Le Duc

Spatial-temporal fractions verification for high-resolution ensemble forecasts

Experiments with two ensemble systems of resolutions 10 km (MF10km) and 2 km (MF2km) were designed to examine the value of cloud-resolving ensemble forecast in predicting precipitation on small spatio-temporal scales. Since the verification was performed on short-term precipitation at high resolution, uncertainties from small-scale processes caused the traditional verification methods to be inconsistent with the subjective evaluation. An extended verification method based on the Fractions Skill Score (FSS) was introduced to account for these uncertainties. The main idea is to extend the concept of spatial neighbourhood in FSS to the time and ensemble dimension. The extension was carried out by recognising that even if ensemble forecast is used, small-scale variability still exists in forecasts and influences verification results. In addition to FSS, the neighbourhood concept was also incorporated into reliability diagrams and relative operating characteristics to verify the reliability and resolution of two systems. The extension of FSS in time dimension demonstrates the important role of temporal scales in short-term precipitation verification at small spatial scales. The extension of FSS in ensemble space is called the ensemble FSS, which is a good representative of FSS for ensemble forecast in comparison with the FSS of ensemble mean. The verification results show that MF2km outperforms MF10km in heavy rain forecasts. In contrast, MF10km was slightly better than MF2km in predicting light rains, suggesting that the horizontal resolution of 2 km is not necessarily enough to completely resolve convective cells.

GPS PWV Assimilation with the JMA Nonhydrostatic 4DVAR and Cloud Resolving Ensemble Forecast for the 2008 August Tokyo Metropolitan Area Local Heavy Rainfalls

On 5th August 2008, scattering local heavy rainfalls occurred at various places over the Tokyo metropolitan area, and five drainage workers were claimed by an abrupt increase of water level. The Japan Meteorological Agency (JMA) operational mesoscale model of the day failed to predict occurrence of the local heavy rainfalls, which were brought about by deep convective cells developed on the unstable atmospheric condition without strong synoptic/orographic forcings. A 11-member mesoscale ensemble prediction with a horizontal resolution of 10 km was conducted using the operational mesoscale analysis of JMA and perturbations of the JMA global one-week ensemble prediction system as the initial condition and the initial and lateral boundary perturbations, but the intense rains exceeding 20 mm/3 h were hardly predicted. A downscaling ensemble forecast experiment with a horizontal resolution of 2 km was conducted using the 6 h forecast of the 10 km ensemble as the initial and boundary conditions. Scattered intense rains were predicted in some ensemble members, but their locations and distribution were insufficient. The total precipitatable water vapor (PWV ) observed by the GNSS Earth Observation Network System (GEONET) of Geospatial Information Authority of Japan showed that the JMA mesoscale analysis given by the hydrostatic Meso-4DVAR underestimated water vapor over the Tokyo metropolitan area. To modify the initial condition, a reanalysis data assimilation experiment was conducted with the JMA’s nonhydrostatic 4DVAR (JNoVA) , where PWV data from GEONET were assimilated 2.5 days with 3-h data assimilation cycles. The 2 km downscale ensemble run from the JNoVA analysis properly predicted the areas of scattering local heavy rains. Threat scores and ROC area skill scores suggest that even in the ensemble prediction, accuracy of initial condition is critical to numerically predict small scale convective rains. Fractions skill scores indicated the value of the cloud resolving ensemble forecast for such the unforced convective rain case.

Mesoscale Hybrid Data Assimilation System based on JMA Nonhydrostatic Model

AbstractThis paper discusses the benefits of using a hybrid ensemble Kalman filter and four-dimensional variational (4D-Var) data assimilation (DA) system rather than a 4D-Var system employing the National Meteorological Center (NMC, now known as NCEP) method (4D-Var-Bnmc) to predict severe weather events. An adjoint-based 4D-Var system was employed with a background error covariance matrix constructed from the NMC method and perturbations in a local ensemble transform Kalman filter system. The DA systems are based on the Japan Meteorological Agency’s nonhydrostatic model. To reduce the sampling noise, three types of implementation (the spatial localization, spectral localization, and neighboring ensemble approaches) were tested. The assimilation of a pseudosingle observation of sea level pressure located at a tropical cyclone (TC) center yielded analysis increments physically consistent with what is expected of a mature TC in the hybrid systems at the beginning of the assimilation window, whereas analogo...

On Cost Functions in the Hybrid Variational–Ensemble Method

AbstractIn the hybrid variational–ensemble data assimilation schemes preconditioned on the square root of background covariance , is a linear map from the model space to a higher-dimensional space. Because of the use of the nonsquare matrix , the transformed cost function still contains the inverse of . To avoid this inversion, all studies have used the diagonal quadratic form of the background term in practice without any justification. This study has shown that this practical cost function belongs to a class of cost functions that come into play whenever the minimization problem is transformed from the model space to a higher-dimension space. Each such cost function is associated with a vector in the kernel of (Ker), leading to an infinite number of these cost functions in which the practical cost function corresponds to the zero vector. These cost functions are shown to be the natural extension of the transformed one from the orthogonal complement of Ker to the full control space.In practice, these cos...