North American Mesoscale Model Research Articles

An evaluation of meteorological data prediction over Washington, D.C.: Comparison of DCNet observations and NAM model outputs

This study presents an example of how outputs of operational and readily-available mesoscale numerical models can be adapted to initialize dispersion calculations within the urban surface roughness layer. Three years of urban meteorological observations from central Washington, DC, are compared against forecast outputs of the North American Mesoscale (NAM) model. NAM wind speed predictions underestimate the observations in light winds and overestimate the measurements in high winds. Average wind directions are consistent. However, an adjustment of the predicted direction of the plume by −20° is needed. The uncertainty associated with this adjustment is large in light NAM wind speed with no evident variation by season. The values of the standard deviation of the wind direction, σθ derived from NAM model outputs underestimate the observations by a small amount (about −1.5 to −2.5°). The results presented here indicate that mesoscale numerical model outputs can provide information adequate for dispersion calculations. However, levels of uncertainty associated with implementation of the suggested procedures increase with decreasing wind speed, causing considerable uncertainty in the implementation of adjustments as low wind speed conditions are approached. Results and recommendations reported here should not be extended to other numerical models or other cities without further testing.

Short-Term Weather Forecast Skill of Artificial Neural Networks

Abstract We evaluate the short-term weather forecast performance of three flavors of artificial neural networks (NNs): feed forward back propagation, radial basis function, and generalized regression. To prepare the application of the NNs to an operational setting, we tune NN hyperparameters using over two years of historical data. Five objective guidance products serve as predictors to the NNs: North American Mesoscale and Global Forecast System model output statistics (MOS) forecasts, the High-Resolution Rapid Refresh (HRRR) model, National Weather Service forecasts, and the National Blend of Models product. We independently test NN performance using 96 real-time forecasts of temperature, wind, and precipitation across 11 U.S. cities made during the WxChallenge, a weather forecasting competition. We demonstrate that all NNs significantly improve short-range weather forecasts relative to the traditional objective guidance aids used to train the networks. For example, 1-day maximum and minimum temperature forecast error is 20%–30% lower than MOS. However, NN improvement over multiple linear regression for short-term forecasts is not significant. We suggest this may be attributed to the small number of training samples, the operational nature of the experiment, and the short forecast lead times. Regardless, our results are consistent with previous work suggesting that applying NNs to model forecasts can have a positive impact on operational forecast skill and will become valuable tools when integrated into the forecast enterprise. Significance Statement We used approximately two years of historical weather data and objective forecasts for a number of cities to tune a series of artificial neural networks (NNs) to forecast 1-day values of maximum and minimum temperature, maximum sustained wind speed, and quantitative precipitation. We compare forecast error against common objective guidance and multiple linear regression. We found that the NNs exhibit about 25% lower error than common objective guidance for temperature forecasting and 50% lower error for wind speed. Our results suggest that NNs will be a valuable contributor to improving weather forecast skill when adopted into the existing forecast enterprise.

The cost of day-ahead solar forecasting errors in the United States

As solar energy contributes an increasing share of total electricity generation, solar forecasting errors become important relative to overall load uncertainty and can add costs to electricity systems. We investigated the costs of day-ahead solar forecast errors across 667 existing solar power plants in the United States (years 2012 through 2019). Our analysis was based on hourly real-time and day-ahead nodal prices. We analyzed two types of solar forecasts: persistence forecasts, a simple approach to forecasting, and a numerical weather prediction forecast, the North American Mesoscale Model (NAM), an improvement over persistence forecasts based on public data and modelling software. We modeled hourly energy forecasts using meteorological forecasts and plant specific characteristics. Hourly plant generation was modeled and debiased with multiple sources of generation records. NAM forecast errors had relatively low costs on average, at no more than $1/MWh in all years except 2016, when costs rose to $1.5/MWh. Even after these error costs, the value of solar was marginally higher when simulating solar participation in day-ahead markets versus participation only in real-time markets. On average, the premium for participating in the day-ahead market, based on NAM forecasts, ranged from −0.5 to 5.2 $/MWh across years. Average error costs were higher in regions with higher solar penetration (i.e., California and New England) compared to regions with low solar penetration. However, California and New England had similar error costs despite higher solar penetration in California, indicating that error costs to date have been only loosely correlated with solar penetration levels.

Local Quantitative Precipitation Forecast with Minimal Data Requirement—An Ensemble Approach

Abstract Operational weather forecasts are routinely performed at convection-allowing resolutions, and thus these forecasts generate weather features that appear to be realistic. However, at times the comparison of the forecast to observations is less favorable, particularly at grid scales. This lack of skill is partly due to the chaotic system underlying the weather. Another problem is that it is impossible to evaluate the risk of making decisions based on deterministic forecasts. However, running global high-resolution ensembles involves substantial computational resources. A 555-m resolution WRF ensemble based on stochastic perturbations of a deterministic forecast of the North American Mesoscale model was created. Observations are used to constrain the ensemble and improve the skill. This method increases the skill for forecasting 60-h accumulated precipitation in five standard, statistical metrics: bias, false alarm ratio, threat score, probability of detection, and success ratio. Furthermore, the ensemble continuous ranked probability score (CRPS) will be compared to a poor man’s ensemble. The forecast error is generally smaller in more than 70% of the case studies performed when compared to nine deterministic model forecasts. The ensemble-enhanced mesoscale system presented can help to determine the most likely scenario without the significant computational requirement of global ensembles and is expected to be useful when global high-resolution ensembles are not available.

Dynamically Optimized Unstructured Grid (DOUG) for Analog Ensemble of numerical weather predictions using evolutionary algorithms

The Analog Ensemble is a statistical technique that generates probabilistic forecasts using a current deterministic prediction, a set of historical predictions, and the associated observations. It generates ensemble forecasts by first identifying the most similar past predictions to the current one, and then summarizing the corresponding observations. This is a computationally efficient solution for ensemble modeling because it does not require multiple numerical weather prediction simulations, but a single model realization.Despite this intrinsic computational efficiency, the required computation can grow very large because atmospheric models are routinely run with increasing resolutions. For example, the North American Mesoscale forecast system contains over 262 792 grid points to generate a 12 km prediction. The North American Mesoscale model generally uses a structured grid to represent the domain, despite the fact that certain physical changes occur non-uniformly across space and time. For example, temperature changes tend to occur more rapidly in mountains than plains.An evolutionary algorithm is proposed to dynamically and automatically learn the optimal unstructured grid pattern. This iterative evolutionary algorithm is guided by Darwinian evolutionary rule generation and instantiation to identify grid vertices. Analog computations are performed only at vertices. Therefore, minimizing the number of vertices and identifying their locations are paramount to optimizing the available computational resources, minimizing queue time, and ultimately achieving better results. The optimal unstructured grid is then reused to guide the predictions for a variety of applications like temperature and wind speed.

Impacts of Local Convective Processes on Rain on the Caribbean Island of Puerto Rico

AbstractThis study aims to determine the impacts of tropical island processes on local convective storms. An analysis of rain events on the island of Puerto Rico between 1 June 2015 and 31 July 2016 showed that local island‐enhanced western storms accounted for 89 of 322 storms. This period is of particular importance for the Caribbean as 2015 was one of the driest years on record. While large‐scale influences such as the El Niño–Southern Oscillation, the North Atlantic Oscillation, African easterly waves, and Saharan dust transport modulate moisture conditions in the region, correlations between precipitation and El Niño–Southern Oscillation (−0.14), North Atlantic Oscillation (−0.42), and Saharan dust (0.1) for 1980–2016 ranged from weak to moderate. Local data for the island of Puerto Rico from weather stations, the Convection, Aerosol, and Synoptic‐Effects in the Tropics field campaign, and the North American Mesoscale model support the initiation or enhancement of convective rain events due to local island processes. In particular, analysis of surface wind speed/direction, convective available potential energy, lifted index, and the bulk Richardson number substantiate local instability due to surface heating, orographic uplift, and sea breeze trade‐wind convergence. These convective forcings along with available precipitable water in excess of 50 mm ultimately led to intense storms despite severe rainfall‐mitigating dust episodes for which aerosol optical thickness exceeded 0.4. These results may have major implications for considering the impacts of local air‐sea‐land interactions on rainfall over other tropical islands.

A Method for Detecting Atmospheric Lagrangian Coherent Structures Using a Single Fixed-Wing Unmanned Aircraft System.

The transport of material through the atmosphere is an issue with wide ranging implications for fields as diverse as agriculture, aviation, and human health. Due to the unsteady nature of the atmosphere, predicting how material will be transported via the Earth’s wind field is challenging. Lagrangian diagnostics, such as Lagrangian coherent structures (LCSs), have been used to discover the most significant regions of material collection or dispersion. However, Lagrangian diagnostics can be time-consuming to calculate and often rely on weather forecasts that may not be completely accurate. Recently, Eulerian diagnostics have been developed which can provide indications of LCS and have computational advantages over their Lagrangian counterparts. In this paper, a methodology is developed for estimating local Eulerian diagnostics from wind velocity data measured by a single fixed-wing unmanned aircraft system (UAS) flying in a circular arc. Using a simulation environment, driven by realistic atmospheric velocity data from the North American Mesoscale (NAM) model, it is shown that the Eulerian diagnostic estimates from UAS measurements approximate the true local Eulerian diagnostics and also predict the passage of LCSs. This methodology requires only a single flying UAS, making it easier and more affordable to implement in the field than existing alternatives, such as multiple UASs and Dopler LiDAR measurements. Our method is general enough to be applied to calculate the gradient of any scalar field.

An Evaluation of Paired Regional/Convection-Allowing Forecast Vertical Thermodynamic Profiles in Warm-Season, Thunderstorm-Supporting Environments

Abstract This study evaluates forecast vertical thermodynamic profiles and derived thermodynamic parameters from two regional/convection-allowing model pairs, the North American Mesoscale Forecast System and the North American Mesoscale Nest model pair and the Rapid Refresh and High Resolution Rapid Refresh model pair, in warm-season, thunderstorm-supporting environments. Differences in bias and mean absolute error between the regional and convection-allowing models in each of the two pairs, while often statistically significant, are practically small for the variables, parameters, and vertical levels considered, such that the smaller-scale variability resolved by convection-allowing models does not degrade their forecast skill. Model biases shared by the regional and convection-allowing models in each pair are documented, particularly the substantial cool and moist biases in the planetary boundary layer arising from the Mellor–Yamada–Janjić planetary boundary layer parameterization used by the North American Mesoscale model and the Nest version as well as the middle-tropospheric moist bias shared by the Rapid Refresh and High Resolution Rapid Refresh models. Bias and mean absolute errors typically have larger magnitudes in the evening, when buoyancy is a significant contributor to turbulent vertical mixing, than in the morning. Vertical thermodynamic profile biases extend over a deep vertical layer in the western United States given strong sensible heating of the underlying surface. The results suggest that convection-allowing models can fulfill the use cases typically and historically met by regional models in operations at forecast entities such as the Storm Prediction Center, a fruitful finding given the proposed elimination of regional models with the Next-Generation Global Prediction System initiative.

WRF Hub-Height Wind Forecast Sensitivity to PBL Scheme, Grid Length, and Initial Condition Choice in Complex Terrain

Abstract This study evaluates the sensitivity of wind turbine hub-height wind speed forecasts to the planetary boundary layer (PBL) scheme, grid length, and initial condition selection in the Weather Research and Forecasting (WRF) Model over complex terrain. Eight PBL schemes available for the WRF-ARW dynamical core were tested with initial conditions sources from the North American Mesoscale (NAM) model and Global Forecast System (GFS) to produce short-term wind speed forecasts. The largest improvements in forecast accuracy primarily depended on the grid length or PBL scheme choice, although the most important factor varied by location, season, time of day, and bias-correction application. Aggregated over all locations, the Asymmetric Convective Model, version 2 (ACM2) PBL scheme provided the best forecast accuracy, particularly for the 12-km grid length. Other PBL schemes and grid lengths, however, did perform better than the ACM2 scheme for individual seasons or locations.

Simulations of a Cold-Air Pool in Utah’s Salt Lake Valley: Sensitivity to Land Use and Snow Cover

Obtaining realistic land-surface states for initial and boundary conditions is important for the numerical weather prediction of many atmospheric phenomena. Here we investigate model sensitivity to land use and snow cover for a persistent wintertime cold-air pool in northern Utah during 1–8 January 2011. A Weather Research and Forecast model simulation using the 1993 United States Geological Survey land-use and North American Mesoscale model reanalysis snow-cover datasets is compared to an improved configuration using the modified 2011 National Land Cover Database and a more realistic representation of snow cover. The improved surface specification results in an increase (decrease) in urban land cover (Great Salt Lake surface area), and changes to the snow-cover initialization, depth, extent, and albedo. The results obtained from the model simulations are compared to observations collected during the Persistent Cold-Air Pool Study. The changes in land use and snow cover and the resulting impacts on the surface albedo and surface heat fluxes contributed to near-surface air temperature increases of 1– $$2\,^{\circ }\hbox {C}$$ in urban areas and decreases of 2– $$4\,^{\circ }\hbox {C}$$ in areas surrounding the Great Salt Lake. Although wind speeds in the boundary layer were overestimated in both simulations, shallow thermally-driven and terrain-forced flows were generally lessened in intensity and breadth in response to the decreased areal extent of the Great Salt Lake and increases in the urban footprint.

Forecasting Building Energy Demands With a Coupled Weather-Building Energy Model in a Dense Urban Environment

Major new metropolitan centers experience challenges during management of peak electrical loads, typically occurring during extreme summer events. These peak loads expose the reliability of the electrical grid on the production and transmission side, while customers may incur considerable charges from increased metered peak demand, failing to meet demand response program obligations, or both. These challenges create a need for analytical tools that can inform building managers and utilities about near future conditions so they are better able to avoid peak demand charges and reduce building operational costs. In this article, we report on a tool and methodology to forecast peak loads at the city scale using New York City (NYC) as a test case. The city of New York experiences peak electric demand loads that reach up to 11 GW during the summertime, and are projected to increase to over 12 GW by 2025, as reported by the New York Independent System Operator (NYISO). The energy forecast is based on the Weather Research and Forecast (WRF) model version 3.5, coupled with a multilayer building energy model (BEM). Urban morphology parameters are assimilated from the New York Primary Land Use Tax-Lot Output (PLUTO), while the weather component of the model is initialized daily from the North American Mesoscale (NAM) model. A city-scale analysis is centered in the summer months of June–July 2015 which included an extreme heat event (i.e., heat wave). The 24-h city-scale weather and energy forecasts show good agreement with the archived data from both weather stations records and energy records by NYISO. This work also presents an exploration of space cooling savings from the use of white roofs as an application of the city-scale energy demand model.

Preprocessing WRF initial conditions for coastal stratocumulus forecasting

The impact of atmospheric liquid water content at model initialization in the Weather Research and Forecasting (WRF) model is explored through the application of two preprocessing schemes. The first scheme, the Well-mixed Preprocessor (WEMPP), was designed and developed based on a well-mixed boundary layer to provide an initial guess at liquid water content when initializing with data from the North American Mesoscale (NAM) model, as liquid water content is not present in NAM output. The second scheme was adapted from a satellite Cloud Data Assimilation (CLDDA) package intended to make the initial model cloud field consistent with observations, using input data from the CIMSS GOES sounder cloud product.Preprocessed simulations were compared against baseline WRF simulations initialized with NAM and the Rapid Refresh (RAP) model (which contains liquid water output), as well as the raw parent model outputs. These intra-day forecasts were validated against both 5-min and 10-min averaged ground station and 30-min (hourly averaged) satellite irradiance observations over the course of a month.Due to their extensive spatial coverage, optical thickness, and reflectivity, stratocumulus (Sc) clouds are responsible for much of the variability in available solar resource at the surface in coastal California, where most rooftop photovoltaic systems are located. Currently, the trend of numerical weather prediction models is to underpredict both the presence and thickness of Sc. Therefore, the validation is conducted for a summer month in southern California, when Sc are most prevalent.Ground station validation showed average improvements by WEMPP in predicting surface irradiance over the baseline NAM (RAP) WRF initializations of 33% (−3%) MBE and 16% (9%) MAE, and by CLDDA of 47% (18%) MBE and 26% (20%) MAE. Additionally, simulations preprocessed by CLDDA were consistently able to outperform 24-h persistence forecast at 3 out of 4 ground stations.Validation against SolarAnywhere® satellite irradiance observations showed that the combination of both preprocessors provided the most improvement in the prediction of Sc spatial coverage, thickness, and lifetime in coastal regions where marine layer stratocumulus is most frequently observed, but cloud cover over the ocean was overestimated by all preprocessors.

Toward Improved Solar Irradiance Forecasts: Introduction of Post-Processing to Correct the Direct Normal Irradiance from the Weather Research and Forecasting Model

Solar electricity production is highly dependent on atmospheric conditions. This study focuses on comparing model forecasts with observations for the period of May–December, 2011. The Weather Research and Forecasting model was run for two nested domains centered on Arizona in order to better capture the complex terrain driven dynamics of the region. The modeling performance from the simulation with the Global Forecast System model output as initial and boundary condition was better, with respect to both direct normal irradiance and global horizontal irradiance, than that with the North American Mesoscale model output. The observed aerosol optical depth is correlated with the water vapor, soil moisture and wind-blown dust and therefore, the aerosol optical depth is parameterized by the modeling outputs for these variables. The aerosol correction factor reduces the relative root mean square error from 12 to 6 %. In cases where dust was transported at high altitude, our algorithm did not correct the bias of direct normal irradiance.

A Regression-Based Approach for Cool-Season Storm Surge Predictions along the New York–New Jersey Coast

AbstractA multilinear regression (MLR) approach is developed to predict 3-hourly storm surge during the cool-season months (1 October–31 March 31) between 1979 and 2012 using two different atmospheric reanalysis datasets and water-level observations at three stations along the New York–New Jersey coast (Atlantic City, New Jersey; the Battery in New York City; and Montauk Point, New York). The predictors of the MLR are specified to represent prolonged surface wind stress and a surface sea level pressure minimum for a boxed region near each station. The regression underpredicts relatively large (≥95th percentile) storm maximum surge heights by 6.0%–38.0%. A bias-correction technique reduces the average mean absolute error by 10%–15% at the various stations for storm maximum surge predictions. Using the same forecast surface winds and pressures from the North American Mesoscale (NAM) model between October and March 2010–14, raw and bias-corrected surge predictions at the Battery are compared with raw output from a numerical hydrodynamic model’s [the Stevens Institute of Technology New York Harbor Observing and Prediction System (SIT-NYHOPS)] predictions. The accuracy of surge predictions between the SIT-NYHOPS output and bias-corrected MLR model at the Battery are similar for predictions that meet or exceed the 95th percentile of storm maximum surge heights.

Analysis of the 1 December 2011 Wasatch Downslope Windstorm

Abstract A downslope windstorm on 1 December 2011 led to considerable damage along a narrow 50-km swath at the western base of the Wasatch Mountains in northern Utah. The strongest surface winds began suddenly at 0900 UTC, primarily in the southern portion of the damage zone. Surface winds reached their peak intensity with gusts to 45 m s−1 at ~1600 UTC, while the strongest winds shifted later to the northern end of the damage swath. The northward shift in strong surface winds relates to the rotation of synoptic-scale flow from northeasterly to easterly at crest level, controlled by an evolving anticyclonic Rossby wave breaking event. A rawinsonde released at ~1100 UTC in the midst of strong (>35 m s−1) easterly surface wind intersected a rotor and sampled the strong inversion that surmounted it. The windstorm’s evolution was further examined via Weather Research and Forecasting Model simulations initialized from North American Mesoscale Model analyses ~54 h before the windstorm onset. The Control model simulation captured core features of the event, including the spatial extent and timing of the strongest surface winds. However, the model developed stronger mountain-wave breaking in the lee of the Wasatch, a broader zone of strong surface winds, and a downstream rotor located farther west than observed. A second simulation, in which the nearby east–west-oriented Uinta Mountains were reduced in elevation, developed weaker easterly flow across the Wasatch during the early stages of the event. This result suggests that the Uinta Mountains block and steer the initial northeasterly flow across the Wasatch.

The Operational Implementation of a Great Lakes Wave Forecasting System at NOAA/NCEP*

Abstract The development of a Great Lakes wave forecasting system at NOAA’s National Centers for Environmental Prediction (NCEP) is described. The system is an implementation of the WAVEWATCH III model, forced with atmospheric data from NCEP’s regional Weather Research and Forecasting (WRF) Model [the North American Mesoscale Model (NAM)] and the National Digital Forecast Database (NDFD). Reviews are made of previous Great Lakes wave modeling efforts. The development history of NCEP’s Great Lakes wave forecasting system is presented. A performance assessment is made of model wind speeds, as well as wave heights and periods, relative to National Data Buoy Center (NDBC) measurements. Performance comparisons are made relative to NOAA’s Great Lakes Environmental Research Laboratory (GLERL) wave prediction system. Results show that 1- and 2-day forecasts from NCEP have good skill in predicting wave heights and periods. NCEP’s system provides a better representation of measured wave periods, relative to the GLERL model in most conditions. Wave heights during storms, however, are consistently underestimated by NCEP’s current operational system, whereas the GLERL model provides close agreement with observations. Research efforts to develop new wave-growth parameterizations and overcome this limitation have led to upgrades to the WAVEWATCH III model, scheduled to become operational at NCEP in 2013. Results are presented from numerical experiments made with the new wave-model physics, showing significant improvements to the skill of NCEP’s Great Lakes wave forecasting system in predicting storm wave heights.

Beyond the Basics: Evaluating Model-Based Precipitation Forecasts Using Traditional, Spatial, and Object-Based Methods

Abstract While traditional verification methods are commonly used to assess numerical model quantitative precipitation forecasts (QPFs) using a grid-to-grid approach, they generally offer little diagnostic information or reasoning behind the computed statistic. On the other hand, advanced spatial verification techniques, such as neighborhood and object-based methods, can provide more meaningful insight into differences between forecast and observed features in terms of skill with spatial scale, coverage area, displacement, orientation, and intensity. To demonstrate the utility of applying advanced verification techniques to mid- and coarse-resolution models, the Developmental Testbed Center (DTC) applied several traditional metrics and spatial verification techniques to QPFs provided by the Global Forecast System (GFS) and operational North American Mesoscale Model (NAM). Along with frequency bias and Gilbert skill score (GSS) adjusted for bias, both the fractions skill score (FSS) and Method for Object-Based Diagnostic Evaluation (MODE) were utilized for this study with careful consideration given to how these methods were applied and how the results were interpreted. By illustrating the types of forecast attributes appropriate to assess with the spatial verification techniques, this paper provides examples of how to obtain advanced diagnostic information to help identify what aspects of the forecast are or are not performing well.

Forecaster Use and Evaluation of Real-Time 3DVAR Analyses during Severe Thunderstorm and Tornado Warning Operations in the Hazardous Weather Testbed

Abstract A weather-adaptive three-dimensional variational data assimilation (3DVAR) system was included in the NOAA Hazardous Weather Testbed as a first step toward introducing warn-on-forecast initiatives into operations. NWS forecasters were asked to incorporate the data in conjunction with single-radar and multisensor products in the Advanced Weather Interactive Processing System (AWIPS) as part of their warning-decision process for real-time events across the United States. During the 2011 and 2012 experiments, forecasters examined more than 36 events, including tornadic supercells, severe squall lines, and multicell storms. Products from the 3DVAR analyses were available to forecasters at 1-km horizontal resolution every 5 min, with a 4–6-min latency, incorporating data from the national Weather Surveillance Radar-1988 Doppler (WSR-88D) network and the North American Mesoscale model. Forecasters found the updraft, vertical vorticity, and storm-top divergence products the most useful for storm interrogation and quickly visualizing storm trends, often using these tools to increase the confidence in a warning decision and/or issue the warning slightly earlier. The 3DVAR analyses were most consistent and reliable when the storm of interest was in close proximity to one of the assimilated WSR-88D, or data from multiple radars were incorporated into the analysis. The latter was extremely useful to forecasters in blending data rather than having to analyze multiple radars separately, especially when range folding obscured the data from one or more radars. The largest hurdle for the real-time use of 3DVAR or similar data assimilation products by forecasters is the data latency, as even 4–6 min reduces the utility of the products when new radar scans are available.

On the Dynamics, Thermodynamics, and Forecast Model Evaluation of Two Snow-Burst Events in Southern Alberta

Abstract Two high-impact convective snowband events (“snow bursts”) that affected Calgary, Alberta, Canada, are examined to better understand the dynamics and thermodynamics of heavy snowbands not associated with lake effects or the cold conveyor belt of synoptic-scale cyclones. Such events are typically characterized by brief, but heavy, periods of snow; low visibilities; and substantial hazards to automobile and aviation interests. Previous literature on these events has been limited to a few case studies across North America, including near the leeside foothills of the U.S. Rockies. The large-scale dynamics and thermodynamics are investigated using the National Centers for Environmental Prediction (NCEP) North American Regional Reanalysis (NARR). Previously, high-resolution convection-explicit Weather Research and Forecasting Model (WRF) simulations have shown some ability to successfully reproduce the dynamics, thermodynamics, and appearance of convective snowbands, with small errors in location and timing. Therefore, WRF simulations are performed for both events, and are evaluated along with the NCEP North American Mesoscale (NAM) model forecasts. Both the NARR and WRF simulations show that while the two snow bursts are similar in appearance, they form as a result of different dynamic and thermodynamic mechanisms. The first event occurs downstream of an upper-tropospheric jet streak in a region of little synoptic-scale ascent, where ageostrophic frontogenesis helps to release conditional, dry symmetric, and inertial instability in an unsaturated environment. The inertial instability is determined to be related to fast flow over upstream high terrain. The second event occurs in a saturated environment in a region of Q-vector convergence (primarily geostrophic frontogenesis), which acts to release conditional, convective, and conditional symmetric instability (CSI).

Precipitation and Temperature Forecast Performance at the Weather Prediction Center

Abstract The role of the human forecaster in improving upon the accuracy of numerical weather prediction is explored using multiyear verification of human-generated short-range precipitation forecasts and medium-range maximum temperature forecasts from the Weather Prediction Center (WPC). Results show that human-generated forecasts improve over raw deterministic model guidance. Over the past two decades, WPC human forecasters achieved a 20%–40% improvement over the North American Mesoscale (NAM) model and the Global Forecast System (GFS) for the 1 in. (25.4 mm) (24 h)−1 threshold for day 1 precipitation forecasts, with a smaller, but statistically significant, 5%–15% improvement over the deterministic ECMWF model. Medium-range maximum temperature forecasts also exhibit statistically significant improvement over GFS model output statistics (MOS), and the improvement has been increasing over the past 5 yr. The quality added by humans for forecasts of high-impact events varies by element and forecast projection, with generally large improvements when the forecaster makes changes ≥8°F (4.4°C) to MOS temperatures. Human improvement over guidance for extreme rainfall events [3 in. (76.2 mm) (24 h)−1] is largest in the short-range forecast. However, human-generated forecasts failed to outperform the most skillful downscaled, bias-corrected ensemble guidance for precipitation and maximum temperature available near the same time as the human-modified forecasts. Thus, as additional downscaled and bias-corrected sensible weather element guidance becomes operationally available, and with the support of near-real-time verification, forecaster training, and tools to guide forecaster interventions, a key test is whether forecasters can learn to make statistically significant improvements over the most skillful of this guidance. Such a test can inform to what degree, and just how quickly, the role of the forecaster changes.