reproduce the afternoon precipitation peak, as (i) they cannot reproduce the gradual growth (typically over 3 hours) of the deep convective cloud layer and (ii) they produce a diurnal cycle of precipitation that is in phase with the diurnal cycle of the convective available potential energy (CAPE) and the convective inhibition (CIN), consistent with the parcel theory and CAPE closure used in the bulk mass-flux scheme. The scheme that links the triggering to the large-scale vertical velocity gets the maximum precipitation at the right time, but this may be artificial as the vertical velocity is enforced in the single-column context.The study is then extended to the global scale using ensembles of 72-hour global forecasts at resolution T511 (40 km), and long-range single 40-day forecasts at resolution T159 (125 km) with the ECMWF generalcirculation model. The focus is on tropical South America and Africa where the diurnal cycle is most pronounced. The forecasts are evaluated against analyses and observed radiosonde data, as well as observed surface and satellite-derived rainfall rates. The ECMWF model version with improved convective trigger produces the smallest biases overall. It also shifts the rainfall maximum to 12 h compared to 9.5 h in the original version. In contrast to the SCM, the vertical-velocity-dependent trigger does not further improve the phase of the diurnal cycle. However, further work is necessary to match the observed 15 h precipitation peak.