Abstract

High-resolution 24-h runs of the Advanced Research version of the Weather Research and Forecasting Model are used to test eight objective methods for estimating convective boundary layer (CBL) depth h, using four planetary boundary layer schemes: Yonsei University (YSU), Mellor–Yamada–Janjic (MYJ), Bougeault–LaCarrere (BouLac), and quasi-normal scale elimination (QNSE). The methods use thresholds of virtual potential temperature Θ_υ, turbulence kinetic energy (TKE), Θ_υ,z , or Richardson number. Those that identify h consistent with values found subjectively from modeled Θ_υ profiles are used for comparisons to fair-weather observations from the 1997 Cooperative Atmosphere–Surface Exchange Study (CASES-97).

The best method defines h as the lowest level at which Θ_υ,z = 2 K km⁻¹, working for all four schemes, with little sensitivity to horizontal grid spacing. For BouLac, MYJ, and QNSE, TKE thresholds did poorly for runs with 1- and 3-km grid spacing, producing irregular h growth not consistent with Θ_υ-profile evolution. This resulted from the vertical velocity W associated with resolved CBL eddies: for W > 0, TKE profiles were deeper and Θ_υ profiles more unstable than for W < 0. For the 1-km runs, 25-point spatial averaging was needed for reliable TKE-based h estimates, but thresholds greater than free-atmosphere values were sensitive to horizontal grid spacing. Matching Θ_υ(h) to Θ_υ(0.05h) or Θ_υ at the first model level were often successful, but the absence of eddies for 9-km grids led to more unstable Θ_υ profiles and often deeper h.

Values of h for BouLac, MYJ, and QNSE, are mostly smaller than observed, with YSU values close to slightly high, consistent with earlier results.

Abstract

Heights of nocturnal boundary layer (NBL) features are determined using vertical profiles from the Advanced Research Weather Research and Forecasting Model (ARW-WRF), and then compared to data for three moderately windy fair-weather nights during the April–May 1997 Kansas-based Cooperative Atmosphere–Surface Exchange Study (CASES-97) to evaluate the success of four PBL schemes in replicating observations. The schemes are Bougeault–LaCarrere (BouLac), Mellor–Yamada–Janjić (MYJ), quasi-normal scale elimination (QNSE), and Yonsei University (YSU) versions 3.2 and 3.4.1. This study’s chosen objectively determined model NBL height h estimate uses a turbulence kinetic energy (TKE) threshold equal to 5% , where TKE′ is relative to its background (free atmosphere) value. The YSU- and MYJ-determined h could not be improved upon. Observed heights of the virtual temperature maximum h _Tvmax and wind speed maximum h _Smax, and the heights h _1wsonde and h _2wsonde, between which the radiosonde slows from ~5 to ~3 m s⁻¹ as it rises from turbulent to nonturbulent air, and thus brackets h, were used for comparison to model results. The observations revealed a general pattern: h _Tvmax increased through the night, and h _Tvmax and h _Smax converged with time, and the two mostly lay between h _1wsonde and h _2wsonde after several hours. Clear failure to adhere to this pattern and large excursions from observations or other PBL schemes revealed excess mixing for BouLac and YSU version 3.2 (but not version 3.4.1) and excess thermal mixing for QNSE under windy conditions. Observed friction velocity was much smaller than model values, with differences consistent with the observations reflecting local skin drag and the model reflecting regional form drag + skin drag.

Abstract

Sources of differences between observations and simulations for a case study using the Noah land surface model–based High-Resolution Land Data Assimilation System (HRLDAS) are examined for sensible and latent heat fluxes H and LE, respectively; surface temperature T_s ; and vertical temperature difference T ₀ − T_s , where T ₀ is at 2 m. The observational data were collected on 29 May 2002, using the University of Wyoming King Air and four surface towers placed along a sparsely vegetated 60-km north–south flight track in the Oklahoma Panhandle. This day had nearly clear skies and a strong north–south soil-moisture gradient, with wet soils and widespread puddles at the south end of the track and drier soils to the north. Relative amplitudes of H and LE horizontal variation were estimated by taking the slope of the least squares best-fit straight line ΔLE/ΔH on plots of time-averaged LE as a function of time-averaged H for values along the track. It is argued that observed H and LE values departing significantly from their slope line are not associated with surface processes and, hence, need not be replicated by HRLDAS. Reasonable agreement between HRLDAS results and observed data was found only after adjusting the coefficient C in the Zilitinkevich equation relating the roughness lengths for momentum and heat in HRLDAS from its default value of 0.1 to a new value of 0.5. Using C = 0.1 and adjusting soil moisture to match the observed near-surface values increased horizontal variability in the right sense, raising LE and lowering H over the moist south end. However, both the magnitude of H and the amplitude of its horizontal variability relative to LE remained too large; adjustment of the green vegetation fraction had only a minor effect. With C = 0.5, model-input green vegetation fraction, and our best-estimate soil moisture, H, LE, ΔLE/ΔH, and T ₀ − T_s , were all close to observed values. The remaining inconsistency between model and observations—too high a value of H and too low a value of LE over the wet southern end of the track—could be due to HRLDAS ignoring the effect of open water. Neglecting the effect of moist soils on the albedo could also have contributed.

Abstract

Fair-weather data from the May–June 2002 International H₂O Project (IHOP_2002) 46-km eastern flight track in southeast Kansas are compared to simulations using the advanced research version of the Weather Research and Forecasting model coupled to the Noah land surface model (LSM), to gain insight into how the surface influences convective boundary layer (CBL) fluxes and structure, and to evaluate the success of the modeling system in representing CBL structure and evolution. This offers a unique look at the capability of the model on scales the length of the flight track (46 km) and smaller under relatively uncomplicated meteorological conditions.

It is found that the modeled sensible heat flux H is significantly larger than observed, while the latent heat flux (LE) is much closer to observations. The slope of the best-fit line ΔLE/ΔH to a plot of LE as a function of H, an indicator of horizontal variation in available energy H + LE, for the data along the flight track, was shallower than observed. In a previous study of the IHOP_2002 western track, similar results were explained by too small a value of the parameter C in the Zilitinkevich equation used in the Noah LSM to compute the roughness length for heat and moisture flux from the roughness length for momentum, which is supplied in an input table; evidence is presented that this is true for the eastern track as well. The horizontal variability in modeled fluxes follows the soil moisture pattern rather than vegetation type, as is observed; because the input land use map does not capture the observed variation in vegetation. The observed westward rise in CBL depth is successfully modeled for 3 of the 4 days, but the actual depths are too high, largely because modeled H is too high. The model reproduces the timing of observed cumulus cloudiness for 3 of the 4 days.

Modeled clouds lead to departures from the typical clear-sky straight line relating surface H to LE for a given model time, making them easy to detect. With spatial filtering, a straight slope line can be recovered. Similarly, larger filter lengths are needed to produce a stable slope for observed fluxes when there are clouds than for clear skies.

Abstract

Fair-weather data along the May–June 2002 International H₂O Project (IHOP_2002) eastern track and the nearby Argonne Boundary Layer Experiments (ABLE) facility in southeast Kansas are compared to numerical simulations to gain insight into how the surface influences convective boundary layer (CBL) structure, and to evaluate the success of the modeling system in replicating the observed behavior. Simulations are conducted for 4 days, using the Advanced Research version of the Weather Research and Forecasting (WRF) model coupled to the Noah land surface model (LSM), initialized using the High-Resolution Land Data Assimilation System (HRLDAS). Because the observations focus on phenomena less than 60 km in scale, the model is run with 1-km grid spacing, offering a critical look at high-resolution model behavior in an environment uncomplicated by precipitation.

The model replicates the type of CBL structure on scales from a few kilometers to ∼100 km, but some features at the kilometer scales depend on the grid spacing. Mesoscale (tens of kilometers) circulations were clearly evident on 2 of the 4 days (30 May and 20 June), clearly not evident on 1 day (22 June), with the situation for the fourth day (17 June) ambiguous. Both observed and modeled surface-heterogeneity-generated mesoscale circulations are evident for 30 May. On the other hand, 20 June satellite images show north-northwest–south-southeast cloud streets (rolls) modulated longitudinally, presumably by tropospheric gravity waves oriented normal to the roll axis, creating northeast–southwest ridges and valleys spaced 50–100 km apart. Modeled cloud streets showed similar longitudinal modulation, with the associated two-dimensional structure having maximum amplitude above the CBL and no relationship to the CBL temperature distribution; although there were patches of mesoscale vertical velocity correlated with CBL temperature. On 22 June, convective rolls were the dominant structure in both model and observations.

For the 3 days for which satellite images show cloud streets, WRF produces rolls with the right orientation and wavelength, which grows with CBL depth. Modeled roll structures appeared for the range of CBL depth to Obukhov length ratios (−z_i /L) associated with rolls. However, sensitivity tests show that the roll wavelength is also related to the grid spacing, and the modeled convection becomes more cellular with smaller grid spacing.