Impact of time sampling on atmospheric energy budget residuals

Diagnostic time-mean budgets of energy and water are evaluated in many atmo spheric process studies. The errors of budget-derived quantities like sub-g ridscale fluxes or diabatic heating are governed by the errors of the budge ts. Here we consider 3D-budgets on the meso-beta scale over Europe. They ar e compiled from analyses of state quantities available from forecast centre s. In the present study we found that the mandatory 6 hours sampling interval between synoptic observations is the main error source for routine time-mea n budgets. The errors have been quantified (i) by first sampling forecast d ata of the German Europamodell every 5 minutes and averaging them over 12 h ours (reference budget), and (ii) by sampling the same data only every 6 ho urs and averaging these also over 12 hours (routine budget). With this meth od we find that routine budgets in single atmospheric meso-beta scale colum ns show relative random errors of typically 200% and systematic errors of u p to 20%, exclusively due to undersampling. Thus routine budgets, if applie d to specific days at individual locations, cannot be expected to yield use ful results, except perhaps for cases with extremely strong signal. Composi ting over several hundreds of columns with similar weather reduces the rand om budget error down to about 50%; this seems to be the best one can achiev e for routine budgets. The systematic error of some budget quantities is ca used by a correlation between the time of occurence of certain processes (m ainly convection) and the sampling times. While this error cannot be reduce d through compositing, we find that it can be crudely estimated by using di fferent time averaging methods. As application for this method we determine sub-gridscale budget quantities over the BALTEX catchment (August-September 1995) for an ensemble of conve ctively active and an ensemble of rain-active columns. For the ensemble mea n profiles we find, in terms of the diagnosed sub-gridscale test quantities diabatic heating and vertical moist enthalpy flux divergence, that their a ccuracy is sufficient to detect statistically significant differences betwe en both ensembles. The diabatic heating is about the same for both ensemble s, while the flux divergence in the convective ensemble is about three time s as large as in the rain ensemble.