Nowcasting: The Promise of New Technologies of Communication, Modeling, and Observation (2024)

The communication revolution.

Until recently a major roadblock to effective nowcasting was the inability to rapidly distribute weather information to the user community—in their homes, offices, schools, while driving or commuting, and during recreational activities. When television and radio broadcasts, supplemented by newspaper weather pages, were the main communication technologies, distribution of real-time weather information was difficult, and often impossible. NOAA Weather Radio, initiated in 1969, provides basic weather information and warnings from approximately 1,000 transmitters around the United States, with Weather Radio receivers found in roughly 20% of U.S. households [T. Buehner, National Weather Service (NWS), 2011, personal communication]. This technology offers excellent coverage over the eastern and central United States, but with significant gaps over the West. In the 1990s, the spread of the Internet, first through dial-up modems and then through hard-wired connections, furnished an approach for providing real-time weather data to fixed locations. During the past decade the extension of the Internet to cell phones and other wireless connections allowed weather graphics (such as radar loops) and textual weather data to reach virtually any location.

But the true breakthrough communication device for weather nowcasting may be the smartphone and the robust data rates increasingly available with them. It would be hard to imagine a better device for distributing weather information and for building portable weather applications. Smartphones (Fig. 1; see title page) generally possess high-resolution screens for easy viewing of complex graphics, as well as high bandwidth communications either through cell phone networks or local wireless (Wi-Fi) links, allowing the distribution of imagery, model output, and other data. Modern smartphones possess substantial computational capacity and most keep track of their current location, either using cell tower triangulation or the Global Positioning System (GPS). Such position information is critical: with it, smartphones can download and display the meteorological information relevant to their surroundings, including location-specific warnings and forecasts.

Both the federal government and private industry are moving aggressively to distribute forecasts and warnings through wireless digital technology. For example, the Federal Emergency Management Agency (FEMA) has begun the nationwide implementation of the Personalized Localized Alerting Network (PLAN) that will provide site-specific warnings of major weather hazards through cell phones and smartphones. As described later, a number of private sector vendors have developed the capabilities to provide local warnings and associated weather information through text messages and smartphone applications (apps).

Improved communication technologies applicable for distributing location-specific information are not limited to smartphones and wireless networks. For example, electronic readerboards have become widespread along many of the nation's roadways (Fig. 2). Such electronic signage could provide warnings about dangerous weather ahead or control the speed of traffic to facilitate safe travel through or around inclement weather, such as heavy precipitation, roadway icing, strong winds, or reduced visibility.

Social media, such as Facebook and Twitter, possess substantial potential for the distribution and collection of nowcasting information. Twitter, with its ability to identify the location of a message (geotag), is ideal for providing short storm reports or other weather information from the field. Furthermore, Twitter can be used to broadcast immediate, terse warnings and forecast information to large groups. Facebook has been used by both the National Weather Service (NWS) and Environment Canada to provide weather warnings, as well as nowcasts. By using RSS (Really Simple Syndication) or SMS (Short Message Service) feeds, users can be notified of and secure real-time weather updates from Facebook or other social media sites.

The weather data revolution.

Effective nowcasting demands a high density of weather information, particularly at the surface. During the past several decades there has been an exponential increase of real-time surface data so that tens of thousands of observations are now available across the United States each hour (National Research Council 2009). In addition to the backbone network run by the Federal Aviation Administration (FAA) and the NWS at U.S. airports, utilities, departments of transportation, air quality agencies, television stations and others have installed weather networks with real-time communication. In addition, many individuals have installed high-quality weather instruments and made the data available through the Internet using services such as the WeatherUnderground. As an illustration, at the University of Washington, data from over seventy networks are collected in real time, with typically 3,000– 4,000 observations gathered each hour for Washington and Oregon alone (Fig. 3). A major development has been the NOAA Meteorological Assimilation Data Ingest System (MADIS) in which data from more than 60,000 sites established by local, state, and federal agencies, as well as numerous private networks, are collected, quality-controlled, archived, and distributed. An important challenge will be the effective use of such a heterogeneous network, with observations of substantially varying quality.

The mesoscale weather data revolution is not limited to surface observations. Instrumented commercial aircraft, reporting through the Aircraft Communications Addressing and Reporting System (ACARS) and Tropospheric Airborne Meteorological Data Reporting (TAMDAR) system now provide numerous soundings at airports around the country during ascent and descent (Fig. 4). Recent work (Moninger et al. 2010) has shown that TAMDAR aircraft observations have significant impacts on RUC analyses and short-range forecasts. The U.S. Doppler radar network [Weather Surveillance Radar-1988 Doppler (WSR-88D)] not only provides reflectivity and Doppler winds but now includes dual-polarization capabilities, allowing determination of precipitation type and improved rainfall estimates. The Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC) GPS-based satellite network, soon to be expanded, provides high-quality vertical soundings around the world, with hundreds over or near North America each day, while ground-based GPS receivers can be used to produce high-resolution three-dimensional water vapor distributions (MacDonald et al. 2002). Finally, the Geostationary Operational Environmental Satellite (GOES-R) will contribute a large increase in the amount of lightning data worldwide through the deployment of the Geostationary Lightning Mapper (GLM). Furthermore, the Advanced Baseline Imager (ABI) on GOES-R will provide full disk coverage from 16 channels with 5-min temporal resolution and “flex modes” that could provide 30-s coverage for mesoscale events.

These new data sources will provide a greatly enhanced capability to describe mesoscale structures over land, as well as improved data availability over the oceans. Such observations will greatly facilitate nowcasting since they provide a dramatically improved mesoscale description of what is happening now and, through extrapolation, data assimilation, and modeling, what will occur during the next few hours.

High-resolution forecasting and mesoscale data assimilation advances.

Data assimilation, the synergistic marriage of observations and models, has advanced rapidly during the past decades. On the synoptic scale, massive increases in the quantity and quality of satellite observations, coupled with advancing data assimilation approaches such as three-dimensional variational data assimilation (3DVAR) (Derber et al. 1991) and four-dimensional variational data assimilation (4DVAR) (Rabier et al. 2000) has led to greatly improved synoptic-scale analyses and forecasts. Such refined synoptic-scale guidance has resulted in improved high-resolution forecasts, since mesoscale models usually receive their initial and boundary conditions from larger-scale models.

On the mesoscale, limited-area operational models have increased greatly in resolution, with many real-time prediction systems now using grid spacings of 4–12 km. Thus, for the first time operational mesoscale models possess or will soon possess the necessary resolution for resolving local features from convection to topographic flows. The skill of these high-resolution operational forecasts has been further enhanced by substantial improvements in model physics (e.g., microphysics and land surface models) as well as rapidly increasing volumes of mesoscale data, as noted previously in this paper.

New ensemble-based data assimilation approaches, such as the ensemble Kalman filter, offer the potential for major improvements in mesoscale data assimilation (Torn and Hakim 2008). Such ensemble data assimilation methods provide flow-dependent background error covariances on the mesoscale that relate different variables and are physically realistic, unlike the static, simplified convariances used in current methods such as 3DVAR. Mesoscale EnKF systems are being tested at a number of U.S. universities (e.g., Zhang et al. 2009; Ancell et al. 2011) and have shown great promise compared to current operational data assimilation/nowcasting systems that use 3DVAR. EnKF and related ensemble-based data assimilation systems (such as hybrid systems that use EnKF covariances for distributing observation influence in space, but apply nudging or variational approaches in time) hold great promise in the more effective use of increasing amounts of mesoscale observations (Liu et al. 2011). Ensemble-based data assimilation also has the advantage of providing both probabilistic analyses and forecasts. As nowcasting matures during the next decade, information about analysis and forecast uncertainty will become more central for a wide variety of products and applications.

As noted earlier, operational high-resolution data assimilation and short-term forecasts are now available in the United States from the RUC system, which includes hourly data assimilation and frequent short-term forecasts on a 13-km grid (Benjamin et al. 2004). During 2012 RUC will be replaced by the more advanced Rapid Refresh (RR) system over an expanded 13-km domain using the Weather Research and Forecasting (WRF) model (Skamarock et al. 2007), and by 2015 with the High-Resolution Rapid Refresh (HRRR) system (Smith et al. 2008) that will downscale RR to 3 km over the United States These developments will lead to nowcasting supported by far better mesoscale analyses and short-term predictions.

The adaptive society revolution.

The advances in computation and communication that make improved nowcasting possible also allow society to react more quickly and effectively to weather challenges. For example, when radar or other observational assets indicate inclement weather over roadways (e.g., heavy precipitation, icing, dust storms), dynamically changing readerboards and speed signs, as well as flow management systems, can control the speed and number of cars. The proposed FAA next-generation (NEXGEN) air traffic control system foresees the use of real-time weather information to enable aircraft to fly more efficiently and safely through the nation's airspace, adapting their routes and speed in response to the changing environment. The coordination of power generation by weather-dependent renewables (e.g., wind and solar) with reserve power sources (e.g., gas turbines or hydro) can be closely controlled in real time based on weather observations and short-term forecasts, while “smart grid” technologies in homes can modify electrical demand. Local municipalities can use short-term forecasts of heavy precipitation to mitigate sewer overflows and to protect vulnerable low-lying areas, while departments of transportation can position trucks and material as well as preparing roadways in advance when snow or icing conditions are imminent. These and many other examples illustrate a key fact: improvements in control and communications now allow industry, government, and the general population to adapt to weather based on real-time information in a way that would have been impossible a decade ago. Even decisions about recreation and ordinary daily tasks (e.g., bicycle commuting) can be informed and improved by enhanced weather guidance provided through new forms of communication.

A change of direction for the National Weather Service.

The unfolding of the nowcasting revolution and the rapid evolution of weather prediction technology suggests a more effective approach for the use of NWS resources and personnel. NWS forecasting operations are on a 6-h cycle, which corresponds with the normal frequency of forecast updates. In most offices, the bulk of forecaster time is spent preparing gridded forecasts out to 168 h at either 6-h or 3-h time resolution (hazards are described at 1-h intervals to 72 h). These grids, prepared at either 5- or 2.5-km grid spacing, can be updated as needed, with forecasters responsible for revising hundreds of grids on a typical shift. When the potential for threatening weather exists, forecasters often put less effort into the grid updates as they prepare special statements, advisories, watches, or warnings as frequently as required.

The basic 6-h forecast update cycle and the tendency to maintain forecast consistency sometimes results in short-term forecasts being at odds with observed weather. Furthermore, important local weather details, which can change rapidly, are sometimes not mentioned or discussed in forecast products, especially during active weather periods. Thus, in quickly changing weather situations, the public and other users often are unaware of significant changes in local weather that could benefit their decision-making. As a result, only highly educated users, familiar with weather technologies and the interpretation of weather observations (radar, satellite, etc.) are in a position to make optimal weather-based decisions.

Not only is short-term forecast information sometimes inadequate, but some studies (e.g., Baars and Mass 2005) suggest that forecaster contributions are typically modest beyond 6–12 h due to the increasing skill of numerical weather prediction and post-processed model output. It appears that humans can improve significantly over raw model output for extreme precipitation events for multi-day forecasts, with only a small improvement over model output statistics (MOS) for non-extreme variables such as maximum temperature (D. Novak, Science and Operations Officer, Hydrometeorological Prediction Center, 2011, personal communication). The ability of humans to interpret satellite and radar imagery and to make useful short-term forecasts is unequalled by any automated system, a situation that should not change soon. Thus, the short-term period (0–3 h) is one in which subjective approaches make the most sense. Beyond 3–6 h, when there is rapid growth in mesoscale uncertainty, probabilistic prediction, mainly dependent on post-processed ensemble forecasts, is clearly the direction the National Weather Service must take, and the value in human intervention in such probabilistic forecasting is uncertain.

An alternative vision of a future National Weather Service forecast office is one in which forecasters spend much of their time on 0–6-h nowcasting, with longer-period predictions transitioning to objective, model-based guidance; the main exception to this approach would be during extreme, highly unusual events. In the new operations schedule, forecasters would provide at least hourly nowcasts of the current weather situation and how the situation was expected to evolve during the next few hours in a variety of formats, including hourly gridded analyses/forecasts through 6 h and prose descriptions. Furthermore, a regular oral/video discussion could be available over the Internet (and accessible through computers, tablets, pads, smartphones, and other units). During particularly fast-changing and significant weather, update frequency would increase, as it does today for tornadic situations. In this approach forecasters would be spending the bulk of their time on what they do best: coupling the extraordinary graphical interpretation capabilities of humans with an understanding of weather systems, and communicating this information to other humans. The transition toward greater forecaster intervention in nowcasting will produce enhanced forecaster situational awareness for short-period, local weather events. This transition should be accompanied by more emphasis on enhanced communication approaches for short-term forecasts, including NWS apps for smartphones and other digital devices.

A reviewer of this manuscript asked why the National Weather Service, supported by public resources, should provide nowcasting services through new technologies, when “the private sector weather industry is already doing this and doing it well.” This can be answered in several ways. First, it is not clear that the private sector is doing this task uniformly well, though there have been some attempts by television outlets and private sector firms to provide some nowcasting services. But in any case, the NWS is now forecasting at all time scales and will continue to do so. As technology changes, the optimal role of humans can and will change, and the means of communication will evolve as well. It is not reasonable to expect that the NWS, by law responsible for providing warning capabilities to the entire nation, should be prevented from using the most modern approaches and technologies in fulfilling its mission.

In an enhanced nowcasting environment, warn-on- forecast (WoF) will be increasingly applied when severe weather is predicted (Stensrud et al. 2009). WoF aims to provide longer lead times for severe convective weather than the realizable limit of ~20 min from warn-on-detection approaches, thereby helping emergency decision makers. The WoF approach requires the ability to continuously update skillful, high-resolution NWP models, a direction consistent with the marriage of data assimilation and high-resolution modeling noted above. In addition, WoF requires the robust, rapid communication infrastructure described earlier.

A new paradigm for the broadcast media.

The trend towards nowcasting should lead to a very different broadcast day for television weathercasters. Television weathercasting is dominated by regular broadcasts during commute periods, lunchtime, and during the late evening. Generally limited to 2–3 minutes, television weathercasts usually provide a broad, but superficial, description of recent weather, short-term local forecasts, and an outlook for the days and week ahead. The only exception to this schedule is during truly severe weather (e.g., tornadoes) when local television stations often go into nowcasting mode, providing continuously updated descriptions of severe storm evolution using radar, spotters, and occasionally traffic helicopters. Such severe-storm nowcasting has proven to be highly effective during several major convective outbreaks (Smith 2010). Local radio stations often provide frequent weather reports, often accompanied by traffic information; updated weather information could be enhanced on such regular segments, including their provision by meteorologists rather than untrained news staff.

As viewers increasingly use automated websites to garner forecast information and probabilistic weather predictions, the latter being difficult to communicate on-air, television weathercasters might well shift to providing frequent (perhaps every half hour) local nowcasts so people could have continuously updated information for planning their lives. Such nowcasts could be available on-air and online through web sites or smartphones. Clearly, the future of commercial weathercasting lies in the seamless integration of broadcasting, Internet, and wireless modes of communication.

Nowcasting applications: Some examples.

The availability of dense networks of mesoscale observations, high-resolution data assimilation and modeling, and high-bandwidth modes of communication makes possible a whole range of powerful approaches for disseminating nowcasting information. This section will review some nowcasting applications available today, with particular attention to those created for the Pacific Northwest, and will discuss potential avenues for innovation.