Search

The potential value of small unmanned aircraft systems (UAS) for monitoring the preconvective environment and providing useful information in real time to weather forecasters for evaluation at a National Weather Service (NWS) Forecast Office are addressed. The general goal was to demonstrate whether a combination of fixed-wing and rotary-wing UAS can provide detailed, accurate, and useful measurements of the boundary layer important for determining the potential for convection initiation (CI). Two field operations were held: a validation study in which the UAS data were compared with collocated measurements made by mobile rawinsondes and ground-based remote sensing systems and a real-time experiment held to evaluate the potential value of the UAS observations in an operationally relevant environment. Vertical profile measurements were made by the rotary-wing UAS at two mesonet sites every 30 min up to 763 m (2500 ft) AGL in coordination with fixed-wing UAS transects between the sites. The results showed the ability of the fixed-wing UAS to detect significant spatial gradients in temperature, moisture, and winds. Although neither of two different types of rotary-wing UAS measurements were able to strictly meet the requirements for sensor accuracy, one of the systems came very close to doing so. UAS sensor accuracy, methods for retrieving the winds, and challenges in assessing the representativeness of the observations are highlighted. Interesting mesoscale phenomena relevant to CI forecasting needs are revealed by the UAS. Issues needing to be overcome for UAS to ever become a NOAA operational observing system are discussed. [ABSTRACT FROM AUTHOR]

The article focuses on usage of artificial intelligence (AI) techniques for weather forecasting services along with management of real-time decision making conditions. Topics discussed include prediction of high-impact weather phenomena through AI techniques; utilization of different approaches for prediction of weather such as support vector machines (SVM); and illustration of big data presence in the AI techniques that impact decision making.

Previous work has shown that the Rapid Refresh (RAP) model severely underrepresents ice pellets in its grid, with a skill near zero and a very low bias. An ice pellet diagnostic upgrade was devised at the Earth System Research Laboratory (ESRL) to resolve this issue. Parallel runs of the experimental ESRL-RAP with the fix and the operational NCEP-RAP without the fix provide an opportunity to assess whether this upgrade has improved the overall performance and the performance of the individual precipitation types of the ESRLRAP. Verification was conducted using the mobile Phenomena Identification Near the Ground (mPING) project. The overall Gerrity skill score (GSS) for the ESRL-RAP was improved relative to the NCEP-RAP at a 3-h lead time but degraded with increasing lead time; the difference is significant at p<0.05. Whether this difference is practically significant for users is unknown. Some improvement was found in the bias and skill scores of ice pellets and snow in the ESRL-RAP, although the model continues to underrepresent ice pellets, while rain and freezing rain were generally the same or slightly worse with the fix. The ESRL-RAP was also found to depict a more realistic spatial distribution of precipitation types in transition zones involving ice pellets and freezing rain. [ABSTRACT FROM AUTHOR]

In winter weather, precipitation type is a pivotal characteristic because it determines the nature of most preparations that need to be made. Decisions about how to protect critical infrastructure, such as power lines and transportation systems, and optimize how best to get aid to people are all fundamentally precipitation-type dependent. However, current understanding of the microphysical processes that govern precipitation type and how they interplay with physics-based numerical forecast models is incomplete, degrading precipitation-type forecasts, but by how much? This work demonstrates the utility of crowd-sourced surface observations of precipitation type from the Meteorological Phenomena Identification Near the Ground (mPING) project in estimating the skill of numerical model precipitation-type forecasts and, as an extension, assessing the current model performance regarding precipitation type in areas that are otherwise without surface observations. In general, forecast precipitation type is biased high for snow and rain and biased low for freezing rain and ice pellets. For both the North American Mesoscale Forecast System and Global Forecast System models, Gilbert skill scores are between 0.4 and 0.5 and from 0.35 to 0.45 for the Rapid Refresh model, depending on lead time. Peirce skill scores for individual precipitation types are 0.7-0.8 for both rain and snow, 0.2-0.4 for freezing rain and freezing rain, and 0.25 or less for ice pellets. The Rapid Refresh model displays somewhat lower scores except for ice pellets, which are severely underforecast, compared to the other models. [ABSTRACT FROM AUTHOR]

Five implicit precipitation-type algorithms are assessed using observed and model-forecast sounding data in order to measure their accuracy and to gauge the effects of model uncertainty on algorithm performance. When applied to observed soundings, all algorithms provide very reliable guidance on snow and rain (SN and RA). However, their skills for ice pellets and freezing rain (IP and FZRA) are comparatively low. Most misclassifications of IP are for FZRA and vice versa. Deeper investigation reveals that no method used in any of the algorithms to differentiate between IP and FZRA allows for clear discrimination between the two forms. The effects of model uncertainty are also considered. For SN and RA, these effects are minimal and each algorithm performs reliably. Conversely, IP and FZRA are strongly impacted. When the range of uncertainty is fully accounted for, their resulting wet-bulb temperature profiles are nearly indistinguishable, leading to very poor skill for all algorithms. Although currently available data do not allow for a thorough investigation, comparison of the statistics from only those soundings that are associated with long-duration, horizontally uniform regions of FZRA shows there are significant differences between these profiles and those that are from more transient, highly variable environments. Hence, a five-category (SN, RA, IP, FZRA, and IP-FZRA mix) approach is advocated to differentiate between sustained regions of horizontally uniform FZRA (or IP) from more mixed environments. [ABSTRACT FROM AUTHOR]

North American Mesoscale Model (NAM) forecasts of low-level temperature and dewpoint during persistent valley cold pools in the Bonneville Basin of Utah are assessed. Stations near the east sidewall have a daytime cold and nighttime warm bias. This is due to a poor representation of the steep slopes on this side of the basin. Basin stations where the terrain is better represented by the model have a distinct warm, moist bias at night. Stations in snow-covered areas have a cold bias for both day and night. Biases are not dependent on forecast lead or validation time. Several potential causes for the various errors are considered in a series of sensitivity experiments. An experiment with 4-km grid spacing, which better resolves the gradient of the slopes on the east side of the basin, yields smaller errors along the east corridor of the basin. The NAM assumes all soil water freezes at a temperature of 273 K. This is likely not representative of the freezing temperature in the salt flats in the western part of the basin, since salt reduces the freezing point of water. An experiment testing this hypothesis shows that reducing the freezing point of soil water in the salt flats leads to an average error reduction between 1.5 and 4 K, depending on the station and time of day. Using a planetary boundary layer scheme that has greater mixing alleviates the cold bias over snow somewhat, but the exact source of this bias could not be determined. [ABSTRACT FROM AUTHOR]

This study assesses forecasts of the preconvective and near-storm environments from the convection-allowing models run for the 2008 National Oceanic and Atmospheric Administration (NOAA) Hazardous Weather Testbed (HWT) spring experiment. Evaluating the performance of convection-allowing models (CAMs) is important for encouraging their appropriate use and development for both research and operations. Systematic errors in the CAM forecasts included a cold bias in mean 2-m and 850-hPa temperatures over most of the United States and smaller than observed vertical wind shear and 850-hPa moisture over the high plains. The placement of airmass boundaries was similar in forecasts from the CAMs and the operational North American Mesoscale (NAM) model that provided the initial and boundary conditions. This correspondence contributed to similar characteristics for spatial and temporal mean error patterns. However, substantial errors were found in the CAM forecasts away from airmass boundaries. The result is that the deterministic CAMs do not predict the environment as well as the NAM. It is suggested that parameterized processes used at convection-allowing grid lengths, particularly in the boundary layer, may be contributing to these errors. It is also shown that mean forecasts from an ensemble of CAMs were substantially more accurate than forecasts from deterministic CAMs. If the improvement seen in the CAM forecasts when going from a deterministic framework to an ensemble framework is comparable to improvements in mesoscale model forecasts when going from a deterministic to an ensemble framework, then an ensemble of mesoscale model forecasts could predict the environment even better than an ensemble of CAMs. Therefore, it is suggested that the combination of mesoscale (convection parameterizing) and CAM configurations is an appropriate avenue to explore for optimizing the use of limited computer resources for severe weather forecasting applications. [ABSTRACT FROM AUTHOR]

This note examines the connection between the probability of precipitation and forecasted amounts from the NCEP Eta (now known as the North American Mesoscale model) and Aviation (AVN; now known as the Global Forecast System) models run over a 2-yr period on a contiguous U.S. domain. Specifically, the quantitative precipitation forecast (QPF)–probability relationship found recently by Gallus and Segal in 10-km grid spacing model runs for 20 warm season mesoscale convective systems is tested over this much larger temporal and spatial dataset. A 1-yr period was used to investigate the QPF–probability relationship, and the predictive capability of this relationship was then tested on an independent 1-yr sample of data. The same relationship of a substantial increase in the likelihood of observed rainfall exceeding a specified threshold in areas where model runs forecasted higher rainfall amounts is found to hold over all seasons. Rainfall is less likely to occur in those areas where the models indicate none than it is elsewhere in the domain; it is more likely to occur in those regions where rainfall is predicted, especially where the predicted rainfall amounts are largest. The probability of rainfall forecasts based on this relationship are found to possess skill as measured by relative operating characteristic curves, reliability diagrams, and Brier skill scores. Skillful forecasts from the technique exist throughout the 48-h periods for which Eta and AVN output were available. The results suggest that this forecasting tool might assist forecasters throughout the year in a wide variety of weather events and not only in areas of difficult-to-forecast convective systems. [ABSTRACT FROM AUTHOR]

A previous study of the mean spatial bias errors associated with operational forecast models motivated an examination of the mechanisms responsible for these biases. One hypothesis for the cause of these errors is that mobile synoptic-scale phenomena are partially responsible. This paper explores this hypothesis using 24-h forecasts from the operational Eta Model and an experimental version of the Eta run with Kain–Fritsch convection (EtaKF). For a sample of 44 well-defined upper-level short-wave troughs arriving on the west coast of the United States, 70% were underforecast (as measured by the 500-hPa geopotential height), a likely result of being undersampled by the observational network. For a different sample of 45 troughs that could be tracked easily across the country, consecutive model runs showed that the height errors associated with 44% of the troughs generally decreased in time, 11% increased in time, 18% had relatively steady errors, 2% were uninitialized entering the West Coast, and 24% exhibited some other kind of behavior. Thus, landfalling short-wave troughs were typically underforecast (positive errors, heights too high), but these errors tended to decrease as they moved across the United States, likely a result of being better initialized as the troughs became influenced by more upper-air data. Nevertheless, some errors in short-wave troughs were not corrected as they fell under the influence of supposedly increased data amount and quality. These results indirectly show the effect that the amount and quality of observational data has on the synoptic-scale errors in the models. On the other hand, long-wave ridges tended to be underforecast (negative errors, heights too low) over a much larger horizontal extent. These results are confirmed in a more systematic manner over the entire dataset by segregating the model output at each grid point by the sign of the 500-hPa relative vorticity. Although errors at grid points with positive relative vorticity are small but positive in the western United States, the errors become large and negative farther east. Errors at grid points with negative relative vorticity, on the other hand, are generally negative across the United States. A large negative bias observed in the Eta and EtaKF over the southeast United States is believed to be due to an error in the longwave radiation scheme interacting with water vapor and clouds. This study shows that model errors may be related to the synoptic-scale flow, and even large-scale features such as long-wave troughs can be associated with significant large-scale height errors. [ABSTRACT FROM AUTHOR]

Rank histograms are a commonly used tool for evaluating an ensemble forecasting system’s performance. Because the sample size is finite, the rank histogram is subject to statistical fluctuations, so a goodness-of-fit (GOF) test is employed to determine if the rank histogram is uniform to within some statistical certainty. Most often, the χ2 test is used to test whether the rank histogram is indistinguishable from a discrete uniform distribution. However, the χ2 test is insensitive to order and so suffers from troubling deficiencies that may render it unsuitable for rank histogram evaluation. As shown by examples in this paper, more powerful tests, suitable for small sample sizes, and very sensitive to the particular deficiencies that appear in rank histograms are available from the order-dependent Cramér–von Mises family of statistics, in particular, the Watson and Anderson–Darling statistics. [ABSTRACT FROM AUTHOR]

The problem of predicting the onset of damaging downburst winds from high-reflectivity storm cells that develop in an environment of weak vertical shear with Weather Surveillance Radar-1988 Doppler (WSR-88D) is examined. Ninety-one storm cells that produced damaging outflows are analyzed with data from the WSR- 88D network, along with 1247 nonsevere storm cells that developed in the same environments. Twenty-six reflectivity and radial velocity–based parameters are calculated for each cell, and a linear discriminant analysis was performed on 65% of the dataset in order to develop prediction equations that would discriminate between severe downburst-producing cells and cells that did not produce a strong outflow. These prediction equations are evaluated on the remaining 35% of the dataset. The datasets were resampled 100 times to determine the range of possible results. The resulting automated algorithm has a median Heidke skill score (HSS) of 0.40 in the 20–45-km range with a median lead time of 5.5 min, and a median HSS of 0.17 in the 45–80-km range with a median lead time of 0 min. As these lead times are medians of the mean lead times calculated from a large, resampled dataset, many of the storm cells in the dataset had longer lead times than the reported median lead times. [ABSTRACT FROM AUTHOR]

From 15 July through 30 September of 2001, an ensemble cloud-scale model was run for the Storm Prediction Center on a daily basis. Each ensemble run consisted of 78 members whose initial conditions were derived from the 20-km Rapid Update Cycle Model, the 22-km operational Eta Model, and a locally run version of the 22-km Eta Model using the Kain–Fritsch convective parameterization. Each ensemble was run over a 160 km × 160 km region and was valid for the 9-h period from 1630 through 0130 UTC. The ensembles were used primarily to provide severe-weather guidance. To that end, model storms with lifetimes greater than 60 min and/or a sustained correlation of at least 0.5 between midlevel updrafts and positive vorticity (the supercell criterion) were considered to be severe-weather indicators. Heidke skill scores, along with the true skill statistic, are between 0.2 and 0.3 when long-lived storms or storms meeting the supercell criteria are used as severe-weather indicators. Equivalent skill scores result when modeled and observed storms are categorized by lifetime and supercell characteristics and compared with expertly interpreted radar data. [ABSTRACT FROM AUTHOR]

As computational capacity has increased, cloud-scale numerical models are slowly being modified from pure research tools to forecast tools. Previous studies that used cloud-scale models as explicit forecast tools, in much the same way as a mesoscale model might be used, have met with limited success. Results presented in this paper suggest that this is due, at least in part, to the nature of cloud-scale models themselves. Results from over 700 cloud-scale model runs indicate that, in some cases, differences in the initial soundings that are smaller than can be measured by the current observing system result in unexpected differences in storm longevity. In other cases, easily measurable differences in the initial soundings do not result in significant differences in storm longevity. There unfortunately appears to be no set of parameters that can be used to determine whether the initial sounding is near some part of the cloud-model parameter space that displays this sensitivity. Because different cloud models share similar philosophies, if not similar design, this sensitivity to initial soundings places a fundamental limit on how well the current slate of cloud-scale models can be expected to perform as explicit forecast tools. Given these results, it is not clear that using state-of-the-art cloud-scale models as explicit forecasting tools is appropriate. However, cloud-model ensembles may help to address some inescapable problems with explicit forecasts from cloud models. The most useful application of cloud-scale models in operational forecasts may be a probabilistic one in which the models are used as members of ensembles, a process that has been demonstrated for models of larger-scale processes. [ABSTRACT FROM AUTHOR]