Browse
Abstract
A prognostic closure is introduced to, and evaluated in, NOAA’s Unified Forecast System. The closure addresses aspects that are not commonly represented in traditional cumulus convection parameterizations, and it departs from the previous assumptions of a negligible subgrid area coverage and statistical quasi-equilibrium at steady state, the latter of which becomes invalid at higher resolution. The new parameterization introduces a prognostic evolution of the convective updraft area fraction based on a moisture budget, and, together with the buoyancy-driven updraft vertical velocity, it completes the cloud-base mass flux. In addition, the new closure addresses stochasticity and includes a representation of subgrid convective organization using cellular automata as well as scale-adaptive considerations. The new cumulus convection closure shows potential for improved Madden–Julian oscillation (MJO) prediction. In our simulations we observe better propagation, amplitude, and phase of the MJO in a case study relative to the control simulation. This improvement can be partly attributed to a closer coupling between low-level moisture flux convergence and precipitation as revealed by a space–time coherence spectrum. In addition, we find that enhanced organization feedback representation and stochastic effects, represented using cellular automata, further enhance the amplitude and propagation of the MJO, and they provide realistic uncertainty estimates of convectively coupled equatorial waves at seasonal time scales. The scale-adaptive behavior of the scheme is also studied by running the global model with 25-, 13-, 9-, and 3-km grid spacing. It is found that the convective area fraction and the convective updraft velocity are both scale adaptive, leading to a reduction of subgrid convective precipitation in the higher-resolution simulations.
Abstract
A prognostic closure is introduced to, and evaluated in, NOAA’s Unified Forecast System. The closure addresses aspects that are not commonly represented in traditional cumulus convection parameterizations, and it departs from the previous assumptions of a negligible subgrid area coverage and statistical quasi-equilibrium at steady state, the latter of which becomes invalid at higher resolution. The new parameterization introduces a prognostic evolution of the convective updraft area fraction based on a moisture budget, and, together with the buoyancy-driven updraft vertical velocity, it completes the cloud-base mass flux. In addition, the new closure addresses stochasticity and includes a representation of subgrid convective organization using cellular automata as well as scale-adaptive considerations. The new cumulus convection closure shows potential for improved Madden–Julian oscillation (MJO) prediction. In our simulations we observe better propagation, amplitude, and phase of the MJO in a case study relative to the control simulation. This improvement can be partly attributed to a closer coupling between low-level moisture flux convergence and precipitation as revealed by a space–time coherence spectrum. In addition, we find that enhanced organization feedback representation and stochastic effects, represented using cellular automata, further enhance the amplitude and propagation of the MJO, and they provide realistic uncertainty estimates of convectively coupled equatorial waves at seasonal time scales. The scale-adaptive behavior of the scheme is also studied by running the global model with 25-, 13-, 9-, and 3-km grid spacing. It is found that the convective area fraction and the convective updraft velocity are both scale adaptive, leading to a reduction of subgrid convective precipitation in the higher-resolution simulations.
Abstract
The case study of a heavy precipitation event associated with the passage of cold front over the Australian Snowy Mountains (ASM) on 3 August 2018 has been examined using the observational data from an intensive field campaign and high-resolution (1 km) Weather Research and Forecasting (WRF) simulation. We divided this event into prefrontal, cold front, and postfrontal periods. The cold front and postfrontal periods were characterized by higher production of graupel, while relatively low graupel was produced in the prefrontal period. Overall, aggregation along with deposition are likely the main growth mechanisms of snow in the prefrontal clouds, while heavy rain was produced below the melting level over windward slopes of the ASM. The simulated melting level is lower compared to the observations, which is consistent with model cold bias. Stronger orographic uplift and frontal forcing were mainly responsible for the enhanced supercooled liquid water (SLW) production over the ASM in the cold front period. A drop in elevation of the freezing level and increase in low-level relative humidity further enhanced the SLW production. The production of graupel through riming processes was highly efficient in the cold front period given the high concentration of ice-phase hydrometeors in the frontal clouds and the development of clouds comprising supercooled liquid water. The orographic updrafts and embedded convection were the main dynamical processes generating postfrontal SLW clouds and graupel. Ice initiation processes were activated once SLW cloud tops reached −15°C level followed by graupel production through riming processes.
Abstract
The case study of a heavy precipitation event associated with the passage of cold front over the Australian Snowy Mountains (ASM) on 3 August 2018 has been examined using the observational data from an intensive field campaign and high-resolution (1 km) Weather Research and Forecasting (WRF) simulation. We divided this event into prefrontal, cold front, and postfrontal periods. The cold front and postfrontal periods were characterized by higher production of graupel, while relatively low graupel was produced in the prefrontal period. Overall, aggregation along with deposition are likely the main growth mechanisms of snow in the prefrontal clouds, while heavy rain was produced below the melting level over windward slopes of the ASM. The simulated melting level is lower compared to the observations, which is consistent with model cold bias. Stronger orographic uplift and frontal forcing were mainly responsible for the enhanced supercooled liquid water (SLW) production over the ASM in the cold front period. A drop in elevation of the freezing level and increase in low-level relative humidity further enhanced the SLW production. The production of graupel through riming processes was highly efficient in the cold front period given the high concentration of ice-phase hydrometeors in the frontal clouds and the development of clouds comprising supercooled liquid water. The orographic updrafts and embedded convection were the main dynamical processes generating postfrontal SLW clouds and graupel. Ice initiation processes were activated once SLW cloud tops reached −15°C level followed by graupel production through riming processes.
Abstract
ERA5 reanalyses and observations of convective clouds and precipitation are used over the northern Gulf of Guinea between 7°W and 3°E to study the influence of ocean surface temperature and the land–sea temperature gradient on Guinea Coast rainfall (GCR) in boreal spring and summer. Seasonal composites are calculated around two dates indexing the onset (T ref) and demise (T end) of the GCR. The T ref date corresponds to the emergence of the equatorial upwelling in boreal spring, which “pushes” the zonal precipitation belt northward against the Guinea coast. The T end date characterizes the emergence of the coastal upwelling in July, which is known to coincide with the beginning of the “Little Dry Season” that lasts until September. Along the Guinea Coast, the diurnal cycle of the air–sea temperature gradient controls precipitation through the land–sea breeze, which explains why precipitation reaches its maximum around noon over the ocean, and in the late afternoon over the continent. The emergence of the Guinea Coast upwelling in July induces a weakening of southerlies on a seasonal scale, and a weaker land breeze on a diurnal scale. It induces a decrease in the convergence of humidity transport across the coast and in coastal oceanic precipitation. Therefore, the GCR is seasonally controlled by the latitude of the maximum tropospheric water vapor content and the annual cycle of the West African monsoon, but the ocean surface temperature is responsible for the abruptness of its onset via the intensification of the equatorial upwelling around the end of May, and possibly of its demise as well via the emergence of the coastal upwelling by early July.
Abstract
ERA5 reanalyses and observations of convective clouds and precipitation are used over the northern Gulf of Guinea between 7°W and 3°E to study the influence of ocean surface temperature and the land–sea temperature gradient on Guinea Coast rainfall (GCR) in boreal spring and summer. Seasonal composites are calculated around two dates indexing the onset (T ref) and demise (T end) of the GCR. The T ref date corresponds to the emergence of the equatorial upwelling in boreal spring, which “pushes” the zonal precipitation belt northward against the Guinea coast. The T end date characterizes the emergence of the coastal upwelling in July, which is known to coincide with the beginning of the “Little Dry Season” that lasts until September. Along the Guinea Coast, the diurnal cycle of the air–sea temperature gradient controls precipitation through the land–sea breeze, which explains why precipitation reaches its maximum around noon over the ocean, and in the late afternoon over the continent. The emergence of the Guinea Coast upwelling in July induces a weakening of southerlies on a seasonal scale, and a weaker land breeze on a diurnal scale. It induces a decrease in the convergence of humidity transport across the coast and in coastal oceanic precipitation. Therefore, the GCR is seasonally controlled by the latitude of the maximum tropospheric water vapor content and the annual cycle of the West African monsoon, but the ocean surface temperature is responsible for the abruptness of its onset via the intensification of the equatorial upwelling around the end of May, and possibly of its demise as well via the emergence of the coastal upwelling by early July.
Abstract
Definition of the tropopause has remained a focus of atmospheric science since its discovery near the beginning of the twentieth century. Few universal definitions (those that can be reliably applied globally and to both common observations and numerical model output) exist and many definitions with unique limitations have been developed over the years. The most commonly used universal definition of the tropopause is the temperature lapse-rate definition established by the World Meteorological Organization (WMO) in 1957 (the LRT). Despite its widespread use, there are recurrent situations where the LRT definition fails to reliably identify the tropopause. Motivated by increased availability of coincident observations of stability and composition, this study seeks to reexamine the relationship between stability and composition change in the tropopause transition layer and identify areas for improvement in a stability-based definition of the tropopause. In particular, long-term (40+ years) balloon observations of temperature, ozone, and water vapor from six locations across the globe are used to identify covariability between several metrics of atmospheric stability and composition. We found that the vertical gradient of potential temperature is a superior stability metric to identify the greatest composition change in the tropopause transition layer, which we use to propose a new universally applicable potential temperature gradient tropopause (PTGT) definition. Application of the new definition to both observations and reanalysis output reveals that the PTGT largely agrees with the LRT, but more reliably identifies tropopause-level composition change when the two definitions differ greatly.
Significance Statement
In this study we provide a review of existing tropopause definitions (and their limitations) and investigate potential improvement in the definition of the tropopause using balloon-based observations of stability and atmospheric composition. This work is motivated by the need for correct identification of the tropopause to accurately assess upper-troposphere–lower-stratosphere processes, which in turn has far-reaching implications for our understanding of Earth’s radiation budget and climate. The result of this research is the creation of a new, universally applicable stability-based definition of the tropopause: the potential temperature gradient tropopause (PTGT).
Abstract
Definition of the tropopause has remained a focus of atmospheric science since its discovery near the beginning of the twentieth century. Few universal definitions (those that can be reliably applied globally and to both common observations and numerical model output) exist and many definitions with unique limitations have been developed over the years. The most commonly used universal definition of the tropopause is the temperature lapse-rate definition established by the World Meteorological Organization (WMO) in 1957 (the LRT). Despite its widespread use, there are recurrent situations where the LRT definition fails to reliably identify the tropopause. Motivated by increased availability of coincident observations of stability and composition, this study seeks to reexamine the relationship between stability and composition change in the tropopause transition layer and identify areas for improvement in a stability-based definition of the tropopause. In particular, long-term (40+ years) balloon observations of temperature, ozone, and water vapor from six locations across the globe are used to identify covariability between several metrics of atmospheric stability and composition. We found that the vertical gradient of potential temperature is a superior stability metric to identify the greatest composition change in the tropopause transition layer, which we use to propose a new universally applicable potential temperature gradient tropopause (PTGT) definition. Application of the new definition to both observations and reanalysis output reveals that the PTGT largely agrees with the LRT, but more reliably identifies tropopause-level composition change when the two definitions differ greatly.
Significance Statement
In this study we provide a review of existing tropopause definitions (and their limitations) and investigate potential improvement in the definition of the tropopause using balloon-based observations of stability and atmospheric composition. This work is motivated by the need for correct identification of the tropopause to accurately assess upper-troposphere–lower-stratosphere processes, which in turn has far-reaching implications for our understanding of Earth’s radiation budget and climate. The result of this research is the creation of a new, universally applicable stability-based definition of the tropopause: the potential temperature gradient tropopause (PTGT).
Abstract
A tropical cyclone (TC) is a powerful, rotating storm that typically originates over warm tropical oceans and creates strong winds and heavy rain; it is usually a natural disaster with respect to human life and property if it moves over land. This work examines effects of varying radiative forcing on the evolution of two typhoon cases—Typhoon Lionrock (2016) and Typhoon Hagibis (2019)—with the Weather Research and Forecasting (WRF) Model. Hagibis was a rapidly intensifying and quickly moving TC, whereas Lionrock gradually developed and was slow moving. Numerous sensitivity experiments in which shortwave and longwave radiative heating rates were modified were conducted. This study examined latent heating and radiative heating for each experiment. Substantial differences between the sensitivity simulation members indicated that radiative effects can strongly influence TC development. The analysis of diabatic heating sources shows that, before eyewall formation, the differential cooling effect, which indicates that longwave cooling rates between cloud clusters and clear sky differ, can promote low-level inflow and increase relative humidity in the cloud clusters. If the initial relative humidity is low, this effect becomes important because, without differential cooling, the relative humidity remains low, which can promote the generation of cold pools that will prevent cyclone development. After eyewall formation, both the change in temperature lapse rate due to a vertical gradient of radiative heating/cooling and the change in the warm core due to radiative heating/cooling can affect the intensity of a TC; however, the net effect may depend on the magnitude of these influences.
Abstract
A tropical cyclone (TC) is a powerful, rotating storm that typically originates over warm tropical oceans and creates strong winds and heavy rain; it is usually a natural disaster with respect to human life and property if it moves over land. This work examines effects of varying radiative forcing on the evolution of two typhoon cases—Typhoon Lionrock (2016) and Typhoon Hagibis (2019)—with the Weather Research and Forecasting (WRF) Model. Hagibis was a rapidly intensifying and quickly moving TC, whereas Lionrock gradually developed and was slow moving. Numerous sensitivity experiments in which shortwave and longwave radiative heating rates were modified were conducted. This study examined latent heating and radiative heating for each experiment. Substantial differences between the sensitivity simulation members indicated that radiative effects can strongly influence TC development. The analysis of diabatic heating sources shows that, before eyewall formation, the differential cooling effect, which indicates that longwave cooling rates between cloud clusters and clear sky differ, can promote low-level inflow and increase relative humidity in the cloud clusters. If the initial relative humidity is low, this effect becomes important because, without differential cooling, the relative humidity remains low, which can promote the generation of cold pools that will prevent cyclone development. After eyewall formation, both the change in temperature lapse rate due to a vertical gradient of radiative heating/cooling and the change in the warm core due to radiative heating/cooling can affect the intensity of a TC; however, the net effect may depend on the magnitude of these influences.
Abstract
Weather regimes (WRs), also known as synoptic types, are defined as recurrent patterns that have been used to categorize variability in atmospheric circulation. However, defining the optimal number of patterns can often be arbitrary, and there are common shortcomings when oversimplifying a wide range of synoptic conditions and weather outcomes. We build on previous work that has defined regional WRs and objectively ascribe an optimal number of once-daily weather patterns for Aotearoa New Zealand (ANZ) using affinity propagation combined with K-means clustering. Nine primary WRs for ANZ were classified based on once-daily geopotential height spatial patterns, but these patterns still retained a wide degree of spatial variability. Subsidiary clusters were subsequently defined within each primary WR by applying affinity propagation and K-means clustering to reveal the largest within-cluster differences based on joint daily temperature and precipitation anomalies. Up to three subsidiary patterns in each of the primary regimes were revealed, with a total of 21 unique daily patterns emerging from the two-tier classification. Subsidiary WRs reveal subtle differences in the location and intensity of regional-scale pressure anomalies, pressure gradients, and wind flow over both main islands that lead to large differences in surface weather anomalies. Impacts of atmospheric variability related to each subsidiary WR are exemplified by different spatial outcomes for rainfall and temperature (including intensity of anomalies) at regional and subregional levels. The approach presented in this study has utility for enhancing prediction of weather outcomes, including extreme weather, and can also be applied more widely over a range of time scales to improve understanding of weather and climate linkages.
Abstract
Weather regimes (WRs), also known as synoptic types, are defined as recurrent patterns that have been used to categorize variability in atmospheric circulation. However, defining the optimal number of patterns can often be arbitrary, and there are common shortcomings when oversimplifying a wide range of synoptic conditions and weather outcomes. We build on previous work that has defined regional WRs and objectively ascribe an optimal number of once-daily weather patterns for Aotearoa New Zealand (ANZ) using affinity propagation combined with K-means clustering. Nine primary WRs for ANZ were classified based on once-daily geopotential height spatial patterns, but these patterns still retained a wide degree of spatial variability. Subsidiary clusters were subsequently defined within each primary WR by applying affinity propagation and K-means clustering to reveal the largest within-cluster differences based on joint daily temperature and precipitation anomalies. Up to three subsidiary patterns in each of the primary regimes were revealed, with a total of 21 unique daily patterns emerging from the two-tier classification. Subsidiary WRs reveal subtle differences in the location and intensity of regional-scale pressure anomalies, pressure gradients, and wind flow over both main islands that lead to large differences in surface weather anomalies. Impacts of atmospheric variability related to each subsidiary WR are exemplified by different spatial outcomes for rainfall and temperature (including intensity of anomalies) at regional and subregional levels. The approach presented in this study has utility for enhancing prediction of weather outcomes, including extreme weather, and can also be applied more widely over a range of time scales to improve understanding of weather and climate linkages.
Abstract
To effectively reduce model bias and improve assimilation quality, we adopt a hybrid adaptive approach of ensemble adjustment Kalman filter (EAKF) and multigrid analysis (MGA), called EAKF-MGA, to implement parameter optimization as follows. For each assimilation cycle, observations are used to adjust the prior ensembles of both state variables and parameters using the EAKF without inflation. Then, the MGA is adaptively triggered to extract multiscale information from the observational residual to innovate the ensemble mean of the state once again. Results of biased twin experiments consisting of a barotropic spectral model and idealized observation systems show that the proposed EAKF-MGA is insensitive to state variance inflation and localization during the parameter optimization process, compared with the EAKF with adaptive inflation. We also find that computational efficiency is another important advantage of the EAKF-MGA for both state estimation and parameter estimation since extremely small ensemble size is allowed, while the EAKF with adaptive inflation does not work anymore. In essence, the EAKF-MGA is designed to estimate and correct systematic errors jointly with model’s state variables. Through alleviating biases, including the model bias caused by the biased parameter and the analysis bias resulting from the sampling noise given the limited ensemble size, it can be guaranteed that the analysis in the EAKF-MGA will be proceeded onward with the standard assumption of the unbiased model background field in modern data assimilation theory to be met.
Abstract
To effectively reduce model bias and improve assimilation quality, we adopt a hybrid adaptive approach of ensemble adjustment Kalman filter (EAKF) and multigrid analysis (MGA), called EAKF-MGA, to implement parameter optimization as follows. For each assimilation cycle, observations are used to adjust the prior ensembles of both state variables and parameters using the EAKF without inflation. Then, the MGA is adaptively triggered to extract multiscale information from the observational residual to innovate the ensemble mean of the state once again. Results of biased twin experiments consisting of a barotropic spectral model and idealized observation systems show that the proposed EAKF-MGA is insensitive to state variance inflation and localization during the parameter optimization process, compared with the EAKF with adaptive inflation. We also find that computational efficiency is another important advantage of the EAKF-MGA for both state estimation and parameter estimation since extremely small ensemble size is allowed, while the EAKF with adaptive inflation does not work anymore. In essence, the EAKF-MGA is designed to estimate and correct systematic errors jointly with model’s state variables. Through alleviating biases, including the model bias caused by the biased parameter and the analysis bias resulting from the sampling noise given the limited ensemble size, it can be guaranteed that the analysis in the EAKF-MGA will be proceeded onward with the standard assumption of the unbiased model background field in modern data assimilation theory to be met.
Abstract
An experimental Warn-on-Forecast System (WoFS) ensemble data assimilation (DA) and prediction system at 1-km grid spacing is developed and tested using two landfalling tropical cyclone (TC) events, one springtime severe thunderstorm event, and one summertime flash flood event. To evaluate the impact of DA at 1-km grid spacing, two experiments are conducted. One experiment, namely, the WoFS-1km, generates 3-h ensemble forecasts from the 1-km WoFS analyses while another experiment, namely, the Downscaled-1km, generates 3-h ensemble forecasts from downscaled 3-km analyses. With 1-km DA, the two landfalling TC events and the summertime event show some improvement in predicting high reflectivity, while the springtime event performs worse. Meanwhile, WoFS-1km is slightly better at predicting heavier precipitation (>20 mm h−1) with lower bias. However, heavy precipitation spatial placement error is only mitigated in one TC event and the summertime event with 1-km DA but is neutral or worse in the other two events. Object-based verification for rotation objects indicates that WoFS-1km performs better in one of the TC events, but worse in the springtime event with lower probability of detection and higher false alarm ratio due to fewer strong rotation objects being generated. The forecast skill of WoFS-1km for the springtime event is degraded mainly because the convective cores do not sufficiently develop as the forecast advances. The conditional benefits from 1-km DA in this study highlights the need for evaluation of a larger sample of convective storm cases and further development of the system.
Abstract
An experimental Warn-on-Forecast System (WoFS) ensemble data assimilation (DA) and prediction system at 1-km grid spacing is developed and tested using two landfalling tropical cyclone (TC) events, one springtime severe thunderstorm event, and one summertime flash flood event. To evaluate the impact of DA at 1-km grid spacing, two experiments are conducted. One experiment, namely, the WoFS-1km, generates 3-h ensemble forecasts from the 1-km WoFS analyses while another experiment, namely, the Downscaled-1km, generates 3-h ensemble forecasts from downscaled 3-km analyses. With 1-km DA, the two landfalling TC events and the summertime event show some improvement in predicting high reflectivity, while the springtime event performs worse. Meanwhile, WoFS-1km is slightly better at predicting heavier precipitation (>20 mm h−1) with lower bias. However, heavy precipitation spatial placement error is only mitigated in one TC event and the summertime event with 1-km DA but is neutral or worse in the other two events. Object-based verification for rotation objects indicates that WoFS-1km performs better in one of the TC events, but worse in the springtime event with lower probability of detection and higher false alarm ratio due to fewer strong rotation objects being generated. The forecast skill of WoFS-1km for the springtime event is degraded mainly because the convective cores do not sufficiently develop as the forecast advances. The conditional benefits from 1-km DA in this study highlights the need for evaluation of a larger sample of convective storm cases and further development of the system.
Abstract
Wind-farm parameterizations in weather models can be used to predict both the power output and farm effects on the flow; however, their correctness has not been thoroughly assessed. We evaluate the wind-farm parameterization of the Weather Research and Forecasting Model with large-eddy simulations (LES) of the wake performed with the same model. We study the impact on the velocity and turbulence kinetic energy (TKE) of inflow velocity, roughness, resolution, number of turbines (one or two), and inversion height and strength. We compare the mesoscale with the LES by spatially averaging the LES within areas correspondent to the mesoscale horizontal spacing: one covering the turbine area and two downwind. We find an excellent agreement of the velocity within the turbine area between the two types of simulations. However, within the same area, we find the largest TKE discrepancies because in mesoscale simulations, the turbine-added TKE has to be highest at the turbine position to be advected downwind. Within the downwind areas, differences between velocities increase as the wake recovers faster in the LES, whereas for the TKE both types of simulations show similar levels. From the various configurations, the impact of inversion height and strength is small for these heights and inversion levels. The highest impact for the one-turbine simulations appears under the low-speed case due to the higher thrust, whereas the impact of resolution is low for the large-eddy simulations but high for the mesoscale simulations. Our findings demonstrate that higher-fidelity simulations are needed to validate wind-farm parameterizations.
Abstract
Wind-farm parameterizations in weather models can be used to predict both the power output and farm effects on the flow; however, their correctness has not been thoroughly assessed. We evaluate the wind-farm parameterization of the Weather Research and Forecasting Model with large-eddy simulations (LES) of the wake performed with the same model. We study the impact on the velocity and turbulence kinetic energy (TKE) of inflow velocity, roughness, resolution, number of turbines (one or two), and inversion height and strength. We compare the mesoscale with the LES by spatially averaging the LES within areas correspondent to the mesoscale horizontal spacing: one covering the turbine area and two downwind. We find an excellent agreement of the velocity within the turbine area between the two types of simulations. However, within the same area, we find the largest TKE discrepancies because in mesoscale simulations, the turbine-added TKE has to be highest at the turbine position to be advected downwind. Within the downwind areas, differences between velocities increase as the wake recovers faster in the LES, whereas for the TKE both types of simulations show similar levels. From the various configurations, the impact of inversion height and strength is small for these heights and inversion levels. The highest impact for the one-turbine simulations appears under the low-speed case due to the higher thrust, whereas the impact of resolution is low for the large-eddy simulations but high for the mesoscale simulations. Our findings demonstrate that higher-fidelity simulations are needed to validate wind-farm parameterizations.
Abstract
In this paper, a new nonlinear forcing singular vector (NFSV) approach is proposed to provide mutually independent optimally combined modes of initial perturbations and model perturbations (C-NFSVs) in ensemble forecasts. The C-NFSVs are a group of optimally growing structures that take into account the impact of the interaction between the initial errors and the model errors effectively, generalizing the original NFSV for simulations of the impact of the model errors. The C-NFSVs method is tested in the context of the Lorenz-96 model to demonstrate its potential to improve ensemble forecast skills. This method is compared with the orthogonal conditional nonlinear optimal perturbations (O-CNOPs) method for estimating only the initial uncertainties and the orthogonal NFSVs (O-NFSVs) for estimating only the model uncertainties. The results demonstrate that when both the initial perturbations and model perturbations are introduced in the forecasting system, the C-NFSVs are much more capable of achieving higher ensemble forecasting skills. The use of a deep learning approach as a remedy for the expensive computational costs of the C-NFSVs is evaluated. The results show that learning the impact of the C-NFSVs on the ensemble provides a useful and efficient alternative for the operational implementation of C-NFSVs in forecasting suites dealing with the combined effects of the initial errors and the model errors.
Significance Statement
A new ensemble forecasting method for dealing with combined effects of initial errors and model errors, i.e., the C-NFSVs, is proposed, which is an extension of the NFSV approach for simulating the model error effects in ensemble forecasts. The C-NFSVs provide mutually independent optimally combined modes of initial perturbations and model perturbations. This new method is tested for generating ensemble forecasts in the context of the Lorenz-96 model, and there are indications that the optimally growing structures may provide reliable ensemble forecasts. Furthermore, it is found that a hybrid dynamical–deep learning approach could be a potential avenue for real-time ensemble forecasting systems when perturbations combine the impact of the initial and the model errors.
Abstract
In this paper, a new nonlinear forcing singular vector (NFSV) approach is proposed to provide mutually independent optimally combined modes of initial perturbations and model perturbations (C-NFSVs) in ensemble forecasts. The C-NFSVs are a group of optimally growing structures that take into account the impact of the interaction between the initial errors and the model errors effectively, generalizing the original NFSV for simulations of the impact of the model errors. The C-NFSVs method is tested in the context of the Lorenz-96 model to demonstrate its potential to improve ensemble forecast skills. This method is compared with the orthogonal conditional nonlinear optimal perturbations (O-CNOPs) method for estimating only the initial uncertainties and the orthogonal NFSVs (O-NFSVs) for estimating only the model uncertainties. The results demonstrate that when both the initial perturbations and model perturbations are introduced in the forecasting system, the C-NFSVs are much more capable of achieving higher ensemble forecasting skills. The use of a deep learning approach as a remedy for the expensive computational costs of the C-NFSVs is evaluated. The results show that learning the impact of the C-NFSVs on the ensemble provides a useful and efficient alternative for the operational implementation of C-NFSVs in forecasting suites dealing with the combined effects of the initial errors and the model errors.
Significance Statement
A new ensemble forecasting method for dealing with combined effects of initial errors and model errors, i.e., the C-NFSVs, is proposed, which is an extension of the NFSV approach for simulating the model error effects in ensemble forecasts. The C-NFSVs provide mutually independent optimally combined modes of initial perturbations and model perturbations. This new method is tested for generating ensemble forecasts in the context of the Lorenz-96 model, and there are indications that the optimally growing structures may provide reliable ensemble forecasts. Furthermore, it is found that a hybrid dynamical–deep learning approach could be a potential avenue for real-time ensemble forecasting systems when perturbations combine the impact of the initial and the model errors.
