Developed for short-term (0-48 h) wind power forecasting purposes, high-resolution meteorological forecasts for Eastern Canada are available from Environment Canada's Numerical Weather Prediction (NWP) model configured on a limited area (GEM-LAM). This paper uses 3 years of forecast data from this model for the region of North Cape (Prince Edward Island, Canada). Although the model resolution is relatively high (2.5 km), statistical analysis and site inspection reveal that the model does not have a sufficiently refined grid to properly represent the meteorological phenomena over this complex coastal site. To cope with such representation error, a generalized Geophysic Model Output Statistics (GMOS) module is developed and applied to reduce the forecast error of the NWP forecasts. GMOS differs from other Model Output Statistics (MOS) that are widely used by meteorological centres in the following aspects: (i) GMOS takes into account the surrounding geophysical parameters such as surface roughness, terrain height, etc., along with wind direction; (ii) GMOS can be directly applied for model output correction without any training.Compared with other methods, GMOS using a multiple grid point approach improves the GEM-LAM predictions root mean squared error by 1-5% for all time horizons and most meteorological conditions. Also, the topographic signature of the forecast error (uneven directional distribution of the forecast error related to the surface characteristics) due to misrepresentation issues is significantly reduced. The NWP forecasts combined with GMOS outperform the persistence model from a 2 h horizon, instead of 3 h using MOS. Finally, GMOS is applied and validated at two other sites located in New Brunswick, Canada. Similar improvements on the forecasts were observed, thus showing the general applicability of GMOS.

Two types of approaches are commonly used for estimating the impact of arbitrary subsets of observations on short-range forecast error. The first was developed for variational data assimilation systems and requires the adjoint of the forecast model. Comparable approaches were developed for use with the ensemble Kalman filter and rely on ensembles of forecasts. In this study, a new approach for computing observation impact is proposed for ensemble–variational data assimilation (EnVar). Like standard adjoint approaches, the adjoint of the data assimilation procedure is implemented through the iterative minimization of a modified cost function. However, like ensemble approaches, the adjoint of the forecast step is obtained by using an ensemble of forecasts. Numerical experiments were performed to compare the new approach with the standard adjoint approach in the context of operational deterministic NWP. Generally similar results are obtained with both approaches, especially when the new approach uses covariance localization that is horizontally advected between analysis and forecast times. However, large differences in estimated impacts are obtained for some surface observations. Vertical propagation of the observation impact is noticeably restricted with the new approach because of vertical covariance localization. The new approach is used to evaluate changes in observation impact as a result of the use of interchannel observation error correlations for radiance observations. The estimated observation impact in similarly configured global and regional prediction systems is also compared. Overall, the new approach should provide useful estimates of observation impact for data assimilation systems based on EnVar when an adjoint model is not available.

It is still common to neglect the spatial error correlations of assimilated observations in numerical weather prediction systems because no practical approach is available to account for them when the number of observations with correlated error is large or when these observations are non‐uniformly distributed. Instead, it is common practice to inflate observation error variances to avoid overfitting large scales and spatially thin observations to reduce error correlations between remaining observations, although both methods generally sacrifice small‐scale information. Inspired by previous work on assimilating the difference between adjacent observations (so‐called spatial difference observations), this study aims at combining direct observations with spatial difference observations in the assimilation to extract both large‐ and smaller‐scale information from observations with spatially correlated errors, while still neglecting these error correlations. Experiments performed in a simplified 1D context over a periodic domain show that the combined approach is numerically equivalent to directly assimilating observations available at every grid point using non‐diagonal observation error covariances based on a first‐order autoregressive correlation function. In a case where observation error correlations have a different structure (e.g. Gaussian), the true observation error correlations are not perfectly taken into account by the combined approach, but it still efficiently extracts information to correct the scales that have the most errors. Combining direct observations with spatial difference observations proves complementary and experimental results show lower analysis errors for both large and intermediate scales. More specifically, while neglecting spatial observation error correlations, the combined approach provides results with lower analysis error than the direct approach, especially when spatial error correlations are large and when direct observations are spatially thinned.

The exigencies of the global community toward Earth system science will increase in the future as the human population, economies, and the human footprint on the planet continue to grow. This growth, combined with intensifying urbanization, will inevitably exert increasing pressure on all ecosystem services. A unified interdisciplinary approach to Earth system science is required that can address this challenge, integrate technical demands and long-term visions, and reconcile user demands with scientific feasibility. Together with the research arms of the World Meteorological Organization, the Young Earth System Scientists community has gathered early-career scientists from around the world to initiate a discussion about frontiers of Earth system science. To provide optimal information for society, Earth system science has to provide a comprehensive understanding of the physical processes that drive the Earth system and anthropogenic influences. This understanding will be reflected in seamless prediction systems for environmental processes that are robust and instructive to local users on all scales. Such prediction systems require improved physical process understanding, more high-resolution global observations, and advanced modeling capability, as well as high-performance computing on unprecedented scales. At the same time, the robustness and usability of such prediction systems also depend on deepening our understanding of the entire Earth system and improved communication between end users and researchers. Earth system science is the fundamental baseline for understanding the Earth’s capacity to accommodate humanity, and it provides a means to have a rational discussion about the consequences and limits of anthropogenic influence on Earth. Without its progress, truly sustainable development will be impossible.

Although many near-surface wind observations are available, very few are assimilated over land mainly due to sub-grid scale topographic interactions with the flow. The main objectives of this study are to understand the impact of near-surface wind observations on the analysis and to point out aspects that need to be improved to make a better use of these observations. A geo-statistical observation operator has been developed to correct for systematic and representativeness errors. Assimilation experiments were performed in a simplified context, assimilating only near-surface wind observations over land in the ensemble-variational data assimilation system developed at Environment Canada. Due to the background-error covariances, the assimilation of near-surface wind observations impacts the lower part of the atmosphere. The resulting correction has been evaluated by verifying the analyses against non-assimilated collocated radiosonde data. This assessment also made it possible to estimate the observation error variance to strike a balance between having an important impact at the surface and maintaining a good vertical fit to upper air observations. Results from 1 month of assimilation experiments show that the geostatistical operator eliminates biases and significantly reduces representativeness errors as well as observation error correlations in the analysis, mainly over complex terrain. Results also show that flow-dependent background error covariances from ensembles provide better vertical information propagation than static error statistics. Overall, the analysis fit to non-assimilated collocated radiosonde observations is improved when assimilating wind observations from surface stations.

High-resolution flow-dependent background error covariances can allow for a better usage of dense observation networks in applications of data assimilation for numerical weather prediction. The generation of high-resolution ensembles, however, can be computationally cost prohibitive. In this study, practical and low-cost ensemble generation methods are presented and compared against both global and regional ensemble Kalman filters (G-EnKF and R-EnKF, respectively). The goal is to provide limited-area deterministic assimilation schemes with higher-resolution flow-dependent background error covariances that perform at least as well as those from the G-EnKF when assimilating the same observations. The low-cost methods are based on short-range regional ensemble forecasts initialized from 1) deterministic analysis plus balanced perturbations (filter free approach) and 2) a simplified ensemble square root filter (S-EnSRF), centered on deterministic analyses. The resulting ensembles from the different approaches are used within a 4D ensemble–variational (4D-EnVar) assimilation system covering most of Canada and the northern United States. Diagnostic results show that the mean is an important component of the ensembles. Results also show that the persistence of the homogeneous characteristics of the perturbations in the filter free approach makes this method unsuited for short assimilation time windows since some error structures take longer to develop. The S-EnSRF approach overcomes this limitation by recycling part of the prior perturbations. Results from 1-month assimilation experiments show that the S-EnSRF and R-EnKF experiments provide forecasts of similar quality to those from G-EnKF. Furthermore, results from precipitation verification indicate that the R-EnKF experiment provides the best precipitation accumulation predictions over 24-h periods.

This study introduces an experimental regional assimilation configuration for a 4D ensemble–variational (4D-EnVar) deterministic weather prediction system. A total of 16 assimilation experiments covering July 2014 are presented to assess both experimental regional climatological background error covariances and updates in the treatment of flow-dependent error covariances. The regional climatological background error covariances are estimated using statistical correlations between variables instead of using balance operators. These error covariance estimates allow the analyses to fit more closely with the assimilated observations than when using the lower-resolution global background error covariances (due to shorter correlation scales), and the ensuing forecasts are significantly improved. The use of ensemble-based background error covariances is also improved by reducing vertical and horizontal localization length scales for the flow-dependent background error covariance component. Also, reducing the number of ensemble members employed in the deterministic analysis (from 256 to 128) reduced computational costs by half without degrading the accuracy of analyses and forecasts. The impact of the relative contributions of the climatological and flow-dependent background error covariance components is also examined. Results show that the experimental regional system benefits from giving a lower (higher) weight to climatological (flow-dependent) error covariances. When compared with the operational assimilation configuration of the continental prediction system, the proposed modifications to the background error covariances improve both surface and upper-air RMSE scores by nearly 1%. Still, the use of a higher-resolution ensemble to estimate flow-dependent background error covariances does not yet provide added value, although it is expected to allow for a better use of dense observations in the future.

This study examines the assimilation of near-surface wind observations over land to improve wind nowcasting and short-term tropospheric forecasts. A new geostatistical operator based on geophysical model output statistics (GMOS) is compared with a bilinear interpolation scheme (Bilin). The multivariate impact on forecasts and the temporal evolution of the analysis increments produced are examined as well as the influence of background error covariances on different components of the prediction system. Results show that Bilin significantly degrades surface and upper-air fields when assimilating only wind data from 4942 SYNOP stations. GMOS on the other hand produces smaller increments that are in better agreement with the model state. It leads to better short-term near-surface wind forecasts and does not deteriorate the upper-air forecasts. The information persists longer in the system with GMOS, although the local improvements do not propagate beyond 6-h lead time. Initial model tendencies indicate that the mass field is not significantly altered when using static error covariances and the boundary layer parameterizations damp the poorly balanced increment locally. Conversely, most of the analysis increment is propagated when using flow-dependent error statistics. It results in better balanced wind and mass fields and provides a more persistent impact on the forecasts. Forecast accuracy results from observing system experiments (assimilating SYNOP winds with all observations used operationally) are generally neutral. Nevertheless, forecasts and analyses from GMOS are more self-consistent than those from both Bilin and a control experiment (not assimilating near-surface winds over land) and the information from the observations persists up to 12-h lead time.