Abstract
Accurate prediction of surface wind speed and temperature is crucial for many sectors. Physical schemes in numerical weather prediction (NWP) and data-driven correction approaches have limitations due to uncertainties in parameterization and lack of robustness, respectively. This study introduces a physics-informed data model, PhyCorNet, which combines a deep learning–based physics emulator (PhyNet) and a subsequent correction network (CorNet). PhyNet imitates the revised surface-layer parameterization scheme [default option of the Weather Research and Forecasting (WRF) Model]. CorNet refines predictions by mitigating the difference between PhyNet and observations. PhyCorNet enables gridded forecasts despite the training data being point based. Therefore, PhyCorNet can be regarded as a rediagnostic scheme for surface wind speed at 10 m (WS10) and temperature at 2 m (T 2). Compared to WRF, it reduced diagnostic root-mean-square errors in the 24-h forecasts for WS10 and T 2 by 40% and 36%, respectively, across China, with unbiased forecasts at almost all sites. PhyCorNet addresses the oversmoothing prediction issue of other deep learning models by providing the ability to represent fine-scale features and perform well in statistically extreme samples. In grid cells without observations for training, PhyCorNet performed much better than WRF, demonstrating the zero-shot learning capability. This study implies that the emulator plus bias correction provided by PhyCorNet could be used as a simple but effective approach to improve the performance of other diagnostic quantities in NWP.
Significance Statement
We developed a new model called PhyCorNet to improve the surface wind speed and temperature forecasts. Existing weather forecast models have limitations in accurately predicting these variables. PhyCorNet first uses a neural network (PhyNet) to mimic an existing physics-based wind and temperature calculation method. Another neural network [correction network (CorNet)] improves the PhyNet forecasts by comparing them with observations. Across China, PhyCorNet reduced 24-h forecast errors for wind speed by 40% and temperature by 36% compared to a conventional model. It performed well under diverse conditions and overcame the oversmoothing issues faced by other machine learning models. Importantly, PhyCorNet outperformed traditional models in areas without historical observations. We suggest that this new framework can also be used to enhance the forecasts of other atmospheric variables.
Abstract
Accurate prediction of surface wind speed and temperature is crucial for many sectors. Physical schemes in numerical weather prediction (NWP) and data-driven correction approaches have limitations due to uncertainties in parameterization and lack of robustness, respectively. This study introduces a physics-informed data model, PhyCorNet, which combines a deep learning–based physics emulator (PhyNet) and a subsequent correction network (CorNet). PhyNet imitates the revised surface-layer parameterization scheme [default option of the Weather Research and Forecasting (WRF) Model]. CorNet refines predictions by mitigating the difference between PhyNet and observations. PhyCorNet enables gridded forecasts despite the training data being point based. Therefore, PhyCorNet can be regarded as a rediagnostic scheme for surface wind speed at 10 m (WS10) and temperature at 2 m (T 2). Compared to WRF, it reduced diagnostic root-mean-square errors in the 24-h forecasts for WS10 and T 2 by 40% and 36%, respectively, across China, with unbiased forecasts at almost all sites. PhyCorNet addresses the oversmoothing prediction issue of other deep learning models by providing the ability to represent fine-scale features and perform well in statistically extreme samples. In grid cells without observations for training, PhyCorNet performed much better than WRF, demonstrating the zero-shot learning capability. This study implies that the emulator plus bias correction provided by PhyCorNet could be used as a simple but effective approach to improve the performance of other diagnostic quantities in NWP.
Significance Statement
We developed a new model called PhyCorNet to improve the surface wind speed and temperature forecasts. Existing weather forecast models have limitations in accurately predicting these variables. PhyCorNet first uses a neural network (PhyNet) to mimic an existing physics-based wind and temperature calculation method. Another neural network [correction network (CorNet)] improves the PhyNet forecasts by comparing them with observations. Across China, PhyCorNet reduced 24-h forecast errors for wind speed by 40% and temperature by 36% compared to a conventional model. It performed well under diverse conditions and overcame the oversmoothing issues faced by other machine learning models. Importantly, PhyCorNet outperformed traditional models in areas without historical observations. We suggest that this new framework can also be used to enhance the forecasts of other atmospheric variables.
Abstract
Hail smaller than 0.75 in. is known to cause economic impacts yet remains understudied due to report biases towards recording larger hail sizes (≤1 in.). In this study, we assembled ground hail reports during 2017–22 from the National Centers for Environmental Information (NCEI) Storm Data, Community Collaborative Rain, Hail and Snow Network (CoCoRaHS), and Meteorological Phenomena Identification Near the Ground (mPING) databases. Then, these reports are collocated with the attributes of radar-derived convective features from the Multi-Radar Multi-Sensor (MRMS) system and fifth major global reanalysis produced by European Centre for Medium-Range Weather Forecasts (ERA5) atmospheric vertical profiles to construct a dataset describing properties of a full spectrum of hailstorms. The characteristics of radar reflectivity and atmospheric profiles are examined for hail of different sizes reported within selected regions over the contiguous United States (CONUS). In addition to the seasonal and diurnal variations, the morphology of convective features shows apparent regional differences from west to east in CONUS. The maximum expected size of hail (MESH) performance against reported hail sizes shows underestimation of hail with significant sizes and overestimation of small hail sizes. ERA5 vertical atmospheric profiles are explored to form relationships between storm environment and hail sizes. In addition to the relationships between wind shear and hail sizes, the roles of low-level relative humidity and freezing level height in regard to hail melting are discussed.
Abstract
Hail smaller than 0.75 in. is known to cause economic impacts yet remains understudied due to report biases towards recording larger hail sizes (≤1 in.). In this study, we assembled ground hail reports during 2017–22 from the National Centers for Environmental Information (NCEI) Storm Data, Community Collaborative Rain, Hail and Snow Network (CoCoRaHS), and Meteorological Phenomena Identification Near the Ground (mPING) databases. Then, these reports are collocated with the attributes of radar-derived convective features from the Multi-Radar Multi-Sensor (MRMS) system and fifth major global reanalysis produced by European Centre for Medium-Range Weather Forecasts (ERA5) atmospheric vertical profiles to construct a dataset describing properties of a full spectrum of hailstorms. The characteristics of radar reflectivity and atmospheric profiles are examined for hail of different sizes reported within selected regions over the contiguous United States (CONUS). In addition to the seasonal and diurnal variations, the morphology of convective features shows apparent regional differences from west to east in CONUS. The maximum expected size of hail (MESH) performance against reported hail sizes shows underestimation of hail with significant sizes and overestimation of small hail sizes. ERA5 vertical atmospheric profiles are explored to form relationships between storm environment and hail sizes. In addition to the relationships between wind shear and hail sizes, the roles of low-level relative humidity and freezing level height in regard to hail melting are discussed.
Abstract
The objective of this study is to understand rainfall processes over tropical islands by identifying synoptic conditions that influence the diurnal cycle of rainfall over western Puerto Rico. Summer rainfall over Puerto Rico is dominated by its afternoon peak, yet there is large variability in its behavior that remains challenging to predict. We use radiosonde and airborne data collected through the NASA Convective Processes Experiment—Aerosols and Winds (CPEX-AW) field campaign (August–September 2021) to achieve our objective, in addition to the network of surface station data over the island. We find that the background wind speed and humidity have strong influences on afternoon rainfall through different mechanisms. A stronger background wind inhibits afternoon rainfall likely by reducing land–sea thermal contrast and weakening sea-breeze convergence over the island. At the same time, an inversion layer often forms with a stronger background wind that further inhibits deep convection. When the background wind is weak and sea breezes are prominent, afternoon rainfall increases exclusively over the island, while limited rainfall appears over the surrounding ocean. However, enhanced rainfall still occurs over the island with weak sea breezes if humidity is high, accompanied by enhanced rainfall over the surrounding ocean due to the offshore movement and development of convective storms. The sources of variability in background wind and humidity are mostly independent, resulting in a wide range of synoptic conditions and associated effects on the island rainfall. This expanded understanding of the mechanisms causing variability of diurnal rainfall can lead to improved forecasts over Puerto Rico and other tropical islands.
Abstract
The objective of this study is to understand rainfall processes over tropical islands by identifying synoptic conditions that influence the diurnal cycle of rainfall over western Puerto Rico. Summer rainfall over Puerto Rico is dominated by its afternoon peak, yet there is large variability in its behavior that remains challenging to predict. We use radiosonde and airborne data collected through the NASA Convective Processes Experiment—Aerosols and Winds (CPEX-AW) field campaign (August–September 2021) to achieve our objective, in addition to the network of surface station data over the island. We find that the background wind speed and humidity have strong influences on afternoon rainfall through different mechanisms. A stronger background wind inhibits afternoon rainfall likely by reducing land–sea thermal contrast and weakening sea-breeze convergence over the island. At the same time, an inversion layer often forms with a stronger background wind that further inhibits deep convection. When the background wind is weak and sea breezes are prominent, afternoon rainfall increases exclusively over the island, while limited rainfall appears over the surrounding ocean. However, enhanced rainfall still occurs over the island with weak sea breezes if humidity is high, accompanied by enhanced rainfall over the surrounding ocean due to the offshore movement and development of convective storms. The sources of variability in background wind and humidity are mostly independent, resulting in a wide range of synoptic conditions and associated effects on the island rainfall. This expanded understanding of the mechanisms causing variability of diurnal rainfall can lead to improved forecasts over Puerto Rico and other tropical islands.
Abstract
On the morning of 21 August 2021, extreme rainfall spurred a flood wave on Trace Creek that ravaged Waverly, Tennessee, causing 19 fatalities. Peak 24-h rainfall of 526 mm was recorded just upstream at McEwen, setting the Tennessee 24-h state rainfall record. A Slight Risk of excessive rainfall and a Flash Flood Watch were issued 16 and 8 h, respectively, before rain began; however, predicting mesobeta scale extreme rainfall remains an elusive skill for models and humans alike. Operational convection-allowing models not only suggested pockets of heavy rain but also displayed 1) peak values generally less than half of those observed, 2) widely ranging solutions, and 3) erroneous similarly heavy rain elsewhere. Future use of storm-scale ensembles which use rapid data assimilation promises to help forecasters anticipate extrema that may only be predictable at shorter time scales. This work will examine compelling forecasts from a retrospective run of the experimental Warn-on-Forecast System (WoFS). The authors, who include National Weather Service forecasters who worked the event, discuss how WoFS and its probabilistic framework could influence services during low-predictability, high-impact flash floods.
Abstract
On the morning of 21 August 2021, extreme rainfall spurred a flood wave on Trace Creek that ravaged Waverly, Tennessee, causing 19 fatalities. Peak 24-h rainfall of 526 mm was recorded just upstream at McEwen, setting the Tennessee 24-h state rainfall record. A Slight Risk of excessive rainfall and a Flash Flood Watch were issued 16 and 8 h, respectively, before rain began; however, predicting mesobeta scale extreme rainfall remains an elusive skill for models and humans alike. Operational convection-allowing models not only suggested pockets of heavy rain but also displayed 1) peak values generally less than half of those observed, 2) widely ranging solutions, and 3) erroneous similarly heavy rain elsewhere. Future use of storm-scale ensembles which use rapid data assimilation promises to help forecasters anticipate extrema that may only be predictable at shorter time scales. This work will examine compelling forecasts from a retrospective run of the experimental Warn-on-Forecast System (WoFS). The authors, who include National Weather Service forecasters who worked the event, discuss how WoFS and its probabilistic framework could influence services during low-predictability, high-impact flash floods.
Abstract
Aerosols affect Earth’s climate both directly and indirectly, which is the largest uncertainty in the assessment of radiative forcings affecting anthropogenic climate change. The standard Aerosol Robotic Network (AERONET) aerosol products have been widely used for more than 30 years. Currently, there is strong community interest in the possibility of determining aerosol composition directly from remote sensing observations. This work presents the results of applying such a recently developed approach by Li et al. to extended datasets of the directional sky radiances and spectral aerosol optical depth (AOD) measured by AERONET for the retrievals of aerosol components. First, the validation of aerosol optical properties retrieved by this component approach with AERONET standard products shows good agreement. Then, spatiotemporal variations of the obtained aerosol component concentration are characterized globally, especially the absorbing aerosol species (black carbon, brown carbon, and iron oxides) and scattering aerosol species (organic carbon, quartz, and inorganic salts). Finally, we compared the black carbon (BC) and dust column concentration retrievals to the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2), products in several regions of interest (Amazon zone, Indo-China Peninsula, North India, southern Africa, sub-Sahel, Gobi Desert, Middle East, Sahara Desert, and Taklamakan Desert) for new insights on the quantitative assessment of MERRA-2 aerosol composition products (R = 0.60–0.85 for BC; R = 0.75–0.90 for dust). The new value-added and long-term aerosol composition product globally is available online (https://doi.org/10.6084/m9.figshare.25415239.v1), which provides important measurements for the improvement and optimization of aerosol modeling to enhance estimation of the aerosol radiative forcing.
Abstract
Aerosols affect Earth’s climate both directly and indirectly, which is the largest uncertainty in the assessment of radiative forcings affecting anthropogenic climate change. The standard Aerosol Robotic Network (AERONET) aerosol products have been widely used for more than 30 years. Currently, there is strong community interest in the possibility of determining aerosol composition directly from remote sensing observations. This work presents the results of applying such a recently developed approach by Li et al. to extended datasets of the directional sky radiances and spectral aerosol optical depth (AOD) measured by AERONET for the retrievals of aerosol components. First, the validation of aerosol optical properties retrieved by this component approach with AERONET standard products shows good agreement. Then, spatiotemporal variations of the obtained aerosol component concentration are characterized globally, especially the absorbing aerosol species (black carbon, brown carbon, and iron oxides) and scattering aerosol species (organic carbon, quartz, and inorganic salts). Finally, we compared the black carbon (BC) and dust column concentration retrievals to the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2), products in several regions of interest (Amazon zone, Indo-China Peninsula, North India, southern Africa, sub-Sahel, Gobi Desert, Middle East, Sahara Desert, and Taklamakan Desert) for new insights on the quantitative assessment of MERRA-2 aerosol composition products (R = 0.60–0.85 for BC; R = 0.75–0.90 for dust). The new value-added and long-term aerosol composition product globally is available online (https://doi.org/10.6084/m9.figshare.25415239.v1), which provides important measurements for the improvement and optimization of aerosol modeling to enhance estimation of the aerosol radiative forcing.
Abstract
Tropical instability waves (TIWs) are oceanic features that form around the equatorial Pacific cold tongue and influence the large-scale circulation and coupled climate variability including El Niño–Southern Oscillation. Local air–sea coupling over TIWs is thought to play an important role in the atmosphere and ocean’s energy and tracer budgets but is not well captured in coarse-resolution models. In this study, we isolate the impacts of TIW thermal (sea surface temperature–driven) and current (surface current–driven) feedbacks by removing TIW signatures in air–sea coupling fields in a high-resolution regional coupled model. The thermal feedback is found to damp TIW temperature variance by a factor of 2, associated both with the direct dependence of surface heat fluxes on SST (∼74%) and indirect impacts on surface winds (∼35%) and air temperature and humidity (∼−9%). These changes lead to cooling of the cold tongue SST by up to 0.1°C through reduced TIW-driven meridional heat fluxes and associated small changes in atmospheric circulation. The current feedback is decomposed into TIW (i.e., mesoscale) and mean (i.e., large-scale) components using separate experiments, with both having distinct impacts on TIWs and the mean state. The mesoscale current feedback reduces TIW eddy kinetic energy (EKE) by 22% through the eddy wind work, while the mean current feedback induces a further reduction of 8% by extracting energy from the mean currents and thus reducing barotropic EKE shear production. An improved understanding of small-scale tropical Pacific processes is needed to address biases in coarse-resolution models that impact their predictions and projections of Pacific climate variability and change.
Significance Statement
Tropical instability waves (TIWs) are oceanic features with ∼1000-km wavelengths that propagate westward on either side of the eastern equatorial Pacific cold tongue. TIWs drive lateral and vertical heat fluxes that impact several aspects of El Niño–Southern Oscillation. While climate models with a moderate, 1/4° ocean resolution can capture some TIW variability, they fail to properly represent many associated processes such as the impact of TIWs on the overlying atmosphere. Using sensitivity studies performed using a high-resolution regional coupled model, we study the impact of TIW air–sea coupling on the eastern Pacific climate system. Increased understanding of small-scale processes from studies such as this is essential to understand and address biases in models used for seasonal climate predictions and projections in the Pacific region.
Abstract
Tropical instability waves (TIWs) are oceanic features that form around the equatorial Pacific cold tongue and influence the large-scale circulation and coupled climate variability including El Niño–Southern Oscillation. Local air–sea coupling over TIWs is thought to play an important role in the atmosphere and ocean’s energy and tracer budgets but is not well captured in coarse-resolution models. In this study, we isolate the impacts of TIW thermal (sea surface temperature–driven) and current (surface current–driven) feedbacks by removing TIW signatures in air–sea coupling fields in a high-resolution regional coupled model. The thermal feedback is found to damp TIW temperature variance by a factor of 2, associated both with the direct dependence of surface heat fluxes on SST (∼74%) and indirect impacts on surface winds (∼35%) and air temperature and humidity (∼−9%). These changes lead to cooling of the cold tongue SST by up to 0.1°C through reduced TIW-driven meridional heat fluxes and associated small changes in atmospheric circulation. The current feedback is decomposed into TIW (i.e., mesoscale) and mean (i.e., large-scale) components using separate experiments, with both having distinct impacts on TIWs and the mean state. The mesoscale current feedback reduces TIW eddy kinetic energy (EKE) by 22% through the eddy wind work, while the mean current feedback induces a further reduction of 8% by extracting energy from the mean currents and thus reducing barotropic EKE shear production. An improved understanding of small-scale tropical Pacific processes is needed to address biases in coarse-resolution models that impact their predictions and projections of Pacific climate variability and change.
Significance Statement
Tropical instability waves (TIWs) are oceanic features with ∼1000-km wavelengths that propagate westward on either side of the eastern equatorial Pacific cold tongue. TIWs drive lateral and vertical heat fluxes that impact several aspects of El Niño–Southern Oscillation. While climate models with a moderate, 1/4° ocean resolution can capture some TIW variability, they fail to properly represent many associated processes such as the impact of TIWs on the overlying atmosphere. Using sensitivity studies performed using a high-resolution regional coupled model, we study the impact of TIW air–sea coupling on the eastern Pacific climate system. Increased understanding of small-scale processes from studies such as this is essential to understand and address biases in models used for seasonal climate predictions and projections in the Pacific region.
Abstract
The turbulent transport of mass, energy, moisture, and momentum between the clouds and the surrounding environment plays a central role in determining the vertical structure of the troposphere. This article investigates the connection between net entrainment, defined here as the net transport of air into individual clouds, and net dilution, defined as the tendencies of passive tracers such as static energy or total water mixing ratio. Entrainment and detrainment rates for 2.6 × 106 individual cloud samples are obtained from a large-eddy simulation of shallow convective boundary layer atmosphere that explicitly calculates the turbulent fluxes across the cloud boundaries. The equations describing the tendencies of cloud mass and tracer concentrations are derived as a function of directly calculated entrainment and detrainment rates of the individual clouds, and used to calculate net entrainment and dilution rates. Directly calculated net entrainment and dilution rates agree well with cloud mass and tracer tendencies and give a dilution time scale of 13 min. In contrast, the traditional bulk-plume approximation overestimates the effect of entrainment and detrainment on the dilution of cloud field properties due to the differential tracer transport through the moist shell surrounding the cloud. Using direct measures of entrainment and detrainment for individual clouds separates different processes that influence the turbulent mass transport between the clouds and their environment.
Abstract
The turbulent transport of mass, energy, moisture, and momentum between the clouds and the surrounding environment plays a central role in determining the vertical structure of the troposphere. This article investigates the connection between net entrainment, defined here as the net transport of air into individual clouds, and net dilution, defined as the tendencies of passive tracers such as static energy or total water mixing ratio. Entrainment and detrainment rates for 2.6 × 106 individual cloud samples are obtained from a large-eddy simulation of shallow convective boundary layer atmosphere that explicitly calculates the turbulent fluxes across the cloud boundaries. The equations describing the tendencies of cloud mass and tracer concentrations are derived as a function of directly calculated entrainment and detrainment rates of the individual clouds, and used to calculate net entrainment and dilution rates. Directly calculated net entrainment and dilution rates agree well with cloud mass and tracer tendencies and give a dilution time scale of 13 min. In contrast, the traditional bulk-plume approximation overestimates the effect of entrainment and detrainment on the dilution of cloud field properties due to the differential tracer transport through the moist shell surrounding the cloud. Using direct measures of entrainment and detrainment for individual clouds separates different processes that influence the turbulent mass transport between the clouds and their environment.
Abstract
The High-Resolution Rapid Refresh (HRRR) model provides hourly updating forecasts of convective-scale phenomena, which can be used to infer the potential for convective hazards (e.g., tornadoes, hail, and wind gusts), across the United States. We used deterministic 2019–20 HRRR, version 4 (HRRRv4), forecasts to train neural networks (NNs) to generate 4-hourly probabilistic convective hazard forecasts [neural network probability forecasts (NNPFs)] for HRRRv4 initializations in 2021, using storm reports as ground truth. The NNPFs were compared to the skill of a smoothed updraft helicity (UH) baseline to quantify the benefit of the NNs. NNPF skill varied by initialization time and time of day but was all superior to the UH forecast. NNPFs valid at hours between 1800 and 0000 UTC were most skillful in aggregate, significantly exceeding the baseline forecast skill. Overnight NNPFs (i.e., valid 0600–1200 UTC) were least skillful, indicating a diurnal cycle in hazard predictability that was present across all HRRRv4 initializations. We explored the sensitivity of HRRRv4 NNPF skill to NN training choices. Including an additional year of 2021 HRRRv4 forecasts for training slightly improved skill for 2022 HRRRv4 NNPFs, while reducing the training dataset size by 40% using only forecasts with storm reports was not detrimental to forecast skill. Finally, NNs trained with 2018–20 HRRRv3 forecasts led to a reduction in NNPF skill when applied to 2021 HRRRv4 forecasts. In addition to documenting practical predictability challenges with convective hazard prediction, these findings reinforce the need for a consistent model configuration for optimal results when training NNs and provide best practices when constructing a training dataset with operational convection-allowing model forecasts.
Significance Statement
Convective hazards, such as hail and tornadoes, are often challenging to predict. To improve hazard predictions, we used machine learning (ML) to generate forecasts of convective hazards across the United States leveraging forecasts of prior events. The ML hazard forecasts were consistently better than a non-ML approach and varied in skill based on the time of day, with nighttime forecasts being particularly challenging. ML forecasts of wind gusts and hail were more skillful than tornadoes. Different strategies of constructing the dataset of prior events led to differences in forecast performance; thus, this work provides recommendations for how to assemble these datasets and train ML models to generate improved forecasts of severe weather events.
Abstract
The High-Resolution Rapid Refresh (HRRR) model provides hourly updating forecasts of convective-scale phenomena, which can be used to infer the potential for convective hazards (e.g., tornadoes, hail, and wind gusts), across the United States. We used deterministic 2019–20 HRRR, version 4 (HRRRv4), forecasts to train neural networks (NNs) to generate 4-hourly probabilistic convective hazard forecasts [neural network probability forecasts (NNPFs)] for HRRRv4 initializations in 2021, using storm reports as ground truth. The NNPFs were compared to the skill of a smoothed updraft helicity (UH) baseline to quantify the benefit of the NNs. NNPF skill varied by initialization time and time of day but was all superior to the UH forecast. NNPFs valid at hours between 1800 and 0000 UTC were most skillful in aggregate, significantly exceeding the baseline forecast skill. Overnight NNPFs (i.e., valid 0600–1200 UTC) were least skillful, indicating a diurnal cycle in hazard predictability that was present across all HRRRv4 initializations. We explored the sensitivity of HRRRv4 NNPF skill to NN training choices. Including an additional year of 2021 HRRRv4 forecasts for training slightly improved skill for 2022 HRRRv4 NNPFs, while reducing the training dataset size by 40% using only forecasts with storm reports was not detrimental to forecast skill. Finally, NNs trained with 2018–20 HRRRv3 forecasts led to a reduction in NNPF skill when applied to 2021 HRRRv4 forecasts. In addition to documenting practical predictability challenges with convective hazard prediction, these findings reinforce the need for a consistent model configuration for optimal results when training NNs and provide best practices when constructing a training dataset with operational convection-allowing model forecasts.
Significance Statement
Convective hazards, such as hail and tornadoes, are often challenging to predict. To improve hazard predictions, we used machine learning (ML) to generate forecasts of convective hazards across the United States leveraging forecasts of prior events. The ML hazard forecasts were consistently better than a non-ML approach and varied in skill based on the time of day, with nighttime forecasts being particularly challenging. ML forecasts of wind gusts and hail were more skillful than tornadoes. Different strategies of constructing the dataset of prior events led to differences in forecast performance; thus, this work provides recommendations for how to assemble these datasets and train ML models to generate improved forecasts of severe weather events.
Abstract
This study uses high-resolution, convection-permitting, dynamically downscaled regional climate simulation output to assess how long-lived, convectively induced, extratropical windstorms known as derechos may change across the CONUS during the 21st century. Three 15-year epochs including a historical period (1990 – 2005), and two, separate late-21st century periods (2085 – 2100) employing intermediate (RCP 4.5) and pessimistic (RCP 8.5) greenhouse gas concentration scenarios are evaluated. A mesoscale convective system (MCS) identification and tracking tool catalogs derecho candidates across epochs using simulated radar reflectivity and maximum 10-meter wind speed as a proxy for near-surface severe wind gusts. Results indicate that MCS-based windstorms, including derechos, are more frequent, widespread, and intense in both future climate scenarios examined for most regions of the central and eastern CONUS. Increases are suggested across all parts of the year, with significant changes in populations concentrated during the early spring and summer months, suggesting the potential for a longer, more extreme MCS windstorm season. This research provides insights for forecasters, emergency managers, and wind-vulnerable stakeholders on how these events may change across the 21st century so that they may mitigate, adapt to, and become resilient against severe convective storm perils.
Abstract
This study uses high-resolution, convection-permitting, dynamically downscaled regional climate simulation output to assess how long-lived, convectively induced, extratropical windstorms known as derechos may change across the CONUS during the 21st century. Three 15-year epochs including a historical period (1990 – 2005), and two, separate late-21st century periods (2085 – 2100) employing intermediate (RCP 4.5) and pessimistic (RCP 8.5) greenhouse gas concentration scenarios are evaluated. A mesoscale convective system (MCS) identification and tracking tool catalogs derecho candidates across epochs using simulated radar reflectivity and maximum 10-meter wind speed as a proxy for near-surface severe wind gusts. Results indicate that MCS-based windstorms, including derechos, are more frequent, widespread, and intense in both future climate scenarios examined for most regions of the central and eastern CONUS. Increases are suggested across all parts of the year, with significant changes in populations concentrated during the early spring and summer months, suggesting the potential for a longer, more extreme MCS windstorm season. This research provides insights for forecasters, emergency managers, and wind-vulnerable stakeholders on how these events may change across the 21st century so that they may mitigate, adapt to, and become resilient against severe convective storm perils.
Abstract
Hail is a significant weather hazard in Canada, but its spatial and temporal distribution is poorly understood. We compiled a Canadian hail report database for 2005–22, containing 7000 unique entries with estimates of the timing and location of the hail reports and estimated hail diameter. We developed a methodology to construct an estimate of the hail climatology across Canada using manual hail observations at airports and a lightning proxy. First, we estimated the probability of hail occurrence at airport locations across the country at any given hour using Bayesian inference. Next, we interpolated in space the probabilities to obtain smooth prior probabilities of hail occurrence at any location in Canada. Then, we refined these probabilities using lightning flash density as a proxy for the likelihood of hail, severe hail (diameter greater than 20 mm), or significant severe hail (diameter greater than 50 mm). Finally, we aggregated the posterior probabilities of hail, severe hail, and significant severe hail over time and space and compared them with the number of reports found in the 2005–22 Canadian hail database. Our results indicate that the posterior probabilities of hail are not consistent with the observed hail reports and suggest that there are many gaps in hail reporting in Canada.
Abstract
Hail is a significant weather hazard in Canada, but its spatial and temporal distribution is poorly understood. We compiled a Canadian hail report database for 2005–22, containing 7000 unique entries with estimates of the timing and location of the hail reports and estimated hail diameter. We developed a methodology to construct an estimate of the hail climatology across Canada using manual hail observations at airports and a lightning proxy. First, we estimated the probability of hail occurrence at airport locations across the country at any given hour using Bayesian inference. Next, we interpolated in space the probabilities to obtain smooth prior probabilities of hail occurrence at any location in Canada. Then, we refined these probabilities using lightning flash density as a proxy for the likelihood of hail, severe hail (diameter greater than 20 mm), or significant severe hail (diameter greater than 50 mm). Finally, we aggregated the posterior probabilities of hail, severe hail, and significant severe hail over time and space and compared them with the number of reports found in the 2005–22 Canadian hail database. Our results indicate that the posterior probabilities of hail are not consistent with the observed hail reports and suggest that there are many gaps in hail reporting in Canada.