Surface Roughness Length Research Articles

The Impact of Vertical Wind Shear on the Outcome of Interactions Between Squall Lines and Cities

AbstractThis study examines how organized lines of deep convective storms can be impacted by a large city with a prominent urban heat island and how low‐level environmental vertical wind shear may influence the outcomes of that interaction. Idealized simulations of squall lines are conducted in which a simplified urban area—defined by perturbations to skin temperature and surface roughness length—is placed in the center of an otherwise horizontally homogeneous domain. Simulations are conducted with three different magnitudes of low‐level vertical wind shear representing “weak,” “medium,” and “strong” shear environments. Results show that storms experience noticeable modification—including enhanced downwind precipitation—after interacting with a prominent urban heat island in all three shear configurations. However, the details of the modification are a function of the shear magnitude. In the medium and strong shear simulations, updrafts are enhanced via increased buoyancy after passing over a prominent urban heat island. In contrast, little updraft strengthening is evident in the weak‐shear simulations. Instead, near‐surface winds are enhanced downwind of the urban heat island due to a more prominent descending rear‐inflow jet.

Numerical simulation of thermal stratification in Lake Qiandaohu using an improved WRF-Lake model

Lake thermal stratification is important for regulating lake environments and ecosystems and is sensitive to climate change and human activity. However, numerical simulation of coupled hydrodynamics and heat transfer processes in deep lakes using one-dimensional lake models remains challenging because of the insufficient representation of key parameters. In this study, Lake Qiandaohu, a deep and warm monomictic reservoir, was used as an example to investigate thermal stratification via an improved parameterization scheme of the Weather Research and Forecast (WRF)-Lake. A comparison with in situ observations demonstrated that the default WRF-Lake model was able to simulate well the seasonal variation of the lake thermal structure. However, the simulations exhibited cold biases in lake surface water temperature (LSWT) throughout the year while generating weaker stratification in summer, thereby leading to an earlier cooling period in autumn. With an improved parameterization (i.e., via determination of initial lake water temperature profiles, light extinction coefficients, eddy diffusion coefficients and surface roughness lengths), the modified WRF-Lake model was able to better simulate LSWT and thermal stratification. Critically, employing realistic initial conditions for lake water temperature is essential for producing realistic hypolimnetic water temperatures. The use of time-dependent light extinction coefficients resulted in a deep thermocline and warm LSWT. Enlarging eddy diffusivity led to stronger mixing in summer and further influenced autumn cooling. The parameterized surface roughness lengths mitigated the excessive turbulent heat loss at the lake surface, improved the model performance in simulating LSWT, and generated a warm mixed layer. This study provides guidance on model parameterization for simulating the thermal structure of deep lakes and advances our understanding of the strength and revolution of lake thermal stratification under seasonal changes.

Observed and Parameterized Roughness Lengths for Momentum and Heat Over Rough Ice Surfaces

AbstractTurbulent heat fluxes, that is, the sensible heat flux and latent heat flux, are important sources/sinks of energy for surface melt over glaciers and ice sheets. Therefore, credible simulations of for example, future Greenland Ice Sheet mass loss need an accurate description of these fluxes. However, the parameterization of surface turbulent heat fluxes in climate models requires knowledge about the surface roughness lengths for momentum, heat and moisture, which are currently either unknown or tuned to indirect observations. In this study we take advantage of a large data set of eddy covariance observations acquired during multiple years and at multiple sites over the Greenland Ice Sheet. These in‐situ observations are used to develop an improved parameterization for the roughness length for momentum, and update the parameterization for the roughness lengths for heat and moisture over rough ice surfaces. The newly derived parameterizations are implemented in a surface energy balance model that is used to compute surface melt. Sensitivity experiments confirm the high sensitivity of surface melt to the chosen roughness length models. The new parameterization models the sensible heat flux to within 10 W m−2, and the cumulative ice ablation within 10% at three out of four sites. Two case studies demonstrate the important contribution of the turbulent heat fluxes to surface ablation. The presented roughness parameterizations can be implemented in climate models to improve the physical representation of surface roughness over rough snow and ice surfaces, which is expected to improve the modeled turbulent heat fluxes and thus surface melt.

도심 녹지공원이 지표흐름에 미치는 영향

In this study, the effects of surface temperature and roughness length of green-park in the urban area on the surface wind was analyzed using a computational fluid dynamics (CFD) model. Three-dimensional configurations and buildings for evaluation on the real urban were numerically constructed based on the geographic information system(GIS) data. The sensitivity experiments with differential heating of the surface temperature in the park area, it is shown that the overall flow pattern was wind direction similar to the inflow(NE). In the upwind area, the wind from the park flowed into the a cluster of building to form a circulating wind and in the downwind area, the flow was separated and the wind speed was decreased because many buildings were concentrated. The changes of roughness length influences a slight effects on the surface wind. In the case of simulation results compared to observations, the wind speed was somewhat overestimated (about +0.07 m/s). It is also suggested that the roughness length and differential heating of surface should be necessary for CFD modeling.

Influence of positive DC electric field on rime icing on the insulator and its experimental verification in the natural environment

AbstractThe electric field imposes a significant effect on ice accretion on the insulator. The impact of an AC electric field on insulator icing in an artificial climate chamber has been addressed. Compared with AC, DC electric field exerts an attraction on supercooled water droplets, which is more detrimental to the power system. Moreover, the icing on energised insulators is rarely studied in the natural environment. In the present research, an electrical, fluid coupled‐field simulation was carried out on the insulator. The force characteristics and motion deviation features of water droplets at different positive DC electric field strengths were studied. On this basis, rime icing on the insulator energised with different positive DC voltages was investigated in the field. Results show that the charged droplet shifts along the electric field line under a positive DC voltage. The deviation degree of the droplet increases with the increment of the voltage and droplet size and decreases with an increase in wind speed. As the voltage rises, the surface roughness, icing amount and icing length of rime ice‐covered insulator increase. Consequently, compared with that without energisation, the local collision coefficient, icing amount and icing length increase by 17.6%, 46.4% and 63.9%, respectively, under a 40 kV DC voltage.

Evaluation of the pattern of spray released from a moving multicopter.

Multicopters are used for releasing particulates seeds, fertilizer and spray. Their low cost and high manoeuvrability make them attractive for spraying in steep terrain and areas where overspray is undesirable. This article describes a model of multicopter wake and its influence on particulate dispersion, which is computationally economical compared to many computational fluid dynamics (CFD) approaches, yet retains reasonable accuracy. A model was successfully implemented in OpenFOAM. It features source terms for the rotor wash, Lagrangian particle tracking, an evaporation model, and a porous medium approach to model the effect of the ground vegetation. Predictions were validated against the field tests of Richardson et al. which used a DJI Agras MG-1 multicopter in three different flights with airspeeds of 3.2-4.9m s-1 , ground speeds of 2.1-2.9m s-1 and cross-wind speeds of 0.04-2.2m s-1 . The effective swath width (30% line separation) was predicted to within one standard deviation. Sensitivity to a rotor rotational speed, flight height, flight velocity, multicopter roll and yaw angles, surface roughness length, plant height and leaf density was checked. In all flight trials, the modelled swath was closest to the experimentally obtained swath when the surface roughness of the fetch was equal to 0.5m (bushes) and the rotational speed of all rotors was equal to 2475 rpm with 0.75R (0.2m) tall plant canopy (grass) introduced to the model. The model showed acceptable validity for flight velocities of ≤2.8-5m s-1 when flight parameters can be approximately estimated. © 2022 The Authors. Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Application and testing of a windblown dust module for a bare riverbed

To accurately predict the occurrence of riverbed dust events, a high-grid resolution (1 km2) riverbed dust module was constructed for this study. Although the module used the built-in CMAQ (Community Multiscale Air Quality model) dust module, the critical parameter values in the module were adjusted according to the experimental data for Taiwan. The results showed that the riverbed dust module can accurately predict the occurrence time of riverbed dust and thus improve the prediction of CMAQ's for the PM10 concentrations of downwind riverbed dust. The mean fractional bias and mean fractional error improved from −77 and 84% to −5 and 57%, respectively. Seven sensitivity-case simulations were performed to understand which parameters and formulas were important. The results are summarized as follows. (1) Selecting an appropriate horizontal sand flux estimation formula/vertical-to-horizontal dust flux ratio estimation formula/surface roughness length value is important because it can significantly affect dust emissions estimation. A local experimental value for the surface roughness length value is preferred. (2) For bare river land, it was acceptable to use the default value (= 1) for the surface roughness correction factor value. (3) In environments with high soil moisture, special attention should be paid to setting the threshold soil moisture value. (4) Different river locations may require different parameter values to be set within the module, and the ideal threshold friction velocity is an important parameter. (5) The estimation and use of gusts can improve the prediction of intense dust emissions; however, the prediction of hourly gusts must be improved.

Scaling and readiness of underlayers for high-numerical aperture extreme ultraviolet lithography

With the introduction of high-numerical aperture (NA) extreme ultraviolet (EUV) lithography, the thickness of layers in the lithographic stack will scale owing to reduced depth of focus and etch budget. The consequence of thinning down underlayers for EUV lithography has been scarcely investigated. In here, we assessed the readiness of nine state-of-the-art underlayers, spin-on and dry deposited, in thickness series down to 4 nm nominal. Preliminarily, the coating quality of these underlayers was evaluated. Thickness uniformity across 300 mm wafer ranged from about ±0.5 nm to < ± 0.05 nm depending on the coating technique employed. Surface roughness of the underlayers varied from as much as 0.63 nm to as low as 0.062 nm but was not impacted by thickness scaling. Film density and total surface energy varied by <10 % with thickness. EUV lithography of dense lines/spaces arrays of pitch 28 nm was carried out using a positive tone chemically amplified photoresist. Dose-to-size and exposure latitude changed by < ± 5 % when thickness of underlayer was decreased. Failure free process was at most 1 nm smaller for thinner underlayer than it was for the thinnest version of each type. The unbiased linewidth roughness increased consistently but limitedly (<5 % ) when thinner underlayers were used, mainly due to a reduction in the correlation length. By calculating the power spectral density of the blanket underlayer we can pinpoint this effect to a reduction of correlation length of the underlayer own surface roughness. Finally, Z-factor calculations demonstrated that overall photoresist performance depended more significantly on the specific underlayer type (±12.6 % ) than it did on underlayer thickness (±8 % ). All these results indicate that most of underlayers investigated had limited impact on the properties as well as the patterning performance when scaling in view of high NA EUV.

Examination of different wall jet and impinging jet concepts to produce large-scale downburst outflow

Thunderstorm downburst winds are a major cause of severe damage to buildings and other infrastructure. The initiative of the NSF-NHERI Wall of Wind (WOW) Experimental Facility to design and develop a downburst simulator was established to open new horizons for multi-hazard engineering research by extending the current capabilities of the facility to enable the simulation of non-synoptic winds. Five different downburst simulator designs have been tested in the 1:15 small-scale replica of the WOW to identify the optimal design. The design concepts tested herein considered both the 2-D impinging jet and the 2-D wall jet simulation methods. The basic design methodology consists of transforming the available atmospheric boundary layer (ABL) wind simulator into downburst winds by adding an external modification device to the exit of the flow management box. A flow characterization comparison among the five contending downburst simulators, along with comparisons to real downbursts and previous literature findings, has been conducted. The study on the effect of surface roughness length on the height of the peak wind velocity showed that the implementation of a 2-D plane wall jet enables large-scale outflows (higher peak velocity height) with high Reynold numbers, which is advantageous in terms of reducing scaling effects. In general, the current research work showed that four downburst simulation methods were suitable for adoption; however, only one downburst simulator was recommended based on the feasibility of construction in the facility. The chosen downburst simulator consisted of a two louver slat system near the bottom, with a blockage at the top. This configuration enables producing a large rolling vortex passing through the testing section, which would serve adequately in the further study of turbulent flow characterization and testing of larger scale test models.

Importance of Parameterizing Lake Surface and Internal Thermal Processes in WRF for Simulating Freeze Onset of an Alpine Deep Lake

AbstractThe alpine lakes widely distributed over the Tibetan Plateau (TP) are not only highly sensitive to climate but also regulate the regional climate. However, the lack of TP alpine lake observations limits the understanding of the lake‐atmosphere interactions. Here, we show the relative importance of parameterizing lake surface and internal thermal processes in describing the lake energy budget and lake‐atmosphere interactions. Based on the in situ observations at Lake Nam Co, a large and deep lake in the TP, we employed the coupled Weather Research and Forecasting with lake (WRF‐Lake) model to clarify their relative roles. Results show that the original model produces cold biases and too early timing of lake freeze onset. The adjustments of parameterizations of lake internal (temperature of maximum water density, extinction coefficient) and lake surface (roughness length) processes significantly improve the ability of the WRF‐Lake in simulating the lake thermal features and the freeze onset timing. The different parameterizations of lake internal thermal processes affect the onset timing of lake freeze mainly by altering the release of turbulent heat fluxes in summer, but the adjustments of lake surface roughness lengths change the turbulent heat fluxes during the unfrozen season from summer through early winter. Accordingly, the simulated lake freeze onset is much more sensitive to the surface roughness length schemes than to the internal thermal processes schemes. As the lake‐atmosphere interactions are very sensitive to the lake freeze onset, our results are critical for understanding and simulating the impact of TP lakes on the regional climate.

Performance analysis of GA, PSO and JA for determining the optimal parameters in friction drilling process

The friction drilling process, a technique for making sheet metal holes, differs from the conventional drilling process. Instead of cutting the workpiece, this process forms bush by using high heat generated at the contact region of the rotating conical tool and work material. The surface quality of the bush is not like a shiny and smooth surface due to the generation of high heat energy. Experimental investigation of friction drilling process has been done for optimization of surface roughness and bushing length by various researchers. This research work attempts application of meta-heuristics such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Jaya Algorithm (JA) and compares the results in order to evaluate their performances. GA, PSO and JA are applied to minimize the surface roughness of the friction drilled holes which depends on parameters viz rotational speed, friction angle and workpiece thickness. Also, these algorithms are applied to maximize bushing length by considering the significant parameters viz friction angle, friction contact area ratio, feed rate and spindle speed. The results are compared for surface roughness value and bushing length signal to noise ratio. The comparative analysis shows that the Jaya algorithm performs more efficiently with quick convergence to the solution compared to other two algorithms. The comparison of PSO with GA shows that, the PSO performs better in terms of convergence to the solution for both the objectives. Further, it has been found that JA shows the robustness in terms of consistency of results for different population sizes and number of iterations.

Effect of Coconut Oil-Based Cutting Fluid on Cutting Performance During Turning With Minimal Fluid Application

In the present work, an attempt was made to use vegetable oil as a cutting fluid during minimal fluid application to make the process more environmentally friendly. Coconut oil was selected as the vegetable oil in this study considering its availability, physical properties, and lubricating ability. The effect of newly formulated coconut oil-based emulsified cutting fluid on cutting performance during turning of hardened AISI 4340 steel with the minimal fluid application was investigated. It was observed that the coconut oil-based cutting fluid offered better cutting performance in terms of tool wear, surface roughness, cutting force, cutting temperature, and tool chip contact length when compared to mineral oil-based cutting fluid and raw coconut oil. It was also observed that the percentage of concentrate in the cutting fluid was to be maintained as 30% to achieve better cutting performance. The use of coconut oil-based cutting fluid holds promise as an environment friendly alternative for mineral oil-based cutting fluid.

Wind Drift in a Homogeneous Equilibrium Sea

Abstract A simple dynamical model is proposed for the near-surface drift current in a homogeneous, equilibrium sea. The momentum balance is formulated for a mass-weighted mean in curvilinear surface-conforming coordinates. Stokes drifts computed analytically for small wave slopes by this approach for inviscid linear sinusoidal and Pollard–Gerstner waves agree with the corresponding Lagrangian means, consistent with a mean momentum balance that determines mean parcel motion. A wave-modified mixing length model is proposed, with a depth-dependent eddy viscosity that depends on an effective ocean surface roughness length z0o, distinct from the atmospheric bulk-flux roughness length z0a, and an additional wave-correction factor ϕw. Kinematic Stokes drift profiles are computed for two sets of quasi-equilibrium sea states and are interpreted as mean wind drift profiles to provide calibration references for the model. A third calibration reference, for surface drift only, is provided by the traditional 3%-of-wind rule. For 10-m neutral wind U10N ≤ 20 m s−1, the empirical z0o ranges from 10−4 to 10 m, while ϕw ranges from 1.0 to 0.1. The model profiles show a shallow log-layer structure at the smaller wind speeds and a nearly uniform near-surface shear at the larger wind speeds. Surface velocities are oriented 10°–20° from downwind for U10N ≤ 10 m s−1 and 20°–35° from downwind for 10 ≤ U10N ≤ 20 m s−1. A small correction to the drag coefficient is implied. The model predictions show a basic consistency with several sets of previously published near-surface current measurements, but the comparison is not definitive.

Effect of various rebar types and crushed glass coating onto BFRP rebar on the bond strength to concrete

This study demonstrates the bond-slip behaviour of steel, BFRP, CFRP and GFRP rebars to concrete obtained from a series of pull-out tests. Results show that CFRP can achieve higher bond strength than steel, whereas BFRP attains half of the bond strength of steel. The bond strength was decreased by 23–28% in steel and BFRP due to an increase in bar diameter. However, the CFRP and GFRP showed a 73–78% reduction in bond strength due to the increase in bar diameter. Steel had the steepest slope in the bond-slip curve, followed by CFRP, BFRP and GFRP. Since BFRP attained optimum performance, the surface coating was applied onto BFRP using natural sand and recycled crushed glass to evaluate the roughness of FRP on the bond performance. Sand and glass coated BFRP demonstrated 37% and 75% higher bond strength compared to uncoated BFRP, while bond stiffness was increased by 14 and 11%, respectively. Compared to the sand coated BFRP, crushed glass coated BFRP exhibited approximately 20% and 10% more bond strength for the 6 mm and 10 mm diameter rebars, respectively. However, the glass coated BFRP exhibited more brittle behaviour compared to its sand coated counterparts. Based on the analytical results, surface roughness and embedment length are found to have a significant influence on the ultimate bond strength (p-value < 0.05) at a 5% level of significance. Additionally, the interaction effect of diameter*embedment and diameter*coating was found to have a significant effect on bond strength.

Evaluation of Eddy Covariance Footprint Models Through the Artificial Line Source Emission of Methane

AbstractThis paper evaluates the performance of footprint models through a line source (LS) system releasing methane (CH4) measured by an eddy covariance (EC) system in a rice‐wheat rotation agro‐ecosystem. Two footprint models namely Kormann and Meixner (KM) and Flux Footprint Prediction (FFP), are used. The “line source width (LW)” was introduced, which could highly impact the model estimation, leading to an error of ∼500% for the FFP and ∼200% for the KM. The surface roughness length (z0) also affects the FFP estimation, which is much weaker than the effect of LW with a 15% error. An overestimation of 217.8% is observed when the LS is close to the EC system, which reduces to 12.2% when the source stays distant. The analysis of different experimental configurations illustrated that the estimated emission calculated based on footprint model improves with increasing emission height, averaging period, and emission rate. The most accurate estimations are achieved by the configurations of canopy level (CL; 5.5% overestimation), 15 min averaging period (1% underestimation), and moderate emission (ME; 1.9% underestimation), respectively. For KM, the CL‐ME‐30 min setup leads to the best performance, while SL (soil level)‐ME‐15 min is the best configuration for the FFP estimation. A multivariate regression model is proposed to provide preliminary guidance for the footprint model application. The designed LS system is valuable for validating the performance of flux measurements in the specific ecosystems (e.g., grassland, sand land, and forest) as well as other greenhouse gases (e.g., N2O and NH3).

Seasonality of Continental Extratropical‐Cyclone Wind Speeds Over Northeastern North America

AbstractThis study investigates the seasonality of near‐surface wind speeds associated with extratropical cyclones (ETCs) over northeastern North America using a global reanalysis data set during 1979–2020. As opposed to most studies that emphasize winter storms, ETCs during the fall exhibit significantly stronger 10‐m winds over this region due to the slightly stronger continental cyclones and significantly weaker low‐level stability during that time of the year. Also, ETCs favor inland lakes and Hudson Bay during the low‐ice‐content fall season, leading to lower surface roughness. Combining these results, we derive simple linear regressions to predict the 10‐m wind speed given three variables: 850‐hPa wind speed, low‐level Richardson number, and surface roughness length. This formula captures the observed seasonality and serves as a valuable tool for cyclone near‐surface wind risk assessment.

Observations of boundary layer wind and turbulence of a landfalling tropical cyclone

This study investigates the atmospheric boundary layer structure based on multiple-level tower observations with a height of 350 m during the landfall of Super Typhoon Mangkhut (2018). Results show a layer of log wind profile outside of the radius of maximum wind speed with a height of 100 m or larger. The log layer height increases with the wind speed. The height of the constant flux layer reaches ~ 300 m for 10-m wind speeds less than 13 m s−1 while this height decreases with the wind speed. Momentum fluxes and turbulent kinetic energy increase with the wind speed at all vertical levels. The drag coefficient and surface roughness length estimated at the tower location have values of 7.3 × 10–3 and 0.09 m, respectively, which are independent of wind speed. The estimated vertical eddy diffusivity and mixing length increase with height up to ~ 160 m and then slowly decrease with height. The vertical eddy diffusivity increases with the wind speed while the vertical mixing length has no dependence on the wind speed. Comparing our results with previous work indicates that the vertical eddy diffusivity is larger over land than over ocean at a given wind speed range.

Surface Sublimation/Evaporation and Condensation/Deposition and Their Links to Westerlies During 2020 on the August‐One Glacier, the Semi‐Arid Qilian Mountains of Northeast Tibetan Plateau

AbstractTo identify the links between glacier sublimation/evaporation, condensation/deposition and westerlies, in situ meteorological data and surface roughness lengths during 2020 were collected on a mid‐latitude and westerlies‐controlled glacier in the northeast TP. Snow or ice sublimation rate was generally believed to be higher during the weather conditions with stronger winds and drier air, while our results showed that the strong westerlies in winter brought the driest and windiest conditions, but did not cause the most sublimation loss due to extremely low surface temperature. The weakening of westerlies intensity and the increased precipitation during summer suppressed the sublimation and evaporation for the low wind speed and high air humidity, and condensation or deposition was more possible during the nighttime of summer together with the radiative cooling on surface and enhancement in atmosphere stability, while annual condensation and deposition were only 1.3 and 14 mm, respectively, which were negligibly small compared to sublimation (115 mm). The westerlies in autumn and spring strengthened sublimation with relatively stronger winds and drier air than summer and higher temperature than winter. Sublimation was more intense in autumn (41 mm/0.4 mm d−1) and spring (35 mm/0.4 mm d−1) than winter (23 mm/0.2 mm d−1) and summer (17 mm/0.2 mm d−1). Therefore, westerlies may not be the most crucial driver of glacier sublimation as its effect may be restricted by low surface temperature during a very cold weather.

The WRF model and the diurnal cycle of orographically-influenced flow: A case study

The Weather Research and Forecasting (WRF) model is widely used in modelling for wind energy purposes. Previous research indicated that under some situations, the WRF model fails to properly represent the diurnal cycle in wind speed. Here we present a case study with a North Sea jet where the WRF-simulated wind diurnal cycle is compared with observations at the Østerild tall mast in Northern Jutland, Denmark. The diurnal cycle in wind speed has different shapes at different heights and the WRF model can replicate the diurnal cycle at higher levels (∼ 250 m a.g.l.) but fails to replicate it near surface. We investigate the effect of using different Planetary Boundary Layer (PBL) and Surface Layer (SL) schemes and compare stability data between model and observations. The results suggest a weak link between the model’s ability to replicate stability, probably linked to incorrect surface roughness length specification, and the model’s ability to replicate diurnal cycle at lower levels.

A measure–correlate–predict model based on neural networks and frozen flow hypothesis for wind resource assessment

In this paper, a measure–correlate–predict (MCP) model based on neural networks (NN) and frozen flow hypothesis, which is abbreviated as the MCPNN-frozen model, is proposed for wind resource assessment and tested using turbulent channel flows with three different surface roughness lengths, i.e., k0=0.001, 0.01, and 0.1 m. The predictions from the MCPNN-frozen model are compared with the real data for different separations (s) between the reference point and the target point. The results show that the correlation coefficients C.C. between the model predictions and real data are roughly higher than 0.5 for small separations s/δ≤3 (where δ is the boundary layer thickness), and the coefficients of determination (R2) are approximately higher than 0.3 when s/δ≤2. The generalization capacity of the MCPNN-frozen model is tested for different roughness lengths and different velocity components. Further analyses show that, even though C.C. and R2 decrease when increasing s, the large-scale variations of velocity fluctuations are well captured by the MCPNN-frozen model especially for the one trained using the data filtered in time. Furthermore, it is found that the model trained using the filtered data without a spanwise offset can well predict the large-scale variations at the target point when the spanwise offsets between the target point and the reference point are small (e.g., 0.1δ and 0.2δ). The proposed model leverages the power of neural networks and physical understanding. Further development of the model for complex scenarios will be carried out in the future work.