Search

Nowcasting based on weather radar uses the current and past observations to make estimations of future radar echoes. There are many types of operationally deployed nowcasting systems, but none of them are currently based on deep learning, despite it being an active area of research in the last few years. This paper explores deep learning models as alternatives to current methods by proposing different architectures and comparing them against some operational nowcasting systems. The methods proposed here, harnessing residual convolutional encoder–decoder architectures, reach a level of performance expected of current systems and in certain scenarios can even outperform them. Finally, some of the potential drawbacks of using deep learning are analyzed. No decay in the performance on a different geographical area from where the models were trained was found. No edge or checkerboard artifact, common in convolutional operations, was found that affects the nowcasting metrics. Significance Statement: Deep learning methods started to become more popular for weather nowcasting, but none is operational. We noticed that most of the studies did not present a benchmark against operational systems, and in many cases, the evaluation methods used were limited to light rain scenarios. Our goal is to propose deep learning models, analyze how they perform against operational methods in many scenarios, and explore some of the potential limitations. We found that our models perform better for low to mild storms. While we showed that common side effects of convolutional operations (a common technique in deep learning) did not impair the performance, we agree with many other authors that the major problem is the smoothing effect that hinders the nowcast of intense storms. [ABSTRACT FROM AUTHOR]

Machine learning algorithms are trained on a 10-yr archive of composite weather radar images in the Swiss Alps to nowcast precipitation growth and decay in the next few hours in moving coordinates (Lagrangian frame). The hypothesis of this study is that growth and decay is more predictable in mountainous regions, which represent a potential source of practical predictability by machine learning methods. In this paper, artificial neural networks (ANN) are employed to learn the complex nonlinear dependence relating the growth and decay to the input predictors, which are geographical location, mesoscale motion vectors, freezing level height, and time of the day. The average long-term growth and decay patterns are effectively reproduced by the ANN, which allows exploring their climatology for any combination of predictors. Due to the low intrinsic predictability of growth and decay, its prediction in real time is more challenging, but is substantially improved when adding persistence information to the predictors, more precisely the growth and decay and precipitation intensity in the immediate past. The improvement is considerable in mountainous regions, where, depending on flow direction, the root-mean-square error of ANN predictions can be 20%–30% lower compared with persistence. Because large uncertainty is associated with precipitation forecasting, deterministic machine learning predictions should be coupled with a model for the predictive uncertainty. Therefore, we consider a probabilistic perspective by estimating prediction intervals based on a combination of quantile decision trees and ANNs. The probabilistic framework is an attempt to address the problem of conditional bias, which often characterizes deterministic machine learning predictions obtained by error minimization. [ABSTRACT FROM AUTHOR]

Classic supercell storms occur in a generally well understood environment characterized by instability and vertical wind shear. Within this broadly favorable environment, large day-to-day variability in environmental parameters may lead to widely varying radar presentation of storms. Of interest here is whether specific storm structures exhibit repeatable characteristics in similar environments and whether radar presentation can be predicted with knowledge of environmental characteristics. Specifically, this paper focuses on (i) updraft characteristics inferred using differential reflectivity ZDR columns, (ii) characteristics of storm-relative inflow inferred using ZDR arcs, and (iii) areal extent and cyclicality of polarimetrically inferred hailfall at low levels. Variability of these radar features is compared among storms in similar environments and among a larger subset of storms across highly varying environments. The similarity of storms in similar and different environments is quantified, and tornadic and nontornadic storms are compared. Associations between inferred updraft, inflow, and hailfall characteristics and environmental variables are discussed. Storm features generally exhibit greater similarity among storms in similar environments than across environments, although exceptions occur. The results indicate that many radar features of classic supercells may be useful to learn about microphysical variability across environments. [ABSTRACT FROM AUTHOR]

Although radial velocity data from Doppler radars can partially resolve some tornadoes, particularly large tornadoes near the radar, most tornadoes are not explicitly resolved by radar owing to inadequate spatiotemporal resolution. In addition, it can be difficult to determine which mesocyclones typically observed on radar are associated with tornadoes. Since debris lofted by tornadoes has scattering characteristics that are distinct from those of hydrometeors, the additional information provided by polarimetric weather radars can aid in identifying debris from tornadoes; the polarimetric tornadic debris signature (TDS) provides what is nearly 'ground truth' that a tornado is ongoing (or has recently occurred). This paper outlines a modification to the hydrometeor classification algorithm used with the operational Weather Surveillance Radar-1988 Doppler (WSR-88D) network in the United States to include a TDS category. Examples of automated TDS classification are provided for several recent cases that were observed in the United States. [ABSTRACT FROM AUTHOR]

The weather radar-based object-oriented convective storm tracking is a standard approach for analyzing and nowcasting convective storms. However, the majority of current storm-tracking algorithms provide nowcasts only in a deterministic fashion with limited ability to estimate the related uncertainties. This paper proposes a method for probabilistic nowcasting of convective storms that addresses the issue of uncertainty of nowcasts. The approach first utilizes a two-dimensional radar-based storm identification and tracking algorithm in conjunction with the Kalman filtering of noisy measurements of storm centroid with the continuous white noise acceleration model. The resulting smoothed estimates of storm centroid and velocity components and their error covariance values are then applied to nowcast the probability of storm occurrence. To verify the approach, 20-60-min nowcasts were computed every 5 min using composite weather radar data in Finland including approximately 22 000 tracked storms. The verification shows that the algorithm is applicable in both deterministic and probabilistic manner. Moreover, the forecast probabilities are consistent with observed frequencies of the storms, especially with 20- and 30-min nowcasts. The accuracy of the probabilistic nowcasts was evaluated through the Brier skill score with respect to the deterministic nowcasts and nowcasts based on observation persistence and sample climatology. The results show that the proposed nowcasting method has an improved accuracy over all of the reference forecast types. [ABSTRACT FROM AUTHOR]