• Atkinson, B. W., , and M. Zhu, 2005: Radar-duct and boundary-layer characteristics over the area of the gulf. Quart. J. Roy. Meteor. Soc., 131, 19231953.

    • Search Google Scholar
    • Export Citation
  • Atkinson, B. W., , and M. Zhu, 2006: Coastal effects on radar propagation in atmospheric ducting conditions. Meteor. Appl., 13, 5362.

  • Atkinson, B. W., , J.-G. Li, , and R. S. Plant, 2001: Numerical modeling of the propagation environment in the atmospheric boundary layer over the Persian Gulf. J. Appl. Meteor., 40, 586603.

    • Search Google Scholar
    • Export Citation
  • Bean, B., , and E. Dutton, 1968: Radio Meteorology. Dover, 435 pp.

  • Brooks, I. M., , and D. P. Rogers, 2000: Aircraft observations of the mean and turbulent structure of a shallow boundary layer over the Persian Gulf. Bound.-Layer Meteor., 95, 189210.

    • Search Google Scholar
    • Export Citation
  • Brooks, I. M., , A. K. Goroch, , and D. P. Rogers, 1999: Observations of strong surface radar ducts over the Persian Gulf. J. Appl. Meteor., 38, 12931310.

    • Search Google Scholar
    • Export Citation
  • Burk, S. D., , and W. T. Thompson, 1995: Mesoscale modeling of refractive conditions in a complex coastal environment. Proc. Conf. on Propagation Assessment in Coastal Environment, Bremerhaven, Germany, AGARD/NATO, 40.1–40.7.

    • Search Google Scholar
    • Export Citation
  • Burk, S. D., , and W. T. Thompson, 1997: Mesoscale modeling of summertime refractive conditions in the Southern California Bight. J. Appl. Meteor., 36, 2231.

    • Search Google Scholar
    • Export Citation
  • Cummings, J. A., 2005: Operational multivariate ocean data assimilation. Quart. J. Roy. Meteor. Soc., 131, 35833604.

  • Davies, T., , M. J. P. Cullen, , A. J. Malcolm, , M. H. Mawson, , A. Staniforth, , A. A. White, , and N. Wood, 2005: A new dynamical core for the Met Office’s global and regional modelling of the atmosphere. Quart. J. Roy. Meteor. Soc., 131, 17591782.

    • Search Google Scholar
    • Export Citation
  • Edwards, J. M., , and A. Slingo, 1996: Studies with a flexible new radiation code. I: Choosing a configuration for a large-scale model. Quart. J. Roy. Meteor. Soc., 122, 689719.

    • Search Google Scholar
    • Export Citation
  • Essery, R., , M. Best, , and P. Cox, 2001: MOSES 2.2 technical documentation. Hadley Centre Tech. Rep. 30, 31 pp.

  • Golding, B. W., 1987: The U.K. Meteorological Office mesoscale model. Bound.-Layer Meteor., 41, 91107.

  • Golding, B. W., 1990: The Meteorological Office mesoscale model. Meteor. Mag., 119, 8196.

  • Gregory, D., , and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability-dependent closure. Mon. Wea. Rev., 118, 14831506.

    • Search Google Scholar
    • Export Citation
  • Haack, T., , and S. T. Burk, 2001: Summertime marine refractivity conditions along coastal California. J. Appl. Meteor., 40, 673687.

  • Haack, T., , C. Wang, , S. Garrett, , A. Glazer, , J. Mailhot, , and R. Marshall, 2010: Mesoscale modeling of boundary layer refractivity and atmospheric ducting. J. Appl. Meteor. Climatol., 49, 24372457.

    • Search Google Scholar
    • Export Citation
  • Lock, A. P., , A. R. Brown, , M. R. Bush, , G. M. Martin, , and R. N. B. Smith, 2000: A new boundary layer mixing scheme. Part I: Scheme description and single-column model tests. Mon. Wea. Rev., 128, 31873199.

    • Search Google Scholar
    • Export Citation
  • Lystad, S., , and T. Tjelta, 1995: High-resolution meteorological grid for clear air propagation modelling in northern coastal regions. Proc. Conf. on Propagation Assessment in Coastal Environment, Bremerhaven, Germany, AGARD/NATO, 41.1–41.12.

    • Search Google Scholar
    • Export Citation
  • Orlanski, I., 1975: A rational subdivision of scales for atmospheric processes. Bull. Amer. Meteor. Soc., 56, 527530.

  • Panofsky, H. A., , and G. W. Brier, 1958: Some Applications of Statistics to Meteorology. The Pennsylvania State University, 224 pp.

  • Reynolds, R. W., , N. A. Rayner, , T. M. Smith, , D. C. Stokes, , and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 16091625.

    • Search Google Scholar
    • Export Citation
  • Stapleton, J., , V. Wiss, , D. Shanklin, , T. Nguyen, , E. Burgess, , W. Thornton, , and T. Brown, 2001: Radar propagation modeling assessment using measured refractivity and directly sensed propagation ground truth—Wallops Island, VA 2000. Tech. Rep. NSWCDD/TR-01/132, 49 pp.

    • Search Google Scholar
    • Export Citation
  • Wilson, D. R., , and S. P. Ballard, 1999: A microphysically based precipitation scheme for the U.K. Meteorological Office Unified Model. Quart. J. Roy. Meteor. Soc., 125, 16071636.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Model domains and locations of Wallops-2000 experiment: (a) 12-km-resolution and 4-km-resolution (grayscale image) model domains and (b) zoom-in image of the dash-lined box in (a). The AB and EF lines indicate where cross-sectional profiles were taken. The black asterisk at the intersection of AB and EF indicates the NPS buoy location, and small white circles within the triangle area indicate HELO coverage during the Wallops-2000 experiment.

  • View in gallery

    Schematic representation of a modified refractivity profile and its characteristics.

  • View in gallery

    Ducting skills from each model calculated from the contingency table (Table 4): N = A + B + X + Y, CFR = (A + Y)/N, HR = A/(A + B), MR = B/(A + B), FAR = X/(X + Y), correct null rate (CNR) = Y/(X + Y), error = (X + B)/N, and DS = A/(A + B) − X/(X + Y). Acronyms are defined in the text.

  • View in gallery

    Surface ducting frequency for observed and simulated ducting events.

  • View in gallery

    The AB cross-sectional profiles at 1600 UTC 1 May 2000 derived from the (a1),(a2) 12 UTC run, (b1),(b2) 00 UTC run, and (c1),(c2) 00 UTC run + SST. Shown are (left) specific humidity (grayscale image) with potential temperature (contours) overplotted and (right) modified refractivity. The label of X is the cross point of the AB and EF lines at the altitude of ~500 m (above NPS). The label |w| indicates Wallops Island.

  • View in gallery

    As in Fig. 5, but at 1600 UTC 4 May 2000.

  • View in gallery

    As in Fig. 5, but for the EF cross-sectional profiles and the right-hand panels show vertical velocity (grayscale image) with westerly winds (contours) overplotted (solid lines for positive values, and dashed lines for negative values).

  • View in gallery

    As in Fig. 7, but at 1600 UTC 4 May 2000.

  • View in gallery

    Tephigrams at 1600 UTC 1 May derived from (top) the 12 UTC (black) and 00 UTC (red) runs and (bottom) the 00 UTC run (red) and 00 UTC run + SST (black).

  • View in gallery

    (left) Surface temperature and wind vectors and (right) distributions of ducting strength as defined in Fig. 2 at 1600 UTC 1 May derived from the (a1),(a2) 12 UTC run, (b1),(b2) 00 UTC run, and (c1),(c2) 00 UTC run + SST. The triangle denotes the study area.

  • View in gallery

    Surface temperature profiles along the EF cross section. The 12 UTC run, 00 UTC run, and 00 UTC run + SST are represented by solid lines, dashed lines, and dash–dotted lines, respectively. Label C indicates Chesapeake Bay, and W indicates Wallops Island.

  • View in gallery

    SST at the NPS site: the black solid line is observations, the dashed line is the ECMWF SST, and the dotted line is the NCODA SST.

  • View in gallery

    Vertical profiles at 1600 UTC 4 May at one of HELO measurement locations (37.58°N, 75.25°W). The solid line is observations; dashed, dash–dotted, and dash–dot–dotted lines are from the 12 UTC run, 00 UTC run, and 00 UTC run + SST, respectively, extracted from model outputs at 1600 UTC.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 43 43 10
PDF Downloads 30 30 7

Effects of Initial and Boundary Conditions of Mesoscale Models on Simulated Atmospheric Refractivity

View More View Less
  • 1 * Joint Centre for Mesoscale Meteorology, Met Office, Reading, United Kingdom
  • 2 Defence Outcomes, Met Office, Exeter, United Kingdom
  • 3 Naval Research Laboratory, Monterey, California
  • 4 Civil Engineering, Faculty of Engineering and Physical Sciences, University of Surrey, Guildford, Surrey, United Kingdom
  • 5 Naval Surface Warfare Center, Dahlgren, Virginia
© Get Permissions
Full access

Abstract

Radar ducting is caused by sharp vertical changes in temperature and, especially, water vapor at the top of the atmospheric boundary layer, both of which are sensitive to variations in the underlying surface conditions, local mesoscale weather, and synoptic weather patterns. High-resolution numerical weather prediction (NWP) models offer an alternative to observation to determine boundary layer (BL) structure and to assess the spatial variability of radar ducts. The benefit of using NWP models for simulating ducting conditions very much depends on the initial state of sea surface temperature (SST) and the model spinup time, both of which have a great impact on BL structure. This study investigates the effects of variation of NWP-model initial conditions and simulation length on the accuracy of simulating the atmosphere’s refractive index over the Wallops Island, Virginia, region, which has pronounced SST variability and complex BL structure. The Met Office Unified Model (MetUM) with horizontal resolution of 4 km (4-km model) was used to simulate the atmospheric environment when observations were made during the Wallops-2000 experiment. Sensitivity tests were conducted in terms of the variability of SST and model spinup time. The evaluation of the model results was focused on low-level moisture, temperature, and surface ducting characteristics including the frequency, strength, and the height of the ducting layer. When provided with more accurate SST and adequate simulation length, the MetUM 4-km model demonstrated an improved ability to predict the observed ducting.

Corresponding author address: Changgui Wang, Met Office, JCMM, Reading, RG6 6BB, United Kingdom. E-mail: chang.wang@metoffice.gov.uk

Abstract

Radar ducting is caused by sharp vertical changes in temperature and, especially, water vapor at the top of the atmospheric boundary layer, both of which are sensitive to variations in the underlying surface conditions, local mesoscale weather, and synoptic weather patterns. High-resolution numerical weather prediction (NWP) models offer an alternative to observation to determine boundary layer (BL) structure and to assess the spatial variability of radar ducts. The benefit of using NWP models for simulating ducting conditions very much depends on the initial state of sea surface temperature (SST) and the model spinup time, both of which have a great impact on BL structure. This study investigates the effects of variation of NWP-model initial conditions and simulation length on the accuracy of simulating the atmosphere’s refractive index over the Wallops Island, Virginia, region, which has pronounced SST variability and complex BL structure. The Met Office Unified Model (MetUM) with horizontal resolution of 4 km (4-km model) was used to simulate the atmospheric environment when observations were made during the Wallops-2000 experiment. Sensitivity tests were conducted in terms of the variability of SST and model spinup time. The evaluation of the model results was focused on low-level moisture, temperature, and surface ducting characteristics including the frequency, strength, and the height of the ducting layer. When provided with more accurate SST and adequate simulation length, the MetUM 4-km model demonstrated an improved ability to predict the observed ducting.

Corresponding author address: Changgui Wang, Met Office, JCMM, Reading, RG6 6BB, United Kingdom. E-mail: chang.wang@metoffice.gov.uk

1. Introduction

The occurrence of anomalous propagation (AP) conditions affects radar signals. Very small variations in the air refractive index n are responsible for AP. Therefore, the analysis of AP is usually done using refractivity N, defined as (n−1) × 106. The refractivity N depends on atmospheric pressure, humidity, and temperature. Using Debye theory (Bean and Dutton 1968), the dependence of refractivity at microwave wavelengths on atmospheric parameters is approximated as
e1
where N is refractivity (dimensionless), T is the air temperature (K), p is atmospheric pressure (hPa), and e is the water vapor pressure (hPa). Taking into account of the curvature of the layer in which refractivity is considered, the above equation can be rewritten as
e2
Equation (2) is referred to as modified refractivity, where z is the height above sea level and R is the mean radius of Earth. The term of (z/R) × 106 accounts for Earth’s curvature. The derived quantity of the modified atmospheric refractivity is referred to as M unit.

Vertical variations in temperature and water vapor content can give rise to a wide range of nonstandard propagation conditions by altering the path of electromagnetic (EM) energy propagating through the atmosphere, as outlined in Table 1. When vertical temperature and humidity gradients yield a negative gradient of modified refractive index, it indicates the presence of ducting conditions. An example of ducting occurs when strong low-level inversions (temperature increases with height) cap very moist air near the ocean surface. Such events are common in coastal regions, by the advection of an air mass from the land over the water, which may result in a large change in temperature and humidity. When warm and dry continental air travels over cooler water, it becomes more stable and suppresses vertical mixing. Heat loss to the surface results in the cooling of the air nearest the surface and forms a stable marine internal boundary layer (MIBL) or marine atmospheric boundary layer (MABL). When a duct occurs at low altitude, propagation loss for radar operating within the duct will usually be much less than for a standard atmosphere, and therefore the range at which low-altitude targets can be detected is increased. Above the duct, the range may be much reduced relative to that for a standard atmosphere, and therefore a radar “hole” exists. Such refractive effects are of particular importance to strike warfare and ship self-defense since failure to consider the formation of ducting may lead to erroneous interpretation of radar returns (or absence of returns).

Table 1.

Summary of refraction types.

Table 1.

Despite understanding the cause of the EM ducting, forecasting propagation conditions remains a significant challenge. These challenges led to a focused effort and increased collaboration by the United States of America–Great Britain–Canada–Australia–New Zealand (ABCANZ) countries to identify the primary weaknesses and limitations of mesoscale modeling systems in forecasting ducting. That research identified five key areas of numerical weather prediction (NWP) systems that contribute significantly to accuracy in simulated and forecast refractivity in coastal regions (Haack et al. 2010); two of these key areas are considered in detail in this paper. Several applications of using mesoscale NWP models for refractivity studies have been investigated in the past few years. For example, Burk and Thompson (1995, 1997) modeled the summertime refractive conditions in the Southern California Bight using a 20-km-horizontal-resolution mesoscale model and found significant diurnal variations of the trapping layer in the MIBL. Their results encouraged the use of refractivity profiles predicted by a mesoscale model as input to EM propagation models for forecasting the propagation environment. Lystad and Tjelta (1995) simulated the refractivity field over a coastal area of Norway and compared it with radiosonde measurements. Their model proved to be useful in predicting the spatial distributions and diurnal variations of refractivity, but it missed the fine vertical structures that are also of critical importance for radio propagation. Brooks et al. (1999) published their observational study of both the surface evaporation and BL ducts that were based on the Ship Antisubmarine Warfare Readiness/Effectiveness Measuring exercise (SHAREM-115) and emphasized that the small-scale variations found in the observations might be required to achieve operationally adequate forecasts of propagation conditions. Atkinson et al. (2001) then simulated the propagation environment in the atmosphere boundary layer over the Persian Gulf using a mesoscale model from the Met Office (Golding 1987, 1990) with a horizontal resolution of 6 km and examined the effects of initializing the model in varying degrees of idealization with the observations taken during the SHAREM-115 experiment (Brooks and Rogers 2000). The existence, location, and surface characteristics of the ducts were captured well, but they found that the level of agreement was sensitive to the initial conditions. A higher-resolution real data modeling study off the U.S. West Coast identified topographic modulation of the MABL as being responsible for variations in the strength and depth of ducted layers (Haack and Burk 2001). More detailed NWP studies over the SHAREM-115 experiment were again carried out by Atkinson and Zhu (2005, 2006) to identify local mesoscale factors affecting propagation, such as sea breeze and coastal configuration. Atkinson and Zhu (2005) in particular investigated the duct characteristics as a function of the distribution of sea surface temperatures (SST), land–sea temperature contrast, orography, and ambient wind at a meso-β (Orlanski 1975) scale and concluded that ambient wind speed was the most important, closely followed by orography, but that the distribution of SST had the smallest effect in their situation.

In 2007, a collaborative radio frequency–infrared propagation working group (referred to in this paper as ABCANZ) established a need for diagnostic and prognostic 4D refractivity modeling capabilities to support littoral naval operations. The ABCANZ working group aims to bring together a diverse group of NWP modelers to focus on naval problems and to develop capabilities that would be exploitable by all ABCANZ countries. Their initial intercomparison study examining the Wallops-2000 Microwave Propagation Measurement Experiment (MPME; Haack et al. 2010) shows that the model performance in predicting ducting occurrence varies greatly. As compared with an observed ducting frequency of 64% during the 7-day period of observations, the modeled ducting frequency ranged from 4% to 43%, and the percentage of correctly simulated ducting/no-ducting events ranged from 37% to 57%. As revealed by the initial study, such differences were related to several factors, including model initial conditions, boundary conditions, and large-scale and surface forcing on mesoscale inner domains, and it subsequently led to the sensitivity analysis carried out initially using the Met Office Unified Model (MetUM; Haack et al. 2010).

In addition to some conclusive points highlighted in the intercomparison study (Haack et al. 2010), this paper presents the results of a sensitivity analysis of the impact of model lead time and the initial state of surface conditions on the accuracy of prediction of refractivity and electromagnetic ducting by the MetUM 4-km model. It not only reveals the importance of the synoptic forcing to the formation of ducts through the model lead time but also shows the impact of mesoscale structure through the interaction between the sea breeze and high pressure subsidence on the occurrence of the ducts and their properties. The model results were evaluated using the dataset from the Wallops-2000 MPME experiment. The measurements in the Wallops-2000 MPPE included low-elevation radar-frequency path-loss and clutter returns as well as meteorological conditions from an observing tower (over land), a buoy located 13 km from shore, a boat, and an instrumented helicopter (HELO). The HELO primarily collected meteorological conditions within the lowest 150 m of the atmosphere, extending up to 65 km from shore, for the period from 28 April to 4 May 2000. The data were collected by the Naval Surface Warfare Center, Dahlgren Division, of the U.S. Navy and were made available through the collaboration among ABCANZ countries. An extensive description of the field campaign and observations gathered during the field experiment was presented by Stapleton et al. (2001).

2. NWP model and setup

The dynamic core of the MetUM implements the fully compressible, nonhydrostatic, deep-atmosphere formulation of the Navier–Stokes equations (Davies et al. 2005). In terms of numerics, the advection of the prognostic variables is treated using a semi-Lagrangian discretization, which follows the air particle trajectories. The exception is density, for which an Eulerian (i.e., grid based) discretization is retained to conserve mass. Semi-implicit time integration is used for the high-frequency terms. The model includes a comprehensive set of parameterizations, including surface (Essery et al. 2001), boundary layer (Lock et al. 2000), mixed-phase cloud microphysics (Wilson and Ballard 1999), and convection (Gregory and Rowntree 1990), with additional downdraft and momentum transport. Edwards and Slingo (1996) implemented the two-stream approximation to radiative transfer and incorporated scattering in both the longwave and shortwave regions of the spectrum in the MetUM for calculating radiative fluxes. The model runs on a rotated latitude–longitude horizontal grid with Arakawa-C staggering and a terrain-following hybrid-height vertical coordinate with Charney–Philips staggering. Because the model is nonhydrostatic, it can be run at a high horizontal resolution (e.g., 4 and 1 km), which will preserve the important features of the topography and coastline that can have a strong impact on the formation of EM ducts and radar holes.

The MetUM was initialized with the operational global analysis (60 vertical levels and approximately 0.5° horizontal grid spacing) from the European Centre for Medium-Range Weather Forecasts (ECMWF). A suite of (one way) nested models was set up to run with grid spacing of 12 and 4 km as shown in Fig. 1a. In general, 70 vertical levels were used for the 4-km model and 38 levels were used for the 12-km model, as listed in Table 2. The 12-km model was used to dynamically downscale the ECMWF global analysis for each day to provide boundary conditions for the 4-km model. Thus, the synoptic conditions are imposed on inner domains by initial conditions at the initialization time (0000 or 1200 UTC) and by boundary conditions that were updated every 6 h.

Fig. 1.
Fig. 1.

Model domains and locations of Wallops-2000 experiment: (a) 12-km-resolution and 4-km-resolution (grayscale image) model domains and (b) zoom-in image of the dash-lined box in (a). The AB and EF lines indicate where cross-sectional profiles were taken. The black asterisk at the intersection of AB and EF indicates the NPS buoy location, and small white circles within the triangle area indicate HELO coverage during the Wallops-2000 experiment.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

Table 2.

Heights of model levels above sea level, assuming zero orography, for the 38-vertical-level set and the 70-level set below 4555.0 m.

Table 2.

The mesoscale orography of the Wallops Island coastal region and strong sea surface temperature gradients between cool shelf waters and the warm Gulf Stream produce complex MIBL structures that often generate vertical refractivity gradients and therefore alter EM propagation. To analyze and simulate refractivity variations affected by complex MIBL structures in the region, three sets of model experiments were carried out and the MetUM was initialized on each day (Table 3). In the first test, the MetUM was initialized using the 1200 UTC ECMWF (short model lead time) operational global analysis; in the second, it was initialized using the 0000 UTC ECMWF analysis (longer model lead time). These two experiments are subsequently referred to as the “12 UTC run” and the “00 UTC run,” respectively. Each of those experiments used the SST from the ECMWF skin temperature analysis performed on their ~0.5° grid from where SST observations were assimilated using four-dimensional variational data assimilation from the National Centers for Environmental Predictions (NCEP) analysis. The main source of NCEP SST is a 1° Reynolds dataset (Reynolds et al. 2002). The third experiment was to replace the ECMWF SST with a much-higher-resolution (4 km) SST analysis obtained from the U.S. Navy Coupled Ocean Data Assimilation multivariate optimum interpolation analysis (NCODA; Cummings 2005). This is referred to as the “00 UTC run + SST.” The NCODA uses the previous 12-h-old analysis as a first guess in a multivariate optimal interpolation scheme and assimilates available satellite altimeter observations and in situ SST, Argo (geostrophic oceanography) floats, and moored buoys. All experiments simulated the atmospheric refractivity conditions on each day between 1200 and 2300 UTC (day time) when the HELO profile measurements were taken during the Wallops-2000 experiment. Therefore, the MetUM initialized at 0000 UTC would have at least 12 h of spinup time to allow mesoscale features to be resolved, whereas the MetUM initialized at 1200 UTC would have less than 1 h to spin up for the HELO profile measurements started at 1200 UTC (Table 3) on each day.

Table 3.

MetUM model experiments initialized on every day. Model lead time: from model initialization to 1200 UTC.

Table 3.

3. Model results

Model evaluation is based upon standard statistics including means, bias, and root-mean-square error (RMSE) along with duct contingency tables for establishing forecast skills as described in the model intercomparison study (Haack et al. 2010). The observations used were near-surface buoy measurements (temperature, specific humidity, and modified refractivity at a height of 4 m and winds at a height of 10 m) and the total of 190 vertical measurements (Fig. 1b) taken from 28 April and 4 May 2000. Statistics for near-surface variables at the buoy were calculated using hourly measurements. In corresponding 4-dimensional model hourly outputs, vertical HELO measurements were interpolated onto model-level heights plus an extra surface level of 1.5 m and the observation time of each profile was rounded to the nearest hour. The model profiles were then extracted at the nearest gridpoint location corresponding to the average latitude and longitude position at the start and end of each HELO descent. For defining the existence of a duct for each profile, it is assumed that its strength (Fig. 2) must be equal to or greater than 1 M unit so that very weak ducts can be ignored in both HELO and model profiles. A schematic representation of a modified refractivity profile with duct characteristics (duct strength, base height, and thickness) is depicted in Fig. 2.

Fig. 2.
Fig. 2.

Schematic representation of a modified refractivity profile and its characteristics.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

In addition, most HELO measurements had high vertical resolution of typically less than 10 m. Therefore, MetUM diagnostics outputs at the level of 1.5 m were introduced when extracting model vertical profiles at HELO locations. The diagnostics at 1.5 m are based on the Monin–Obukhov similarity theory, which was coded in the MetUM. The diagnostics at this level were not used in the intercomparison paper (Haack et al. 2010) because of the difficulty in obtaining the data at the same level from the other NWP models that were used. For resolving surface-based ducts, meteorological conditions at 1.5 m are essential because the formation of a strong surface duct will need a good definition of the surface conditions, such as skin temperature and humidity. As a result, the observed ducting frequency will be changed in this paper relative to that in Haack et al. (2010) because an extra level was introduced (see sections 3a and 3b). The statistics of HELO profiles presented in this paper are slightly different from those presented in Haack et al. (2010) because of rounding differences and conversion to mks units.

To investigate how the model lead time and the SST forcing affect the forecasting of ducts, the analysis was focused on the results for 1 and 4 May because synoptic conditions (high pressure subsidence) on these days were more favorable to radar ducts. In particular, 1 May was an alongshore case in which a warm southerly flow interacted with subsidizing dry air to create a shallow MABL and 4 May was an onshore case in which upwind SST created a deep MABL capped by the large-scale subsidence (Haack et al. 2010).

a. Impact of model lead time

During the Wallops-2000 experiment, most measurements were taken after 1200 UTC (0700 local time)—most important, the HELO profiles. For the ABCANZ Wallops-2000 model intercomparison (Haack et al. 2010), the 12 UTC run from the MetUM (Table 3) was analyzed but produced limited mesoscale structure. The shorter the model simulation length is, the less error is expected in capturing the large-scale structure. A short model simulation length gives less time for small-scale structure to spin up, however. This section presents the impact of model simulation length on the results of the 4-km model. To this end, the MetUM 4-km model was then run starting at 0000 UTC, which is referred to as the 00 UTC run (Table 3), on each day during the Wallops-2000 experiment, and its results were compared with the initial 12 UTC run (Table 3), with both runs being evaluated against a total of 190 HELO profile measurements.

At the HELO sites, the mean ducting characteristics (Table 4) and forecasting skills (Fig. 3) from both runs reveal that the 12-h-earlier initialization led to predictions of a higher frequency of thicker and stronger ducts. In comparing with the 12 UTC run, the number of correctly forecast ducting events (75) is doubled; apart from significant increases of the ducting thickness (DTK) and ducting base height (DBH), the ducting strength (DST) also increased to 4.33 M units from 3.88 M units. The ducting forecast skill in the 00 UTC run (Fig. 3), on the basis of the ducting contingency table (Panofsky and Brier 1958) (Table 5) also improved. When compared with 190 observations, the 00 UTC run simulated the existence of ducts on 96 occasions, or 51% of the total. Correct forecasting of the presence or absence of ducts occurred on 111 occasions, or 58% of the total [correct forecast rate (CFR)]. Ducts were correctly forecast 75 out of 133 times, or 56% [hit rate (HR)] of the total observed ducts (133). In the 12 UTC run, only 80 of the simulated ducts and no ducts were correct, with 37 correct predictions of the existence of ducts and 43 correct predictions for no ducts, yielding 42% as the CFR. Of the 133 observed ducts, the ducting HR only achieved 28% . Figure 3 also shows that the 00 UTC run presented a lower error and a lower miss rate (MR) but had a higher false-alarm rate (FAR). Nevertheless, the higher HR in the 00 UTC run is not just because the model is forecasting more ducting events but is because it is getting these extra events in the right place, thus raising the discrimination score (DS = HR − FAR) to 20%.

Table 4.

Contingency table of ducting events from each model simulation, with 4-km resolution. Total number of observed HELOs is 190. Here, A indicates a correct forecast of ducting, Y indicates a correct forecast of no ducting, B indicates the number of ducting events that were found in the observations but were not predicted in the model, and X is the number of ducting events that were predicted in the model but were not found in the observations. It is assumed that each profile is only counted as one duct event if it exhibits ducts.

Table 4.
Fig. 3.
Fig. 3.

Ducting skills from each model calculated from the contingency table (Table 4): N = A + B + X + Y, CFR = (A + Y)/N, HR = A/(A + B), MR = B/(A + B), FAR = X/(X + Y), correct null rate (CNR) = Y/(X + Y), error = (X + B)/N, and DS = A/(A + B) − X/(X + Y). Acronyms are defined in the text.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

Table 5.

Ducting statistics at HELO paths (triangle area in Fig. 1) computed from each model simulation where ducting was observed from HELO measurements.

Table 5.

The improvement in the 00 UTC run relative to the 12 UTC run varies on each day during the Wallops-2000 experiment. The daily simulated observed duct/no-duct event frequency (mainly between 1200 and 2300 UTC each day) shown in Fig. 4 indicates that earlier initialization (00 UTC run) had a noticeable impact on ducting events on 29 April and from 2 to 4 May but not much impact for the other dates. The overall statistics for simulated observed-duct events in the 00 UTC run were in a closer agreement with the observations. For example, the 00 UTC run simulated the same observed frequency of ducting events as was computed from HELO measurements on 2 and 3 May, which were 46% and 100%, respectively. On 1 and 4 May, the ducting frequency increased to 23% and 28% from 14% and 0% predicted at 1200 UTC. The model oversimulated observed ducts on 29 April, however.

Fig. 4.
Fig. 4.

Surface ducting frequency for observed and simulated ducting events.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

By examining the cross-sectional (AB line in Fig. 2) profiles at 1600 UTC on 1 and 4 May shown in Figs. 5 and 6, it is clear that the earlier model run showed a layer of dry air (water vapor < ~3 g kg−1) above the sea surface (at an altitude of ~500 m on 1 May and ~700 m on 4 May) closer toward or across the coastline. On 1 May, the tongue of the dry-air layer in the 00 UTC run just reached the coast of Wallops Island whereas it is ~120 km away from the coastline in the 1200 UTC model. On 4 May, the tongue of the dry-air layer in the 00 UTC run advanced across the coastline whereas it did not quite reach to the coastline in the 12 UTC run. The change in the dry-layer structure resulted in the increase of humidity gradients within 200 km of the coast in the 00 UTC run, thus leading to the formation of an elevated ducting layer (Figs. 5b2 and 6b2).

Fig. 5.
Fig. 5.

The AB cross-sectional profiles at 1600 UTC 1 May 2000 derived from the (a1),(a2) 12 UTC run, (b1),(b2) 00 UTC run, and (c1),(c2) 00 UTC run + SST. Shown are (left) specific humidity (grayscale image) with potential temperature (contours) overplotted and (right) modified refractivity. The label of X is the cross point of the AB and EF lines at the altitude of ~500 m (above NPS). The label |w| indicates Wallops Island.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

Fig. 6.
Fig. 6.

As in Fig. 5, but at 1600 UTC 4 May 2000.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

The presence of dry air and the formation of a ducting layer (Figs. 5 and 6) can be explained using the cross-sectional profiles (Figs. 7 and 8) along EF (Fig. 1). From Figs. 7 and 8, it is clear that the dry-air layers on these two days were introduced by the high pressure subsidence (Figs. 7b2 and 8b2). On 1 May, Fig. 7 shows that fast-moving southerlies cut into the subsiding air and caused a sheet of dry air (Fig. 6) along the AB (Fig. 1) section that moved slowly between the cool southeasterly flow near the surface and the warm southwesterly flow aloft (label X at a height of 400 m above the location of the buoy shows the intersection between AB and EF). Because of the change of model initialization, the synoptic forcing was modified; thus the movements of the warm southwesterly flow aloft and the subsiding dry air on the east side were corrected. In the 00 UTC run, the warm southwesterly flow aloft advanced faster; the tongue of the subsiding air reached the position of X (Figs. 5b and 7b). In the 12 UTC run, the movement of the warm southwesterly flow aloft was not as fast as that in the 00 UTC run and the subsiding air was also ~50 km away from position of X (Figs. 5a and 7a). The interaction of the synoptic forcing with the mesoscale sea breeze was also changed. The abrupt lifting of the MABL (from ~200 to ~500 m) at 500 km away from position A (Fig. 7) in the 00 UTC run enhanced humidity gradients and led to the formation of elevated ducts (Fig. 5b2) near the coast, which was not seen in the 12 UTC run. A stronger water vapor inversion in the 00 UTC run at ~975 hPa is also shown in Fig. 9 and led to the formation of a trapping layer at an altitude of 500 m between ~50 and ~150 km offshore as depicted in Fig. 5b2. Figure 10b2 depicts the horizontal distribution of ducting strength at 1600 UTC and shows that the 00 UTC run simulated the elevated ducts (triangle in Fig. 10) identified in Fig. 5b2. In contrast, the 12 UTC run failed to simulate any ducts over the same location.

Fig. 7.
Fig. 7.

As in Fig. 5, but for the EF cross-sectional profiles and the right-hand panels show vertical velocity (grayscale image) with westerly winds (contours) overplotted (solid lines for positive values, and dashed lines for negative values).

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

Fig. 8.
Fig. 8.

As in Fig. 7, but at 1600 UTC 4 May 2000.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

Fig. 9.
Fig. 9.

Tephigrams at 1600 UTC 1 May derived from (top) the 12 UTC (black) and 00 UTC (red) runs and (bottom) the 00 UTC run (red) and 00 UTC run + SST (black).

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

Fig. 10.
Fig. 10.

(left) Surface temperature and wind vectors and (right) distributions of ducting strength as defined in Fig. 2 at 1600 UTC 1 May derived from the (a1),(a2) 12 UTC run, (b1),(b2) 00 UTC run, and (c1),(c2) 00 UTC run + SST. The triangle denotes the study area.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

On 4 May, the synoptic-scale subsidence developed a strong cap layer and the upwind warm SST (Fig. 8b2) from the Gulf Stream created a deep, moist mixed layer, but the SST from the Gulf Stream in the 00 UTC run were slightly lower than those in the 12 UTC run, as shown in Fig. 11. Accompanied by the stronger local sea-breeze return flow in the 00 UTC run, a strong elevated ducting layer at ~750 m was formed across the model domain (Fig. 6b2). A thin surface ducting layer (Fig. 6b2) over the study region also developed in the 00 UTC run. Along with slightly weaker synoptic-scale subsidence (Fig. 8a1) due to later initialization, the earlier arrival of moist air to the study region from the Gulf Stream (Fig. 8b1) mixing with warm air from land that was simulated by the12 UTC run reduced the humidity gradients and thus prevented the formation of the elevated and surface ducts over the study region (Fig. 6b1).

Fig. 11.
Fig. 11.

Surface temperature profiles along the EF cross section. The 12 UTC run, 00 UTC run, and 00 UTC run + SST are represented by solid lines, dashed lines, and dash–dotted lines, respectively. Label C indicates Chesapeake Bay, and W indicates Wallops Island.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

The better agreement between the 00 UTC run and the observations can also be seen from the meteorological parameter statistics over all HELO profiles (Table 6). The results from the 00 UTC run showed closer agreement in humidity and modified refractivity with observations than did those from the 12 UTC run. At each of the heights measured, the 00 UTC run gave a prediction with a smaller absolute bias value when compared with the 12 UTC run. There is, however, less distinction with the potential temperature.

Table 6.

Vertical-profile statistics at HELO paths coverage (triangle area in Fig. 1) for levels H at 5, 22, 45, and 112 m for specific humidity (g kg−1), potential temperature (K), and modified refractivity (M units) from each model simulation for the total 190 HELO paths. The first number is the mean, the second is the bias, and the third is the RMSE (mean/bias/RMSE).

Table 6.

At the Naval Postgraduate School (NPS) buoy site, the time-series statistics (Table 7) did not show large differences between the 12 UTC run and the 00 UTC run. An improvement can still be seen from the 00 UTC run. The bias in air temperature is greater than 1 K, but the relative humidity (RH) and wind speed in the 00 UTC run were much closer to the observed means.

Table 7.

Near-surface-variable statistics at the NPS buoy (asterisk in Fig. 1) for temperature T, RH, wind speed (WSP), and SST computed from each model simulation and for NPS buoy measurements. The first data column shows observations, and the other columns show each run’s bias (bias = model output − observation) and the RMSE (bias/RMSE). The statistics are sampled from 1200 UTC 28 Apr to 0000 UTC 5 May 2000.

Table 7.

b. Impact of changes of the SST

To assess the impact of SST on the radar ducting characteristics and using the 00 UTC run, the 12-km model was initialized from the ECMWF operational global analysis but the SST in the 4-km model were updated each day at 0000 UTC using NCODA SST (00 UTC run + SST). The 4-km-model results were then evaluated against the observations in terms of time series and vertical profiles at the NPS buoy and HELO paths and mean ducting characteristics.

Figure 10c1 shows that the warmest Gulf Stream current within the 4-km-model domain, with an SST above 292 K, is located ~250 km offshore Wallops Island. This warmest Gulf Stream current is not captured by the ECMWF SST, as shown in Fig. 10b1. The NCODA SST also depicts cold coastal water and a warmer Gulf Stream, in contrast with the weaker gradients in the ECMWF SST. By taking 1600 UTC on 1 and 4 May as examples (Fig. 11), it can be seen that the difference between the NCODA and ECMWF SST reached ~7 K along section EF. Further, the NCODA SST analysis presents a more realistic warming trend over the experiment period, whereas the ECMWF SST displays a nearly constant warm bias. Figure 12 depicts the SST time series from ECMWF (dashed line), NCODA SST (dotted line), and observations (solid line) at the NPS buoy site, showing the decreasing warm bias in the ECMWF field with time. On average, the NCODA SST at the NPS buoy site was about 2 K lower than the ECMWF SST. An SST difference over the model domain led to the change of modeled results, especially changes to the internal boundary layer (IBL). The lower surface temperature over the study region produced a more stable IBL, which then led to the formation of stronger surface ducting.

Fig. 12.
Fig. 12.

SST at the NPS site: the black solid line is observations, the dashed line is the ECMWF SST, and the dotted line is the NCODA SST.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

The statistics of the vertical profiles at HELO profile locations in Table 6 show that the 00 UTC run + SST predicted lower temperature and specific humidity values when compared with the 00 UTC run—in particular, near the surface. At the level of 5 m, the absolute bias in potential temperature was −1.7 K, and its RMSE increased to 1.9 from 1.2 K. In response to the lower SST within 100 km of the coast, the near-surface inversion became stronger and shallower, which led to less vertical mixing and drier air above the inversion, as shown in Fig. 7 at 1600 UTC 1 May. Therefore, the mean specific humidity was substantially lower on average and was in much better agreement with the observations. The bias and RMSE at 5 m were −0.01 and 0.5 g kg−1, respectively, whereas for the 00 UTC run they were 0.50 and 1.0 g kg−1, respectively. Hence the 00 UTC run + SST produced better agreement in specific humidity (lower than that from the 00 UTC run), but there was a disparity in potential temperature (too cold). The 00 UTC run predicted better agreement in potential temperature at the surface (below 50 m), but its bias got larger above the level of 112 m. The specific humidity that was too high in the 00 UTC run was much closer to observations in the 00 UTC run + SST. Figure 13 shows an example of vertical profiles on 4 May 2000. The potential temperature bias at 1600 UTC from the 00 UTC run was higher than that from the 00 UTC run + SST at levels above 50 m. In comparison with Fig. 8b2, Fig. 8c2 shows that the 00 UTC run + SST produced a stronger sea-breeze circulation below the inversion capping layer; therefore the return flow from the land improved the stratification well above the surface layer. In addition, Fig. 13 also demonstrates that the shape of the temperature profile from the 00 UTC run + SST is more consistent with the observations despite the large bias at all observed levels. Because atmospheric ducting depends upon the vertical gradients, improved profile shapes allowed the 00 UTC run + SST to attain much better agreement in modified refractivity when compared with those computed from observations (Table 6 and Fig. 13).

Fig. 13.
Fig. 13.

Vertical profiles at 1600 UTC 4 May at one of HELO measurement locations (37.58°N, 75.25°W). The solid line is observations; dashed, dash–dotted, and dash–dot–dotted lines are from the 12 UTC run, 00 UTC run, and 00 UTC run + SST, respectively, extracted from model outputs at 1600 UTC.

Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-012.1

The above findings from the 00 UTC run + SST indicate that the magnitude of the vertical stratification over the Wallops Island region was greatly increased by the lower coastal SST, as well as by the much warmer Gulf Stream, which had influence during periods of onshore flow. These SST features in the high-resolution NCODA SST analysis were much more pronounced and showed larger temporal variations than those in the ECMWF SST field. For example, at 1600 UTC 1 May 2000, the coastal NCODA SST was ~2 K lower and the Gulf Stream SST was ~7 K higher along the cross section of EF as compared with the ECMWF SST (Fig. 11). Such a large contrast in the 00 UTC run + SST led to the change of IBL. Near the coast, a very strong and shallow MIBL formed (Fig. 7) as a result of the lower SST and enhanced subsidence, leading to stronger surface ducts, as shown in Fig. 5. Toward the Gulf Stream, the elevated ducting layer was mixed away because of a higher SST, which produced a deep BL and more mixing. Therefore, the change of SST modified the distribution of specific humidity and temperature as well as refractivity profiles (Table 5) near and above the surface over the study region.

An increase in the number of surface ducts in the 00 UTC run + SST over the study region improved model skills in terms of the ducting contingency table (Table 4 and Fig. 3). For 190 profiles, the observed ducting occurrence was 70% while the modeled ducting occurrence increased to 70% (133 ducting events) in the 00 UTC run + SST. In correspondence, the correct forecast and hit rates went up to 76% and 82%, respectively, and the forecast error was reduced to 26%. The daily observed event frequency data (Fig. 4) show that the improvement in the 00 UTC run + SST mainly occurred on 1 and 4 May when high pressure subsidence was established over the region. Both days had flow originating over the warm Gulf Stream water to the south and east. The frequencies on these two days climbed up to 95% and 70% respectively; therefore, the overall observed surface-based duct frequency rose to 95% from 82%, slightly higher than that in the observations (90%). The 00 UTC run + SST overpredicted surface ducting events. This is because the lower surface SST from NCODA (Fig. 11) produced a more stable BL when air advected from land over the sea, increasing the chance of the formation of stronger surface ducting.

Figures 5c1 and 6c1 show the change of potential temperature, specific humidity, and ducting layers along the cross section of AB (Fig. 1) at 1600 UTC 1 and 4 May 2000, representing different background wind directions aloft. On 1 May, the SST produced stronger synoptic subsidence that brought drier and warmer air above the MABL. As a result, vertical mixing was suppressed by the strong inversion induced by subsidence and low coastal SST, producing more surface-based ducting, as shown in Fig. 5c2.

For 4 May, as shown in Fig. 6c, the stability of the MABL was also increased but not as strongly as on 1 May, as shown in Fig. 5c. The case of onshore flow demonstrated the effect of the high SST within the Gulf Stream on IBL development and the formation of ducts. Figure 6c clearly shows a stronger surface inversion and a much thicker surface layer duct as compared with Fig. 6b. The HELO profile at 1600 UTC 4 May 2000 (Fig. 13) verified the occurrence of the surface ducting in the 00 UTC run + SST and was closer to the observations.

Another difference evident in Fig. 6c is the lack of elevated ducting at about 700 m in the 00 UTC run. The disappearance of the elevated ducting layer was due to more mixing at the top of the BL associated with a stronger land-breeze circulation. The increased land breeze resulted from the increase in the temperature gradient between the land and sea surface that introduced warmer air from the Gulf Stream over the study area, thus enhancing mixing at the top of the BL. Note the very different stratification aloft at the southeastern end of the cross section. The measurements taken on 4 May 2000 were all below this height, however, and so cannot be used to validate the refractive layers above 500 m. As a consequence of the increased frequency of the surface ducting and elimination of some elevated ducting events that appeared in the 00 UTC run, the total mean duct base height dropped to 0.5 m from the 15 m simulated in the 00 UTC run (Table 5). The stable BL produced in the 00 UTC run + SST also yielded slightly stronger and thicker ducts (Table 5).

The increase in ducting is also evident from the map of ducting strength at 2300 local time (1600 UTC) 1 May 2000 in Fig. 10c2. Figure 10c1 also shows that the NCODA SST in the 00 UTC run + SST within the observation area were lower while the Gulf Stream current in the 00 UTC run + SST was also closer to Wallops Island than that in the 00 UTC run. The stronger temperature gradients between the land and sea together with lower coastal SST and warmer air from the Gulf Stream altered the thermodynamic structure and enhanced the development of a more stable MABL over the observation area. Such an SST change consequently led to the formation of the surface ducts predominately over the colder-coastal-SST region spreading from the edge of the Gulf Steam toward the coastline of Wallops Island. Higher SST near the Gulf Stream (from ~250 km offshore) created more mixing (Fig. 7), however, thus destroying the elevated ducting layer that appeared in the 00 UTC run. The 00 UTC run failed to capture key thermal gradient features of the Gulf Stream in the southeastern portion of the model domain. Such a distinct difference suggested that the initial SST in the MetUM model have played an important role in producing more realistic near-surface and elevated boundary mixing. These differences served to change the coastal MABL and ducting structure, revealing that the refractivity over the Wallops Island region can be strongly influenced by the Gulf Stream located approximately ~250 km offshore.

The sensitivity analysis revealed that the lower SST (from NCODA) contributed to a higher duct hit rate because of enhanced sea breeze, which resulted in more surface ducts. The analysis of model lead time in section 3a also showed the contribution of lower SST from the model initialization to the higher duct hit rate in the 00 UTC run: lower SST at 00 UTC over the model domain was initiated from the ECMWF global analysis, which was modified through ECMWF model data assimilation. Figures 58 (panels a and b) and Fig.11 suggested that the difference between the 12 UTC and 00 UTC runs is contributed to by the synoptic-scale forcing introduced by the intialization. The cross point X of the AB and EF lines shown in Figs. 58 indicates that the subsidence layer is drier and closer to Wallops Island in the 00 UTC run. The SST also played a part in changing the humidity gradients near the surface. Higher upwind SST from the Gulf Stream (Fig. 11) in the 1200 UTC run introduced a deeper MABL and led to more mixing, which in turn reduced humidity gradients, as shown in the case at 1600 UTC 1 May (Fig. 7a1). At the NPS buoy site (black asterisk in Fig. 1), the SST in the 00 UTC run was just 0.2 K lower than that in the 12 UTC run on average, as shown in Table 7. The SST differences between each model run over the cross section of EF get larger toward the Gulf Stream, however, as shown in Fig. 11, which depicts SST sections at 1600 UTC 1 and 4 May along line EF. The SST in the 00 UTC run was ~2.5 K lower on 1 May and ~1.5 K on 4 May at ~470 km (at the right end of the cross section) from the coast. As a result, the larger temperature difference between solar heating of the coastal land and offshore waters in the 0000 UTC enhanced the sea breeze and increased surface ducts.

The intermodel comparison study (Haack et al. 2010) stated that both the degree of subsidence and the timing of large-scale synoptic transitions may be involved in creating model errors on 1 May, whereas small-scale variability and spatial patterns in SST are a key factor in setting up the multilayered MABL and its corresponding ducts on 4 May. Among the four NWP models, the Coupled Ocean–Atmosphere Mesoscale Prediction System (COAMPS) was the only one that captured strong surface ducts over the observation area because of its continuous run starting from 25 April accompanied by 12-hourly data assimilation and its use of more accurate SST from NCODA. The 1200 UTC run from the MetUM produced only weak ducts and in the wrong locations. The current findings reveal that errors in the MetUM presented in the interrun comparison were introduced by a lack of accurate SST forcing and insufficient model spinup time. The combination of an earlier model start and improved SST has enhanced the degree of subsidence and the timing of large-scale synoptic transitions in the MetUM, thus leading to the improvement in capturing the observed ducts. The examples of both 1 and 4 May have proved that an earlier-starting MetUM simulated stronger low-level subsidence and more accurate SST, enabling it to produce more accurate near-surface mixing and sea-breeze conditions. In comparing the simulated ducting events with the observed ducting frequency of 70%, it is seen that such changes increased ducting events to 58% and 76% for the earlier run and the use of NCODA SST, respectively, from the 42% simulated in the 12 UTC run. The ducting correct hit rate jumped to 83% from 28% when the MetUM started 12 h earlier together with the NCODA SST forcing.

4. Concluding remarks

High-resolution NWP models used for predicting ducting and refractivity conditions were rigorously evaluated in Haack et al. (2010), with an extensive dataset collected over 7 days during the Wallops-2000 MPME experiment. The characteristics of ducts occurring at the locations of HELO paths were examined statistically, indicating a high occurrence rate of surface ducting in most of the models. In this paper, sensitivity studies were carried out with the MetUM 4-km model with regard to changes in model lead time and SST. The findings not only provide clear identification that SST is the most significant input in the prediction of atmospheric ducting characteristics but also reveal that enough model lead time is necessary to allow high-resolution features to spin up, in terms of the information entering the model domain at the boundaries and the initial conditions. Large-scale forcing of SST has a strong influence on the model performance of near-surface mixing, stability, and the formation of sea breezes. The correct initialization time is also critical to get correct synoptic conditions, which have great impact on the atmospheric refractivity in terms of the vertical and horizontal positions of ducts. When coupled with correct initial and boundary conditions, the MetUM 4-km model demonstrated an improved capability to simulate the observed EM ducting within the lowest 500 m of the atmosphere:

  • Twelve-hour-earlier initialization in the 0000 UTC model run allowed for 12 h of mesoscale structure to develop on the inner grids before the start of the validation time period. The difference between results from the 00 UTC run and the 12 UTC run would also be partly from having different initial fields and SST because of the different initial time.
  • Using an earlier initial time also changed the synoptic situation slightly—in particular, the bottom base of the subsiding dry air over the study region. This altered the inversion cap layer, subsequently changing the occurrence of the ducts.
  • The use of the NCODA analyzed SST in the MetUM 4-km model improved the simulated ducting events significantly. This is because the NCODA SST captured the cold coastal SST and strength and position of the Gulf Stream. Both of these SST features led to more accurate coastal thermal gradients, land-/sea-breeze circulations, and IBL development near the Wallops Island area. As a result, the SST changed the stability and vertical mixing in the MABL, creating more accurate temperature and humidity profiles. In terms of the modified refractivity, the 00 UTC run + SST showed good agreement with HELO measurements.
  • The use of NCODA SST also shows that lower SST increased the occurrence of surface ducts, and higher SST have potential to destroy both surface and elevated ducts because of vertical mixing.

Acknowledgments

We are grateful to anonymous reviewers whose suggestions helped to shape up the manuscript. Financial support for this work was provided by the UK Ministry of Defence. The Wallops-2000 experiment data were provided through the ABCANZ model intercomparison collaboration supported by the U.S. Office of Naval Research, Program Element 0602271N.

REFERENCES

  • Atkinson, B. W., , and M. Zhu, 2005: Radar-duct and boundary-layer characteristics over the area of the gulf. Quart. J. Roy. Meteor. Soc., 131, 19231953.

    • Search Google Scholar
    • Export Citation
  • Atkinson, B. W., , and M. Zhu, 2006: Coastal effects on radar propagation in atmospheric ducting conditions. Meteor. Appl., 13, 5362.

  • Atkinson, B. W., , J.-G. Li, , and R. S. Plant, 2001: Numerical modeling of the propagation environment in the atmospheric boundary layer over the Persian Gulf. J. Appl. Meteor., 40, 586603.

    • Search Google Scholar
    • Export Citation
  • Bean, B., , and E. Dutton, 1968: Radio Meteorology. Dover, 435 pp.

  • Brooks, I. M., , and D. P. Rogers, 2000: Aircraft observations of the mean and turbulent structure of a shallow boundary layer over the Persian Gulf. Bound.-Layer Meteor., 95, 189210.

    • Search Google Scholar
    • Export Citation
  • Brooks, I. M., , A. K. Goroch, , and D. P. Rogers, 1999: Observations of strong surface radar ducts over the Persian Gulf. J. Appl. Meteor., 38, 12931310.

    • Search Google Scholar
    • Export Citation
  • Burk, S. D., , and W. T. Thompson, 1995: Mesoscale modeling of refractive conditions in a complex coastal environment. Proc. Conf. on Propagation Assessment in Coastal Environment, Bremerhaven, Germany, AGARD/NATO, 40.1–40.7.

    • Search Google Scholar
    • Export Citation
  • Burk, S. D., , and W. T. Thompson, 1997: Mesoscale modeling of summertime refractive conditions in the Southern California Bight. J. Appl. Meteor., 36, 2231.

    • Search Google Scholar
    • Export Citation
  • Cummings, J. A., 2005: Operational multivariate ocean data assimilation. Quart. J. Roy. Meteor. Soc., 131, 35833604.

  • Davies, T., , M. J. P. Cullen, , A. J. Malcolm, , M. H. Mawson, , A. Staniforth, , A. A. White, , and N. Wood, 2005: A new dynamical core for the Met Office’s global and regional modelling of the atmosphere. Quart. J. Roy. Meteor. Soc., 131, 17591782.

    • Search Google Scholar
    • Export Citation
  • Edwards, J. M., , and A. Slingo, 1996: Studies with a flexible new radiation code. I: Choosing a configuration for a large-scale model. Quart. J. Roy. Meteor. Soc., 122, 689719.

    • Search Google Scholar
    • Export Citation
  • Essery, R., , M. Best, , and P. Cox, 2001: MOSES 2.2 technical documentation. Hadley Centre Tech. Rep. 30, 31 pp.

  • Golding, B. W., 1987: The U.K. Meteorological Office mesoscale model. Bound.-Layer Meteor., 41, 91107.

  • Golding, B. W., 1990: The Meteorological Office mesoscale model. Meteor. Mag., 119, 8196.

  • Gregory, D., , and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability-dependent closure. Mon. Wea. Rev., 118, 14831506.

    • Search Google Scholar
    • Export Citation
  • Haack, T., , and S. T. Burk, 2001: Summertime marine refractivity conditions along coastal California. J. Appl. Meteor., 40, 673687.

  • Haack, T., , C. Wang, , S. Garrett, , A. Glazer, , J. Mailhot, , and R. Marshall, 2010: Mesoscale modeling of boundary layer refractivity and atmospheric ducting. J. Appl. Meteor. Climatol., 49, 24372457.

    • Search Google Scholar
    • Export Citation
  • Lock, A. P., , A. R. Brown, , M. R. Bush, , G. M. Martin, , and R. N. B. Smith, 2000: A new boundary layer mixing scheme. Part I: Scheme description and single-column model tests. Mon. Wea. Rev., 128, 31873199.

    • Search Google Scholar
    • Export Citation
  • Lystad, S., , and T. Tjelta, 1995: High-resolution meteorological grid for clear air propagation modelling in northern coastal regions. Proc. Conf. on Propagation Assessment in Coastal Environment, Bremerhaven, Germany, AGARD/NATO, 41.1–41.12.

    • Search Google Scholar
    • Export Citation
  • Orlanski, I., 1975: A rational subdivision of scales for atmospheric processes. Bull. Amer. Meteor. Soc., 56, 527530.

  • Panofsky, H. A., , and G. W. Brier, 1958: Some Applications of Statistics to Meteorology. The Pennsylvania State University, 224 pp.

  • Reynolds, R. W., , N. A. Rayner, , T. M. Smith, , D. C. Stokes, , and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 16091625.

    • Search Google Scholar
    • Export Citation
  • Stapleton, J., , V. Wiss, , D. Shanklin, , T. Nguyen, , E. Burgess, , W. Thornton, , and T. Brown, 2001: Radar propagation modeling assessment using measured refractivity and directly sensed propagation ground truth—Wallops Island, VA 2000. Tech. Rep. NSWCDD/TR-01/132, 49 pp.

    • Search Google Scholar
    • Export Citation
  • Wilson, D. R., , and S. P. Ballard, 1999: A microphysically based precipitation scheme for the U.K. Meteorological Office Unified Model. Quart. J. Roy. Meteor. Soc., 125, 16071636.

    • Search Google Scholar
    • Export Citation
Save