The cool-roof strategy is an effective way of mitigating severe urban heat islands, and its impacts on urban flow and air quality deserve in-depth investigations. This study examines the effects of cool roofs on turbulent coherent structures and ozone air quality in Seoul, South Korea using the Weather Research and Forecasting-Community Multiscale Air Quality (WRF-CMAQ) model with a 50 m horizontal grid spacing. Cool roofs decrease the daily average air temperature, planetary boundary layer (PBL) height, and wind speed in the urban area of Seoul by 0.80°C, 230 m, and 0.17 m s⁻¹, respectively. Due to the lowered air temperature by cool roofs, the sea breeze and convective structures weaken and eddies at the PBL top appear less frequently. Since high O3 concentration air flows into Seoul by the sea breeze, the weakened sea breeze decreases the daily average O3 concentration near the surface by 3.3 ppb. Air at lower level is transported upward across the PBL top by convective structures and eddies at the PBL top. The transported air at lower level has lower O3 concentration and higher concentrations of O3 precursors than air at upper level. Therefore, the weakened convective structures and less-frequent appearance of eddies at the PBL top by cool roofs weaken the upward transport of O3 precursors at lower level across the PBL top. As a result, the chemical production of O3 and O3 concentration slightly above the PBL top are decreased. An integrated process rate analysis shows that cool roofs weaken the effects of turbulent coherent structures on O3 concentration.

Wind flows strongly affect the land-surface exchange between forest and the overlying atmosphere. They can develop into inertial sublayer (ISL) and roughness sublayer (RSL). While ISL flows are well characterized by the flux-gradient theory, RSL characteristics are less understood. A series of wind tunnel experiments are conducted to investigate the RSL dynamics over the vegetation canopy fabricated by arrays of reduced-scale tree models. The constant-temperature (CT) hot-wire anemometry (HWA) is used to measure the boundary-layer flows developed over different configurations of tree models. The drag coefficient, representing aerodynamic resistance induced by the forest canopy, is in the range of 4.0×10³ - 7.6×10^-3. A new analytical model, which is verified by the wind tunnel data, is developed to predict the mean wind profile in both RSL and ISL. The turbulence characteristics and motion scales in RSL and ISL are further examined by conditional sampling based on quadrant analysis and frequency spectrum. The evolution of ejection Q2 and sweep Q4 contributions as well as the high- and low-frequency eddies in RSL and ISL will be reported in details.

Current understandings of the turbulent transport in urban roughness sublayer are mainly based on the studies of flows above idealized urban geometries like two-dimensional street canyons and building block arrays. However, real urban morphology is much more complex than idealized urban geometries, the turbulent transport characteristic of a real urban roughness sublayer can be substantially different from those above idealized urban geometries. For this reason, this study is aimed to investigate the effects of realistic urban geometry on the flow and turbulence characteristics of the roughness sublayer above using high-resolution building-resolved large-eddy simulation (LES). The open-source computational fluid dynamics code OpenFOAM is used to perform the LESs. The LES method is first validated against benchmark wind tunnel measurements of the mean wind and turbulence within and above a staggered array of cubes. A grid sensitivity test is performed based on three different mesh resolutions to ensure grid independency of the LES results. After, the LES code is applied to simulate the flow above realistic urban geometries. Three selected surfaces, with areas approximately equal to 1km * 0.5km, are prepared using the high-resolution digital building maps of three cities (Guangzhou, Shenzhen and Zhuhai) in Guangdong province of China. LESs are then performed to simulate the turbulent flows above the selected urban surfaces in neutral conditions with a spatial resolution of about 3m in all directions. Using the simulation results, the mean wind and turbulence characteristics such as the sectional drag forces, the Reynolds stresses and the dispersive shear stresses are compared with those of the idealized urban canopies. Based on the comparisons, the rationality of using the results of idealized urban geometries to predict or parametrize real urban flows will be discussed.

Mean wind profiles within urban canopies are estimated using the methods of vortex dynamics. Assuming that the velocity is controlled by intense layers of vorticity near solid boundaries, spatially averaged velocity profiles are calculated from the solution of a 3-D Poisson equation, i.e. from the numerical Green's function. Good agreement for the streamwise and spanwise velocity profiles is obtained for a unit aspect ratio street canyon and different winds. It is shown that the method generalises to step-up, step-down and deep street canyons and can be calibrated using single-point measurements taken above the roof level. RMS errors are in the range of 20% inside the canopy, which is a significant improvement over idealised solutions such as the log and exponential profiles. This approach, which offers enormous computational savings over CFD, can be applied to urban canopy parameterisation and fast dispersion models.

  The effect of an abrupt change in surface roughness on atmospheric flows is investigated through the large-eddy simulation (LES) with the one-equation subgrid-scale (SGS) model. Flow responses to surface transitions of smoother-to-rougher and rougher-to-smoother are examined. Unlike previous analytical solutions, the roughness sublayers (RSLs) are explicitly resolved by solid boundary in the form of sinusoidal wavy walls at the bottom. Their influence on the growth of internal boundary layer (IBL), which characterizes the flow adaptation to the downstream new surface, is also studied. The modeling results show that, after the increase (decrease) in surface roughness, the flows decelerate (accelerate) and the downward momentum flux increases (decreases). In the near field after surface discontinuity, the flows are not yet in equilibrium with the bottom surface. In the far field, the flows are fully developed, exhibiting self-similarity in the streamwise direction. Budget analysis of momentum flux is carried out to elucidate the transport mechanism. Apparently, advection plays a critical role in the near field after surface discontinuity. In the far field, on the other hand, shear production and pressure diffusion dominate, adjusting themselves along with IBL development. Besides, over the rougher surface, turbulent diffusion is augmented and energy-carrying scales shift to higher-frequency regime. The shift accomplished substantially in the RSL, indicating the limitation of modeling surface roughness.

Micrometeorological conditions in urban areas are strongly influenced by the ground-level heating that induces stratifications of air temperature and alters the dynamics of wind flows. Here we numerically investigate how different levels of unstable stratification caused by ground heating affect the turbulent flows over an idealized urban area based on the finite-volume library OpenFOAM, with a focus on the turbulent flows above the canopy layer. In this study, aligned cubical roughness elements are explicitly resolved with a plan area fraction of 0.25 while five Richardson numbers are evaluated to systematically analyze the dynamical responses. Profiles of turbulent kinetic energy and Reynolds stresses reveal that turbulent flows around the altitude of z/h=3 are strongly affected by the presence of unstable stratifications, where z and h​ are the height of the vertical position and roughness elements respectively. As the stratification level increases, skewness and kurtosis of streamwise and vertical fluctuating velocities are enhanced, implying a modulation of thermal stratifications to the turbulent organized structures above the building arrays. Quadrant analysis reveals that ejection events are the most sensitive quadrant components to the increasing of unstable stratification levels, whose fractional occurrence and contribution together with the average strength are intensified. Finally, we study the two-point correlation of fluctuating velocity under different stratification levels, in which the sweep-centered and ejection-centered spatial structures are discussed respectively.

The effects of inhomogeneous urban turbulence on aerosol dynamical processes (i.e. deposition, coagulation and condensation) are examined using large-eddy simulation. A comprehensive urban CFD model (PALM) with a sectional aerosol module (SALSA) has been modified to accommodate different sources and applied to a street canyon. The sensitivity of aerosol dynamical processes to source type, emission size distribution, source location, and flow structure is considered. Coagulation is maximised inside coherent vortices and away from the source region, while deposition is maximised inside corner vortices near the ground. The effects of background concentration, source strength, and local heating are also assessed. A simple multiple-box model that captures the key fluid dynamical and microphysical processes is discussed.

In this study, the statistical relationship between tropical upper-tropospheric troughs (TUTTs) and the initiation of summertime tropical-depression type disturbances (TDDs) over the western and central North Pacific is investigated.  By applying a spatiotemporal filter to the 34-year record of brightness temperature and using JRA-55 reanalysis products, TDD-initiation events are detected and classified as trough-related (TR) or non-trough-related (non-TR).  Contrary to the conventional assumption that TDDs originate primarily in the lower-troposphere, our results reveal that approximately 30\% of TDDs in the 10˚N-20˚N latitude ranges are generated under the influence of a TUTT.  Lead-lag composite analysis of both TR- and non-TR-TDDs clarifies that TR-TDDs occur under relatively dry and less convergent large-scale conditions in the lower-troposphere, which is unfavorable to TDD initiation.  The horizontal and vertical structure of the wave activity flux reveals southward and downward propagation of wave energy in the upper troposphere that converges at the mid-troposphere around the region where TR-TDDs occur, suggesting the existence of extratropical forcing.  Further, the roles of dynamic forcings associated with the TUTT and diabatic heating associated with cumulus convection on the TR-TDD-initiation are analyzed using the quasi-geostrophic omega equation. The results revealed that moistening in the mid-to-upper troposphere takes place in association with the sustained dynamical ascent at the southeast side of the TUTT, which precedes the occurrence of deep convective heating.  Through the moistening process associated with TUTTs, low-level perturbations, otherwise under unfavorable conditions, can become deeper and more organized, leading to the initiation of synoptic-scale TDDs. 

Three series of 3D cloud-resolving simulations of tropical atmosphere driven by prescribed sea surface temperature (SST) are performed using the System for Atmospheric Modeling (SAM) to investigate how large-scale circulations interact with tropical convection and precipitation. The zonal distribution of SST is in the shape of Gaussian function with warm center in the middle of domain.  Climatology precipitation are classified into 3 types: type-1 has a central single peak; type-2 has symmetric double peaks; type-3 has a single peak off the center. The location of convection region oscillates around the heating center periodically and the convection tend to become more organized as SST gradient increases. For each case, composite analysis is carried out to extract the states when convections develop in the center and off the center. It shows precipitation change from type-1 to type-2 when SST amplitude increase to 10K in the simulations with smaller domain. Precipitation change from type-1 to type-3 in simulations with quadrupled domain length as SST gradient increases. In type-1 cases, evaporation and precipitation are weaker when convection deviate from the center. In type-2 cases, a local high pressure above the BL convergence center suppresses convection and precipitation is more active on either flank of heating center. Deviate convection is unstable due to larger consumption of vapor than generation of it. In type-3 cases, deviation of surface convergence induces larger evaporation due to a positive feedback of Wind Induced Surface Heat Exchange (WISHE), which maintains convection off the center.

Geostationary Environmental Monitoring Spectrometer (GEMS) is a hyperspectral UV-vis sensor on board the Geostationary Korea Multi-Purpose Satellite-2B (GEO-KOMPSAT-2B) for monitoring air quality in East Asia, which had launched in February 2020. The spectral coverage of GEMS is 300–500 nm at 0.6 nm full-width at half-maximum (FWHM). Since an enormous amount of data from the hyperspectral sensor (e.g., GEMS, providing more than one thousand wavelengths at a pixel), channel selection studies have been conducted for each instrument and forecasting model for efficient data transmission storage, and assimilation. In this study, following the mission of the GEMS, we perform a preliminary test of channel selection using simulated data containing the characteristics of the GEMS, to aim for the improvement of state vectors of trace gases. To find the optimal channel set, we employ a genetic algorithm that uses the evolution process concept through natural selection and mutation mechanism. Degree of freedom for the signal (DFS) is used as a figure of merit to determine the optimal channel set. The results will propose several optimal channel sets according to the number of channels and discuss the efficiency of each set.

The ensemble data assimilation system expresses the model uncertainties by ensemble spread, that is, the standard deviation of ensemble background error covariance. The ensemble spread generally suffers from underestimation problems due to the limited ensemble size, sampling errors, model errors, etc. To solve this problem in terms of the model errors, recent studies proposed stochastic perturbation schemes to increase the ensemble spreads by adding random forcing to model tendencies or variables. In this study, we focus on the near-surface uncertainties which are affected by interactions between the surface and atmospheric processes. Although the surface variables are crucial to provide various fluxes as the lower boundary condition to the atmosphere, their uncertainties were not much addressed yet. We employed the stochastically perturbed parameterization scheme (SPPS) for the surface variables within the Weather Research and Forecasting (WRF) ensemble system. As a testbed experiment with the newly-developed WRF ensemble-SPPS system, we perturbed soil temperature and soil moisture in the Noah Land Surface Model, and sea surface temperature to improve the lower-atmospheric performance over the land and ocean, respectively. To determine the optimal random forcing parameters used in perturbation, we employed a global optimization algorithm — the micro-genetic algorithm (micro-GA). Our results depicted positive impacts on the ensemble spread.

Low-level jets (LLJs) act as an important carrier of moisture transport that significantly affect heavy rainfall. In South China, three branches of LLJs coexist including a boundary layer jet in Beibu Gulf (BLJ-BG), a boundary layer jet in South China Sea (BLJ-SCS) and a synoptic-system-related LLJ in South China (SLLJ-SC). So far, the roles in moisture transports of those LLJs are not clear. In the present study, based on a Lagrangian backward tracking model and high-resolution ERA5 reanalysis, moisture sources and sinks owing to the three LLJs are investigated. The differences and similarities in moisture transports of LLJs events and Non-LLJ events are further studied. In general, the moisture transported by the LLJs mainly comes from the Pacific Ocean, South China Sea, the Bay of Bengal and eastern China. As the pre-rainy season in South China progresses (from April to June), the moisture paths gradually decrease from the east and north side, while established from the west side. The moisture trajectories of BLJ-BG tend to pass over Indo-china Peninsula, but those of BLJ-SCS tend to go around it. Moisture tracks of BLJs from the east side are more in May than in June, while those of SLLJs are the opposite. Moisture along the trajectories increases from April to June. The moisture sources and sinks are further identified based on water vapor changes in the moisture trajectories. With the advance of the monsoon, moisture sources of BLJ-BG events shift from the South China Sea to Bay of Bengal, while local moisture transports for BLJ-SCS. Comparing moisture paths and relevant moisture changes in Non-LLJs with LLJs, the moisture sources and sinks in Non-LLJ events in May are different from those in LLJ events, but similar in June when monsoon outbreaks. It implies that moisture transports mainly rely on LLJs before monsoon outbreak. 

This study investigates the satellite-observed convective systems during 14 events of Northwest Pacific Monsoon Trough onset between 2001 and 2016, focusing on their lifetimes and contribution to regional precipitation. The convective systems are identified as the spatially contiguous precipitation objects in the GPM-IMERG (0.1-km resolution, 30-min sampling), and their lifetimes are determined by the iterative tracking algorithm, which examines the area overlapped by systems in the consecutive time steps. Using the monsoon trough index based on relative vorticity at 850 hPa, the 20 days before and after the establishment dates of monsoon trough each year are defined as the pre- and post-onset periods, respectively. The probability function of system lifetimes reveals that the occurrence of the long-lived systems (> 2 day) increases by about two times after the monsoon trough onset, while that of the short-lived systems (≤2 day in time) remains similar. For the spatial distribution of occurrence and precipitation contribution, the systems existing less than one day have most hotspots over land and coastal regions, and the system persisting 1-2 days tend to develop on the ocean regions; the spatial distribution does not show significant changes between pre- and post-onset periods. The hotspots of long-lived systems shifts from coastal regions of East Asia islands in the pre-onset period to the northwest pacific in the post-onset period, and their contribution to precipitation significantly increases from 26.8% to 42.5%. The environment favors the convection organization, such as sea surface temperature, column water vapor, and lower-tropospheric wind shear will be examined and discussed. The current results potentially provide a useful metrics to evaluate the convection lifecycle simulated in the cloud-permitting atmospheric models.

To improve the predictability of weather/climate models, a prediction system capable of simulating the land surface-atmosphere interaction is essential. In land surface models (LSMs), the parameter values are applied differently depending on the land surface type. Previous studies reported that the Noah LSM underestimated the snow-related variables such as snow albedo, snow depth, and snow cover, compared to actual observations. In this study, among the variables of Noah LSM, we optimize several parameters, related to snow albedo, using the genetic algorithm (GA) and satellite (MODIS) data: the parameters to be optimized include 1) the threshold value of the amount of snow with full coverage, 2) the distribution shape coefficient related to the maximum albedo of new snowfall, and 3) the maximum albedo coefficient. We also propose the MODIS data processing method to extract representative snow albedo values, rather than the point (pixel) values, for different land cover types in a 10 km by 10 km area around a model gridpoint ⸺ the representative values are used to calculate the fitness function in GA. It turns out that the snow albedo simulation by Noah LSM has improved with the parameter values optimized using the representative values, compared with those optimized using the point values. We expect to see further improvement in the weather/climate simluations using the coupled land surface-atmosphere model (e.g., WRF-Noah LSM) by implementing the optimized parameter values related to snow albedo. 

Aerosol cloud radiation interaction is one of the foremost uncertainties in our radiative calculation studies. Among the many parameters which affect the clouds, aerosol particle size remains the difficult to parameterize. Here we try to understand the aerosol particle size influenced changes in the cloud radiative forcing using satellite retrieved data. Indo Gangetic Plain (IGP) known for the increased pollution episode has been chosen for the study. Size resolved aerosol optical depth (AOD) has been retrieved from Level 3 NASA Terra based Multiangle Imaging Spectro Radiometer (MISR) for the IGP region from 2001 to 2017.  Shortwave and Longwave radiation data at the surface and Top of the Atmosphere (TOA) for the same period have also been retrieved from the Clouds and the Earth’s Radiant Energy System (CERES) dataset. AOD for different size ranges small (radius <0.35 μm) and large particles (radius > 0.7 μm) is retrieved at a spatial resolution of 0.5º x 0.5º under the green spectral band from MISR. Shortwave Cloud Radiative Forcing (SWCRF), LongWave Cloud Radiative Forcing (LWCRF) and NET Cloud Radiative Forcing (NETCRF) at surface and TOA have been calculated using the CERES datasets. We have analysed 17 years of data for the Pre-Monsoon season (MAM) at the study area. Co-located grids of two parameters have been used for the percentile correlation studies in order to avoid the mixing up of data points.  SWCRF, LWCRF and NETCRF at the surface and TOA has been analysed as function of AOD under small and large size ranges. Our analysis has revealed a size sensitive difference in CRF. AOD-Small revealed a better association with CRF than AOD- Large and will be discussed in detail in the paper.

In remote sensing, bad pixels are known as defected pixels occurred during the calibration process converting measured digital counts to radiances. To replace the bad pixels in the process, measurements from the neighbors on a two-dimensional detector are mostly used for interpolating the erroneous pixels. When the bad pixel area is too large, however, it tends to make high interpolation error especially for a scene having high spatial variability, such as cloud edges. A similar issue has been raised during the in-orbit test of Geostationary Environment Monitoring Spectrometer (GEMS), because of bad pixels occurred in the form of a cluster which was found during the on-ground test. As GEMS observes the Earth by moving a scan mirror from east to west to cover the full field of regard (FOR) every hour during daytime, a horizontal discontinuity line appears in radiance data. It significantly affects to the retrieval of some cloud and aerosol properties as the affected spectral range is overlapped with the O₂-O₂ absorption lines. In order to reduce the effect caused by bad pixel error in the retrieval process, we propose a new method to fill in the missing radiances using spectral relations between radiances. We compare multi-variate linear regression and artificial neural network for estimating linear and non-linear relations between radiances at different wavelengths. The measured normal spectra adjacent to the cluster of bad pixels are used by dividing each spectrum to the input and output radiances. The output radiances are observed at the same wavelengths of bad pixels (484-491 nm). Results show that the multi-variate linear regression shows better results with the prediction error of 0.5%, even with nonlinearity caused by atmospheric interactions. For the replacement of bad pixels, we will test further whether the regression could show similar results at strong absorption lines.  

In urban areas, traffic emission confined by complex buildings results in a high spatial variability in fine particulate matters (PM_2.5). This study investigated the effect of buildings on PM_2.5 in the New Taipei City area by applying a street-canyon model, called the Graz Lagrangian Model, to simulate three-dimensional pollution distribution in 5-m resolution. A detailed building-height dataset was used to construct the urban topography. Traffic flow derived from the automatic traffic analysis system and automobile emission factors from the Taiwan Emission Data System were applied to estimate the PM_2.5 emissions in the street canyons. The model performance was validated using three-dimensional research-grade observations, which also provided the background PM_2.5 concentration. The correlation coefficient, mean fractional bias, and mean fractional error of the simulation against observation are 0.65±0.07, −0.16±0.17%, and 0.30±0.07%, respectively.  The simulations reveal that urban PM_2.5 due to traffic emission typically declines as the wind speed increases. Regardless of the wind direction, the concentration of pollutant increments are more than double between the different conditions of the wind speed higher than 1.5 m/s and lower than 1.5 m/s. For both step-down canyon and step-up canyon, the higher pollutant increments are on the windward side, especially at low wind speeds. However, the pollutant increment distributes evenly when wind speed is 0.5 m/s for the step-up canyon. Furthermore, PM_2.5 increment tends to exist at the entrance to the alley rather than the center of the road. PM_2.5 increment also decreases vertically, with a 5^th-floor to 1^st-floor ratio of about 0.35±0.12 near the center of main road regardless of the wind speeds. However, such ratio in the allies can differ significantly depending on the front-building configuration and wind speed. These results demonstrate highly variant exposure levels within the urban street canyon environment.

A severe afternoon thunderstorm (ATS) system developed within the Taipei Basin on 14 June 2015, which produced intense rainfall (with a rainfall rate of 131 mm h ⁻¹) and urban-scale flooding. A control simulation (CNTL) using the Weather Research and Forecasting (WRF) model with the horizontal grid size nested down to 500 m was performed to capture reasonably well the onset of the sea breeze, the merger of convective cells, and the development and evolution of the afternoon thunderstorm system. A mid-level dry layer with the layer-mean relative humidity of 35% occurred at 500–700 hPa for the Banchiao (near Taipei) sounding at 00 UTC (08 LST). A set of four numerical sensitivity experiments with the increase or decrease of mid-level relative humidity of 10% and 20% were conducted, and simulation results were compared with those from the CNTL. Sensitivity experiments showed that a drier layer at middle levels would result in stronger cold pool, more organized and deeper convection, stronger updraft, more graupel and snow particles, stronger net latent heating above the melting level, and much larger area of the potential flooding region [> 40 mm (30 min)^-1]. Statistics from 200 backward trajectories indicated that 37% of air parcels within the cold pool came from the middle levels (3–6 km height). Domain-accumulated rainfall was not positively related to the potential of flooding. Finally, numerical experiments found a nonlinear response of simulated convection intensity to the mid-level moisture content in the environment.

Springtime precipitation over Southern China is largely associated with mesoscale convective systems (MCSs), but the climatological characteristics of springtime MCSs over Southern China are still unclear. We developed an objective MCS diagnosis algorithm using RADAR reflectivity images and analyzed the occurrences, structural characteristics, tracks, and spatiotemporal patterns of springtime MCSs over Southern China during 2009 to 2019. We found that MCS occurrences increased from 103±82 hr in March to 337±225 hr in May. The interannual variation of MCSs occurrences was strongly correlated to the interannual variation of springtime precipitation. Most MCSs over Southern China were linear in morphology and moved from west to east. Springtime MCSs over Southern China occurred most frequently in central-south of Guangdong Province, followed by Guangxi Province. In March, MCSs were more frequent in the daytime; in April and May MCSs were more likely to occur in the afternoon and nighttime, with the peak occurrences in the early morning. Moreover, the correlations between surface PM_2.5 or PM₁₀ concentrations and MCSs from 2013 to 2019 showed that MCSs were more likely to occur over Southern China in March under polluted conditions, while in April MCS occurrences were significantly reduced under polluted conditions. The negative correlation between surface particulate matter concentrations and MCS occurrences in April was consistent with our previous study, which showed that anthropogenic aerosols may suppress MCS occurrences by stabilizing the lower atmosphere by exerting direct and indirect radiative forcings.

In this study, an ensemble-based sensitivity analysis (ESA) on the extreme-rainfall event along the northern coast of Taiwan on 2 June 2017 in the Mei-yu season is performed using the results from 45 forecast members with a grid size of 2.5-5 km. An quasi-stationary rainband associated with the front produced localized rainfall up to 635 mm in 12 h (0000-1200 LST 2 June), causing serious flooding and inundation near the northern tip of Taiwan. With a relatively large spread (i.e., standard deviation, or SD), small ensemble mean (~130 mm), and low probability of heavy rainfall in northern Taiwan, the ensemble indicates a lower predictability there, compared to the topographic rainfall over the mountains. However, the ESA allows for identification of several contributing factors to heavy rainfall in northern Taiwan in a quantitative manner as given below.With their impact given in change of (areal-mean) rainfall amount per one SD increase, these factors include: (1) surface frontal position and moving speed (-16.00 mm per 5 km h^-1), (2) position of 700-hPa wind-shift line (+12.59 mm per 0.4° latitude), (3) environmental moisture amount near the surface front (+11.73 mm per 0.92 g kg^-1 in mixing ratio), (4) timing and location of frontal low-pressure disturbance (+11.03 mm per 1.38° longitude), and (5) frontal intensity (+9.58 mm per 3 K across 0.5° in contrast of equivalent potential temperature). While many of the factors identified are interconnected, they tend to increase the local rainfall through enhanced near-surface convergence and its duration along the northwestern coast of Taiwan (over the area immediately upstream).

In this study, a total of 24 idealized CReSS simulations are performed to study the impacts of different orientation moving speed of a straight-line Meiyu front on the rainfall over northern Taiwan, with (horizontally) uniform and steady southwesterly flow ahead of the front. Using geostrophic wind relationship, three-dimensional (3-D) idealized fields are constructed from gridded analyses to represent pre-frontal and post-frontal environments, and the two are combined to provide initial and boundary conditions (IC/BCs) based on a prescribed slope, orientation, and moving speed of the front. With eight different orientation angles, every 10° from -20° (WNW-ESE) to 0° (E-W) then to +50° (NE-SW), and three different moving speed (from fast, medium, to slow, at 0.5°, 0.375°, and 0.25° per 3 h, respectively), these fields are then fed into the Cloud-Resolving Storm Simulator (CReSS) for high-resolution simulations at a horizontal grid size of 2 km. At the same moving speed, our results indicate that the rainband at the leading edge of the advancing Meiyu front would yield a higher rain rate when the front’s orientation is near 30°, but a longer rainfall duration (over northern Taiwan) toward -20°. On the other hand, with the same orientation, the faster-moving fronts give a higher rainfall intensity but a shorter duration. As a result, two types of higher rainfall total amount over northern Taiwan are found: (1) a slow-moving front with an alignment near 0°, and this gives the highest accumulation among all our experiments and is close to the actual event (the one on 2 June 2017) upon which the environmental fields are based; and (2) a faster front with an orientation near 20°-30°, which produces a few hours of rainfall with shorter duration but higher intensity. 

A stationary line-shaped precipitation system (SLPS), which is one of mesoscale convective systems (MCSs), is a typical heavy-rain-producing weather system formed during a warm season in Japan. In this study, two SLPS cases were studied, using high-resolution numerical experiments: 1 September 2015 over the Kinki district and 7 July 2018 over the Tokai district, respectively. The first half of the study investigated the SLPS event over the Kinki district. In the numerical experiments, the SLPS was formed in a low-level convergence zone of the westerly with the warm and humid south-southwesterly from the Kii Channel. New convective cells generated over the north of Awaji Island and are propagated northeastward by the middle-level southwesterly. This cell formation process was repeated and resulted in the formation of the SLPS. The sensitivity experiments for the orography around the occurrence area of the SLPS indicated that the orography was not an essential factor for the formation of the SLPS in this event. The latter part of this study investigates the SLPS event over the Tokai district. The simulation showed that a warm and humid southerly was present to the south of the Baiu front. The humid air was lifted by the windward slopes of the highlands and convective cells developed. They were moved with developing to the northeast by the middle-level southwesterly. During the SLPS formation, secondary convective lines were generated on the southeast side of SLPS. This process was maintained due to environmental conditions such as stagnation of the Baiu front. In the sensitivity experiments of the orography of the highlands, this study found that the total precipitation amounts of the SLPS were less than a half of that of the simulation experiment. This indicates that the topography triggered and reinforced convective cells, which results in the formation and rainfall enhancement of SLPS.

A Stationary Line-shaped Precipitation System (SLPS) is one of the primary sources of localized heavy rainfall. In Shikoku island, a SLPS oriented in a south to north direction that generating from the Muroto Peninsula (mountainous area of the southeastern tip of Shikoku Island) is frequently observed (hereafter, referred to as the Muroto Line). The Muroto Line often stagnates for equal or more than several hours and causes heavy rainfall. In this study, we analyze 17 Muroto Lines that brought about heavy rain equal or more than 2 hours to investigate their formation mechanism and occurrence environment by using observation data from 2004 to 2020. We find that wind direction below 500 m is between easterly and south-southeasterly, and veers to southerly around 3 km. The Muroto Lines generate around the ridge of northern Muroto Peninsula with a height of 700 – 1100 m, which is higher than the LFC (~700 m) observed in these events. They indicate that such vertical wind structure and orographic forcing should be essential to form the Muroto Line. In addition, the Muroto Lines can be divided into two types by their dominant echo-top height except one case. The "Low-top” cases (dominant echo-top height is lower than 10 km) occur under abundant water vapor flux (250 – 600 g m^-2 s^-1) around the surface and wind speed below 6 km is almost stronger than 15 m s^-1. In addition, dry air (RH < 30%) exists above a height of 500 hPa in several cases. On the other hand, the "Hight-top" (that is higher than 10 km) cases are characterized by relatively low water vapor flux (150 – 350 g m^-2 s^-1), weaker wind speed (< 15 m s^-1 in most cases), and wetter (RH > 40%) above a height of 500 hPa. 

In this study, we investigated the impact of the scale-aware convective parameterization scheme (CPS) on a heavy precipitation simulation across the gray zone using the Weather Research and Forecasting (WRF) model. We selected Kain-Fritsch (KF) and Multi-scale Kain-Fritsch (MSKF) schemes as non-scale-aware and scale-aware convective parameterization schemes, respectively. The MSKF scheme uses the scale-aware parameter, which modulates the convective available potential energy (CAPE) timescale and entrainment process in the KF scheme as a function of the horizontal grid spacing.  This study showed that simulation of convection only with the microphysics parameterization scheme (MPS) possibly causes an unreasonable overestimation of precipitation in the gray zone largely because atmospheric instability is not adequately reduced. On the other hand, the CPS without scale awareness in the gray zone exaggerated the convection and caused distortion of synoptic fields that led to the erroneous simulation of heavy precipitation at high resolution. In contrast, the MSKF scheme with scale-awareness improved simulated precipitation due to a smooth transition from the role of CPS to that of MPS by removing atmospheric instability in the gray zone. In addition, the sensitivity experiments showed that the shorter CAPE timescale and decreased entrainment process respectively resulted in the fast development and exaggeration of convective activities. These parameters modulated by the scale-aware MSKF scheme can play an important role in the balanced effect between the CPS and MPS in the gray zone by controlling the entrainment rate and CAPE timescale.

Correct representation of convective storms is key for Numerical Weather Prediction (NWP) models to perform satisfactorily over tropics, which is a challenging task. Challenges arises mostly due to the issues in predicting initiation of the storms, which gets further complicated due to storm’s interaction with its surrounding that often leads to either splitting of the original storm or causes it to merge with the other. Storms that undergo split and/or merge in their lifetime are termed as complex storms and constitute more than 50% of storms observed in the vicinity of Singapore. In in this study we use the Thunderstorm Identification Tracking Analysis and Nowcasting (TITAN) software (Dixon and Wiener, 1993) to track such complex storms in both model and observation (radar) for a few months. Storm tracks will then be used to derive the life cycle of a typical complex storm in observation and two model versions of SINGV (Huang et al., 2019): a convection-permitting at 1.5km horizontal resolution and a storm-resolving at 300m horizontal resolution. Focus will be on extracting relevant information to improve our understanding of storm’s behaviour before and after split (and merge), which will then be used to evaluate the two model versions to see what is gained by storm-resolving simulations.  

The aviation industry is heavily investing in electric vertical takeoff and landing (eVTOL) aircraft for aerial ridesharing across metropolitan areas to avoid frequent gridlock in surface transportation.  Urban environments, however, pose unique weather challenges that these emerging aerial operations will have to deal with.  This presentation will highlight the dynamic nature of urban environments and associated wind and turbulence hazards for Dallas, Texas, one of the cities considered for early eVTOL flight demonstrations.

A case of severe low-level windshear at the Hong Kong International Airport was investigated in this study using large eddy simulation (LES) with a spatial resolution of 40 m.  The computer simulation results were compared with the actual observations at the airport, including Doppler Light Detection And Ranging (LIDAR) systems and surface automatic weather stations.  The LES was found to be capable of reproducing the tiny vortices associated with the terrain-disrupted airflow.  Some interesting features of surface temperature, relative humidity and vertical velocity were also observed in the computer simulation.  The simulated and observed meteorological parameters had satisfactory comparison results, though the micro-scale variability of such parameters still could not be captured by the high resolution LES.  Finally, the computer-simulated headwind profiles appeared to have skills in capturing the low-level windshear reports from the pilots.  The study was novel in the use of the rather large domain for LES simulation and in the application to low-level windshear alerting.

For aviation safety it is necessary for the numerical weather prediction (NWP)-based forecast system to estimate turbulence intensity as a function of eddy dissipation rate (EDR), a standard turbulence metric by the International Civil Aviation Organization. Turbulence diagnostics derived from NWP model outputs are converted to the EDR-scale by using lognormal distributions of both model-based diagnostic and EDR observation. This method uses mean and standard deviations of the lognormal distributions of both turbulence diagnostics and observed EDR, which is called the lognormal mapping technique (LMT). In this study, low-level turbulence forecast system was developed using Korean Meteorological Administration Post Processing (KMAPP) that is statistically downscaled from Local Data Assimilation and Prediction System domain with 1 km of horizontal grid spacing to the KMAPP domain with 100 m over South Korea. Statistics (mean and standard deviation) of ten turbulence diagnostics were obtained from the KMAPP datasets in three months (Sep – Nov 2017), and statistics of observed EDR from the 3D sonic anemometer measurements equipped in the 300 m tower at Boseong Meteorological Observatory (BMO) for 1-month (Oct 2018). Low-level EDR forecasts from the KMAPP data (hereafter KMAPP-EDR) were calculated by the LMT, which can capture a detailed structure of downslope windstorm case on the lee-side of the eastern mountainous region in the Korean Peninsula. KMAPP-EDR was evaluated by directly comparing the time series of KMAPP-EDR and observed EDR at BMO in October 2018. It is found that KMAPP-EDR reasonably well predicts the diurnal variations in the observed EDR with mean absolute error and root mean squared error ranging from 0.08 m^-2/3 s^-1 to 0.1 m^-2/3 s^-1.Acknowledgement: This research is supported by the Korean Meteorological Administration Research and Development Program (KMI2020-01910), and by the National Research Foundation of Korea Research and Development Program (NRF- 2019R1I1A2A1060035).

Turbulence in the free atmosphere (z=3–30 km) is estimated in terms of thickness of turbulence layer (THTL) and eddy dissipation rate (ε), based on Thorpe method using high vertical-resolution radiosonde data for 6 years (January 2012–December 2017) in USA. Turbulence occurs more frequently in the troposphere than in the stratosphere, with log₁₀ε ranging from -4 to 0 (from -4 to -0.5) m² s^-3 in the troposphere (stratosphere). THTL is also much larger in the troposphere with the maximum of 4,000 m, than in the stratosphere with the maximum of 2,000 m. However, layer-mean of log₁₀ε is larger in the stratosphere, due to much smaller number of turbulence cases than in the troposphere. For better representation of layer-mean turbulence, a new quantity, named layer-mean effective ε (EE), is suggested in this study, by combining ε and THTL. EE is much larger in the troposphere than in the stratosphere, with the largest in z=6–9 km. Potential sources of the observed turbulence are examined by analyzing four turbulence indices: Brunt-Vaisala frequency, vertical wind shear (VWS), orographic gravity-wave drag (OGWD), and precipitation using ERA5 reanalysis. N² shows clear seasonal variations with large in DJF (JJA) and small in JJA (DJF) in z=3–15 (15–21) km. Strong VWS appears in eastern USA in DJF (JJA) in z=9–12 (12–18) km. Non-zero OGWD appears mostly in Rocky and Appalachian mountain regions, and its magnitude increases with height due to density reduction, while precipitation is concentrated in western coast and mid-eastern USA. The correlations between EE and turbulence indices show that (i) N² is negatively correlated while precipitation is positively correlated through all altitude ranges and most regions, (ii) VWS is positively correlated under strong static-stability conditions, and (iii) OGWD is positively correlated in western mountainous regions in z=15–21 km.

The goal of this work is to provide an overview of the microphysical measurements and a relationship between fog visibility (Vis) and eddy dissipation rate (EDR). Observations were collected during the C-FOG (Toward Improving Coastal Fog Prediction) field project in the fall of 2018. The project took place along eastern Canada’s (Nova Scotia, NS and Newfoundland, NL) coastlines and open water environments from August-October 2018, where environmental conditions play an important role for late-season fog formation. C-FOG is designed to advance understanding of liquid phase fog formation, development, and dissipation in coastal environments to improve fog predictability and monitoring. Vis, 3D wind speed (U_3d), and turbulence along coastlines are the most critical weather-related parameters affecting marine transportation and aviation operations. The numerical weather prediction (NWP) models have significant difficulties for Vis prediction due to small-scale dynamical processes and, for this reason, they need a simplified Vis parameterization applicable to various microphysical schemes. In our analysis, microphysical observations are summarized first and then they are, together with 3D-wind measurements, used for fog intensity (Vis) evaluation. Analysis suggests that detailed microphysical observations collected at the supersites and aboard the Research Vessel (R/V) Hugh R. Sharp are useful for developing microphysical parameterizations that focus on EDR. We find that increasing EDR can lead to fog dissipation quickly as indicated by increasing Vis. In this presentation, fog Vis prediction as a function of EDR over the coastal environments will be presented and challenges with measurement issues and scales will be emphasized.

Sea fog event over the Eastern Yellow Sea on 15–16 April 2012 was reproduced in a Weather Research and Forecasting (WRF) simulation with high-resolution to investigate the roles of physical processes and synoptic-scale flows on advection fog with sea surface warming. Initially, longwave radiative cooling (LRC) with negative sensible heat flux (SHF) due to the turbulence after sunset triggered a formation of cloud at the surface under the moist advection due to a southerly wind. This is a conventional type of advection fog. At night, continuous cooling due to longwave radiation and SHF near the surface modulated the change of the SHF from negative to positive, resulting in a drastic increase in the latent heat flux (LHF) that provided sufficient moisture at lower atmosphere (self-moistening). This is a favorable condition for advection fog with sea surface heating (ssH). Enhanced turbulent mixing driven by a buoyancy force increased the depth of the sea fog with a gradual rise in the marine atmospheric boundary layer (MABL) height, even at nighttime. In addition, cold advection with a prevailing northerly wind at the top of the MABL led to a drastic increase in turbulent mixing and the MABL height, which resulted in rapid growth of the height of sea fog due to vertical diffusion. After sunrise, shortwave radiative warming in the fog layers offsetting the LRC near the surface weakened thermal instability, which contributed to the reduction in the MABL height, even during the daytime. In addition, dry advection of northerly wind induced dissipation of the fog via evaporation. An additional sensitivity test of sea surface salinity showed weaker and shallower sea fog than the control due to the decrease in both the LHF and local self-moistening. This work was funded by the Korea Meteorological Administration Research and Development Program under Grant KMI2020-01910. 

A method to predict areas of possible aircraft icing based on model gridded output is presented. A complex parameter to describe meteorological conditions which contribute to aircraft icing is calculated at each point (i, j, k) of a three-dimensional grid in the form of I = p (q - q_min), where p is pressure, q is mixing ratio, q_min is a threshold value of q. It is assumed that phase transition of moisture is possible only if a value of q exceeds threshold value of q_min=aQ(t, p), where Q(t, p) is saturation mixing ratio at a temperature t and pressure p and a is a chosen threshold value of relative humidity. It is expected that icing occurs if temperature is not above zero Celsius then maximum amount of water vapor which is potentially possible for phase transition at pressure p does not exceed (1-a)Q(0, p). Considering that product of atmospheric pressure and maximum amount of water vapor which is possible for phase transition at zero Celsius is almost constant then maximum value of I can be determined for selected meteorological conditions. In this study, selected meteorological conditions (temperature interval and value of a) which contribute to icing are based on observational data, results of other studies and took into consideration uncertainties of model forecasts. Area of possible icing covers grid points where values of I exceed zero. Icing intensity is determined by a value of I; range of positive values of I is divided into intervals which correspond to light, moderate and severe icing. It is not necessary to define these intervals of I for each flight level individually. Described method is being tested on numerical results with 15-km horizontal grid spacing of the Weather Research and Forecasting (WRF) - Advanced Research WRF (WRF-ARW).

Model ensemble predictions on monthly, seasonal and longer timescales are found to contain only small signal-to-noise ratios, exhibit little predictability and are therefore unable to predict their own ensemble members in and around the North Atlantic sector. The signal-to-noise paradox arises when we then use the same ensembles to predict the observed evolution of climate.  Providing the ensemble size is large enough, paradoxically high skill is found for the North Atlantic and climate over surrounding land regions.
This talk summarises the latest evidence and hypotheses for the cause of this unsolved problem. We show that it is not a simple matter of predictions containing too much variability. Instead, the predictable variance in climate models appears to be too weak by a factor of two or more. We also show that this is common to different climate models and common to different timescales from subseasonal to multidecadal.  Finally we discuss some of the implications of this intriguing problem.

It’s difficult to detect improvements in the Madden-Julian Oscillation (MJO) in state-of-the-art coupled climate models because of substantial internal variability in the system. For example the impact of climate model resolution on the fidelity of the MJO is unclear, even when three 65-year coupled simulations are available for each version of the model: the variability between ensemble members is large compared to the difference between resolutions. To proceed we need to quantify the dependency of the MJO on low frequency drivers. This enables us regress these dependencies out, and reveal underlying improvements. In this presentation I discuss a probabilistic approach for quantifying the impact of low frequency drivers on the MJO. We model the MJO using a Markov Chain, and evaluate the dependency of the observed transition probabilities on drivers such as ENSO. We then apply this approach to simulations from the EC-Earth climate model, to assess the impact of resolution and model parametrisation schemes on the representation of the MJO.

El Niño-Southern Oscillation (ENSO) influences the Tropical North Atlantic (TNA), leading to anomalous sea surface temperatures (SST) that peak in boreal spring. Various mechanisms have been invoked to explain this teleconnection, including tropical and extratropical pathways, contributing to anomalous trade winds and static stability over the TNA region. The TNA region is of great interest, as the time lag in the influence of ENSO on the TNA may provide predictability in the Atlantic region beyond the ENSO teleconnections during peak ENSO season. However, the overall linearity with respect to ENSO’s strength and phase for the ENSO-TNA teleconnection remains unclear. We use reanalysis data to confirm that the SST anomaly (SSTA) in the TNA exhibits nonlinearity with respect to Pacific SSTA, as further increases in El Niño magnitudes cease to create further increases of the TNA SSTA. We create novel indices to further show that the tropical pathway to the TNA is more linear than the extratropical pathway. The North Atlantic Oscillation (NAO) and location of the tropical Pacific SSTA further modulate the extratropical pathway, but this modulation does not fully explain the nonlinearity in TNA SSTA. We confirm the results by using phase 5 of the Coupled Model Intercomparison Project (CMIP5). As neither extratropical nor tropical pathways can fully explain the nonlinearity, external factors are likely at play. Further analysis shows that the TNA SSTA is highly influenced by the preconditioning of the tropical Atlantic SSTs. This preconditioning is further associated with the NAO through the North Atlantic SST-tripole pattern. Overall, our results show that the ENSO-TNA teleconnection is nonlinear, helping to clarify relevant external factors for the overall TNA SSTA impact, thus allowing for better TNA SSTA predictions following future ENSO events. 

^{Changes in the activity of summer high-temperature extremes (HTEs) over many regions are linked with El Niño-Southern Oscillation. However, how the HTEs respond to different flavors of El Niño remains unclear. Using observational analyses, we show that summer HTEs in China during eastern Pacific (EP) and central Pacific (CP) types of El Niño exhibit different spatial patterns. EP (CP) El Niño is significantly correlated with summer HTE frequency over South China (Yangtze River valley, YRV). These relationships can be attributed to the modulation of the western North Pacific subtropical high (WNPSH) and East Asian jet stream. EP (CP) El Niño events are accompanied by westward (eastward) extension of the WNPSH, and equatorward (poleward) movement of the jet stream over East Asia (Pacific). During EP (CP) El Niño episodes, the westward (eastward) and equatorward (poleward) displacements of upper-tropospheric anticyclone over the western North Pacific coincide with subsidence over South China (YRV).}

Here we provide the first presentation of a 100-member large ensemble suite of calculations with the Community Earth system model version 2 (CESM2).  The runs are performed with historical/SSP3-7.0 forcing over 1850-2100, with the initialization procedure including a combination of micro- and macro-perturbations. The study is motivated by the question of how climate variability will evolve as a forced response under sustained emissions over the 21st century.  While much attention has been focused on how particular internal modes of variability such as the El Niño-Southern Oscillation (ENSO) and the Madden-Julian Oscillation (MJO) respond to climate change, our results shows that greenhouse warming perturbs disparate components of the climate system across a broad range of timescales.  Analysis of power spectra of a variety of variables throughout the Earth system often reveal changes across the entire spectrum in response to anthropogenic forcing.  This includes but is not limited to precipitation in the equatorial Pacific over a broad range of timescales, surface air temperature over the Arctic, the strength of the Atlantic Meridional Overturning Circulation (AMOC), and surface chlorophyll concentrations over the subpolar North Atlantic and the Southern Ocean. Analyses of changes in the atmospheric teleconnection strength indicate that ENSO is not the principal driver in changes in interannual variability in precipitation over the extra-tropics over the 21st century. Additionally, it is found that there are significant forced phenological changes to the growing season length of terrestrial ecosystems in the Northern Hemisphere over the 21st century. In summary, the CESM2 large ensemble reveals that changes in natural Earth system variability are ubiquitous, and are manifested in changes in amplitude, frequency, phase, teleconnections, and patterns. 

We study the forced response of the teleconnection between the El Niño–Southern Oscillation (ENSO) and the Indian summer monsoon (ISM) in a multi-model ensemble of initial conditional ensembles under historical forcing and future (RCP8.5 or SSP585) forcing scenarios. The forced response of the teleconnection, or a characteristic of it, is defined as the time dependence of a correlation coefficient evaluated over the ensemble. We consider the temporal variability of spatial averages and that with respect to dominant spatial modes in the sense of Maximal Covariance Analysis, Canonical Correlation Analysis and Empirical Orthogonal Function analysis across the ensemble. A further representation of the teleconnection that we define here takes the point of view of the predictability of the spatiotemporal variability of the Indian summer monsoon. On the one hand, we find it a typical feature under moderate, historical and early 21st century, global warming that the teleconnection is strengthening or nondecreasing. In a regression analysis framework we identify three drivers of the teleconnection change: (i) ENSO variability; (ii) ENSO-ISM coupling strength; (iii) noise strength; and attribute any strengthening of the teleconnection to (i)-(ii), namely, increasing ENSO variability and ENSO-ISM coupling strength. Under strong RCP8.5/SSP585 scenario forcing, on the other hand, the noise (iii) also increases in most models, and typically ENSO variability (i) starts to decline in a nonmonotonic, nonlinear fashion. However, because of a diversity in the response of the coupling strength (ii), and the delicate balance of the drivers' influence, there is more inter-model diversity in the forced change of the ENSO-ISM teleconnection under strong late 21st century forcing. 

The Madden Julian Oscillation (MJO) has profound impacts on the extratropical circulations and extreme weather events. Eastward propagation of MJO may be disrupted or completely blocked by the Maritime Continent (MC).We survey the global impact of this MJO disruption by the MC barrier effect. We categorize the MJO propagation based on whether the MJO precipitation over the MC during extended boreal winter seasons: successfully propagating across MC (MJO-C) and being blocked by MC (MJO-B). Compositing atmospheric circulation upon these two categories reveals that tropical precipitation anomalies of MJO-C are stronger and more coherent than that of MJO-B, while their phase speed and lifetime are comparable. The MJO-C and MJO-B cases excite distinct extratropical responses due to their diabatic heating in the deep tropics. Mid-latitude circulation displays stronger and long-lasting negative geopotential anomalies in the northern Pacific Ocean 5-14 days after phases 7-8 of MJO-C, but weak anomalies for MJO-B. The upper-level Rossby wave source excited by MJO-C is significantly stronger, especially during phases 4-5 due to the vigorous convection over eastern Indian Ocean and MC. The extratropical water vapor transport of MJO-B and MJO-C differs markedly after phase 3. The PNA (Pacific-North American) and NAO (North Atlantic Oscillation) teleconnection patterns both show significantly positive values about 5-20 days after phase 6 of MJO-C, but only marginally positive for MJO-B. The surface air temperature at high-latitude of MJO-B and MJO-C are also significantly different. Much weaker temperature anomalies are associated with MJO-B than MJO-C. We attribute these differences between MJO-C and MJO-B primarily to their sources and less to Rossby wave refraction by the mean flow.