Wind field reconstruction with adaptive random Fourier features

We investigate the use of spatial interpolation methods for reconstructing the horizontal near-surface wind field given a sparse set of measurements. In particular, random Fourier features is compared with a set of benchmark methods including kriging and inverse distance weighting. Random Fourier features is a linear model β ( x ) = ∑ k = 1 K β k e i ω k x approximating the velocity field, with randomly sampled frequencies ω k and amplitudes β k trained to minimize a loss function. We include a physically motivated divergence penalty | ∇ ⋅ β ( x ) | 2 , as well as a penalty on the Sobolev norm of β . We derive a bound on the generalization error and a sampling density that minimizes the bound. We then devise an adaptive Metropolis–Hastings algorithm for sampling the frequencies of the optimal distribution. In our experiments, our random Fourier features model outperforms the benchmark models.

Scaling Feature of the Ocean Surface Wind Field

<p>The ocean surface wind plays a crucial role in the air-sea exchanges of momentum, heat, and mass, consequently is vital to the controlling of weather and climate. Due to the extremely large range of scales of the motion of the wind field, e.g., flow structures from millimeters to thousands of kilometers, the multiscale dynamics are known to be relevant. In this work, with the help of a Wiener-Khinchine theorem-based Fourier power spectrum estimator, the scaling features of the wind field provided by several satellites, i.e., QuikSCAT, Metop-A, -B, and -C, Haiyang-2B, and China France Oceanography SATellite (CFOSAT), is examined. Power-law scaling behavior is evident in the ranges of 100 to 3000 km with a scaling exponent β varying from 5/3 to 3. The global distributions and seasonal variations of the scaling exponent β have also been considered. The results show that due to the energetic convective activities in the low-latitude zones, the scaling exponents β in these regions are closer to the value of 5/3. As for the mid-latitudes, the values of β are close to 2 and independent of the variation of longitude. Concerning the seasonal variations, for most regions, the scaling exponents measured in winter are larger than those in summer. Furthermore, the seasonal variations of β in low-latitudes are stronger than those in the mid-latitudes. Our preliminary results indicate that all satellites provide a consistent scaling feature of the ocean surface wind field.</p>

Forecasting Hurricanes using Large-Ensemble Output

<p>This paper describes the development of a model framework for Forecasts of Hurricanes using Large-ensemble Outputs (FHLO). Computationally inexpensive, FHLO quantifies the forecast uncertainty of a particular tropical cyclone (TC) through O(1000) ensemble members. The model framework consists of three components: (1) a track model that generates synthetic tracks from the TC tracks of an ensemble numerical weather prediction (NWP) model, (2) an intensity model that predicts the intensity along each synthetic track, and (3) a TC wind field model that estimates the time-varying twodimensional surface wind field. In this framework, we consider the evolution of a TC’s intensity and wind field as though it were embedded in a timeevolving environmental field. The environmental fields are derived from the forecast fields of ensemble NWP models, leading to probabilistic forecasts of track, intensity, and wind speed that incorporate the flow-dependent uncertainty. Each component of the model is evaluated using four years (2015- 2018) of TC forecasts in the Atlantic and Eastern Pacific basins. We show that the synthetic track algorithm can generate tracks that are statistically similar to those of the underlying global ensemble models. We show that FHLO produces competitive intensity forecasts, especially when considering probabilistic verification statistics. We also demonstrate the reliability and accuracy of the probabilistic wind forecasts. Limitations of the model framework are also discussed.</p>

Sea Surface Roughness Observed by High Resolution Radar

Changes in the sea surface roughness are usually associated with a change in the sea surface wind field [...]