Abstract. Environment Canada has been developing a community environmental modelling system (Modélisation Environmentale Communautaire -MEC), which is designed to facilitate coupling between models focusing on different components of the earth system. The ultimate objective of MEC is to use the coupled models to produce operational forecasts. MESH (MEC -Surface and Hydrology), a configuration of MEC currently under development, is specialized for coupled land-surface and hydrological models. To determine the specific requirements for MESH, its different components were implemented on the Laurentian Great Lakes watershed, situated on the Canada-US border. This experiment showed that MESH can help us better understand the behaviour of different land-surface models, test different schemes for producing ensemble streamflow forecasts, and provide a means of sharing the data, the models and the results with collaborators and end-users. This modelling framework is at the heart of a testbed proposal for the Hydrologic Ensemble Prediction Experiment (HEPEX) which should allow us to make use of the North American Ensemble Forecasting System (NAEFS) to improve streamflow Correspondence to: A. Pietroniro (al.pietroniro@ec.gc.ca) forecasts of the Great Lakes tributaries, and demonstrate how MESH can contribute to a Community Hydrologic Prediction System (CHPS).

This paper presents an assessment of the operational system used by the Meteorological Service of Canada for producing near-real-time precipitation analyses over North America. The Canadian Precipitation Analysis (CaPA) system optimally combines available surface observations with numerical weather prediction (NWP) output in order to produce estimates of precipitation on a 15-km grid at each synoptic hour (0000, 0600, 1200, and 1800 UTC). The validation protocol used to assess the quality of the CaPA has demonstrated the usefulness of the system for producing reliable estimates of precipitation over Canada, even in areas with few or no weather stations. The CaPA is found to be better in autumn, spring, and winter than in summer. This is because of the difficulty in correctly producing convective precipitation in the NWP because of the low spatial resolution of the meteorological model. An investigation of the quality of the precipitation analyses in the 15 terrestrial ecozones of Canada indicates the need to have a sufficient number of observations (at least ~1.17 stations per 10 000 km2) in order to produce a precipitation analysis that is significantly better than the raw NWP product. Improvements of the CaPA system by including provincial networks as well as radar and satellite information are expected in the future.

When evaluating the reliability of an ensemble prediction system, it is common to compare the root-mean-square error of the ensemble mean to the average ensemble spread. While this is indeed good practice, two different and inconsistent methodologies have been used over the last few years in the meteorology and hydrology literature to compute the average ensemble spread. In some cases, the square root of average ensemble variance is used, and in other cases, the average of ensemble standard deviation is computed instead. The second option is incorrect. To avoid the perpetuation of practices that are not supported by probability theory, the correct equation for computing the average ensemble spread is obtained and the impact of using the wrong equation is illustrated.

No abstract

SUMMARYEnsembles of meteorological forecasts can both provide more accurate long-term forecasts and help assess the uncertainty of these forecasts. No single method has however emerged to obtain large numbers of equiprobable scenarios from such ensembles. A simple resampling scheme, the 'best member' method, has recently been proposed to this effect: individual members of an ensemble are 'dressed' with error patterns drawn from a database of past errors made by the 'best' member of the ensemble at each time step. It has been shown that the best-member method can lead to both underdispersive and overdispersive ensembles. The error patterns can be rescaled so as to obtain ensembles which display the desired variance. However, this approach fails in cases where the undressed ensemble members are already overdispersive. Furthermore, we show in this paper that it can also lead to an overestimation of the probability of extreme events. We propose to overcome both difficulties by dressing and weighting each member differently, using a different error distribution for each order statistic of the ensemble. We show on a synthetic example and using an operational ensemble prediction system that this new method leads to improved probabilistic forecasts, when the undressed ensemble members are both underdispersive and overdispersive.