The Warn-on-Forecast System (WoFS) is a rapidly updating storm-scale ensemble that employs rapid assimilation of radar and satellite data to enable near-term (0–6 h) probabilistic forecasts of individual thunderstorms. WoFS development was motivated by needs in severe weather watch and warning operations. A similar need to confidently predict individual thunderstorms and their hazards extends to other applications. The National Severe Storms Laboratory (NSSL) and National Weather Service (NWS) Aviation Weather Center (AWC) conducted a four-week demonstration during summer 2024 to explore the possible application of WoFS to aviation weather forecasting. Two regional WoFS domains were run in the vicinity of major airports in the contiguous United States on 19 days in June and July. NWS forecasters at AWC, Center Weather Service Units (CWSU), and Weather Forecast Offices (WFO) provided feedback through testbed activities, a real-time and asynchronous chat room, shift shadowing, and post-event surveys. The demonstration revealed common aviation-related forecast tasks regarding convective weather, and the WoFS products that were most frequently used to aid those tasks. These themes emerged through the various feedback mechanisms and included several illustrative real-world use cases. The demonstration also identified potential hindrances to operational use, such as the cadence of WoFS forecasts relative to existing workflows at AWC. Staffing challenges and workload at CWSUs also hindered the ability of their meteorologists to provide feedback, suggesting future data-gathering efforts involving CWSUs should be co-designed and deliberate.
Synoptic-scale frontal boundaries are understood to be a source of intensification and severe weather production for supercell thunderstorms, owing to associated convergence and baroclinically generated horizontal vorticity that can be tilted into the updraft. However, boundaries also are associated with strong spatial gradients in environmental quantities; separately, these variations also are known to influence storm intensity, longevity, and severe weather production. It is unclear whether the boundary circulation and associated vorticity or the rapid changes in the near-storm environment more significantly influence the known enhancements for supercells near boundaries. Thus, the research presented herein explores the contribution of a rapidly changing background environment consistent with a supercell’s dwell time on a frontal boundary (without the associated convergence or boundary circulation) via idealized simulations, with a long-term goal of understanding and better anticipating supercell behavior near boundaries. The simulations used base-state environments rooted in an observed supercell-stationary boundary event on 29 May 2011. Representative environments were generated from model analyses on the warm-side, cold-side, and on the boundary itself. Idealized model experiments in CM1 tested each of these environments either fixed over time (control simulations), or varying over time via base-state substitution (BSS). Different types of boundary interactions were replicated through changes to the length of the transition between warm, cold, and boundary environments, as well as testing storm initiation on either side of the boundary (i.e., warm-to-cold versus cold-to-warm). While the control simulations indicated that the fixed boundary environment produced the strongest and longest lived supercell, all BSS experiments led to supercell dissipation. Overall, these simulations suggest that the boundary itself (i.e., associated convergence and vorticity) is an important contributor to the maintenance and severe weather production during supercell-boundary interactions. The context of these findings with regards to prior research and implications for operational forecasting also are discussed.
Coastal flooding is increasingly impacting local communities throughout the Chesapeake Bay, including in Annapolis, MD. Local stakeholders rely on water level forecasts from the National Weather Service (NWS) Baltimore/Washington Weather Forecast Office (WFO) to prepare and take the appropriate mitigation actions when coastal flooding is forecasted. This study evaluates the accuracy of NWS Baltimore/Washington WFO water level forecasts associated with forecasted coastal flooding events in Annapolis, Maryland, from 2018 through 2022. The results show a systematic high bias in the mean and median water level forecast errors for such events. The forecast errors are the largest during the 12 to 36 hour forecast window, which is the period when local stakeholders start taking protective actions. The median and mean water level forecast errors across all forecast windows for Annapolis associated with forecasted coastal flooding events, along with the large forecast error spread, suggest that the current deterministic approach is not providing local stakeholders with the detailed forecast information necessary to take appropriate actions during coastal flooding events. Given the lack of model guidance within the Chesapeake Bay needed for probabilistic water level forecasting, an alternative approach would be to assess water level forecast uncertainty for specific forecast windows based on prior forecast errors. This approach would be similar to the tropical cyclone track cone used by the NWS National Hurricane Center.
Colormaps are central to the design and interpretation of meteorological graphics, yet the color combinations historically used by the United States National Weather Service (NWS) often diverge from recommendations in the visualization literature. This study had three objectives: (1) to review best practices for using color in meteorology, summarized in a flowchart of actionable recommendations; (2) to assess how graphics from the National Centers for Environmental Prediction (NCEP; n=160) align with these recommendations; and (3) to gather insights from 14 NCEP personnel on current practices and future directions. We found that most graphics use the recommended colormap class for the data type, but fewer than half apply the recommended color combination. Many instead rely on multi-hue spectral scales with large numbers of bins, a legacy practice reinforced by familiarity and operational constraints but one that raises concerns about accessibility, perceptual uniformity, and interpretability. Only 49 of 160 graphics in our inventory were distinguishable for people with common forms of color vision deficiency. Beyond these broad patterns, we also observed limited use of single-hue sequential palettes and wide variation in color choices across NCEP offices. We close with a discussion of what these findings imply for current and future products, emphasizing the challenges of reconciling literature-based recommendations with operational realities. While we support empirically informed design decisions, we recognize that they must be balanced against operational constraints, and we argue that progress will require sustained dialogue, applied research, and iterative improvements to ensure forecast graphics remain accessible, scientifically grounded, and operationally effective.
In 2022, we implemented a new version of the "City Forecasts" contest in our Synoptic Meteorology II undergraduate capstone at the University of South Alabama. The contest, along with student-led current weather discussions, is a way for students to apply the theory learned in lecture. The new contest is Microsoft Excel (R)-based, and introduced Student Consensus, the National Blend of Models (NBM), and Model Output Statistics (MOS) from the Global Forecast System (GFS) and North American Mesoscale (NAM) model as participating "contestants." Forecast parameters were temperature, dewpoint, 12-h max/min temperature, wind direction and speed, cloud cover, ceiling and visibility flight categories, precipitation type, 12-h accumulated precipitation, 12-h probability of precipitation, 12-h frontal passage (yes/no), and 12-h thunderstorm observed (yes/no). From an initial time of 1200 UTC, students produced 12-, 24-, 36-, and 48-h forecasts in lab for a total of eight to nine cities within the contiguous United States during the semester, chosen weekly based on expected weather and geographic region. Although the spreadsheets greatly shortened student feedback times, introducing the four automated competitors made the contest more challenging, as only four out of 28 students have beaten Consensus and all three model products over the three years of this study. Among the model contestants, NAM MOS had the best performance in 2022, while NBM had the best performance in 2023 and 2024. In addition to the enhanced student learning opportunities, we learned that while the NBM was the best model product overall during this period, the MOS products also provided added value to our students.
In early 1976, I was a graduate student at the University of Wisconsin-Madison and just had a paper published in Monthly Weather Review (Uccellini, 1975) that was derived from my master’s thesis (under Professor Charles E. Anderson) on large-scale gravity waves capable of initiating convective storms. I was working toward my PhD degree under Professor Donald R. Johnson on the “vertical coupling” of upper and lower-level jet streaks within the indirect transverse circulation found in the exit region of the upper-level jet streak. The coupling process between upper- and low-level jets is related to a mutual mass-momentum adjustment process throughout the troposphere within the jet-streak exit region that could then be applied to describe the vertical shear and differential moisture transports from the lower to middle troposphere that work to destabilize the well-known pre-convective environment prior to the development of severe local storm systems (Uccellini and Johnson 1979). Much of this work on the vertical coupling of upper- and lower level jets was built off the dynamic equations applied in both isobaric (P) and isentropic (θ) vertical coordinates, which necessitated the ability to map the atmospheric circulation patterns in both coordinate systems.
This year marks the 50th anniversary of the National Weather Association’s National Weather Digest. First published in 1976, the Digest provided a forum for operational meteorologists to share case studies, diagnostic tools, and practical insights. It is both fitting and inspiring that our anniversary celebration returns to the very first article ever published: Dr. Louis W. Uccellini’s “Operational Diagnostic Applications of Isentropic Analysis.”
Determining fire fuel status remains a challenge across the Great Plains of the United States, where grasses and shrubs are the primary land covers and fuels for fires. This study compared two methods of estimating the Curing Index, an indicator of the moisture level of vegetation, that is included in the Grassland Fire Danger Index (GFDI): i) a satellite method using data acquired by the Geostationary Operational Environmental Satellite (GOES) Advanced Baseline Imager and ii) a method using a Growing Season Index (GSI) that includes observed meteorological conditions to estimate plant growth. Curing Index (CI) values were calculated for 2021 and 2022 at 47 Remote Automated Weather Stations (RAWS) that had dominant land cover types of either grassland or shrubland vegetation. Differences in the Curing Index of >50% were observed at various times from spring through autumn between the two methods. Differences in Curing Index between the two methods resulted in differences in GFDI, which changed the fire danger rating category within the National Fire Danger Rating System. The greatest differences in grassland and shrubland GFDI values included greater values of GFDI in the summer months when GSI was used to compute the CI. Also, during both years, the CI computed with the GSI displayed an autumn green-up during a dry early autumn that was not observed by satellite data. The early autumn green-up displayed by the GSI-based CI, when vegetation is normally in senescence, was attributed to a decrease in the vapor pressure deficit component of the GSI.
Flooding is one of the most hazardous weather-related phenomena in the United States, with flash flooding being a particularly dangerous form of flooding generally caused by short-duration, high-intensity precipitation. In April 2013 the National Weather Service (NWS) Weather Prediction Center introduced the Mesoscale Precipitation Discussion (MPD) product to address the forecast challenges associated with these events across the lower 48 states. This study evaluates the frequency, spatial distribution, and storm types associated with short-term forecasts of potential flash flood events through the lens of an MPD catalog for a period between 2013 and 2023. The spatial and temporal distribution of MPDs illustrates the likely storm types that most frequently produce short-duration, high-intensity precipitation over CONUS including atmospheric rivers (ARs), convective storms, fronts, and monsoons. Analysis of MPDs with AR keywords shows the highest frequencies in the western United States, whereas analysis of MPDs associated in space and time with AR features in AR detection tools (i.e., tARget, AR Scale) shows much higher frequencies over the southeastern United States. These results suggest that AR-related forecast diagnostics and the environments they describe may be useful for ingredients-based operational forecasting of flash flood conditions for various storm event types in locations beyond the West Coast of the United States. Additionally, this study highlights the ingredients-based forecasting methodology, supports the recent efforts by the NWS to quantify skill of the MPD product, and contributes to the objectives of NOAA's Precipitation Prediction Grand Challenge.
Weather prediction over complex terrain consistently poses one of the most significant challenges to forecasters and numerical models. One particularly unique location in this regard is Canaan Valley, West Virginia, with a floor elevation higher than any other large valley east of the Mississippi River. It is especially susceptible to cold air drainage, in which cooler, denser air sinks into the valley, forming a cold pool with temperatures much lower than the surrounding valley rim. The National Blend of Models (NBM), a critical foundation to operational forecasting, does not resolve these cold pool events and instead equates valley floor temperatures with those over surrounding higher elevations. This issue leads to a significant warm bias that exceeds 15 degrees C in the most extreme cases. A preliminary study connected this warm bias to non-linear quality control (NLQC) within the UnRestricted Mesoscale Analysis (URMA), which serves as ground truth for NBM bias correction. This research expands on this hypothesis by presenting four new case studies of Canaan Valley temperature patterns, taking place on 22 January, 4 February, 5 May, and 2 June 2023. Three out of four cases featured a pronounced cold pool at the valley floor, with warm biases observed in both NBM output and URMA analyses. It is shown that a low weight in NLQC was a likely cause of the warm bias, as there were no other recorded quality control issues, and lower values of the weighting function are directly associated with strong differences from the background field. To improve this discrepancy and subsequent NBM biases, algorithms within NLQC should allow for higher weights over complex terrain regions prone to strong temperature gradients.
During the afternoon of 20 May 2019 through the early morning hours the next day, 42 tornadoes occurred in Oklahoma with four rated EF2 on the Enhanced Fujita scale. This event was notable because forecast environmental parameters for tornadoes were so exceptionally extreme and several sources of convectionallowing model guidance depicted up to nine supercells in the warm sector moving across central Oklahoma during the afternoon and evening suggesting an historic tornado outbreak. While the tornado report coverage, intensities, and cumulative path lengths justified the "High Risk" issued by the Storm Prediction Center, the warm sector supercells did not occur as expected and thus the magnitude and severity of the event did not reach historic levels. This study presents an analysis of this event and diagnosis of the factors that led to the most impactful forecast errors in the operational High-Resolution Rapid Refresh (HRRR) model. Errors in the HRRR model could be traced to westward-displacement errors of 700 hPa mesoscale vorticity disturbances over western Texas, which contributed overestimates of warm sector areal coverage, insolation, instability, moisture, and sustained mesoscale ascent in west-central Oklahoma during the afternoon compared to observations. The position errors in the vorticity disturbances were linked to the separate development in the HRRR forecasts of a large mesoscale convective vortex with a mesoscale convective system in the Oklahoma Panhandle and southern Kansas. More generally, this study highlights the importance of mesoscale vorticity disturbances that influence downstream convective initiation by modulating the mesoscale environment embedded in an otherwise favorable synoptic-scale severe weather environment.
The National Weather Service (NWS) uses radar observations as the primary data source for tornado warning decision making, as they are the only tool available to safely and remotely identify the storm-scale rotation that is commonly associated with tornado formation. However, effective radar operation necessitates scanning at an angle above the horizon to avoid ground clutter from obstacles such as trees, buildings, and small hills. A natural consequence of this requirement is that the radar's sample volume becomes progressively higher above the ground with increasing distance from the radar. This effect becomes detrimental to the prediction and detection of tornadoes and the precursory near-ground rotation that indicates tornadogenesis is imminent, especially when tornadoes form in a non-descending manner. The inability of the radar to detect a low-level mesocyclone or a near-ground Tornadic Vortex Signature (TVS) may render radar data inadequate when an NWS Weather Forecast Office (WFO) is considering whether or not to warn on a potentially tornadic storm. However, to date, there have been no concentrated efforts to investigate the direct relationship between radar observational capabilities and tornado warning accuracy. This study analyzes the presence of TVSs, tornadogenesis times and locations, warning issuance, and lead time of tornado warnings issued by the NWS. Furthermore, comparisons are made between tornadoes that occurred "near" (0-50 km) and "far" (50-160 km) from the radar to investigate the relationships between tornado warning issuance, TVS detection, storm mode, and distance between the tornado and the radar. Three Midwest NWS WFOs were selected: Charleston, West Virginia, Wilmington, Ohio. and Indianapolis, Indiana, due to the somewhat equal fractions of Quasi-Linear Convective System (QLCS) and supercell tornadoes, with results compared to data from Norman, Oklahoma, because of its notable tornado frequency. Results suggest that TVSs are better detected and tornadoes are more accurately warned at closer distances to Weather Surveillance Radar-1988 Doppler (WSR-88D) radars and WFOs in the combined sample of all surveyed WFOs. Furthermore, TVS detection, warning issuance, and lead time are found to be dependent on storm mode. However, results for individual WFOs vary.
Emergency managers (EMs) serve as critical conduits of weather information, providing forecasts and information from the National Weather Service (NWS) and other sources to support decision-making and public safety efforts. This study investigates the role EMs in the United States play in gathering and disseminating weather forecast information before hazardous weather events. Using data from a nationwide survey of EMs, this paper examines EMs' information-sharing practices, their preferred sources, and the communication channels they use. Although EMs almost universally rely on local NWS Weather Forecast Offices, they also access information from a wide array of other sources via several different communication channels to access this information. When disseminating forecast information, the preferred communication methods of EMs vary between internal partners (e.g., first responders and jurisdictional leadership) and the public. Notably, social media emerges as the primary tool for public dissemination, while email dominates internal communications. Furthermore, most EMs report minimal modification of the forecast information they receive before passing it along. This study highlights the integral role EMs play in the weather information ecosystem, providing meteorologists who are serving as originators of forecasts with a deeper understanding of how one of their core partners is acting to share and amplify messages across their jurisdiction.
Basic isentropic analysis techniques are discussed and comparisons with pressure surface analyses presented to illustrate several advantages of mapping synoptic and sub-synoptic scale flow patterns in the isentropic framework. Since detailed isentropic maps are not presently available to operational meteorologists, a method for generating simple isentropic charts on an operational basis, utilizing the standard 850 mb, 700 mb, and 500 mb pressure maps, is presented. Isentropic maps constructed with the operational technique are then related to conventional surface and upper air analyses to demonstrate the diagnostic capability of this approach.
On the afternoon of 7 August 2023, a linear storm system moving through Knox County, Tennessee, spawned one EF-2 tornado amidst a county-wide swath of straight-line wind damage. The area was under a tornado watch and a severe thunderstorm warning. The tornado was unique in the county in that it was the first August tornado recorded and the first significant tornado in 40 years. We used a mixed-methods approach, including a thematic analysis of discussions with broadcast meteorologists (n=4) and descriptive statistics from a public survey (n=286), to examine how broadcast meteorologists in Knoxville interpreted and communicated the tornado threat and how the public engaged with their forecasts. A notable finding is that some broadcast meteorologists missed some warning details, such as the tornado-possible tag on the severe thunderstorm warning, when juggling forecasting and broadcasting through multiple media, and would prefer such forecast information to be detailed in the United States National Weather Service (NWS) Chat along with more uncertainty information. Broadcast meteorologists described that they typically rely heavily on their own forecasts and past experiences during an event and perceive that the public does not take severe thunderstorms seriously. Survey results raised the possibility that people who interacted with forecasts from television stations may not have accessed forecast details and may have instead relied on push notifications and other methods that left them to develop their own opinions on the forecast. As the use of forecasts by broadcast meteorologists shifts from television-based to app-and social media-based, future studies should seek to better understand how the public interacts with television station forecasts.
The National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/ NCAR) 20th Century Reanalysis (20CR) is used to estimate synoptic-scale atmospheric conditions associated with historic accounts of two very large wildfires, one of which reportedly burned >24 281 km2 (>6 million acres) on the southern Great Plains of New Mexico and Texas in March 1906. The reanalysis reveals atmospheric features consistent with recent wildfire outbreaks and megafires in the region. Fire perimeter mapping of modern megafires that occurred under similar meteorological conditions in the same geographical region were modified to approximate the reported dimensions of the March 1906 fires. Connecting the past to the present provides historical precedence and context for the escalation of southern Great Plains wildfire outbreaks and megafires observed in the region during recent decades. Fire, land, and emergency management officials, as well as fire forecasters, can use this proxy reanalysis as a reference point for planning. It illustrates a potential worst-case scenario for the plains fire regime without modern land use practices and fire suppression.
The weather domain has given a lot of attention towards numeracy, meaning a person’s ability to understand numerical information, for predicting how accurately different groups interpret forecast information. However, these forecasts are often presented graphically with a map, yet not many studies include a geographic equivalent to numeracy in their measures. Additionally, despite the cognitive load on participants, much of the literature measures numeracy and spatial cognition with mathematical tests. This study provides justification from interdisciplinary literature for using subjective scales to measure spatial cognition, or the ability to understand maps, along with numeracy to predict accuracy. Using a new wind speed exceedance threshold (WSET) tropical cyclone graphic, we tested the relationships between numeracy, spatial cognition, and accuracy in a public survey (n = 624) and with nineteen experts (emergency managers and meteorologists) from Louisiana and Florida. Numeracy independently predicted accuracy of interpretation, with spatial cognition having a small, statistically nonsignificant effect. This study provides empirical support for measuring numeracy and spatial cognition using subjective scales, as well as an accuracy measure for a new tropical forecast graphic. Future studies should continue to measure both numeracy and spatial cognition when assessing how individual characteristics impact forecast graphic interpretation.
Cyclogenesis in the lee of the Rocky Mountains frequently induces strong southerly winds on the southern Great Plains. From late fall through early spring, initial low-level moisture advection within this flow is sometimes insufficient to preclude problematic fire spread. When preceded by surface high pressure and light winds—ideal for intentional burning—this dry return flow (DRF) can lead to escapes of pre-existing fire. The emergence of wildfires in DRF is most prevalent where timber fuel types exist on the southern Great Plains. Red flag warnings have historically misrepresented threats in DRF because wildfires evolve from pre-existing fires in conditions that burners and forecasters perceive as relatively benign. Owing to the reason that fire-effectiveness of DRF is dependent upon adverse spread of pre-existing fire, the pattern is most impactful during innocuous states of vegetative fuel dryness and in the absence of government-issued burn bans. With these inherent weather and fuel-based limitations of the fire environment, DRF fire episodes tend to be less volatile than southern Great Plains wildfire outbreaks associated with midlatitude cyclones. DRF fire episodes, however, are a challenge for land/fire/emergency management agencies given their propensity for high volumes of fire dispatches that can overwhelm response capabilities. This paper presents reanalysis composites of atmospheric features, as well as fuelscape characteristics, associated with 21 fire-effective DRF events in Oklahoma, Kansas, and Texas between 2016 and 2020. Operational prediction of a subsequent DRF episode in 2022 is shown and a mitigation messaging template to communicate rapid changes in DRF fire environments is proposed.
Accurately distinguishing and warning convective cells in the outer rainbands of landfalling tropical cyclones (TC) that will produce tornadoes (TCTORs) presents considerable challenges for forecasters. To enhance warning efforts, this study compares the near-cell environments between tornadic cells (including both warning hits and misses) and warned nontornadic cells (false alarms) in landfalling tropical cyclones in the contiguous United States from 2013 to 2020. For each cell, RAP analysis gridpoint proximity soundings were obtained to represent the near-cell environment and compared between tornadic and nontornadic cells. Warning skill with respect to various factors such as time of day, position relative to the TC center, distance from the nearest radar, and distance from the coast and year is also analyzed. Finally, this study characterizes the spatiotemporal distribution of sounding-derived convective parameters for TCTORs. Findings from this study indicate slight differences in the near cell environments between tornadic and nontornadic cells. Kinematic parameters, specifically the 0–6 km shear and effective storm-relative helicity, are more effective in distinguishing between tornadic and nontornadic near-cell environments than thermodynamic parameters. This distinction becomes most pronounced when comparing near-cell environments conducive to F/EF1+ tornadoes with nontornadic environments. Warning skill has improved with time over the analyzed period but tends to decrease farther from the TC center and the nearest NEXRAD WSR-88D. TC environments exhibit spatiotemporal variability across all environmental parameters. Convective instability is most prominent in the southeastern quadrant and increases with distance from the TC center. In contrast, shear and storm-relative-helicity is largest near the TC center and within the northeastern quadrant. The potential for near-storm environment parameters to discriminate between tornadic and nontornadic cells also varies spatiotemporally.
