Browse
Abstract
Cool season (September–May) extreme temperature and precipitation events are frequently tied to surface cyclones and anticyclones, which are modulated on the synoptic scale by the state and evolution of the upper-tropospheric jet stream. This study adopts a jet-centered approach to classify the prevailing large-scale flow pattern into jet regimes based on the leading modes of variability of the North Pacific jet (NPJ) and the North Atlantic jet (NAJ), respectively. The characteristics of these joint NPJ–NAJ regimes are subsequently examined using composite analysis to identify large-scale flow environments conducive to the occurrence of anomalous temperatures and precipitation across North America. The analysis reveals that composite large-scale flow environments associated with each joint NPJ–NAJ regime can be approximated as a linear combination of the separate large-scale environments that characterize each NPJ regime and NAJ regime independently. Furthermore, knowledge of the joint NPJ–NAJ regime provides more precision regarding the relative likelihood and spatial coverage of anomalous temperatures and precipitation than would be obtained from consideration of the NPJ or NAJ regime in isolation. The frequencies of each joint NPJ–NAJ regime can also be modulated by the occurrence of sudden stratospheric warmings, with increased frequencies of an equatorward-shifted NAJ and a retracted NPJ during the 30-day period following a warming event. The results from the present study demonstrate that knowledge of the joint NPJ–NAJ regime exhibits the potential to inform forecasts of anomalous temperatures and precipitation at medium-range and subseasonal time scales.
Significance Statement
The development of cool season temperature and precipitation extremes are modulated by the state and evolution of the upper-tropospheric jet stream. Therefore, this study adopts a jet-centered approach to quantify the extent to which temperature and precipitation extremes over North America are related to the coevolution of the North Pacific and North Atlantic segments of the jet stream. The analysis demonstrates that the relative likelihood and spatial coverage of temperature and precipitation extremes varies significantly across North America based on the combined state of the North Pacific and North Atlantic jets. The jet-centered approach utilized in this study exhibits the potential to inform operational medium-range (6–10 day) and subseasonal forecasts of temperature and precipitation across North America.
Abstract
Cool season (September–May) extreme temperature and precipitation events are frequently tied to surface cyclones and anticyclones, which are modulated on the synoptic scale by the state and evolution of the upper-tropospheric jet stream. This study adopts a jet-centered approach to classify the prevailing large-scale flow pattern into jet regimes based on the leading modes of variability of the North Pacific jet (NPJ) and the North Atlantic jet (NAJ), respectively. The characteristics of these joint NPJ–NAJ regimes are subsequently examined using composite analysis to identify large-scale flow environments conducive to the occurrence of anomalous temperatures and precipitation across North America. The analysis reveals that composite large-scale flow environments associated with each joint NPJ–NAJ regime can be approximated as a linear combination of the separate large-scale environments that characterize each NPJ regime and NAJ regime independently. Furthermore, knowledge of the joint NPJ–NAJ regime provides more precision regarding the relative likelihood and spatial coverage of anomalous temperatures and precipitation than would be obtained from consideration of the NPJ or NAJ regime in isolation. The frequencies of each joint NPJ–NAJ regime can also be modulated by the occurrence of sudden stratospheric warmings, with increased frequencies of an equatorward-shifted NAJ and a retracted NPJ during the 30-day period following a warming event. The results from the present study demonstrate that knowledge of the joint NPJ–NAJ regime exhibits the potential to inform forecasts of anomalous temperatures and precipitation at medium-range and subseasonal time scales.
Significance Statement
The development of cool season temperature and precipitation extremes are modulated by the state and evolution of the upper-tropospheric jet stream. Therefore, this study adopts a jet-centered approach to quantify the extent to which temperature and precipitation extremes over North America are related to the coevolution of the North Pacific and North Atlantic segments of the jet stream. The analysis demonstrates that the relative likelihood and spatial coverage of temperature and precipitation extremes varies significantly across North America based on the combined state of the North Pacific and North Atlantic jets. The jet-centered approach utilized in this study exhibits the potential to inform operational medium-range (6–10 day) and subseasonal forecasts of temperature and precipitation across North America.
Abstract
Meso-γ-scale (2–20 km) local heavy rain (LHR) can cause fatalities through the sudden rise of rivers and flooding of roads. To help prevent this loss of life, we developed prediction methods for these types of meteorological hazards. We assimilated ground-based cloud radar (Ka-band radar) data that can capture cloud droplets before raindrops form and attempted to predict LHR with a cloud resolving numerical weather prediction (NWP) model. High-temporal (1-min interval) three-dimensional cloud radar data obtained through special observation were assimilated using a water vapor nudging method in the pre-rain stage of an LHR-causing cumulonimbus. While rainfall was not predicted by the NWP model without assimilation, LHR was predicted approximately 20 min after the conclusion of cloud radar data assimilation cycling. Results suggest that NWP with cloud radar data assimilation in the pre-rain stage has great potential for predicting LHR, and can lead to an early evacuation warning and subsequent evacuation of vulnerable populations.
Significance Statement
The development of prediction methods for local (within several kilometers) heavy rain (LHR) is important because LHR events can cause deaths through the sudden rise of rivers and flooding of roads by rapidly developing (≤30 min) rain clouds. This study aims to develop a method for predicting LHR even before it begins to rain, which has been difficult to date. Using a technique called data assimilation, which integrates observation and simulation, we developed a method for assimilating cloud radar observations that can capture cloud droplets before raindrops form. As a result, we succeeded in predicting LHR before rainfall commenced. By extending and applying this research, early evacuation of vulnerable populations during LHR is possible.
Abstract
Meso-γ-scale (2–20 km) local heavy rain (LHR) can cause fatalities through the sudden rise of rivers and flooding of roads. To help prevent this loss of life, we developed prediction methods for these types of meteorological hazards. We assimilated ground-based cloud radar (Ka-band radar) data that can capture cloud droplets before raindrops form and attempted to predict LHR with a cloud resolving numerical weather prediction (NWP) model. High-temporal (1-min interval) three-dimensional cloud radar data obtained through special observation were assimilated using a water vapor nudging method in the pre-rain stage of an LHR-causing cumulonimbus. While rainfall was not predicted by the NWP model without assimilation, LHR was predicted approximately 20 min after the conclusion of cloud radar data assimilation cycling. Results suggest that NWP with cloud radar data assimilation in the pre-rain stage has great potential for predicting LHR, and can lead to an early evacuation warning and subsequent evacuation of vulnerable populations.
Significance Statement
The development of prediction methods for local (within several kilometers) heavy rain (LHR) is important because LHR events can cause deaths through the sudden rise of rivers and flooding of roads by rapidly developing (≤30 min) rain clouds. This study aims to develop a method for predicting LHR even before it begins to rain, which has been difficult to date. Using a technique called data assimilation, which integrates observation and simulation, we developed a method for assimilating cloud radar observations that can capture cloud droplets before raindrops form. As a result, we succeeded in predicting LHR before rainfall commenced. By extending and applying this research, early evacuation of vulnerable populations during LHR is possible.
Abstract
Pulse severe storms are single-cell thunderstorms that produce severe wind and/or severe hail for a brief period of time. These storms pose a major warm season forecasting problem since forecasters presently do not have sufficient guidance to know which, if any, of the cells that are observed will become severe. The empirical Probability of Severe (ProbSevere) model, developed by the Cooperative Institute for Meteorological Satellite Studies (CIMSS), fuses real-time data to produce short-term (0–60 min), statistically derived probabilistic forecasts of thunderstorm intensity. This study evaluates the ability of ProbSevere to predict pulse severe storms in the southeast United States. ProbSevere objects fitting the usual definition of a pulse severe environment were matched with severe events from Storm Data to create a dataset of ProbSevere objects that corresponded to pulse severe thunderstorms. A null dataset consisted of objects in pulse severe environments that did not match with a severe event. Results reveal that ProbSevere’s probabilities are small to moderate at the times corresponding to pulse severe events. While probabilities of nonsevere storms are generally smaller, there are a large number of outliers. Lightning flash rate is the only predictor relevant to this study that correlates strongly with increasingly favorable pulse storm probabilities. We conclude that ProbSevere provides forecasters only limited guidance as to whether a pulse severe event will soon occur. Developing a version of ProbSevere specifically for pulse severe storms would likely lead to better predictability for this mode of convection.
Abstract
Pulse severe storms are single-cell thunderstorms that produce severe wind and/or severe hail for a brief period of time. These storms pose a major warm season forecasting problem since forecasters presently do not have sufficient guidance to know which, if any, of the cells that are observed will become severe. The empirical Probability of Severe (ProbSevere) model, developed by the Cooperative Institute for Meteorological Satellite Studies (CIMSS), fuses real-time data to produce short-term (0–60 min), statistically derived probabilistic forecasts of thunderstorm intensity. This study evaluates the ability of ProbSevere to predict pulse severe storms in the southeast United States. ProbSevere objects fitting the usual definition of a pulse severe environment were matched with severe events from Storm Data to create a dataset of ProbSevere objects that corresponded to pulse severe thunderstorms. A null dataset consisted of objects in pulse severe environments that did not match with a severe event. Results reveal that ProbSevere’s probabilities are small to moderate at the times corresponding to pulse severe events. While probabilities of nonsevere storms are generally smaller, there are a large number of outliers. Lightning flash rate is the only predictor relevant to this study that correlates strongly with increasingly favorable pulse storm probabilities. We conclude that ProbSevere provides forecasters only limited guidance as to whether a pulse severe event will soon occur. Developing a version of ProbSevere specifically for pulse severe storms would likely lead to better predictability for this mode of convection.
Abstract
The relationship between convective parameters derived from ERA5 and cloud-to-ground (CG) lightning flashes from the PERUN network in Poland was evaluated. All flashes detected between 2002 and 2019 were divided into intensity categories based on a peak 1-min CG lightning flash rate and were collocated with proximal profiles from ERA5 to assess their climatological variability. Thunderstorms in Poland are the most frequent in July, between 1400 and 1600 UTC and over the southeastern parts of the country. The highest median of most unstable convective available potential energy (MUCAPE) for CG lightning flash events is from June to August, between 1400 and 1600 UTC (around 900 J kg−1), whereas patterns in 0–6-km wind shear [deep-layer shear (DLS)] are reversed, with the highest median values during winter and night (around 25 m s−1). The best overlap of MUCAPE and DLS (MUWMAXSHEAR parameter) is in July–August, typically between 1400 and 2000 UTC with median values of around 850 m2 s−2. Thunderstorms in Poland are the most frequent in MUCAPE below 1000 J kg−1, and DLS between 8 and 15 m s−1. Along with increasing MUCAPE and DLS, peak CG lightning flash rates increase as well. Compared to DLS, MUCAPE is a more important parameter in forecasting any lightning activity, but when these two are combined together (MUWMAXSHEAR) they are more reliable in distinguishing between thunderstorms producing small and high CG lightning flash rates. Our results also indicate that higher CG lightning flash rates result in thunderstorms more frequently associated with severe weather reports (hail, tornado, wind).
Significance Statement
Each year severe thunderstorms produce considerable material losses and lead to deaths across central Europe; thus, a better understanding of local storm climatologies and their accompanying environments is important for operational forecasters, emergency managers, and risk estimation. In this research we address this issue by analyzing 18 years of lightning intensity data and collocated atmospheric environments. Thunderstorms in Poland are the most frequent in July between 1400 and 1600 UTC and form typically in environments with low atmospheric instability and moderate vertical shear of the horizontal wind. The probability for storms producing intense lightning increases when both of these environmental parameters reach higher values.
Abstract
The relationship between convective parameters derived from ERA5 and cloud-to-ground (CG) lightning flashes from the PERUN network in Poland was evaluated. All flashes detected between 2002 and 2019 were divided into intensity categories based on a peak 1-min CG lightning flash rate and were collocated with proximal profiles from ERA5 to assess their climatological variability. Thunderstorms in Poland are the most frequent in July, between 1400 and 1600 UTC and over the southeastern parts of the country. The highest median of most unstable convective available potential energy (MUCAPE) for CG lightning flash events is from June to August, between 1400 and 1600 UTC (around 900 J kg−1), whereas patterns in 0–6-km wind shear [deep-layer shear (DLS)] are reversed, with the highest median values during winter and night (around 25 m s−1). The best overlap of MUCAPE and DLS (MUWMAXSHEAR parameter) is in July–August, typically between 1400 and 2000 UTC with median values of around 850 m2 s−2. Thunderstorms in Poland are the most frequent in MUCAPE below 1000 J kg−1, and DLS between 8 and 15 m s−1. Along with increasing MUCAPE and DLS, peak CG lightning flash rates increase as well. Compared to DLS, MUCAPE is a more important parameter in forecasting any lightning activity, but when these two are combined together (MUWMAXSHEAR) they are more reliable in distinguishing between thunderstorms producing small and high CG lightning flash rates. Our results also indicate that higher CG lightning flash rates result in thunderstorms more frequently associated with severe weather reports (hail, tornado, wind).
Significance Statement
Each year severe thunderstorms produce considerable material losses and lead to deaths across central Europe; thus, a better understanding of local storm climatologies and their accompanying environments is important for operational forecasters, emergency managers, and risk estimation. In this research we address this issue by analyzing 18 years of lightning intensity data and collocated atmospheric environments. Thunderstorms in Poland are the most frequent in July between 1400 and 1600 UTC and form typically in environments with low atmospheric instability and moderate vertical shear of the horizontal wind. The probability for storms producing intense lightning increases when both of these environmental parameters reach higher values.
Abstract
A scale-aware convective parameterization based on the Tiedtke scheme is developed and tested in the Weather Research and Forecasting (WRF) Model and the Model for Prediction Across Scales (MPAS) for a few convective cases at grid sizes in the ranges of 1.5–4.5 km. These tests demonstrate that the scale-aware scheme effectively reduces the outcome of deep convection by decreasing the convective portion of the total surface precipitation. When compared to the model runs that use microphysics without the cumulus parameterization at these grid sizes, the modified Tiedtke scheme is shown to improve some aspects of the precipitation forecasts. When the scheme is applied on a variable mesh in MPAS, it handles the convection across the mesh transition zones smoothly.
Significance Statement
Representing convection accounting for variations in the size of grid mesh is crucial in numerical models with variable resolutions, and in precipitation events where convection is not well depicted even by a model mesh of a few kilometers. Many convective parameterizations have already considered this grid-size dependency. This paper fills a gap by applying the same concept to a different convective parameterization, and evaluating it in a few precipitation forecast scenarios.
Abstract
A scale-aware convective parameterization based on the Tiedtke scheme is developed and tested in the Weather Research and Forecasting (WRF) Model and the Model for Prediction Across Scales (MPAS) for a few convective cases at grid sizes in the ranges of 1.5–4.5 km. These tests demonstrate that the scale-aware scheme effectively reduces the outcome of deep convection by decreasing the convective portion of the total surface precipitation. When compared to the model runs that use microphysics without the cumulus parameterization at these grid sizes, the modified Tiedtke scheme is shown to improve some aspects of the precipitation forecasts. When the scheme is applied on a variable mesh in MPAS, it handles the convection across the mesh transition zones smoothly.
Significance Statement
Representing convection accounting for variations in the size of grid mesh is crucial in numerical models with variable resolutions, and in precipitation events where convection is not well depicted even by a model mesh of a few kilometers. Many convective parameterizations have already considered this grid-size dependency. This paper fills a gap by applying the same concept to a different convective parameterization, and evaluating it in a few precipitation forecast scenarios.
Abstract
The High-Resolution Rapid Refresh (HRRR) is a convection-allowing implementation of the Advanced Research version of the Weather Research and Forecasting (WRF-ARW) Model with hourly data assimilation that covers the conterminous United States and Alaska and runs in real time at the NOAA/National Centers for Environmental Prediction (NCEP). Implemented operationally at NOAA/NCEP in 2014, the HRRR features 3-km horizontal grid spacing and frequent forecasts (hourly for CONUS and 3-hourly for Alaska). HRRR initialization is designed for optimal short-range forecast skill with a particular focus on the evolution of precipitating systems. Key components of the initialization are radar-reflectivity data assimilation, hybrid ensemble-variational assimilation of conventional weather observations, and a cloud analysis to initialize stratiform cloud layers. From this initial state, HRRR forecasts are produced out to 18 h every hour, and out to 48 h every 6 h, with boundary conditions provided by the Rapid Refresh system. Between 2014 and 2020, HRRR development was focused on reducing model bias errors and improving forecast realism and accuracy. Improved representation of the planetary boundary layer, subgrid-scale clouds, and land surface contributed extensively to overall HRRR improvements. The final version of the HRRR (HRRRv4), implemented in late 2020, also features hybrid data assimilation using flow-dependent covariances from a 3-km, 36-member ensemble (“HRRRDAS”) with explicit convective storms. HRRRv4 also includes prediction of wildfire smoke plumes. The HRRR provides a baseline capability for evaluating NOAA’s next-generation Rapid Refresh Forecast System, now under development.
Significance Statement
NOAA’s operational hourly updating, convection-allowing model, the High-Resolution Rapid Refresh (HRRR), is a key tool for short-range weather forecasting and situational awareness. Improvements in assimilation of weather observations, as well as in physics parameterizations, have led to improvements in simulated radar reflectivity and quantitative precipitation forecasts since the initial implementation of HRRR in September 2014. Other targeted development has focused on improved representation of the diurnal cycle of the planetary boundary layer, resulting in improved near-surface temperature and humidity forecasts. Additional physics and data assimilation changes have led to improved treatment of the development and erosion of low-level clouds, including subgrid-scale clouds. The final version of HRRR features storm-scale ensemble data assimilation and explicit prediction of wildfire smoke plumes.
Abstract
The High-Resolution Rapid Refresh (HRRR) is a convection-allowing implementation of the Advanced Research version of the Weather Research and Forecasting (WRF-ARW) Model with hourly data assimilation that covers the conterminous United States and Alaska and runs in real time at the NOAA/National Centers for Environmental Prediction (NCEP). Implemented operationally at NOAA/NCEP in 2014, the HRRR features 3-km horizontal grid spacing and frequent forecasts (hourly for CONUS and 3-hourly for Alaska). HRRR initialization is designed for optimal short-range forecast skill with a particular focus on the evolution of precipitating systems. Key components of the initialization are radar-reflectivity data assimilation, hybrid ensemble-variational assimilation of conventional weather observations, and a cloud analysis to initialize stratiform cloud layers. From this initial state, HRRR forecasts are produced out to 18 h every hour, and out to 48 h every 6 h, with boundary conditions provided by the Rapid Refresh system. Between 2014 and 2020, HRRR development was focused on reducing model bias errors and improving forecast realism and accuracy. Improved representation of the planetary boundary layer, subgrid-scale clouds, and land surface contributed extensively to overall HRRR improvements. The final version of the HRRR (HRRRv4), implemented in late 2020, also features hybrid data assimilation using flow-dependent covariances from a 3-km, 36-member ensemble (“HRRRDAS”) with explicit convective storms. HRRRv4 also includes prediction of wildfire smoke plumes. The HRRR provides a baseline capability for evaluating NOAA’s next-generation Rapid Refresh Forecast System, now under development.
Significance Statement
NOAA’s operational hourly updating, convection-allowing model, the High-Resolution Rapid Refresh (HRRR), is a key tool for short-range weather forecasting and situational awareness. Improvements in assimilation of weather observations, as well as in physics parameterizations, have led to improvements in simulated radar reflectivity and quantitative precipitation forecasts since the initial implementation of HRRR in September 2014. Other targeted development has focused on improved representation of the diurnal cycle of the planetary boundary layer, resulting in improved near-surface temperature and humidity forecasts. Additional physics and data assimilation changes have led to improved treatment of the development and erosion of low-level clouds, including subgrid-scale clouds. The final version of HRRR features storm-scale ensemble data assimilation and explicit prediction of wildfire smoke plumes.
Abstract
The High-Resolution Rapid Refresh (HRRR) is a convection-allowing implementation of the Advanced Research version of the Weather Research and Forecast (WRF-ARW) Model that covers the conterminous United States and Alaska and runs hourly (for CONUS; every 3 h for Alaska) in real time at the National Centers for Environmental Prediction. The high-resolution forecasts support a variety of user applications including aviation, renewable energy, and prediction of many forms of severe weather. In this second of two articles, forecast performance is documented for a wide variety of forecast variables and across HRRR versions. HRRR performance varies across geographical domain, season, and time of day depending on both prevalence of particular meteorological phenomena and the availability of both conventional and nonconventional observations. Station-based verification of surface weather forecasts (2-m temperature and dewpoint temperature, 10-m winds, visibility, and cloud ceiling) highlights the ability of the HRRR to represent daily planetary boundary layer evolution and the development of convective and stratiform cloud systems, while gridded verification of simulated composite radar reflectivity and quantitative precipitation forecasts reveals HRRR predictive skill for summer and winter precipitation systems. Significant improvements in performance for specific forecast problems are documented for the upgrade versions of the HRRR (HRRRv2, v3, and v4) implemented in 2016, 2018, and 2020, respectively. Development of the HRRR model data assimilation and physics paves the way for future progress with operational convective-scale modeling.
Significance Statement
NOAA’s operational hourly updating convection-allowing model, the High-Resolution Rapid Refresh (HRRR), is a key tool for short-range weather forecasting and situational awareness. Improvements in assimilation of weather observations, as well as in physics parameterizations, has led to improvements in simulated radar reflectivity and quantitative precipitation forecasts since the initial implementation of HRRR in September 2014. Other targeted development has focused on improved representation of the diurnal cycle of the planetary boundary layer, resulting in improved near-surface temperature and humidity forecasts. Additional physics and data assimilation changes have led to improved treatment of the development and erosion of low-level clouds, including subgrid-scale clouds. The final version of HRRR features storm-scale ensemble data assimilation and explicit prediction of wildfire smoke plumes.
Abstract
The High-Resolution Rapid Refresh (HRRR) is a convection-allowing implementation of the Advanced Research version of the Weather Research and Forecast (WRF-ARW) Model that covers the conterminous United States and Alaska and runs hourly (for CONUS; every 3 h for Alaska) in real time at the National Centers for Environmental Prediction. The high-resolution forecasts support a variety of user applications including aviation, renewable energy, and prediction of many forms of severe weather. In this second of two articles, forecast performance is documented for a wide variety of forecast variables and across HRRR versions. HRRR performance varies across geographical domain, season, and time of day depending on both prevalence of particular meteorological phenomena and the availability of both conventional and nonconventional observations. Station-based verification of surface weather forecasts (2-m temperature and dewpoint temperature, 10-m winds, visibility, and cloud ceiling) highlights the ability of the HRRR to represent daily planetary boundary layer evolution and the development of convective and stratiform cloud systems, while gridded verification of simulated composite radar reflectivity and quantitative precipitation forecasts reveals HRRR predictive skill for summer and winter precipitation systems. Significant improvements in performance for specific forecast problems are documented for the upgrade versions of the HRRR (HRRRv2, v3, and v4) implemented in 2016, 2018, and 2020, respectively. Development of the HRRR model data assimilation and physics paves the way for future progress with operational convective-scale modeling.
Significance Statement
NOAA’s operational hourly updating convection-allowing model, the High-Resolution Rapid Refresh (HRRR), is a key tool for short-range weather forecasting and situational awareness. Improvements in assimilation of weather observations, as well as in physics parameterizations, has led to improvements in simulated radar reflectivity and quantitative precipitation forecasts since the initial implementation of HRRR in September 2014. Other targeted development has focused on improved representation of the diurnal cycle of the planetary boundary layer, resulting in improved near-surface temperature and humidity forecasts. Additional physics and data assimilation changes have led to improved treatment of the development and erosion of low-level clouds, including subgrid-scale clouds. The final version of HRRR features storm-scale ensemble data assimilation and explicit prediction of wildfire smoke plumes.
Abstract
A short-range regional, two-way coupled atmosphere–ocean–ice model has been recently developed in an attempt to improve, among other things, quantitative precipitation forecasts (QPFs) over southern Ontario, Canada, by incorporating air–lake interaction over the Great Lakes region. Here, we attempt to 1) assess the impact of the air–lake coupling on daily QPFs, as verified against the Canadian Precipitation Analysis and independent observations, over southern Ontario during the period of June 2016–May 2017; and 2) diagnose major physical processes governing the QPF differences between the coupled and uncoupled models by relating precipitation to those processes at the air–water interface and above. Results indicate that the coupled model tends to reduce the area-averaged and monthly averaged daily QPF biases and standard deviations in 5 months of October, November, and December 2016, and April and May 2017, but increase and deteriorate precipitation biases during the summer months. Most of the deteriorations occur during the daytime, while improvements are observed during the nighttime (in 7 of 12 months). During the daytime, slight improvements appear in 2 months. A further diagnosis indicates that the daily QPF differences between the two models are highly correlated with the differences of their sensible and latent heat fluxes. The maximum (minimum) difference of sensible (latent) heat flux in August 2016 (December 2016) is in phase with the maximum (minimum) difference of the two-model daily QPFs. The daily QPF differences in the other months are also controlled by the differences of vertically integrated water vapor flux convergence, and surface temperature.
Abstract
A short-range regional, two-way coupled atmosphere–ocean–ice model has been recently developed in an attempt to improve, among other things, quantitative precipitation forecasts (QPFs) over southern Ontario, Canada, by incorporating air–lake interaction over the Great Lakes region. Here, we attempt to 1) assess the impact of the air–lake coupling on daily QPFs, as verified against the Canadian Precipitation Analysis and independent observations, over southern Ontario during the period of June 2016–May 2017; and 2) diagnose major physical processes governing the QPF differences between the coupled and uncoupled models by relating precipitation to those processes at the air–water interface and above. Results indicate that the coupled model tends to reduce the area-averaged and monthly averaged daily QPF biases and standard deviations in 5 months of October, November, and December 2016, and April and May 2017, but increase and deteriorate precipitation biases during the summer months. Most of the deteriorations occur during the daytime, while improvements are observed during the nighttime (in 7 of 12 months). During the daytime, slight improvements appear in 2 months. A further diagnosis indicates that the daily QPF differences between the two models are highly correlated with the differences of their sensible and latent heat fluxes. The maximum (minimum) difference of sensible (latent) heat flux in August 2016 (December 2016) is in phase with the maximum (minimum) difference of the two-model daily QPFs. The daily QPF differences in the other months are also controlled by the differences of vertically integrated water vapor flux convergence, and surface temperature.
Abstract
Recently, the use of machine learning in meteorology has increased greatly. While many machine learning methods are not new, university classes on machine learning are largely unavailable to meteorology students and are not required to become a meteorologist. The lack of formal instruction has contributed to perception that machine learning methods are “black boxes” and thus end-users are hesitant to apply the machine learning methods in their everyday workflow. To reduce the opaqueness of machine learning methods and lower hesitancy toward machine learning in meteorology, this paper provides a survey of some of the most common machine learning methods. A familiar meteorological example is used to contextualize the machine learning methods while also discussing machine learning topics using plain language. The following machine learning methods are demonstrated: linear regression, logistic regression, decision trees, random forest, gradient boosted decision trees, naïve Bayes, and support vector machines. Beyond discussing the different methods, the paper also contains discussions on the general machine learning process as well as best practices to enable readers to apply machine learning to their own datasets. Furthermore, all code (in the form of Jupyter notebooks and Google Colaboratory notebooks) used to make the examples in the paper is provided in an effort to catalyze the use of machine learning in meteorology.
Abstract
Recently, the use of machine learning in meteorology has increased greatly. While many machine learning methods are not new, university classes on machine learning are largely unavailable to meteorology students and are not required to become a meteorologist. The lack of formal instruction has contributed to perception that machine learning methods are “black boxes” and thus end-users are hesitant to apply the machine learning methods in their everyday workflow. To reduce the opaqueness of machine learning methods and lower hesitancy toward machine learning in meteorology, this paper provides a survey of some of the most common machine learning methods. A familiar meteorological example is used to contextualize the machine learning methods while also discussing machine learning topics using plain language. The following machine learning methods are demonstrated: linear regression, logistic regression, decision trees, random forest, gradient boosted decision trees, naïve Bayes, and support vector machines. Beyond discussing the different methods, the paper also contains discussions on the general machine learning process as well as best practices to enable readers to apply machine learning to their own datasets. Furthermore, all code (in the form of Jupyter notebooks and Google Colaboratory notebooks) used to make the examples in the paper is provided in an effort to catalyze the use of machine learning in meteorology.
