Weather Forecasting

Abstract Model Output Statistics (MOS) is an objective weather forecasting technique which consists of determining a statistical relationship between a predictand and variables forecast by a numerical model at some projection time(s). It is, in effect, the determination of the “weather related” statistics of a numerical model. This technique, together with screening regression, has been applied to the prediction of surface wind, probability of precipitation, maximum temperature, cloud amount, and conditional probability of frozen precipitation. Predictors used include surface observations at initial time and predictions from the Subsynoptic Advection Model (SAM) and the Primitive Equation model used operationally by the National Weather Service. Verification scores have been computed, and, where possible, compared to scores for forecasts from other objective techniques and for the official forecasts. MOS forecasts of surface wind, probability of precipitation, and conditional probability of frozen precipi...

Abstract Differences of opinion exist among forecasters—and between forecasters and users—regarding the meaning of the phrase “good (bad) weather forecasts.” These differences of opinion are fueled by a lack of clarity and/or understanding concerning the nature of goodness in weather forecasting. This lack of clarity and understanding complicates the processes of formulating and evaluating weather forecasts and undermines their ultimate usefulness. Three distinct types of goodness are identified in this paper: 1) the correspondence between forecasters’ judgments and their forecasts (type 1 goodness, or consistency), 2) the correspondence between the forecasts and the matching observations (type 2 goodness, or quality), and 3) the incremental economic and/or other benefits realized by decision makers through the use of the forecasts (type 3 goodness, or value). Each type of goodness is defined and described in some detail. In addition, issues related to the measurement of consistency, quality, and value ar...

Traditional weather forecasting has been built on a foundation of deterministic modeling--start with initial conditions, put them into a supercomputer model, and end up with a prediction about future weather. But as Gneiting and Raftery discuss in their Perspective, a new approach--ensemble forecasting--was introduced in the early 1990s. In this method, up to 100 different computer runs, each with slightly different starting conditions or model assumptions, are combined into a weather forecast. In concert with statistical techniques, ensembles can provide accurate statements about the uncertainty in daily and seasonal forecasting. The challenge now is to improve the modeling, statistical analysis, and visualization technologies for disseminating the ensemble results.

We take a simple time-series approach to modeling and forecasting daily average temperature in U.S. cities, and we inquire systematically as to whether it may prove useful from the vantage point of participants in the weather derivatives market. The answer is, perhaps surprisingly, yes. Time-series modeling reveals both strong conditional mean dynamics and conditional variance dynamics in daily average temperature, and it reveals sharp differences between the distribution of temperature and the distribution of temperature surprises. The approach can easily be used to produce not only short-horizon point forecasts, but also the long-horizon density forecasts of maximal relevance in weather derivatives contexts. We produce and evaluate both, with some success. We conclude that additional inquiry into nonstructural weather forecasting methods will likely prove useful in weather derivatives contexts.

Abstract. Numerical models that combine weather forecasting and atmospheric chemistry are here referred to as chemical weather forecasting models. Eighteen operational chemical weather forecasting models on regional and continental scales in Europe are described and compared in this article. Topics discussed in this article include how weather forecasting and atmospheric chemistry models are integrated into chemical weather forecasting systems, how physical processes are incorporated into the models through parameterization schemes, how the model architecture affects the predicted variables, and how air chemistry and aerosol processes are formulated. In addition, we discuss sensitivity analysis and evaluation of the models, user operational requirements, such as model availability and documentation, and output availability and dissemination. In this manner, this article allows for the evaluation of the relative strengths and weaknesses of the various modelling systems and modelling approaches. Finally, this article highlights the most prominent gaps of knowledge for chemical weather forecasting models and suggests potential priorities for future research directions, for the following selected focus areas: emission inventories, the integration of numerical weather prediction and atmospheric chemical transport models, boundary conditions and nesting of models, data assimilation of the various chemical species, improved understanding and parameterization of physical processes, better evaluation of models against data and the construction of model ensembles.

Weather forecasting is a canonical predictive challenge that has depended primarily on model-based methods. We explore new directions with forecasting weather as a data-intensive challenge that involves inferences across space and time. We study specifically the power of making predictions via a hybrid approach that combines discriminatively trained predictive models with a deep neural network that models the joint statistics of a set of weather-related variables. We show how the base model can be enhanced with spatial interpolation that uses learned long-range spatial dependencies. We also derive an efficient learning and inference procedure that allows for large scale optimization of the model parameters. We evaluate the methods with experiments on real-world meteorological data that highlight the promise of the approach.

It has been suggested [1] that the resolution of the information paradox for evaporating black holes is that the holes are surrounded by firewalls, bolts of outgoing radiation that would destroy any infalling observer. Such firewalls would break the CPT invariance of quantum gravity and seem to be ruled out on other grounds. A different resolution of the paradox is proposed, namely that gravitational collapse produces apparent horizons but no event horizons behind which information is lost. This proposal is supported by ADS-CFT and is the only resolution of the paradox compatible with CPT. The collapse to form a black hole will in general be chaotic and the dual CFT on the boundary of ADS will be turbulent. Thus, like weather forecasting on Earth, information will effectively be lost, although there would be no loss of unitarity.

Abstract In the literature, the use of analogues for short-range weather forecasting has practically been discarded. This is because no good matches for today's extratropical large-scale flow patterns can be found in a 30-year data library. We propose here a limited-area model approach for Analogue-Forecasting (AF). In order to make a 12-hour AF valid at a target point, we require analogy in initial states only over a circle with radius of about 900 km. On a limited area there are usually several good analogues, sometimes to within observational error. Different historical analogues may be used at different target points. The usefulness of the limited area approach is first demonstrated with some examples. We then present verification statistics of 3000 12-hour 500-mb height point forecasts in the Northern Hemisphere winter at 38°N, 80°W (over West Virginia, U.S.A.). In order to beat persistence at 12 hours at this point we need an analogue which differs by about 40 geopotential meters or less from the ba...

We reconstruct the global structure and kinematics of coronal mass ejections (CMEs) using coordinated imaging and in situ observations from multiple vantage points. A forward modeling technique, which assumes a rope-like morphology for CMEs, is used to determine the global structure (including orientation and propagation direction) from coronagraph observations. We reconstruct the corresponding structure from in situ measurements at 1 AU with the Grad-Shafranov method, which gives the flux-rope orientation, cross section, and a rough knowledge of the propagation direction. CME kinematics (propagation direction and radial distance) during the transit from the Sun to 1 AU are studied with a geometric triangulation technique, which provides an unambiguous association between solar observations and in situ signatures; a track fitting approach is invoked when data are available from only one spacecraft. We show how the results obtained from imaging and in situ data can be compared by applying these methods to the 2007 November 14-16 and 2008 December 12 CMEs. This merged imaging and in situ study shows important consequences and implications for CME research as well as space weather forecasting: (1) CME propagation directions can be determined to a relatively good precision as shown by the consistency between different methods; (2) the geometric triangulation technique shows a promising capability to link solar observations with corresponding in situ signatures at 1 AU and to predict CME arrival at the Earth; (3) the flux rope within CMEs, which has the most hazardous southward magnetic field, cannot be imaged at large distances due to expansion; (4) the flux-rope orientation derived from in situ measurements at 1 AU may have a large deviation from that determined by coronagraph image modeling; and (5) we find, for the first time, that CMEs undergo a westward migration with respect to the Sun-Earth line at their acceleration phase, which we suggest is a universal feature produced by the magnetic field connecting the Sun and ejecta. The importance of having dedicated spacecraft at L4 and L5, which are well situated for the triangulation concept, is also discussed based on the results.

Data‐driven approaches, most prominently deep learning, have become powerful tools for prediction in many domains. A natural question to ask is whether data‐driven methods could also be used to predict global weather patterns days in advance. First studies show promise but the lack of a common data set and evaluation metrics make intercomparison between studies difficult. Here we present a benchmark data set for data‐driven medium‐range weather forecasting (specifically 3–5 days), a topic of high scientific interest for atmospheric and computer scientists alike. We provide data derived from the ERA5 archive that has been processed to facilitate the use in machine learning models. We propose simple and clear evaluation metrics which will enable a direct comparison between different methods. Further, we provide baseline scores from simple linear regression techniques, deep learning models, as well as purely physical forecasting models. The data set is publicly available at https://github.com/pangeo‐data/WeatherBench and the companion code is reproducible with tutorials for getting started. We hope that this data set will accelerate research in data‐driven weather forecasting.