NWP Model Research Articles

Weather Prediction Using Machine Learning Classifiers

This study evaluates machine learning algorithms to increase weather predicting accuracy. However, traditional NWP models are computationally time-consuming and may fail to capture the nonlinear complexities inherent in meteorological observations. This research studies three machine learning algorithms—XGBoost, Gradient Boosting, and Random Forest. For the prediction of future conditions based on detailed patterns of past weather, these methods are adopted. The XGBoost model recorded an accuracy rate of 82.25% which is higher than the Random Forest and Gradient Boosting models. All models performed well in normal weather conditions such as "sun" and "rain," but were least competent at predicting less common phenomena like "drizzle," "fog," and "snow." This demonstrates the challenge of making predictions when unbalanced data sets are used in weather forecasting exercises. The incorporation of such machine learning techniques with conventional NWP models has the potential to result in improved prediction performance in various applications.

Changes of Intense Extratropical Cyclone Deepening Mechanisms in a Warmer Climate in Idealized Simulations

Abstract To understand how extratropical cyclones (ETCs) may change in a warmer climate, we conduct idealized baroclinic life cycle simulations using the Icosahedral Nonhydrostatic (ICON)–NWP model with varied initial conditions. With respect to a present-day climate, two experiments are highlighted: a 4-K uniform warming and a more realistic late twenty-first-century warming pattern projected by a Coupled Model Intercomparison Project, version 6 (CMIP6), model. Different ETC deepening mechanisms, especially by diabatic processes, are quantified via the pressure tendency equation analysis, and the horizontal model resolution dependency is examined by contrasting coarse-grid (80 km) and convection-permitting (2.5 km) simulations. While our simulated ETCs are primarily baroclinically driven, dominated by the horizontal warm-air advection in the air column above the surface low, such an effect only strengthens by 10% in both warming experiments. However, the direct contribution of diabatic heating to surface pressure drop almost doubles, which likely feeds back positively to horizontal warm-air advection. Although their combined response to warming is pronounced, it is largely offset by the strengthened adiabatic cooling (17%) due to enhanced upward motions in warmer and moister ETCs, leading to a marginal ETC deepening at maturity (lowers by ∼1.5–4 hPa). Nevertheless, the near-surface impacts strongly increase, particularly the local extreme precipitation (up to 56%). The convection-permitting and the coarse-grid simulations show qualitatively consistent ETC responses to global warming. We suggest that the systematically weaker ETCs (with higher central pressure) in 2.5-km compared to 80-km simulations might be primarily caused by model uncertainty in representing the convective-diabatic heating over the warm front near the cyclone core.

Representation of Blowing Snow and Associated Visibility Reduction in an Operational High-Resolution Weather Model

Abstract Blowing snow is a hazard for motorists because it may rapidly reduce visibility. Numerical weather prediction models in the United States do not capture the movement of snow once it reaches the ground, but visibility reductions due to blowing snow can be diagnosed based on model-predicted land surface and environmental conditions that correlate with blowing snow occurrence. A recently developed diagnostic framework for forecasting blowing snow concentration and the associated visibility reduction is applied to High-Resolution Rapid Refresh (HRRR) and Rapid Refresh Forecast System (RRFS) model output including surface snow conditions to predict surface visibility reduction due to blowing snow. Twelve blowing snow events around Wyoming from 2018 to 2023 are examined. The analysis shows that visibility reductions due to blowing snow tend to be overpredicted, caused by the initial assumption of full driftability of the snowpack. This study refines the aging of the blowing snow reservoir with two methods. The first method estimates driftability based on time-varying snow density from the Rapid Update Cycle land surface model (RUC LSM) used in the HRRR and experimental RRFS models and is evaluated in a real-time context with the RRFS model. The second, complementary method diagnoses snowpack driftability using a process-based approach that requires data for recent snowfall, wind speed, and skin temperature. Compared to the full driftability assumption, this method shows limited improvements in forecasting skill. To improve model-based diagnosis of visibility reduction due to blowing snow, empirical work is needed to determine the relation between snowpack driftability and the recent history of snowfall and other weather conditions. Significance Statement Blowing snow presents a significant hazard to motorists and airport operations through sometimes very rapid and intense reductions in visibility, yet little predictive guidance exists for blowing snow. This study aims to improve the prediction of blowing snow occurrence and associated surface visibility reduction using diagnostics from an operational high-resolution weather model. One key challenge regards the question of driftability of the snowpack. This study evaluates two approaches to quantify driftability in terms of visibility reduction due to blowing snow and acknowledges that more measurements are needed to improve the representation of blowing snow physics in NWP models.

Evaluation of near‐surface and boundary‐layer meteorological conditions that support cold‐fog formation using Cold Fog Amongst Complex Terrain field campaign observations

AbstractCold fog refers to a type of fog that forms when the temperature is below 0°C. It can be composed of liquid, ice, and mixed‐phase fog particles. Cold fog happens frequently over mountainous terrain in the cold season, but it is difficult to predict. Using observations from the Cold Fog Amongst Complex Terrain (CFACT) field campaign conducted in Heber Valley, Utah, in the western United States during January and February of 2022, this study investigates the meteorological conditions in the surface and boundary layers that support the formation of wintertime ephemeral cold fog in a local area of small‐scale mountain valleys. It is found that fog formation is susceptible to subtleties in forcing conditions and is supported by several factors: (1) established high pressure over the Great Basin with associated local clear skies, calm winds, and a stable boundary layer; (2) near‐surface inversion with saturation near the surface and strong moisture gradient in the boundary layer; (3) warm (above‐freezing) daytime air temperature with a large diurnal range, accompanied with warm soil temperatures during the daytime; (4) a period of increased turbulence kinetic energy (above 0.5 m2·s−2), followed by calm conditions throughout the fog's duration; and (5) supersaturation with respect to ice. Then, the field observations and identified supporting factors for fog formation were utilized to evaluate high‐resolution (˜400 m horizontal grid spacing) Weather Research and Forecasting (WRF) model simulations. Results show that the WRF model accurately simulates the mesoscale conditions facilitating cold‐fog formation but misses some critical surface and atmospheric boundary conditions. The overall results from this paper indicate that these identified factors that support fog formation are vital to accurately forecasting cold‐fog events. At the same time, they are also critical fields for the NWP model validation.

Common Community Physics Package: Fostering Collaborative Development in Physical Parameterizations and Suites

Abstract The Common Community Physics Package (CCPP) is a state-of-the-art infrastructure designed to facilitate community-wide development of atmospheric physics parameterizations, support their interoperability among different modeling centers, and enable the transition of research to operations in NWP and climate modeling. The CCPP consists of two elements: the Physics (a repository of parameterizations) and the Framework (an infrastructure for interfacing the parameterizations with host models). The CCPP is a community resource: Its latest release has 23 primary parameterizations, which can be organized into six supported suites. It is distributed with a single-column model to facilitate physics development and experimentation. The Developmental Testbed Center provides support to users and developers. A key aspect of the CCPP is its interoperability, that is, its ability to be used by multiple host models. This enables synergistic collaboration among groups dispersed over various institutions and working on various models. In this article, we provide an overview of the CCPP and how it is being used in two leading modeling systems. The CCPP is part of the Unified Forecast System (UFS), is included in the NOAA operational Hurricane Analysis and Forecast System (HAFS) version 1, and is slated for use in all upcoming NOAA global and limited-area UFS applications for operations. Similarly, the CCPP has been integrated into the Navy Environmental Prediction System Using a Nonhydrostatic Engine (NEPTUNE) model and is undergoing testing for the upcoming transition to operations. These experiences make physics interoperability a reality and open the doors for much broader collaborative efforts on ESM development.

Evaluating Day-Ahead Solar Radiation Forecasts from ICON, GFS, and MeteoFrance Global NWP Models

Evaluating Day-Ahead Solar Radiation Forecasts from ICON, GFS, and MeteoFrance Global NWP Models

Use of CAMS near Real-Time Aerosols in the HARMONIE-AROME NWP Model

Near real-time aerosol fields from the Copernicus Atmospheric Monitoring Services (CAMS), operated by the European Centre for Medium-Range Weather Forecasts (ECMWF), are configured for use in the HARMONIE-AROME Numerical Weather Prediction model. Aerosol mass mixing ratios from CAMS are introduced in the model through the first guess and lateral boundary conditions and are advected by the model dynamics. The cloud droplet number concentration is obtained from the aerosol fields and used by the microphysics and radiation schemes in the model. The results show an improvement in radiation, especially during desert dust events (differences of nearly 100 W/m2 are obtained). There is also a change in precipitation patterns, with an increase in precipitation, mainly during heavy precipitation events. A reduction in spurious fog is also found. In addition, the use of the CAMS near real-time aerosols results in an improvement in global shortwave radiation forecasts when the clouds are thick due to an improved estimation of the cloud droplet number concentration.

Lead‐time‐continuous statistical postprocessing of ensemble weather forecasts

AbstractNumerical weather prediction (NWP) ensembles often exhibit biases and errors in dispersion, so they need some form of postprocessing to yield sharp and well‐calibrated probabilistic predictions. The output of NWP models is usually at a multiplicity of different lead times and, even though information is often required on this range of lead times, many postprocessing methods in the literature are applied either at a fixed lead time or by fitting individual models for each lead time. However, this is (1) computationally expensive because it requires the training of multiple models if users are interested in information at multiple lead times and (2) prohibitive because it restricts the data used for training postprocessing models and the usability of fitted models. This article investigates the lead‐time dependence of postprocessing methods in the idealized Lorenz'96 system as well as temperature and wind‐speed forecast data from the Met Office Global and Regional Ensemble Prediction System (MOGREPS‐G). The results indicate that there is substantial regularity between the models fitted for different lead times and that one can fit models that are lead‐time‐continuous that work for multiple lead times simultaneously by including lead time as a covariate. These models achieve similar and, in small data situations, even improved performance compared with the classical lead‐time‐separated models, whilst saving substantial computation time.

Precursors of hazard due to super cyclone AMPHAN for Kolkata, India from surface observations

The track, intensity, and associated hazards of a cyclone are mostly pre- dicted using NWP models, satellites, and radar. Though the cyclones originate and strengthen in the ocean, they cause devastation in the populated land area over which they ultimately pass. Over the years, the accuracy of cyclone prediction has improved a lot. Yet, there is some uncertainty in the accurate prediction of track, intensity and associated hazards. In this article, we have studied super cyclone AMPHAN and its hazards for Kolkata, India. Here, we have proposed a new scheme for improving the forecast accuracy for cyclone distance, associated wind, and hazard for lead time up to 12-24 hours ahead based on curve fitting techniques and extrapolation using surface observational data. For the prediction of distance of the system from the concerned station and corresponding gusty wind speed, the accuracy of the proposed scheme is found to be better than the existing operational forecast and various reputed NWP models.

Numerical investigation of the Pedrógão Grande pyrocumulonimbus using a fire to atmosphere coupled model

Understanding the development of fire-generated thunderstorms in mega fire events is important given their high impact on the evolution of the fire fronts, where the fire spread becomes highly unpredictable and difficult to suppress. This study aims to investigate numerically the influence of strong pyro-convective activity on the local atmospheric conditions by means of a numerical simulation based on the coupled Meso-NH/ForeFire code. To our knowledge, it is the first time that the effect of wildfire spread on the local atmospheric conditions is accounted explicitly in a high-resolution NWP model to investigate pyro-convection activity. More specifically, we study numerically the Portuguese Pedrógão Grande mega fire, which was one of the most destructive and deadliest wildfire hazards affecting the Mediterranean region in the recent years. The spatio-temporal propagation of the wildfire was assigned a priori on the basis of the official investigation's reports, while the impact of the forced fire evolution and of the ensuing heat and water vapour emissions on the local atmospheric conditions is accounted explicitly. The simulation, configured with very-high spatial and temporal resolutions, was capable of resolving the intense convective column reaching the upper troposphere and the fast development of the associated cloud system. The numerical fire produced intense updraughts with vertical velocities above 15 m/s, whereas the associated pyroCb cloud was composed by five different hydrometeor species along the main convective column and reached an altitude of 10 km. It is remarkable that the numerical experiment reproduced phenomena occurring at a fine scale related to cloud microphysics, such as very-localized outflows. This study, based on a coupled numerical simulation, was capable of illustrating in detail the development of a pyroCb cloud from strong pyro-convective activity.

Measurements and Model Improvement: Insight into NWP Model Error Using Doppler Lidar and Other WFIP2 Measurement Systems

Abstract Doppler-lidar wind-profile measurements at three sites were used to evaluate NWP model errors from two versions of NOAA’s 3-km-grid HRRR model, to see whether updates in the latest version 4 reduced errors when compared against the original version 1. Nested (750-m grid) versions of each were also tested to see how grid spacing affected forecast skill. The measurements were part of the field phase of the Second Wind Forecasting Improvement Project (WFIP2), an 18-month deployment into central Oregon–Washington, a major wind-energy-producing region. This study focuses on errors in simulating marine intrusions, a summertime, 600–800-m-deep, regional sea-breeze flow found to generate large errors. HRRR errors proved to be complex and site dependent. The most prominent error resulted from a premature drop in modeled marine-intrusion wind speeds after local midnight, when lidar-measured winds of greater than 8 m s−1 persisted through the next morning. These large negative errors were offset at low levels by positive errors due to excessive mixing, complicating the interpretation of model “improvement,” such that the updates to the full-scale versions produced mixed results, sometimes enhancing but sometimes degrading model skill. Nesting consistently improved model performance, with version 1’s nest producing the smallest errors overall. HRRR’s ability to represent the stages of sea-breeze forcing was evaluated using radiation budget, surface-energy balance, and near-surface temperature measurements available during WFIP2. The significant site-to-site differences in model error and the complex nature of these errors mean that field-measurement campaigns having dense arrays of profiling sensors are necessary to properly diagnose and characterize model errors, as part of a systematic approach to NWP model improvement. Significance Statement Dramatic increases in NWP model skill will be required over the coming decades. This paper describes the role of major deployments of accurate profiling sensors in achieving that goal and presents an example from the Second Wind Forecast Improvement Program (WFIP2). Wind-profile data from scanning Doppler lidars were used to evaluate two versions of HRRR, the original and an updated version, and nested versions of each. This study focuses on the ability of updated HRRR versions to improve upon predicting a regional sea-breeze flow, which was found to generate large errors by the original HRRR. Updates to the full-scale HRRR versions produced mixed results, but the finer-mesh versions consistently reduced model errors.

Automatic cumulonimbus and towering cumulus identification based on the Italian weather radar network data

AbstractWe present an automatic cumulonimbus and towering cumulus cloud identification method, based on the Italian weather radar network data coupled with a NWP model and ground station data. In particular, we focus on its aeronautical applications as a support for the aerodrome automated weather observation systems and METAR AUTO reports. In addition to the significant convective cloud detection and discrimination, we also present a new methodology for convective cloud coverage and ceiling estimations.

A study on some dynamical aspects of Uttarakhand heavy rainfall events

In recent years, heavy rainfall events have been increasing over the Uttarakhand region. Improvement in the prediction of such events crucially dependent on the inclusion of the physical & dynamical processes responsible for such events, in the NWP model. This again, in turn depends on the understanding of such processes. In this study an attempt has been made to understand parts of these processes and some of the dynamical aspects of these heavy rainfall events. For this different important derived NWP products, viz., differential vorticity advection (DVA), differential thermal advection (DTA), Differential moisture advection (DMA), Precipitable water (PW), non-dimensional stability index (NDSI) have been computed using ECMWF high-resolution gridded reanalysis data sets. Heavy rainfall events are defined using IMD high resolution gridded daily rainfall data set. Preliminary analysis revealed that there was a steady increase in DVA, decrease in DTA, increase in PW and decrease in DMA before the heavy rainfall event. An enhanced DVA results in an enhancement in LLC, a decrease in DTA along with a decrease in DMA results in an enhancement of lapse rate. Combined effect of these results in the increase in the low-level convergence at Uttarakhand region along with the rising motion are the major dynamical processes resulted in the heavy rainfall event.

Test of Power Transformation Function to Hydrometeor and Water Vapor Mixing Ratios for Direct Variational Assimilation of Radar Reflectivity Data

Abstract Assimilating radar reflectivity into convective-scale NWP models remains a challenging topic in radar data assimilation. A primary reason is that the reflectivity forward observation operator is highly nonlinear. To address this challenge, a power transformation function is applied to the WRF Model’s hydrometeor and water vapor mixing ratio variables in this study. Three 3D variational data assimilation experiments are performed and compared for five high-impact weather events that occurred in 2019: (i) a control experiment that assimilates reflectivity using the original hydrometeor mixing ratios as control variables, (ii) an experiment that assimilates reflectivity using power-transformed hydrometeor mixing ratios as control variables, and (iii) an experiment that assimilates reflectivity and retrieved pseudo–water vapor observations using power-transformed hydrometeor and water vapor mixing ratios (qυ) as control variables. Both qualitative and quantitative evaluations are performed for 0–3-h forecasts from the five cases. The analysis and forecast performance in the two experiments with power-transformed mixing ratios is better than the control experiment. Notably, the assimilation of pseudo–water vapor with power-transformed qυ as an additional control variable is found to improve the performance of the analysis and short-term forecasts for all cases. In addition, the convergence rate of the cost function minimization for the two experiments that use the power transformation is faster than that of the control experiments. Significance Statement The effective use of radar reflectivity observations in any data assimilation scheme remains an important research topic because reflectivity observations explicitly include information about hydrometeors and also implicitly include information about the distribution of moisture within storms. However, it is difficult to assimilate reflectivity because the reflectivity forward observation operator is highly nonlinear. This study seeks to identify a more effective way to assimilate reflectivity into a convective-scale NWP model to improve the accuracy of predictions of high-impact weather events.

Interactive 3D Visual Analysis of Weather Prediction Data Reveals Midlatitude Overshooting Convection during the CIRRUS-HL Field Experiment

Abstract In summer 2021, microphysical properties and climate impact of high- and midlatitude ice clouds over Europe and the North Atlantic were studied during the Cirrus High Latitude (CIRRUS-HL) airborne field campaign. The related forecasting and flight planning tasks provided a testbed for interactive 3D visual analysis. Operational analyses and forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) were visualized with the open-source software “Met.3D.” A combination of traditional 2D displays with innovative 3D views in the interactive visualization framework facilitated rapid and comprehensive exploration of the NWP data. By this means, the benefit of interactive 3D visual forecast products in the routine flight planning procedure was evaluated. Here, we describe the use of 3D tropopause and cloud visualizations during a convective event over the Alps, which became one of the CIRRUS-HL observation targets. For the planning of the research flight on 8 July 2021, our analysis revealed that simulated strong convective updrafts locally disturb the tropopause and inject ice water across the dynamical tropopause into the lower stratosphere. The presented example provides a novel 3D perspective of convective overshooting in a global NWP model and its impact on the tropopause and lower stratosphere. The case study shall encourage the atmospheric science community to further evaluate the use of modern 3D visualization capabilities for NWP analysis.

A case study of exceptionally heavy rainfall event over Uttarakhand, India on 18th October, 2021 and its forecasting

The unprecedented rainfall observed over Uttarakhand on 18th October 2021 caused landslides, debris flow and floods over the Kumaun region and adjoining districts of the Garhwal region of Uttarakhand, which resulted in huge damage to life, agriculture, transport, tourism and other sectors. The synoptic and dynamic study of the current event showed the movement of the Low-Pressure Area over central India resulting in the strong southeasterly winds (Atmospheric River) over Indo-Gangetic planes from the Bay of Bengal from 17th to 19th October. The interaction and blocking of the Atmospheric River by the deep trough of eastward-moving Western Disturbance (WD) caused extreme rainfall over Uttarakhand. However, the X-band Doppler Weather Radar and 123 Automatic Weather/raingauge Stations data suggest that the hourly rainfall rate was of light to moderate intensity (10-20 mm/h) over most of the area and at most of the time. The rainfall rate was extremely intense (50-100 m/hour) for around 1-hour duration in 7 stations of Udham Singh Nagar, Champawat, Nainital and Pauri districts. Unlike the June 2013 extremely heavy rainfall event over Uttarakhand which impacted the whole Uttarakhand state, the present event was concentrated over the Kumaun region and the highest ever 24-hours accumulated rainfall was observed on 18th October, 2021 in Kumaon region of Uttarakhand. The expected rainfall as well as the impact of the event over Uttarakhand was forecasted 5 days in advance with good accuracy based on the synoptic analysis and NWP model guidance. The predictability of the IMD-GFS (T-1534) NWP model was found to be up to 10 days for this extreme rainfall event.

Wind observations from hot‐air balloons and the application in an NWP model

AbstractIn this paper, we report on a wind observation method based on the movement of hot‐air balloons (HABs). A quality assessment was carried out by comparing against wind observations at the meteorological tower of Cabauw in the Netherlands during May–September 2018, and the obtained standard deviations in error were and for the measured zonal and meridional wind components, respectively. Subsequent comparison against short‐term forecasts of the HARMONIE‐AROME model showed a standard deviation of 2.5 ms−1 for the wind vector difference. From the HAB observation set, a case was selected with a rapidly changing wind field belonging to a small intensifying depression. The HAB wind observation was applied in data assimilation as a proof of principle for a single‐observation experiment. It is shown that in a complex baroclinic situation, the model state is slightly improved.

A short-term wind power prediction method based on dynamic and static feature fusion mining

Wind power is a kind of time-varying time series with fluctuation characteristics. To take full advantage of the time-varying value provided by wind power fluctuations, a short-term wind power prediction method based on dynamic and static feature fusion mining is proposed. First, three statistical features are manually constructed to characterize the dynamic fluctuation of wind speed, these features provide more valuable patterns for the input data. Then, we construct a residual network structure that incorporates the bidirectional gate recurrent unit, and incorporate temporal and spatial attention mechanisms in the network structure. This network structure is used to train the wind power prediction model, which has great advantages in reducing the degradation and overfitting problems caused by increasing the depth of the network. Finally, a wind power prediction index is proposed to quantify the proportion of NWP link error and modeling link error in the total error. Simulation experiments were conducted on a wind farm with an installed capacity of 400.5 MW in Jilin Province, China, and the predicted NRMSE is 0.1581.

SwinRDM: Integrate SwinRNN with Diffusion Model towards High-Resolution and High-Quality Weather Forecasting

Data-driven medium-range weather forecasting has attracted much attention in recent years. However, the forecasting accuracy at high resolution is unsatisfactory currently. Pursuing high-resolution and high-quality weather forecasting, we develop a data-driven model SwinRDM which integrates an improved version of SwinRNN with a diffusion model. SwinRDM performs predictions at 0.25-degree resolution and achieves superior forecasting accuracy to IFS (Integrated Forecast System), the state-of-the-art operational NWP model, on representative atmospheric variables including 500 hPa geopotential (Z500), 850 hPa temperature (T850), 2-m temperature (T2M), and total precipitation (TP), at lead times of up to 5 days. We propose to leverage a two-step strategy to achieve high-resolution predictions at 0.25-degree considering the trade-off between computation memory and forecasting accuracy. Recurrent predictions for future atmospheric fields are firstly performed at 1.40625-degree resolution, and then a diffusion-based super-resolution model is leveraged to recover the high spatial resolution and finer-scale atmospheric details. SwinRDM pushes forward the performance and potential of data-driven models for a large margin towards operational applications.

A Single-Column Method to Identify Sea and Land Breezes in Mesoscale-Resolving NWP Models

Abstract One of the most prominent mesoscale phenomena in the coastal zone is the sea-breeze/land-breeze circulation. The pattern and its implications for the weather in coastal areas are well described, and with mesoscale-resolving operational NWP models the circulation can be captured. In this study, a straightforward method to identify sea and land breezes based on the change in wind direction in the column above a grid point on the coastline is presented. The method was tested for southern Sweden using archived output from the HARMONIE-AROME model with promising results, describing both the seasonal and diurnal cycles well. In areas with a complex coastline, such as narrow straits, the concept of the land–sea breeze becomes less clear, and several ways to address this problem for the suggested method are discussed. With an operational index of the sea and land breezes, the forecaster can better understand and express the weather situation and add value for people in the coastal zone. Further, the indices can be used to study systematic biases in the model and to create climatologies of the sea and land breezes. Significance Statement A wind pattern that is frequently occurring in the coastal zone is the sea-breeze/land-breeze circulation, and the purpose of this study is to test a new method to automatically identify sea breezes and land breezes in weather forecasts. Knowing if a sea breeze or a land breeze is occurring is helpful for the operational weather forecaster in understanding the weather situation. It can also be used to study systematic model behavior, for example, errors in the forecast temperature during sea-breeze conditions. The method has been tested for seven coastal sites in Sweden and shows promising results both in case studies and multiyear statistics.