Journal of Meteorological Research最新文献

Numerical weather prediction of wind speed requires statistical postprocessing of systematic errors to obtain reliable and accurate forecasts. However, use of postprocessing models is often undesirable for extreme weather events such as gales. Here, we propose a postprocessing algorithm based on a gale-aware deep attention network to simultaneously improve wind speed forecasts and gale area warnings. Specifically, the algorithm includes both a galeaware loss function that focuses the model on potential gale areas, and an observation station supervision strategy that alleviates the problem of missing extreme values caused by data gridding. The effectiveness of the proposed model was verified by using data from 235 wind speed observation stations. Experimental results show that our model can produce wind speed forecasts with a root-mean-square error of 1.1547 m s⁻¹, and a Hanssen–Kuipers discriminant score of 0.517, performance that is superior to that of the other postprocessing algorithms considered.

At kilometer and sub-kilometer resolutions, known as the numerical gray zone for boundary layer turbulence, the atmospheric boundary layer turbulence becomes partially resolved and partially subgrid-scale (SGS) in a numerical model, thus requiring scale-adaptive turbulence schemes. Such schemes are often built by modifying the existing parameterizations, either the planetary boundary layer (PBL) schemes or the large-eddy simulation (LES) closures, to produce the right SGS turbulent fluxes at gray zone resolutions. However, the underlying forcings responsible for the changes in the vertical turbulent fluxes are largely ignored in these approaches. This study follows the original approach of Wyngaard (2004) and analyzes the turbulent buoyancy and momentum flux budgets, to gain a better understanding of the variations of flux forcings at gray zone resolutions. The investigation focuses on the pressure covariance term, which is one of the most dominant terms in the budget equations. By using the coarse-grained LES of a dry convective boundary layer (CBL) case as reference, two widely-used pressure covariance models are evaluated and optimized across the gray zone resolution range. The optimized linear model is further evaluated a priori against another dry CBL case with a different bulk stability, and a shallow-cumulus-topped boundary layer case. The model applies well to both cases, and notably shows good performance for the cloud layer. Based on the analysis of the flux forcings and the optimized pressure model, a scale-adaptive turbulence model for the gray zone is derived from the steady-state flux budgets.

The basic terrain-following (BTF) coordinate simplifies the lower boundary conditions of a numerical model but leads to numerical error and instability on steep terrain. Hybrid terrain-following (HTF) coordinates with smooth slopes of vertical layers (slopeVL) generally overcome this difficulty. Therefore, the HTF coordinate becomes very desirable for atmospheric and oceanic numerical models. However, improper vertical layering in HTF coordinates may also increase the incidence of error. Except for the slopeVL of an HTF coordinate, this study further optimizes the HTF coordinate focusing on the thickness of vertical layers (thickVL). Four HTF coordinates (HTF1–HTF4) with similar slopeVL but different vertical transition methods of thickVL are designed, and the relationship between thickVL and numerical errors in each coordinate is compared in the classic idealized thermal convection [two-dimensional (2D) rising bubble] experiment over steep terrain. The errors of potential temperature θ and vertical velocity w are reduced most, by approximately 70% and 40%, respectively, in the HTF1 coordinate, with a monotonic increase in thickVL according to the increasing height; however, the errors of θ increased in all the other HTF coordinates, with nonmonotonic thickVLs. Furthermore, analyses of the errors of vertical pressure gradient force (VPGF) show that due to the interpolation errors of thickVL, the inflection points in the vertical transition of thickVL induce the initial VPGF errors; therefore, the HTF1 coordinate with a monotonic increase in thickVL has the smallest errors among all the coordinates. More importantly, the temporal evolution of VPGF errors manifests top-type VPGF errors that propagate upward gradually during the time integration. Only the HTF1 and HTF4 coordinates with a monotonic increase in thickVL near the top of the terrain can suppress this propagation. This optimized HTF coordinate (i.e., HTF1) can be a reference for designing a vertical thickVL in a numerical model.

The passage of tropical cyclones induces ocean surface cooling through vertical mixing, upwelling, and surface heat loss. The dependence of tropical cyclone-induced ocean surface cooling on the intensity and translation speed of tropical cyclones has been documented in previous studies. The present study investigates the latitudinal and seasonal variations in tropical cyclone-induced ocean surface cooling in the tropical western North Pacific based on data for the 2001–2020 period. Our analysis focuses on the open ocean (0°–25°N, 130°E–180°) to reduce the interference of coastal topography so that the obtained results better represent the influences of the intensity and translation speed of tropical cyclones. Our analysis confirms the dependence on the intensity and translation speed of tropical cyclone-induced cooling. The new findings are as follows. First, the time to reach the maximum cooling increases with the magnitude of the maximum cooling. Second, the magnitude of ocean surface cooling increases with latitude in the tropical region for tropical cyclones with different intensities and translation speeds. Third, the ocean surface cooling is larger in summer and autumn than in spring for tropical cyclones with different intensities and translation speeds. Fourth, the dependence of ocean surface cooling on the translation speed is more obvious at higher latitudes in the tropics and less apparent in spring. These new findings add to the existing knowledge of the impacts of tropical cyclone intensity and translation speed on ocean surface cooling.

The challenges of applying deep learning (DL) to correct deterministic numerical weather prediction (NWP) biases with non-Gaussian distributions are discussed in this paper. It is known that the DL UNet model is incapable of correcting the bias of strong winds with the traditional loss functions such as the MSE (mean square error), MAE (mean absolute error), and WMAE (weighted mean absolute error). To solve this, a new loss function embedded with a physical constraint called MAE_MR (miss ratio) is proposed. The performance of the UNet model with MAE_MR is compared to UNet traditional loss functions, and statistical post-processing methods like Kalman filter (KF) and the machine learning methods like random forest (RF) in correcting wind speed biases in gridded forecasts from the ECMWF high-resolution model (HRES) in East China for lead times of 1–7 days. In addition to MAE for full wind speed, wind force scales based on the Beaufort scale are derived and evaluated. Compared to raw HRES winds, the MAE of winds corrected by UNet (MAE_MR) improves by 22.8% on average at 24–168 h, while UNet (MAE), UNet (WMAE), UNet (MSE), RF, and KF improve by 18.9%, 18.9%, 17.9%, 13.8%, and 4.3%, respectively. UNet with MSE, MAE, and WMAE shows good correction for wind forces 1–3 and 4, but negative correction for 6 or higher. UNet (MAE_MR) overcomes this, improving accuracy for forces 1–3, 4, 5, and 6 or higher by 11.7%, 16.9%, 11.6%, and 6.4% over HRES. A case study of a strong wind event further shows UNet (MAE_MR) outperforms traditional post-processing in correcting strong wind biases.

Abstract

In this study, idealized simulations are conducted to investigate potential influences of solar radiation on the tropical cyclone (TC) recurvature at higher latitudes. Results indicate that TC track is sensitive to the seasonal variation of radiative forcing at higher latitudes. In the absence of a background flow, TCs at higher latitudes tend to recurve (remain northwestward) in the cold (warm) season. This feature is an additional aspect of the so-called intrinsic recurvature property of TC movement at high latitude. Physically, the greater meridional gradient of temperature in the cold season due to solar radiative forcing would induce a larger thermal wind, which affects the upper-level anticyclonic circulation and associated outflow. The structure changes of TC, mainly at upper-levels, modulate the steering flow for TC, leading to a higher probability of TCs at higher latitudes to recurve in the cold season than in the warm season.

Abstract

We investigated the sensitivity of the size of a tropical cyclone (TC) to warming or cooling sea surface temperatures (SST) in its outer region by simulating the SST beyond a radius of 200 km from the TC center. Sensitivity experiments showed that an increased SST outside the core region of the TC had a negative effect on its size. Warming in the outer region contributed to the local enhancement of the latent heat flux from sea surface, which promoted the development of small-scale convection and warmed the lower and midtroposphere. This warming altered the local pressure gradient force in the upper and lower troposphere in such a way that it weakened the secondary circulation of the TC and led to suppression of the spiral rainbands outside the eyewall. Further analysis showed that the outward-propagating rainband structure favored an increase in the size of the TC. The diabatic heat released by the rainbands induced an inflow at lower levels, facilitating expansion of the TC. The greater the distance of the rainbands from the center of the TC, given the same amplitude of diabatic heating, the stronger the forced inflow, resulting in a faster increase in the size of the TC.