Numerical Weather Prediction Model Forecasts Research Articles

Operational Aviation Icing Forecast Algorithm for the Korea Meteorological Administration

Abstract In this study, we developed and evaluated the Korean Forecast Icing Potential (K-FIP), an in-flight icing forecast system for the Korea Meteorological Administration (KMA) based on the simplified forecast icing potential (SFIP) algorithm. The SFIP is an algorithm used to postprocess numerical weather prediction (NWP) model forecasts for predicting potential areas of icing based on the fuzzy logic formulations of four membership functions: temperature, relative humidity, vertical velocity, and cloud liquid water content. In this study, we optimized the original version of the SFIP for the global NWP model of the KMA through three important updates using 34 months of pilot reports for icing as follows: using total cloud condensates, reconstructing membership functions, and determining the best weight combination for input variables. The use of all cloud condensates and the reconstruction of these membership functions resulted in a significant improvement in the algorithm compared with the original. The weight combinations for the KMA’s global model were determined based on the performance scores. While several sets of weights performed equally well, this process identified the most effective weight combination for the KMA model, which is referred to as the K-FIP. The K-FIP demonstrated the ability to successfully predict icing over the Korean Peninsula using observations made by research aircraft from the National Institute of Meteorological Sciences of the KMA. Eventually, the K-FIP icing forecasts will provide better forecasts of icing potentials for safe and efficient aviation operations in South Korea. Significance Statement In-flight aircraft icing has posed a threat to safe flights for decades. With advances in computing resources and an improvement in the spatiotemporal resolutions of numerical weather prediction (NWP) models, icing algorithms have been developed using NWP model outputs associated with supercooled liquid water. This study evaluated and optimized the simplified forecast icing potential, an NWP model–based icing algorithm, for the global model of the Korean Meteorological Administration (KMA) using a long-term observational dataset to improve its prediction skills. The improvements shown in this study and the SFIP implemented in the KMA will provide more informative predictions for safe and efficient air travel.

The G-IV Inner Circumnavigation: A Story of Successful Organic Interactions Between Research and Operations at NOAA

Abstract This study describes the research-to-operations process leading to a recent change in tropical cyclone (TC) reconnaissance sampling patterns as well as observing-system experiments that evaluated the impact of that change on numerical weather prediction model forecasts of TCs. A valuable part of this effort was having close, multipronged connections between the TC research and operational TC prediction communities at the National Oceanic and Atmospheric Administration (NOAA). Related to this work, NOAA’s Atlantic Oceanographic and Meteorological Laboratory (AOML) and National Hurricane Center (NHC) have a long history of close collaboration to improve TC reconnaissance. Similar connections between AOML and NOAA’s Environmental Modeling Center (EMC) also laid a foundation for the observing-system experiments conducted here. More specifically, AOML and NHC collaborated in 2018 to change how NHC uses NOAA’s Gulfstream-IV (G-IV) jet during TC synoptic surveillance missions. That change added a second circumnavigation at approximately 1.5° from TC centers, when possible. Preliminary experiments suggest that the change improved track forecasts, though the intensity results are more mixed. Despite the somewhat small sample size over a 3-yr period, the track improvement does agree with prior work. This effort has led to additional work to more fully examine G-IV sampling strategies.

Exploring the Use of Machine Learning to Improve Vertical Profiles of Temperature and Moisture

Abstract Vertical profiles of temperature and dewpoint are useful in predicting deep convection that leads to severe weather, which threatens property and lives. Currently, forecasters rely on observations from radiosonde launches and numerical weather prediction (NWP) models. Radiosonde observations are, however, temporally and spatially sparse, and NWP models contain inherent errors that influence short-term predictions of high impact events. This work explores using machine learning (ML) to postprocess NWP model forecasts, combining them with satellite data to improve vertical profiles of temperature and dewpoint. We focus on different ML architectures, loss functions, and input features to optimize predictions. Because we are predicting vertical profiles at 256 levels in the atmosphere, this work provides a unique perspective at using ML for 1D tasks. Compared to baseline profiles from the Rapid Refresh (RAP), ML predictions offer the largest improvement for dewpoint, particularly in the middle and upper atmosphere. Temperature improvements are modest, but CAPE values are improved by up to 40%. Feature importance analyses indicate that the ML models are primarily improving incoming RAP biases. While additional model and satellite data offer some improvement to the predictions, architecture choice is more important than feature selection in fine-tuning the results. Our proposed deep residual U-Net performs the best by leveraging spatial context from the input RAP profiles; however, the results are remarkably robust across model architecture. Further, uncertainty estimates for every level are well calibrated and can provide useful information to forecasters.

Verification of operational numerical weather prediction model forecasts of precipitation using satellite rainfall estimates over Africa

AbstractRainfall is an important variable to be able to monitor and forecast across Africa, due to its impact on agriculture, food security, climate‐related diseases and public health. Numerical weather models (NWMs) are an important component of this work, due to their complete spatial coverage, high resolution and ability to forecast into the future. In this study, the spatio‐temporal skill of short‐term forecasts of rainfall across Africa from 2016 through 2018 is evaluated. Specifically, the European Centre for Medium‐Range Weather Forecasts (ECMWF) and the National Centers for Environmental Prediction‐Global Forecast System (NCEP‐GFS) forecast models are verified by Rainfall Estimates 2.0 (RFE2) and African Rainfall Climatology Version 2 (ARC2), which are fused products of satellite and in situ observations and are commonly used in analysis of African rainfall. Model rainfall forecasts show good consistency with the satellite rainfall observations in spatial distribution over Africa on the seasonal timescale. Evaluation metrics of daily and weekly forecasts show high spatial and seasonal variations over the African continent, including a strong link to the location of the inter‐tropical convergence zone (ITCZ) and topographically enhanced precipitation. The rainfall forecasts at 1 week aggregation time are improved against daily forecasts.

Evaluating the Economic Impacts of Improvements to the High-Resolution Rapid Refresh (HRRR) Numerical Weather Prediction Model

Abstract Forecasts from numerical weather prediction (NWP) models play a critical role in many sectors of the U.S. economy. Improvements to operational NWP model forecasts are generally assumed to provide significant economic savings through better decision-making. But is this true? Since 2014, several new versions of the High-Resolution Rapid Refresh (HRRR) model were released into operation within the National Weather Service. Practically, forecasts have an economic impact only if they lead to a different action than what would be taken under an alternative information set. And in many sectors, these decisions only need to be considered during certain weather conditions. We estimate the economic impacts of improvements made to the HRRR, using 12-h wind, precipitation, and temperature forecasts in several cases where they can have “economically meaningful” behavioral consequences. We examine three different components of the U.S. economy where such information matters: 1) better integration of wind energy resources into the electric grid, 2) increased worker output due to better precipitation forecasts that allow workers to arrive to their jobs on time, and 3) better decisions by agricultural producers in preparing for freezing conditions. These applications demonstrate some of the challenges in ascertaining the economic impacts of improved weather forecasts, including highlighting key assumptions that must be made to make the problem tractable. For these sectors, we demonstrate that there was a marked economic gain for the United States between HRRR versions 1 and 2 and a smaller, but still appreciable economic gain between versions 2 and 3.

Machine Learning Applications to Dust Storms: A Meta-Analysis

Dust storms are natural hazards that affect both people and properties. Therefore, it is important to mitigate their risks by implementing an early notification system. Different methods are used to predict dust storms, such as observing satellite images, analyzing meteorological data, and using numerical weather prediction model forecasts. However, recent studies have shown that machine learning algorithms have higher capacities to predict dust storms in less time and with fewer processing operations compared to numerical weather models. This paper conducted a meta-analysis review to examine studies that addressed the areas associated with the application of machine learning to dust storm prediction. It aims to compare the applied models and the types of data used in the literature under study. Given that the location of a dust storm event is essential, the properties of dust storms are discussed in relation to the region. The output classes and the various performance metrics observed in each reviewed paper are also summarized. Subsequently, the present paper offers a detailed analysis highlighting the capabilities of machine learning models in predicting dust storms. The analysis shows two main categories: early detection and dust storm prediction. Most models used for dust storm early detection from satellite images are support vector machines (SVM). In contrast, the most used models for dust storm prediction are SVM and random forests that predict the occurrence of dust storms from meteorological data. Finally, the paper highlights the challenges and future trends in the field, illustrating the potential directions for applying deep learning algorithms and providing long-range predictions with assessments of dust storm duration and intensity.

How Can Microwave Observations at 23.8 GHz Help in Acquiring Water Vapor in the Atmosphere over Land?

Due to concerns about radio frequency interference from emerging telecommunications technology, there have been intensive discussions on the changes to the international Radio Regulations around 24 GHz at the recent World Radiocommunication Conference in 2019. Although the sensitivity to total precipitable water (TPW) at 23.8 GHz over land is small, and in some cases close to zero, state-of-the-art retrieval systems with no dependence on real-time ancillary data (e.g., numerical weather prediction (NWP) model forecasts), such as the National Oceanic and Atmospheric Administration (NOAA) operational Microwave Integrated Retrieval System (MiRS), have been producing reliable TPW products over both ocean and land, which implies that the microwave channel at 23.8 GHz is providing valuable information on TPW over land as well as over ocean. The contradiction between the zero or near-zero sensitivity and practical performance over land raises questions for the remote sensing community and public users of such data. In this study, we examine the underlying physics and include mathematical explanations, which address and clarify the apparent contradiction. The channel at 23.8 GHz is a direct measurement and indispensable for its combined use with microwave temperature and moisture sounding channels.

Objective identification of thunderstorm gust fronts in numerical weather prediction models for fire weather forecasting

Abrupt changes in wind direction and speed can dramatically impact wildfire development and spread, endangering firefighters. A frequent cause of such wind shifts is outflow from thunderstorms and organised convective systems; thus, their identification and prediction present critical challenges for fire weather forecasters. Here, we develop a methodology and implement it in a software tool that can identify and depict convective outflow boundaries in high-resolution numerical weather prediction (NWP) models to provide guidance for fire weather forecasting. The tool can process model output, objectively identify gust fronts, and graphically display detected gust fronts and similar boundaries in NWP model forecasts. The tool is demonstrated with output from the Weather Research and Forecasting (WRF) model from the operational High-Resolution Rapid Refresh (HRRR) forecasting system and from a WRF ensemble run at the National Center for Atmospheric Research that can provide probabilistic information about model-predicted gust fronts. The tool can identify outflow boundaries in model forecasts of convective events occurring in both simple and complex terrain, both with and without concurrent wildfire activity. With accurate underlying model forecast output, the tool can reliably reveal areas of potential gust front activity and thus provide valuable guidance to incident meteorologists and command personnel.

An Operational Multi-Radar Multi-Sensor QPE System in Taiwan

AbstractOver the last two decades, the Central Weather Bureau of Taiwan and the U.S. National Severe Storms Laboratory have been involved in a research and development collaboration to improve the monitoring and prediction of river flooding, flash floods, debris flows, and severe storms for Taiwan. The collaboration resulted in the Quantitative Precipitation Estimation and Segregation Using Multiple Sensors (QPESUMS) system. The QPESUMS system integrates observations from multiple mixed-band weather radars, rain gauges, and numerical weather prediction model fields to produce high-resolution (1 km) and rapid-update (10 min) rainfall and severe storm monitoring and prediction products. The rainfall products are widely used by government agencies and emergency managers in Taiwan for flood and mudslide warnings as well as for water resource management. The 3D reflectivity mosaic and QPE products are also used in high-resolution radar data assimilation and for the verification of numerical weather prediction model forecasts. The system facilitated collaborations with academic communities for research and development of radar applications, including quantitative precipitation estimation and nowcasting. This paper provides an overview of the operational QPE capabilities in the Taiwan QPESUMS system.

Improved Severe Weather Forecasts Using LEO and GEO Satellite Soundings

AbstractIt is shown here that improvements in numerical weather prediction (NWP) model forecasts of hazardous weather can be obtained by assimilating profile retrievals obtained in real time from combined direct broadcast system (DBS) polar satellite hyperspectral and geostationary satellite multispectral radiance data. Results of NWP model forecasts are shown for two recent tornado outbreak cases: 1) the 3 March 2019 tornado outbreak over the southeast United States and 2) the tornado outbreak that occurred across Illinois, Indiana, and Ohio during the night of 27 May and the morning of 28 May 2019, and 3) the 4 March 2019 severe precipitation event that occurred in southeast China. Improvements in both quantitative precipitation forecasts (QPFs) and predictions of the location of tornado occurrence are obtained. It is also shown that geostationary satellite hyperspectral soundings [i.e.,Fengyun-4A(FY-4A) Geosynchronous Interferometric Infrared Sounder (GIIRS)] further improve hazardous precipitation forecasts when used, in addition to the combined polar hyperspectral and geostationary multispectral satellite profile data, to initialize the numerical forecast model. The lowest false alarm rate (FAR) and the highest probability of detection (POD) and critical success index (CSI) scores are achieved when assimilating atmospheric profile retrievals obtained by combining all the available satellite high-vertical-resolution hyperspectral radiance measurements with geostationary satellite high-spatial-resolution and high-temporal-resolution multispectral radiance measurements.

Comparative Assessment of Various Machine Learning‐Based Bias Correction Methods for Numerical Weather Prediction Model Forecasts of Extreme Air Temperatures in Urban Areas

AbstractForecasts of maximum and minimum air temperatures are essential to mitigate the damage of extreme weather events such as heat waves and tropical nights. The Numerical Weather Prediction (NWP) model has been widely used for forecasting air temperature, but generally it has a systematic bias due to its coarse grid resolution and lack of parametrizations. This study used random forest (RF), support vector regression (SVR), artificial neural network (ANN) and a multi‐model ensemble (MME) to correct the Local Data Assimilation and Prediction System (LDAPS; a local NWP model over Korea) model outputs of next‐day maximum and minimum air temperatures ( and ) in Seoul, South Korea. A total of 14 LDAPS model forecast data, the daily maximum and minimum air temperatures of in‐situ observations, and five auxiliary data were used as input variables. The results showed that the LDAPS model had an R2 of 0.69, a bias of −0.85 °C and an RMSE of 2.08 °C for forecast, whereas the proposed models resulted in the improvement with R2 from 0.75 to 0.78, bias from −0.16 to −0.07 °C and RMSE from 1.55 to 1.66 °C by hindcast validation. For forecasting , the LDAPS model had an R2 of 0.77, a bias of 0.51 °C and an RMSE of 1.43 °C by hindcast, while the bias correction models showed R2 values ranging from 0.86 to 0.87, biases from −0.03 to 0.03 °C, and RMSEs from 0.98 to 1.02 °C. The MME model had better generalization performance than the three single machine learning models by hindcast validation and leave‐one‐station‐out cross‐validation.

Prediction of the August 2018 heavy rainfall events over Kerala with high‐resolution NWP models

AbstractThe southern Indian state of Kerala experienced exceptionally high rainfall during August 2018, which led to devastating floods in many parts of the state. Prediction and early warning of severe weather events in vulnerable areas is crucial for disaster management agencies in order to protect life and property. In recent years, state‐of‐the‐art numerical weather prediction (NWP) models have been used operationally to predict rainfall over different spatial and temporal scales. In the present paper, predictions based on the National Centre for Medium Range Weather Forecasting (NCMRWF) models (NCUM, NCUM‐R and NEPS) are assessed over Kerala to demonstrate the capabilities of high‐resolution models. It is found that the deterministic NWP model (NCUM and NCUM‐R) forecasts are accurate at shorter lead times (up to Day 3) mainly in terms of timing and, to some extent, intensity. At higher lead times (beyond Day 3), the ensemble‐based probabilistic forecasts are useful and actionable.

Comparison of short‐period daytime convective rainfall accumulations with total column precipitable water: Derivation of an operational forecasting technique

AbstractA major forecasting challenge is predicting the likely intensity of rainfall in convective situations in which intense short‐period rainfall can lead to surface water flooding. The study demonstrates a useful correlation between precipitable water and 1 hr gauge‐derived rainfall accumulations in situations with deep instability and relatively slack flow in depth. The correlation forms the basis of a forecasting technique, independent of numerical weather prediction (NWP) model forecasts of rainfall, being applied at the UK's Met Office Flood Forecasting Centre (FFC).

The Second Wind Forecast Improvement Project (WFIP2): Observational Field Campaign

AbstractThe Second Wind Forecast Improvement Project (WFIP2) is a U.S. Department of Energy (DOE)- and National Oceanic and Atmospheric Administration (NOAA)-funded program, with private-sector and university partners, which aims to improve the accuracy of numerical weather prediction (NWP) model forecasts of wind speed in complex terrain for wind energy applications. A core component of WFIP2 was an 18-month field campaign that took place in the U.S. Pacific Northwest between October 2015 and March 2017. A large suite of instrumentation was deployed in a series of telescoping arrays, ranging from 500 km across to a densely instrumented 2 km × 2 km area similar in size to a high-resolution NWP model grid cell. Observations from these instruments are being used to improve our understanding of the meteorological phenomena that affect wind energy production in complex terrain and to evaluate and improve model physical parameterization schemes. We present several brief case studies using these observations to describe phenomena that are routinely difficult to forecast, including wintertime cold pools, diurnally driven gap flows, and mountain waves/wakes. Observing system and data product improvements developed during WFIP2 are also described.

Aircraft Icing Study Using Integrated Observations and Model Data

Abstract Light (LGT) to moderate (MOD) aircraft icing (AI) is frequently reported at Cold Lake, Alberta, but forecasting AI has been a big challenge. The purpose of this study is to investigate and understand the weather conditions associated with AI based on observations in order to improve the icing forecast. To achieve this goal, Environment and Climate Change Canada in cooperation with the Department of National Defence deployed a number of ground-based instruments that include a microwave radiometer, a ceilometer, disdrometers, and conventional present weather sensors at the Cold Lake airport (CYOD). A number of pilot reports (PIREPs) of icing at Cold Lake during the 2016/17 winter period and associated observation data are examined. Most of the AI events were LGT (76%) followed by MOD (20%) and occurred during landing and takeoff at relatively warm temperatures. Two AI intensity algorithms have been tested based on an ice accumulation rate (IAR) assuming a cylindrical shape moving with airspeed υa of 60 and 89.4 m s−1, and the Canadian numerical weather prediction model forecasts. It was found that the algorithms IAR2 with υa = 89.4 m s−1 and IAR1 with υa = 60 m s−1 underestimated (overestimated) the LGT (MOD) icing events, respectively. The algorithm IAR2 with υa = 60 m s−1 appeared to be more suitable for forecasting LGT icing. Over all, the hit rate score was 0.33 for the 1200 UTC model run and 0.6 for 0000 UTC run for both algorithms, but based on the individual icing intensity scores, the IAR2 did better than IAR1 for forecasting LGT icing events.

Impacts of a University-led, On-demand Sounding Program on Human and Numerical Weather Prediction Model Forecasts in an Upper-air Observation Hole

Impacts of a University-led, On-demand Sounding Program on Human and Numerical Weather Prediction Model Forecasts in an Upper-air Observation Hole

Himawari-8 Atmospheric Motion Vectors - Operational Generation and Assimilation

In October 2014 the Japanese Meteorological Agency (JMA) launched the new generation geostationary satellite Himawari-8. This satellite provides ten minute imagery in sixteen wavebands over the Asian and Australasian region. The imagery has been navigated, calibrated and subsequently used in the Bureau of Meteorology (BoM) to generate Atmospheric Motion Vectors (AMVs) over the full earth disk viewed from the satellite every ten minutes. Each vector has been error characterised and assigned an expected error. In preparation for the operational assimilation of the ten minute data, these high temporal and spatial resolution data were used with the BoM operational database to provide forecasts from the next generation operational forecast model ACCESS APS2 using 4D Var. Results from these tests indicate these locally generated Himawari-8 ten minute AMVs are of high density and quality and have the potential to improve numerical weather prediction (NWP) model initialisation and forecasts. The forecasts undertaken include cases associated with extreme weather. The results also provided the appropriate times, data selection and application methods for the effective use of these high temporal resolution data. As a result of these studies these wind data were approved for inclusion in the BoMs operational database and are used in operational forecasting.

Himawari-8 Atmospheric Motion Vectors - Operational Generation and Assimilation

In October 2014 the Japanese Meteorological Agency (JMA) launched the new generation geostationary satellite Himawari-8. This satellite provides ten minute imagery in sixteen wavebands over the Asian and Australasian region. The imagery has been navigated, calibrated and subsequently used in the Bureau of Meteorology (BoM) to generate Atmospheric Motion Vectors (AMVs) over the full earth disk viewed from the satellite every ten minutes. Each vector has been error characterised and assigned an expected error. In preparation for the operational assimilation of the ten minute data, these high temporal and spatial resolution data were used with the BoM operational database to provide forecasts from the next generation operational forecast model ACCESS APS2 using 4D Var. Results from these tests indicate these locally generated Himawari-8 ten minute AMVs are of high density and quality and have the potential to improve numerical weather prediction (NWP) model initialisation and forecasts. The forecasts undertaken include cases associated with extreme weather. The results also provided the appropriate times, data selection and application methods for the effective use of these high temporal resolution data. As a result of these studies these wind data were approved for inclusion in the BoMs operational database and are used in operational forecasting.

Critical weather situations for renewable energies – Part B: Low stratus risk for solar power

Accurately predicting the formation, development and dissipation of fog and low stratus (LS) still poses a challenge for numerical weather prediction (NWP) models. Errors in the low cloud cover NWP forecasts directly impact the quality of photovoltaic (PV) power prediction. On days with LS, day-ahead forecast errors of Germany-wide PV power frequently lie within the magnitude of the balance energy and thus pose a challenge for maintaining grid stability. An indication in advance about the possible occurrence of a critical weather situation such as LS would represent a helpful tool for transmission system operators (TSOs) in their day-to-day business. In the following, a detection algorithm for low stratus risk (LSR) is developed and applied as post-processing to the NWP model forecasts of the regional non-hydrostatic model COSMO-DE, operational at the German Weather Service. The aim of the LSR product is to supply day-ahead warnings and to support the decision making process of the TSOs. The quality of the LSR is assessed by comparing the computed regions of LSR occurrence with a satellite based cloud classification product from the Nowcasting Satellite Facility (NWCSAF). The results show that the LSR provides additional information that should in particular be useful for risk adverse users.

Nowcasting of deep convective clouds and heavy precipitation: Comparison study between NWP model simulation and extrapolation

An evaluation of convective cloud forecasts performed with the numerical weather prediction (NWP) model COSMO and extrapolation of cloud fields is presented using observed data derived from the geostationary satellite Meteosat Second Generation (MSG). The present study focuses on the nowcasting range (1–5h) for five severe convective storms in their developing stage that occurred during the warm season in the years 2012–2013. Radar reflectivity and extrapolated radar reflectivity data were assimilated for at least 6h depending on the time of occurrence of convection. Synthetic satellite imageries were calculated using radiative transfer model RTTOV v10.2, which was implemented into the COSMO model. NWP model simulations of IR10.8μm and WV06.2μm brightness temperatures (BTs) with a horizontal resolution of 2.8km were interpolated into the satellite projection and objectively verified against observations using Root Mean Square Error (RMSE), correlation coefficient (CORR) and Fractions Skill Score (FSS) values. Naturally, the extrapolation of cloud fields yielded an approximately 25% lower RMSE, 20% higher CORR and 15% higher FSS at the beginning of the second forecasted hour compared to the NWP model forecasts. On the other hand, comparable scores were observed for the third hour, whereas the NWP forecasts outperformed the extrapolation by 10% for RMSE, 15% for CORR and up to 15% for FSS during the fourth forecasted hour and 15% for RMSE, 27% for CORR and up to 15% for FSS during the fifth forecasted hour. The analysis was completed by a verification of the precipitation forecasts yielding approximately 8% higher RMSE, 15% higher CORR and up to 45% higher FSS when the NWP model simulation is used compared to the extrapolation for the first hour. Both the methods yielded unsatisfactory level of precipitation forecast accuracy from the fourth forecasted hour onward.