• Beesley, J. A., , C. S. Bretherton, , C. Jakob, , E. L. Andreas, , J. M. Interieri, , and T. A. Uttal, 2000: A comparison of cloud and boundary layer variables in the ECMWF forecast model with observations at the Surface Heat Budget of the Arctic Ocean (SHEBA) ice camp. J. Geophys. Res., 105 , 1233712349.

    • Search Google Scholar
    • Export Citation
  • Briegleb, B. P., 1992: Delta-eddington approximation for solar radiation in the NCAR Community Climate Model. J. Geophys. Res., 97 , 76037612.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., , and G. F. Herman, 1985: Relationships between large-scale heat and moisture budgets and the occurrence of Arctic stratus clouds. Mon. Wea. Rev., 113 , 14411457.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., , W. B. Rossow, , D. Randall, , and J. L. Schramm, 1996: Overview of Arctic cloud and radiation characteristics. J. Climate, 9 , 17311764.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., and Coauthors, 2000: FIRE Arctic Clouds Experiment. Bull. Amer. Meteor. Soc., 81 , 529.

  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121 , 764787.

  • Grell, G. A., , J. Dudhia, , and D. R. Stauffer, 1995: A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 138 pp.

    • Search Google Scholar
    • Export Citation
  • Holtslag, A. M., , and B. A. Boville, 1993: Local versus nonlocal boundary-layer diffusion in a global climate model. J. Climate, 6 , 18251842.

    • Search Google Scholar
    • Export Citation
  • Key, J., 2000: The Cloud and Surface Parameter Retrieval (CASPR) System for polar AVHRR data user's guide. Space Science and Engineering Center, University of Wisconsin—Madison, 62 pp.

    • Search Google Scholar
    • Export Citation
  • Matrosov, S. Y., , A. V. Korolev, , and A. J. Heymsfield, 2002: Profiling cloud ice mass and particle characteristic size from Doppler radar measurements. J. Atmos. Oceanic Technol., 19 , 10031018.

    • Search Google Scholar
    • Export Citation
  • Mlawer, E. J., , S. J. Taubman, , P. D. Brown, , M. J. Iacono, , and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 , 1666316682.

    • Search Google Scholar
    • Export Citation
  • Morrison, H., , M. D. Shupe, , and J. A. Curry, 2003: Modeling clouds observed at SHEBA using a bulk microphysics parameterizations implemented into a single-column model. J. Geophys. Res.,108, 4255, doi:10.1029/2002JD002229.

    • Search Google Scholar
    • Export Citation
  • Persson, P. O. G., , C. W. Fairall, , E. L. Andreas, , P. S. Guest, , and D. K. Perovich, 2002: Measurements near the Atmospheric Surface Flux Group tower at SHEBA: Near-surface conditions and surface energy budget. J. Geophys. Res.,107, 8045, doi:10.1029/2000JC000705.

    • Search Google Scholar
    • Export Citation
  • Pinto, J. O., , H. C. Morrison, , and J. A. Curry, 2000: Advection profiles inferred from radiosonde data for use in single column model simulations at SHEBA. Preprints, Fifth Int. Symp. on Tropospheric Profiling: Needs and Technology, Adelaide, Australia, 217–219.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., , and D. G. Cripe, 1999: Alternative methods for specification of observed forcing in single-column models and cloud system models. J. Geophys. Res., 104 , 2452724545.

    • Search Google Scholar
    • Export Citation
  • Schramm, J. L., , M. M. Holland, , J. A. Curry, , and E. E. Ebert, 1997: Modeling the thermodynamics of a sea ice thickness distribution. 1. Sensitivity to ice thickness resolution. J. Geophys. Res., 102 , 2307923091.

    • Search Google Scholar
    • Export Citation
  • Shupe, M. D., , T. Uttal, , S. Y. Matrosov, , and A. S. Frisch, 2001: Cloud water contents and hydrometeor sizes during the FIRE Arctic Clouds Experiment. J. Geophys. Res., 106 , 1501515028.

    • Search Google Scholar
    • Export Citation
  • Stamnes, K., , R. G. Ellingson, , J. A. Curry, , J. E. Walsh, , and B. D. Zak, 1999: Review of science issues and deployment strategies for the North Slope of Alaska/Adjacent Arctic Ocean (NSA/AAO) ARM site. J. Climate, 12 , 4663.

    • Search Google Scholar
    • Export Citation
  • Uttal, T., and Coauthors, 2002: The Surface Heat Budget of the Arctic Ocean. Bull. Amer. Meteor. Soc., 83 , 255275.

  • Westwater, E. R., , Y. Han, , M. D. Shupe, , and S. Y. Matrosov, 2001: Analysis of integrated cloud liquid and precipitable water vapor retrievals from microwave radiometers during SHEBA. J. Geophys. Res., 106 , 3201932030.

    • Search Google Scholar
    • Export Citation
  • Zhang, M. H., , and J. L. Lin, 1997: Constrained variational analysis of sounding data based on column-integrated budgets of mass, heat, moisture, and momentum: Approach and application to ARM measurements. J. Atmos. Sci., 54 , 15031524.

    • Search Google Scholar
    • Export Citation
  • Zhang, M. H., , S. Xie, , R. T. Cederwall, , and J. J. Yio, 2001: Description of the ARM operational objective analysis system. Tech. Note ARM TR-005, 19 pp.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Time series of temperature: observed, TOBS (dot); ECMWF model forecast, TEC (solid line); and a simple time integration of the ECMWF total tendencies, TINT (dotted line), at approximately 860 m

  • View in gallery

    Histogram distribution of 1-h changes in the horizontal advection tendency for three model levels between 600 and 700 mb from successive runs of the operational ECMWF model for the month of Jul. Distributions are for all times (dashed) and end-of-the-day times (i.e., 35-h forecast valid at 2300 UTC minus 12-h forecast valid at 0000 UTC on the following day) (solid)

  • View in gallery

    Comparison of observed TOA downwelling SW flux with the observed column radiative flux divergence, RAD: (a) Apr, (b) May, (c) Jun, and (d) Jul. The solid line is a least squares fit to the data

  • View in gallery

    The correction term in the (a) temperature advection and (b) water vapor mixing ratio advection in Jul

  • View in gallery

    Scatterplot of the corrected and unmodified ECMWF (a) temperature and (b) water vapor mixing ratio advection for Jul

  • View in gallery

    Time series of the observed (left) temperature and (right) water vapor mixing ratio for Apr–Jul

  • View in gallery

    Time series of errors (predicted minus observed) in the modeled temperature profiles for the (left) baseline and (right) corrected simulations

  • View in gallery

    Time series of errors (predicted minus observed) in the modeled water vapor mixing ratio profiles for the (left) baseline and (right) corrected simulations.

  • View in gallery

    Rms error in the modeled (top) temperature and (bottom) water vapor mixing ratio profiles for the baseline (solid) and corrected (dotted) simulations.

  • View in gallery

    Time series of modeled baseline (dotted), corrected (dashed), and observed (solid) surface downwelling (a) shortwave and (b) longwave radiative flux

  • View in gallery

    Temperature differences between simulations forced with advections calculated using the baseline radiative flux divergence and the (a) baseline radiative flux divergence minus 28 W m−2 and (b) baseline radiative flux divergence plus 28 W m−2

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 25 25 2
PDF Downloads 8 8 0

A New Approach for Obtaining Advection Profiles: Application to the SHEBA Column

View More View Less
  • 1 Program in Atmospheric and Oceanic Sciences, University of Colorado, Boulder, Colorado
© Get Permissions
Full access

Abstract

Time-averaged vertically integrated 3D advections are inferred from heat and moisture budgets obtained from observations at the Surface Heat Budget of the Arctic Ocean (SHEBA) field experiment for April, May, June, and July. Advection was a source of heat and moisture in the column budgets during the time period, balanced mostly by precipitation and radiative cooling. These inferred advections are used to evaluate and correct the 3D temperature and water vapor advection profiles obtained from operational forecasts of the ECMWF model. Advections from the ECMWF model are generally too warm and moist, particularly in July. These biases lead to overpredictions of temperature and water vapor mixing ratio, often exceeding 12 K and 50%, respectively, in monthlong single-column model simulations. A correction algorithm is developed that constrains the ECMWF advections to the observed column budgets, thereby eliminating a first-order source of error in the advective forcing. The approach described here differs from other constrained analysis techniques since it does not require a spatial network of observed or analyzed fields. Simulations forced with the corrected advections show significant improvements in the modeled temperature and water vapor profiles and precipitation. It is demonstrated that using the new observationally constrained advection profiles allows for a less ambiguous evaluation of the model's physical parameterizations.

Corresponding author address: Hugh Morrison, Program in Atmospheric and Oceanic Sciences, University of Colorado, Boulder, CO 80309. Email: hugh@monsoon.colorado.edu

Abstract

Time-averaged vertically integrated 3D advections are inferred from heat and moisture budgets obtained from observations at the Surface Heat Budget of the Arctic Ocean (SHEBA) field experiment for April, May, June, and July. Advection was a source of heat and moisture in the column budgets during the time period, balanced mostly by precipitation and radiative cooling. These inferred advections are used to evaluate and correct the 3D temperature and water vapor advection profiles obtained from operational forecasts of the ECMWF model. Advections from the ECMWF model are generally too warm and moist, particularly in July. These biases lead to overpredictions of temperature and water vapor mixing ratio, often exceeding 12 K and 50%, respectively, in monthlong single-column model simulations. A correction algorithm is developed that constrains the ECMWF advections to the observed column budgets, thereby eliminating a first-order source of error in the advective forcing. The approach described here differs from other constrained analysis techniques since it does not require a spatial network of observed or analyzed fields. Simulations forced with the corrected advections show significant improvements in the modeled temperature and water vapor profiles and precipitation. It is demonstrated that using the new observationally constrained advection profiles allows for a less ambiguous evaluation of the model's physical parameterizations.

Corresponding author address: Hugh Morrison, Program in Atmospheric and Oceanic Sciences, University of Colorado, Boulder, CO 80309. Email: hugh@monsoon.colorado.edu

1. Introduction

The Surface Heat Budget of the Arctic Ocean (SHEBA) field experiment was designed to provide a comprehensive dataset for studying the sea ice–albedo and cloud–radiative feedbacks over the Arctic Ocean (Uttal et al. 2002) using a single-column approach. Observations were made on a multiyear ice floe to characterize the 1D heat and moisture budgets for the ocean, sea ice, and atmosphere. Because of the prohibitive cost of obtaining sounding data at several remote sites surrounding the SHEBA field site, the advective tendencies required to close these budgets were not observed. To address this need, advective tendencies of temperature and water vapor from operational forecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF) model were archived for use in single-column-model (SCM) simulations of SHEBA.

The adequacy of the ECMWF SHEBA dataset for SCM studies and parameterization testing has not been rigorously analyzed up to this point. C. S. Bretherton et al. (2000, personal communication) showed that the monthly averaged physical and advective tendencies were generally in balance, indicating that there was no large systematic drift in the model variables. However, this balance does not guarantee that errors in the advective tendencies are small, as biases in the physical and advective tendencies may offset. Pinto et al. (2000) found that the advective tendencies obtained from successive forecasts often showed significant disagreement, indicating uncertainties associated with the advective forcing dataset.

In this paper, we demonstrate deficiencies in using the ECMWF SHEBA temperature and water vapor advections to directly force SCM simulations. We develop a technique to nudge the predicted ECMWF advective tendencies toward observed time-averaged vertically integrated heat and moisture budgets. This approach is similar to that of Zhang and Lin (1997) and Zhang et al. (2001), who developed a technique to constrain analyzed fields so that column budgets of mass, momentum, heat, and moisture are conserved. Their method was applied to the Atmospheric Radiation Measurement Program (ARM) Cloud and Radiation Testbed (CART) site to produce dynamically and thermodynamically consistent advective tendencies and vertical velocities for column modeling studies. Their method relies on data supplied by an observational network of surface stations and sounding sites surrounding the ARM CART (Zhang et al. 2001). In contrast, the new technique that we have developed requires only a column of observed data and an initial guess at the advection profiles that may be provided by an operational NWP model. In this paper, we describe the method developed to constrain modeled advective tendencies to the observations. We then show how using this new method allows for a less ambiguous evaluation of model parameterizations in simulations performed over the Arctic Ocean.

2. ECMWF model tendencies

Profiles of the predicted temperature, T, and water vapor mixing ratio, q, were archived from successive runs of the ECMWF model version 13R4 for the SHEBA year. The ECMWF data have 31 vertical levels and nominal 60-km horizontal resolution for the grid corresponding with the SHEBA ship track (the simulations were improved by the ECMWF in 2000). The hourly time tendencies (total, physical, and advective) of T and q for each day were taken from the 12–35-h forecast that had been initialized at 1200 UTC on the preceding day. The 12–35-h forecast period was chosen so that errors resulting from the model initialization (e.g., data-poor regions) were minimized. A continuous time series of T and q tendency profiles was developed by concatenating the 12–35-h forecast for each day.

A key aspect of this concatenated time series is that the modeled T and q profiles are maintained close to the observations by the reinitialization of each forecast. Soundings launched at SHEBA were routinely used in the initialization of the model (Beesley et al. 2000). Despite the inclusion of these data, the predicted profiles of temperature and moisture quickly develop biases, likely because of the lack of radiosonde observation (raob) data upstream of the SHEBA site. The time series of predicted temperature, TEC, at ∼860 mb (Fig. 1) demonstrates how reinitializing keeps the simulated temperature close to the observed value. The difference in T and q between hours 12 and 35 of the previous day's forecast may be thought of as an initialization nudging term. Time integration of the concatenated time series of total tendencies demonstrates the importance of initialization nudging in maintaining TEC close to observations (see Fig. 1). The temperature given by time integration of the total tendencies, TINT, agrees with TEC during the first 24-h period but then abruptly deviates at the start of the second day. The divergence of the two time series is caused by an accumulation of errors imbedded in the total tendency due to biases in both the model physics and the predicted synoptic pattern.

The ECMWF 3D T and q advections may be evaluated by checking for self-consistency within the concatenated time series of tendencies. Changes in the horizontal advective tendency with time may be used to reveal inconsistencies in the time series. The frequency distribution of 1-h time changes in the horizontal advection tendency for three levels between 600 and 700 mb from the entire month of July shows a Gaussian shape with a peak near zero (Fig. 2). A subset of these time changes in horizontal advection, which includes only the differences between the 35-h forecast valid at 2300 UTC and the 12-h forecast valid at 0000 UTC the next day, reveals inconsistencies in the concatenated time series. If there were no systematic “drift” in the advective tendencies between the two forecasts, the frequency distribution of this subset would be similar to that of the full dataset (i.e., Gaussian with a mean near zero). However, the frequency distribution of this subset reveals a skewed distribution with a mean value of 3.8 K day−1, indicating a systematic warm “drift” in the 36-h forecasts of horizontal temperature advection. This warm bias is evident in the 3D advections as well, as is discussed in the remaining sections.

3. Observational data used in column budgets

Meteorological observations gathered at SHEBA are used to analyze the heat and moisture budgets, derive column advections of heat and water vapor, force and initialize the SCM simulations, and evaluate the model results. SHEBA was coordinated with the First International Satellite Cloud Climatology Project (ISCCP) Regional Experiment Arctic Clouds Experiment (FIRE–ACE; Curry et al. 2000) and ARM (Stamnes et al. 1999). Data were collected from a variety of in situ and remote instruments over an annual cycle (1 October 1997 to 1 October 1998), focusing on surface energy and sea ice mass balance (Uttal et al. 2002). ARM provided several surface-based instruments to measure clouds and radiation, while research aircraft provided in situ and remote measurements of atmospheric and surface properties as part of FIRE-ACE.

Rawinsondes measuring temperature, humidity, and winds were launched from the ice station four times per day during the period of interest. A Nipher-shielded snow gauge system measured total accumulations of precipitation on a daily basis (or as new precipitation warranted). The precipitation data were corrected by the SHEBA Project Office to take into account various factors of the high-latitude environment (e.g., blowing snow, sublimation).

Surface heat and radiative fluxes, temperature, pressure, and albedo were directly measured at the Atmospheric Surface Flux Group (ASFG) tower (Persson et al. 2002). Top-of-the-atmosphere (TOA) radiative fluxes for the 5-km pixel associated with the SHEBA site were derived from Advanced Very High Resolution Radiometer (AVHRR) measurements as described by Key (2000). These radiative fluxes are only available twice per day.

Cloud properties were retrieved from a combination of microwave radiometer and cloud radar measurements. Liquid water path (LWP) was estimated from microwave radiometer brightness temperatures as described in Westwater et al. (2001). Ice water path (IWP) was retrieved from 35-GHz cloud radar as described by Shupe et al. (2001). LWP retrievals have an instantaneous error of 25 g m−2. Retrievals of ice water content showed a relative deviation of 55% compared to in situ measurements made during late April near SHEBA (Matrosov et al. 2002). Uncertainties in the retrievals are reduced by averaging over longer time periods.

4. Observed heat and moisture budgets

Time-averaged vertically integrated (TAVI) 3D T and q advections are derived from observed heat and moisture budgets in order to evaluate and correct the ECMWF advective tendencies. We start with the “advective” form of the scalar conservation equation:
i1520-0493-132-3-687-e1
Here q is the water vapor mixing ratio [although (1) describes any scalar quantity], V is the horizontal wind vector, p is the pressure, ω is the vertical pressure velocity, and P is the physical tendency of the variable. For water vapor, the physics term P is given by
i1520-0493-132-3-687-e2
where ∂qc/∂t is the total time tendency of the condensed water mixing ratio, qc, ∂ϕ/∂t is the precipitation rate, and prime quantities denote subgrid fluctuations. For simplicity, we have ignored the tendency in qc due to advection and turbulent diffusion. Here, (1) and (2) correspond with large-scale atmospheric fields so that the grid-scale quantities are assumed to be domain averages.
Combining (1) and (2), dividing by the acceleration of gravity, g, and the time interval, Δt, and integrating over dt from 0 to Δt and dp from 0 to ps, the surface pressure
i1520-0493-132-3-687-e3
Because the relative variation of pressure with time is much less than that of q or qc, pressure is assumed to be an independent variable, so that (3) may be solved as an iterated integral. This allows (3) to be expressed in terms of quantities that were measured at SHEBA. Integrating (3) over dt gives
i1520-0493-132-3-687-e4
Here qi and qf are the initial and final values of water vapor mixing ratio, qc,i and qc,f are the initial and final values of the condensed water mixing ratio, ϕt is the total precipitation, and δ is the time-averaged tendency in q due to subgrid vertical motion.
Using the definition of precipitable water vapor, W, and condensed water path, CWP,
i1520-0493-132-3-687-e5
Equation (4) may be expressed in terms of these quantities. The vertically integrated total precipitation, PREC, is the total mass of water per unit area leaving the column at the surface because of precipitation. The vertically integrated turbulent diffusion tendency is simply the time-averaged surface latent heat flux, LH, divided by the latent heat of vaporization, Lυ, to convert it to a mass flux. Equation (4) may then be written as
i1520-0493-132-3-687-e7
where Wadv refers to the TAVI 3D advection of water vapor,
i1520-0493-132-3-687-e8
Equation (7) is an expression for the TAVI moisture budget. Values of Wadv are calculated from (7), using observations of W, CWP, PREC, and LH from SHEBA, calculated each month (i.e., Δt = 1 month) for the April–July period of interest.
The TAVI heat budget, equivalent to (7), is
i1520-0493-132-3-687-e9
where Ls is the latent heat of sublimation, PRECs is the total liquid-equivalent snowfall at the surface, PRECr is the total rainfall at the surface, SH is the time-averaged surface turbulent sensible heat flux, and RAD is the time-averaged radiative flux divergence of the column. Values of RAD must be derived empirically since TOA values were only available twice per day (solar noon and 6 h prior). A model was developed to relate RAD to the downwelling shortwave flux at TOA, SWd (see Fig. 3):
βdxy
where β is the slope, x is the sample mean of SWd, and y is the sample mean RAD, where RAD is given by
i1520-0493-132-3-687-e11
Here TOAu is the time-averaged upwelling flux at the top of the atmosphere, SURu is the time-averaged upwelling flux at the surface, and SURd is the time-averaged downwelling flux at the surface. Values of β, x, and y are calculated for each month during the period of interest. Monthly averaged values of SWd, calculated directly using hourly zenith angles and a solar constant of 1366.5 W m−2, are used in (10) to calculate RAD.
The vertically integrated enthalpy, H, is expressed as
i1520-0493-132-3-687-e12
where cp is the isobaric specific heat capacity for dry air. The TAVI 3D temperature advection, Hadv, is expressed as
i1520-0493-132-3-687-e13
Here R is the gas constant for dry air. Note that we convert the temperature advection to a heat flux in (13) for consistency with the other heat budget terms in (9). Hereafter, time averaging and vertical integration of the advective, physical, or total temperature tendency implies conversion to a heat flux. Because the adiabatic compression/expansion term, RωT/cpp, results from large-scale vertical motions supplied by the dynamic forcing, it is included in (13). This is consistent with the total adiabatic forcing available in the ECMWF SHEBA dataset. Values of Hadv are calculated from (9), using observations of H, LWP, IWP, PRECr, PRECs, SH, and RAD obtained at SHEBA, calculated each month (i.e., Δt = 1 month) for the April–July period of interest.

Values of the budget terms in (7) and (9) are shown in Table 1. The monthly moisture budgets are dominated by precipitation and advection, while the monthly heat budgets are dominated by radiative cooling and advection. There is an overall warming and moistening in April, May, and June, and a slight cooling and drying in July. The SHEBA atmospheric column acts as a sink for both heat and moisture during this time period, as the TAVI advections of temperature and water vapor are positive, balanced mostly by radiative cooling and precipitation. These results are similar to those reported by Curry and Herman (1985) for June 1980 over the Beaufort Sea region. Surface turbulent exchange of heat and water vapor is relatively small compared with the other budget terms. Cloud water is much less important in the budgets and may be ignored without greatly affecting the derived values of Hadv and Wadv.

5. Evaluation and correction of the ECMWF advections

The ECMWF 3D T and q advections are used to calculate modeled TAVI advections that are compared with observed TAVI advections, determined as discussed in the previous section. The difference between the modeled and observed TAVI advections is used to develop a correction term that may be used in SCM studies at SHEBA.

a. Evaluation of the ECMWF advections

Errors in the ECMWF TAVI advective tendencies, qADVerr and TADVerr, may be directly calculated from
i1520-0493-132-3-687-e14
Here Wadv and Hadv are the TAVI advections derived from the observed budgets in the previous section, and Wadv,EC and Hadv,EC are TAVI advections calculated from the ECMWF 3D T and q advection profiles following (8) and (13). Errors in the ECMWF TAVI advections can also be expressed as
i1520-0493-132-3-687-e16
where qTOTerr and TTOTerr are the errors in the TAVI total tendencies, and TPHYerr and qPHYerr are the errors in the TAVI physical tendencies. Errors qTOTerr and TTOTerr are calculated by the following expressions:
i1520-0493-132-3-687-e18
where the total time tendency errors are calculated from the observed profiles
i1520-0493-132-3-687-e20
Here q0,24 and T0,24 are the observed profiles of water vapor mixing ratio and temperature at the beginning and end of each 24-h ECMWF forecast period (see section 2), and (∂q/∂t)EC and (∂T/∂t)EC are the profiles of total time tendencies associated with the q and T provided by the ECMWF, averaged over the same forecast period. Thus, (∂q/∂t)err and (∂T/∂t)err are profiles of the daily average errors in the total time tendency of q and T.

Values of the terms in (14)–(17) are calculated for each month (i.e., Δt = 1 month) for the April–July period of interest (Table 2). The TAVI ECMWF advections have a warm, moist bias relative to the TAVI advections inferred from observations. The TAVI ECMWF physical tendencies have a cold, dry bias in May, June, and July that tends to offset errors in the advective tendencies of T and q, maintaining the approximate balance in the total tendencies described by C. S. Bretherton et al. (2000, personal communication). However, errors in the TAVI advective and physical tendencies of temperature do not offset in April, resulting in a large positive bias in the TAVI total tendency of temperature in this month.

b. Correction of the ECMWF advections

In this section we develop a correction term for the ECMWF advections that is a function of time and height. In order to do this, we must generate equations that contain time and height dependences for each term in (16) and (17). The vertical and temporal distribution of the total tendency error (qTOT and TTOT) is obtained using (18)–(21). Errors in the TAVI ECMWF advections may be related to the time- and height-varying advections following
i1520-0493-132-3-687-e22
where qADVEC(t, p) and TADVEC(t, p) are the time- and height-varying ECMWF advections, and qADV′(t, p) and TADV′(t, p) are the “true” time- and height-varying advections.
Determination of the vertical and temporal distribution of the physical tendency error is more problematic because it depends on observed quantities, such as the condensation and freezing rates, that were not well characterized during SHEBA. To avoid this complication, we assume that the temporal and spatial variation in the physics tendency error is proportional to the magnitude of the physical tendency predicted by the model:
i1520-0493-132-3-687-e24
where (∂q/∂t)err,phy(t, p) and (∂T/∂t)err,phy(t, p) are the errors in the ECMWF physical tendencies as a function of time and vertical level, and ζW(t, p) and ζH(t, p) are vertical and temporal weights calculated from the magnitude of the ECMWF water vapor mixing ratio physical tendency, |(∂q/∂t)phyec|, and temperature physical tendency, |(∂T/∂t)phyec|:
i1520-0493-132-3-687-e26

The assumption of proportionality between the physical tendency error and the magnitude of the physical tendencies predicted by the ECMWF model forecast is crucial in calculating the vertical distribution of the physical tendency error. This is particularly true in determining the correction profiles for the water vapor advection tendency, which varies by several orders of magnitude over the depth of the atmospheric column. It is noted that this assumption is not strictly valid and may be of particular concern when there are errors in the predicted cloud field since clouds produce a large perturbation in the physical tendencies. Fortunately, the formation, dissipation, and vertical distribution of clouds are generally well predicted by the ECMWF model for the SHEBA column despite biases in the predicted water paths and phase (Beesley et al. 2000). The primary limitation of this assumption is that the weights cannot account for changes in the sign of the physical tendency errors either vertically or temporally.

Combining (18), (19), (22), (23), (24), and (25) with (16) and (17), multiplying by gΔt (and dividing by cp for the temperature equations), differentiating with respect to time and pressure, and solving for qADV′(t, p) and TADV′(t, p) yields the following expressions for the corrected advective tendencies:
i1520-0493-132-3-687-e28
The advection correction term is given by the difference between the time-varying physical (rhs, term 1) and total tendency error (rhs, term 2) profiles. Use of this correction term in monthlong SCM simulations constrains the predicted T and q profiles to satisfy the observed column budgets; however, uncertainties associated with the partitioning of the TAVI physical tendency error may degrade the simulation in some situations, particularly over short time scales. Time–height plots of the correction term for T and q advection in July (Fig. 4) reveal its complex vertical and temporal structure. The two time scales evident in the correction term result from the use of daily-average total tendency errors and hourly variations in the physical tendency errors. The corrected and uncorrected 3D advections may differ substantially at any given time and vertical level, as evidenced by Fig. 5. The mean biases indicate that the ECMWF forecasted advections were too warm and moist in July. The pattern of the scatter indicates that the largest errors were obtained where the initial predicted advective tendency was small (i.e., near 0).

6. SCM simulation results

Simulations of the SHEBA atmospheric column are conducted using the Arctic single-column model (ARCSCM; Morrison et al. 2003) forced with both the corrected and uncorrected ECMWF advective tendencies. We briefly describe ARCSCM and then discuss the model results and implications for evaluating simulations forced with the corrected advections.

ARCSCM was developed by Morrison et al. (2003) for the purpose of evaluating physical parameterizations and understanding the complex thermodynamic interactions that occur in a column over multiyear sea ice in the Arctic. Prognostic variables in ARCSCM include temperature and water vapor, cloud liquid water, cloud ice, snow, and rain mixing ratios. The time tendencies of temperature, T, and water vapor mixing ratio, q, are given by the following equations:
i1520-0493-132-3-687-e30
where V is the horizontal wind vector, ω is the pressure vertical velocity, p is the pressure, R is the gas constant for air, cp is the specific heat at constant pressure, (∂T/∂t)RAD is the radiative heating rate, (∂T/∂t)TUR and (∂q/∂t)TUR are the tendencies due to turbulent diffusion, (∂q/∂t)CL is the water vapor tendency resulting from cloud microphysical processes, and (∂T/∂t)LAT is the latent heating rate. The adiabatic compression temperature tendency is incorporated into the specified advective forcing. The simulations represent a scale similar to the nominal grid spacing used by the ECMWF model, which is about 60 km. We also use the same number of levels (31) used in the ECMWF model in order to avoid uncertainties of interpolating highly variable advective tendencies.

Shortwave radiative transfer is treated using the two-stream delta-Eddington method following Briegleb (1992). Longwave radiative transfer is given by the Rapid Radiative Transfer Model (RRTM; Mlawer et al. 1997). The boundary layer parameterization is a first-order nonlocal scheme following Holtslag and Boville (1993). Sensible and latent surface turbulent fluxes over pack ice are calculated from Schramm et al. (1997). Cloud and precipitation physics is divided into two categories based upon the separation between large-scale (resolvable) and subgrid-scale processes. Resolvable-scale microphysical processes are parameterized following Grell et al. (1995). Subgrid convective clouds are parameterized following Grell (1993).

Monthlong simulations are performed with ARCSCM for April, May, June, and July using the original ECMWF advections and a new set of advections corrected using the method outlined in the previous section. We will refer to the results of these two sets of simulations as “baseline” and “corrected,” respectively. Simulations are initialized using profiles of T and q that were constructed from rawinsonde and tower data. The 3D wind field (required for the calculation of surface turbulent fluxes and vertical diffusion) is specified from successive operational forecasts of the ECMWF model.

The observations show an overall warming and moistening of the atmosphere from April to July (Fig. 6). However, the warming and moistening that occurred was often interrupted by brief periods of cooling and/or drying (e.g., 8–11 May, 20–25 June) that reflected changes in synoptic-scale circulation patterns. Early- and mid-April temperatures in the lower atmosphere were unseasonably warm. A notable feature is the presence of inversions in both the T and q profiles in the lower troposphere during much of the period. These inversions are a dominant feature of the climate in the Arctic and as such have been extensively studied [for a review see Curry et al. (1996)]. More detailed accounts of the meteorological conditions that occurred during the 4-month period are given by Curry et al. (2000).

Time–height error plots show that the baseline results are generally too warm and moist, with the exception of the June simulation (Figs. 7 and 8). These results are in general agreement with the warm, moist bias found in the TAVI ECMWF advective tendencies (see Table 2). The sign and magnitude of these biases are not systematic, as they vary both vertically and temporally in a complex manner. For example, in April, the largest temperature errors occur within the boundary layer during the first 3 weeks, but then propagate upward into the middle troposphere during the last week of the simulation. In July, a cold bias develops near the surface and grows in depth and magnitude after 20 July.

Plots of rms T and q error show that, with only a few exceptions, the biases are reduced across the vertical profile in each month using the corrected advections (Fig. 9). Time–height T and q error plots for the baseline and corrected simulations show that the reduction of error occurs throughout the course of the simulation, although the magnitude of the correction often appears to be too small (Figs. 7 and 8). A notable exception occurs in July, when the corrections appear to be too large, resulting in profiles that are too cold and dry below about 600 mb. The time–height error plots indicate that this cold, dry bias develops almost immediately and grows in magnitude and depth with time, possibly indicating that a positive feedback is occurring in the model. It is shown in section 7 that this evolving cold bias is likely associated with biases in the simulated cloud properties. A similar cold bias develops in the corrected June simulation, although it is much smaller in magnitude.

Biases in T and q in the corrected simulations may be attributed to problems with the vertical and temporal distribution of the correction terms in addition to the model physics. Nonetheless, the overall reduction of biases regardless of their sign [e.g., the moist (dry) bias in the first (second) half of April] suggests that the vertical and temporal distribution of the advections is improved with the correction procedure. Reducing model drift in T and q allows us to more effectively evaluate the physical parameterizations implemented in ARCSCM, as is discussed further in section 7.

Significant biases associated with the predicted cloud water and precipitation are apparent in the baseline simulations (Fig. 10; Tables 3 and 4). The monthly mean LWP is underpredicted in April and May and substantially overpredicted in July. The IWP is overpredicted in both the baseline and corrected simulations with the exception of April. Improvements in the modeled cloud water using the corrected advections are somewhat less apparent than improvements in the T and q profiles, except in July when the large biases in the baseline LWP are significantly reduced. The fact that the underprediction of the LWP in April and May is accentuated in the corrected run likely demonstrates a deficiency of the cloud microphysical parameterization. This will be discussed further in the following section. The correction term nudges the modeled monthly accumulated precipitation closer to the observations in all months; however, large positive biases remain in June and July as a result of uncertainties in the correction term and/or the model physics.

Biases in the modeled surface radiative fluxes are consistent with biases in the condensed water paths. The downwelling shortwave (SW) flux is overpredicted in April and May and substantially underpredicted in July. The downwelling longwave (LW) flux exhibits the opposite trend. Observed and modeled downwelling radiative fluxes at the surface for a period in July are shown in Fig. 10. Note that clear-sky periods indicated by the observed downwelling LW flux at the surface (i.e., when the flux decreases to ∼250 W m−2) are not captured by the model; both the baseline and corrected simulations predict continuous cloud cover during the time period, resulting in the continuously elevated downwelling LW flux in Fig. 10. However, when clouds are evident the predicted downwelling LW flux is improved in the corrected simulation as the clouds are optically thinner (Table 3). Similarly, biases in the downwelling SW flux tend to be reduced using the corrected advections (note the large improvements on clear-sky days 198, 202, and 207), although large biases are still present.

7. Discussion

Nudging the ECMWF advections toward the observed TAVI advections eliminates a first-order source of error in the model and subsequently reduces model drift. By constraining model drift we allow the physical parameterizations to respond to a more realistic set of conditions. Since we are nudging the advection terms and not T and q directly, we are allowing the profiles to vary more freely in response to the physical parameterizations. Direct nudging of the T and q profiles toward observations, which is often done in SCM experiments, introduces a term with no corresponding physical process and limits feedbacks between the parameterizations. If the nudging of T and q is coupled to the model physics and treated as an advective term [e.g., the “relaxation forcing” method discussed in Randall and Cripe (1999)], this problem is circumvented, but then model drift is allowed to feed back on the forcing, resulting in large errors in parameterized processes driven by rates (e.g., precipitation) (Randall and Cripe 1999). The corrected advections presented here are constrained only to the observations and do not depend on the modeled profiles. Use of these advections reduces simultaneously model drift and tendency inputs (i.e., rates) to drive the model physics (as evidenced by the significantly improved precipitation). Minimizing both sources of error is necessary in order to evaluate parameterizations that depend on the state variables themselves and their associated tendencies.

Using the corrected advections to reduce uncertainty in evaluating model parameterizations is demonstrated by an analysis of the simulated cloud water paths. A review of the monthly mean LWP obtained from the two sets of simulations reveals two interesting results (see Table 3). In spring (i.e., April and May), the simulated LWP in the corrected run is degraded compared to the baseline values. This result demonstrates how use of the corrected advective tendencies improves our ability to detect deficiencies in the model physics. Errors in the modeled T and q profiles and precipitation are reduced in April and May in the corrected simulations compared to baseline (see Figs. 7 and 8). The more realistic profiles are colder and drier and hence result in smaller values of LWP since less water vapor is available for condensation, and the production of ice is generally favored at lower temperatures in this particular microphysical scheme (Morrison et al. 2003). Even though the T and q profiles and precipitation are improved in the corrected simulations, the LWP bias is increased, thereby revealing deficiencies in the model that would otherwise appear to be less substantial. In contrast, use of the corrected advections improves the simulated LWP in July, not only in the monthly averaged value (Table 3), but in the time series as well (not shown). Use of the corrected advections therefore suggests that much of the substantial LWP bias in the July baseline simulation is associated with the advective forcing rather than the model physics since the T and q profiles and precipitation are simultaneously improved along with the LWP.

To study sensitivities and feedbacks in a SCM, the response or signal associated with the parameterized physical processes must be distinguishable from that imparted by errors in the advective forcing. This is illustrated with July data in which the large, warm bias in the advective tendencies overwhelms the cold bias in the radiative flux divergence term. A similar cold bias in the radiative term is seen in the corrected simulation, although it is smaller in magnitude (Table 5). Excessive radiational cooling in the simulations is concomitant with an overprediction of cloud fraction, LWP, and IWP. Thus, deficiencies in the model parameterizations (i.e., deficiencies in the microphysics scheme resulting in poorly predicted cloud properties) tend to nudge the predicted profiles toward a cold bias in both the baseline and corrected simulations. This cold bias is reflected in the temperature profile of the corrected simulation; however, in the baseline simulation, the signal is completely overwhelmed by excessive warm advection, resulting in profiles that are significantly too warm (see Fig. 7). Similar biases in the cloud fraction and column radiative flux divergence appear in June (not shown). In this month, the ECMWF TAVI T and q advections are close to observed (see Table 2); correspondingly, the cold bias in the lower troposphere is present in both the baseline and corrected simulations (Fig. 7).

The accuracy of the corrected advections is dependent not only upon assumptions used in the correction method, but also upon the accuracy and representativeness of the observations. Since the column radiative flux divergence and precipitation flux are the largest terms in the observed heat and moisture budgets, respectively, relative uncertainty in these terms will most influence the results. In particular, we expect the largest errors to be associated with the radiative flux divergence since this quantity was not directly observed and was instead derived using a correlation model (see section 4). An error analysis is conducted to determine the influence of uncertainties in the radiative flux divergence. We estimate this uncertainty to be about ±28 W m−2 for July based upon the rms difference between the measurements and the regression line (see Fig. 3d). Two sets of temperature advection profiles are calculated using this estimate of uncertainty. Figure 11 shows time–height plots of the difference in temperature predicted using the baseline corrected advections and advections determined using radiative flux divergences of the baseline value ±28 W m−2. The simulations exhibit moderate sensitivity using the different values of temperature advection. The maximum temperature differences are about ±6 K, but they are often less than ±3 K. The sensitivity to error in the radiative flux divergence is much smaller than that exhibited in the simulation using corrected advective tendencies (see Fig. 7). July simulations forced with two datasets of water vapor advection calculated by assuming a 50% error in the observed precipitation results in generally less than a ±15% relative difference in the water vapor mixing ratio. This range is also much smaller than the differences between the baseline and corrected simulations.

Since the corrected advections are constrained by the observed TAVI heat and moisture budgets, there is uncertainty associated with instantaneous values of the corrections because of the assumptions required to distribute the corrections vertically and temporally. In principle, the corrected advections may be constrained to observed budgets averaged over time scales as short as about 1 day (determined from the frequency of the measurements). However, uncertainties associated with the observations increase as the number of data points is reduced as a consequence of shorter time-averaging periods (particularly for the radiative flux divergence term). In addition, differences in the horizontal scale associated with the measurements and the ECMWF model become important as the time scale is reduced. To examine the influence of the averaging time scale on the corrected advections, we successively reduced the time scale and saw little influence on the model results down to time scales as short as a week. Errors in the ECMWF advective tendencies grew larger as the averaging time scale was reduced. ECMWF TAVI T and q advections averaged over one week periods in May were significantly more biased than the monthly averaged quantities. We note that this increased bias may also reflect additional uncertainty in the observed TAVI advections as the time period for averaging the observations is reduced.

8. Summary and conclusions

The 3D advective tendencies obtained from operational forecasts of the ECMWF model have been evaluated using observations collected during SHEBA. Time-averaged vertically integrated advective tendencies were inferred from monthly averaged quantities for column heat and moisture budgets. Advection dominated both budgets during the spring-to-summer transition, with the main sink terms being radiative flux divergence and precipitation. The column-integrated advections from the ECMWF model exhibit a warm and moist bias from April through July, with the largest biases occurring in July. However, it is noted that using monthly averages to determine the sign and magnitude of these biases may be somewhat misleading. In May, for example, biases in the monthly average column temperature advections are small because of the balance between a strong cold bias from 16 to 20 May and an overall warm bias in the rest of the month.

A correction algorithm was developed to improve the forcing dataset for single-column modeling at SHEBA, resulting in an observationally constrained version of the ECMWF advections. Since only a single column of data is analyzed, the approach developed here is inherently less robust than the ARM CART variational analysis described by Zhang et al. (2001). The primary advantage of the new analysis is that it does not require information on the horizontal distributions of temperature and water vapor or the 3D wind field as the variational approach does. We envision this approach to be used to improve modeled advections for SCM studies when only a column of observations are available.

The new method uses a correction term that is found by differencing the total tendency error and the physical tendency error. The total tendency errors are determined by comparing profiles of the predicted ECMWF model total tendencies to the observed 24-h changes in the T and q profiles. In contrast, the physical tendency error in the ECMWF model forecasts can only be determined as a time-averaged vertically integrated quantity. In order to distribute the error in time and space, we assume that the sign of the physical tendency error is constant with height and has a magnitude that is proportional to the predicted physical tendency. This assumption is the primary source of uncertainty in the corrected advection profiles. A second source of uncertainty arises from measurements errors and issues regarding the representativeness of the observations.

The corrected advections were tested in ARCSCM for four different monthlong simulations beginning in April. Results indicate that using the corrected advections significantly reduces model drift while allowing T and q profiles to vary freely in response to the physical parameterizations. This allows for a more stringent evaluation of the physical parameterizations than can be achieved in simulations that use direct observational nudging of the T and q profiles because it allows feedbacks between the thermodynamic profiles and the parameterized physics. Use of the corrected advections reveals that the liquid water path is substantially underpredicted in April and May, likely reflecting difficulties in the treatment of cold cloud processes (i.e., mixed-phase clouds).

This correction algorithm may be applied to any column of a forecast or analysis where the necessary measurements are available. Its utility depends upon the quality of observations and the background forecast or analysis. The results presented here are dependent upon the physical parameterizations in ARCSCM, which are different from those used by the ECMWF model. SCMs with different physical parameterizations could potentially produce widely varying results using the constrained advections. Because of uncertainty associated with the temporal distribution of the correction term, the constrained advections are best suited for longer-term simulations such as coupled climate simulations or studies requiring the statistical evaluation of parameterizations.

Acknowledgments

This research is supported by the NSF SHEBA Grant OPP-0084225. We are grateful to C. Jakob, S. de Roode, and C. Bretherton for providing the ECMWF column dataset for SHEBA. Retrieved cloud properties were provided by M. Shupe and the NOAA Environmental Technology Laboratory. Precipitation and rawinsonde data were obtained from the University of Washington Applied Physics Laboratory. Surface measurements were obtained from the SHEBA Atmospheric Surface Flux Group. APP data was kindly made available by J. Key.

REFERENCES

  • Beesley, J. A., , C. S. Bretherton, , C. Jakob, , E. L. Andreas, , J. M. Interieri, , and T. A. Uttal, 2000: A comparison of cloud and boundary layer variables in the ECMWF forecast model with observations at the Surface Heat Budget of the Arctic Ocean (SHEBA) ice camp. J. Geophys. Res., 105 , 1233712349.

    • Search Google Scholar
    • Export Citation
  • Briegleb, B. P., 1992: Delta-eddington approximation for solar radiation in the NCAR Community Climate Model. J. Geophys. Res., 97 , 76037612.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., , and G. F. Herman, 1985: Relationships between large-scale heat and moisture budgets and the occurrence of Arctic stratus clouds. Mon. Wea. Rev., 113 , 14411457.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., , W. B. Rossow, , D. Randall, , and J. L. Schramm, 1996: Overview of Arctic cloud and radiation characteristics. J. Climate, 9 , 17311764.

    • Search Google Scholar
    • Export Citation
  • Curry, J. A., and Coauthors, 2000: FIRE Arctic Clouds Experiment. Bull. Amer. Meteor. Soc., 81 , 529.

  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121 , 764787.

  • Grell, G. A., , J. Dudhia, , and D. R. Stauffer, 1995: A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 138 pp.

    • Search Google Scholar
    • Export Citation
  • Holtslag, A. M., , and B. A. Boville, 1993: Local versus nonlocal boundary-layer diffusion in a global climate model. J. Climate, 6 , 18251842.

    • Search Google Scholar
    • Export Citation
  • Key, J., 2000: The Cloud and Surface Parameter Retrieval (CASPR) System for polar AVHRR data user's guide. Space Science and Engineering Center, University of Wisconsin—Madison, 62 pp.

    • Search Google Scholar
    • Export Citation
  • Matrosov, S. Y., , A. V. Korolev, , and A. J. Heymsfield, 2002: Profiling cloud ice mass and particle characteristic size from Doppler radar measurements. J. Atmos. Oceanic Technol., 19 , 10031018.

    • Search Google Scholar
    • Export Citation
  • Mlawer, E. J., , S. J. Taubman, , P. D. Brown, , M. J. Iacono, , and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 , 1666316682.

    • Search Google Scholar
    • Export Citation
  • Morrison, H., , M. D. Shupe, , and J. A. Curry, 2003: Modeling clouds observed at SHEBA using a bulk microphysics parameterizations implemented into a single-column model. J. Geophys. Res.,108, 4255, doi:10.1029/2002JD002229.

    • Search Google Scholar
    • Export Citation
  • Persson, P. O. G., , C. W. Fairall, , E. L. Andreas, , P. S. Guest, , and D. K. Perovich, 2002: Measurements near the Atmospheric Surface Flux Group tower at SHEBA: Near-surface conditions and surface energy budget. J. Geophys. Res.,107, 8045, doi:10.1029/2000JC000705.

    • Search Google Scholar
    • Export Citation
  • Pinto, J. O., , H. C. Morrison, , and J. A. Curry, 2000: Advection profiles inferred from radiosonde data for use in single column model simulations at SHEBA. Preprints, Fifth Int. Symp. on Tropospheric Profiling: Needs and Technology, Adelaide, Australia, 217–219.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., , and D. G. Cripe, 1999: Alternative methods for specification of observed forcing in single-column models and cloud system models. J. Geophys. Res., 104 , 2452724545.

    • Search Google Scholar
    • Export Citation
  • Schramm, J. L., , M. M. Holland, , J. A. Curry, , and E. E. Ebert, 1997: Modeling the thermodynamics of a sea ice thickness distribution. 1. Sensitivity to ice thickness resolution. J. Geophys. Res., 102 , 2307923091.

    • Search Google Scholar
    • Export Citation
  • Shupe, M. D., , T. Uttal, , S. Y. Matrosov, , and A. S. Frisch, 2001: Cloud water contents and hydrometeor sizes during the FIRE Arctic Clouds Experiment. J. Geophys. Res., 106 , 1501515028.

    • Search Google Scholar
    • Export Citation
  • Stamnes, K., , R. G. Ellingson, , J. A. Curry, , J. E. Walsh, , and B. D. Zak, 1999: Review of science issues and deployment strategies for the North Slope of Alaska/Adjacent Arctic Ocean (NSA/AAO) ARM site. J. Climate, 12 , 4663.

    • Search Google Scholar
    • Export Citation
  • Uttal, T., and Coauthors, 2002: The Surface Heat Budget of the Arctic Ocean. Bull. Amer. Meteor. Soc., 83 , 255275.

  • Westwater, E. R., , Y. Han, , M. D. Shupe, , and S. Y. Matrosov, 2001: Analysis of integrated cloud liquid and precipitable water vapor retrievals from microwave radiometers during SHEBA. J. Geophys. Res., 106 , 3201932030.

    • Search Google Scholar
    • Export Citation
  • Zhang, M. H., , and J. L. Lin, 1997: Constrained variational analysis of sounding data based on column-integrated budgets of mass, heat, moisture, and momentum: Approach and application to ARM measurements. J. Atmos. Sci., 54 , 15031524.

    • Search Google Scholar
    • Export Citation
  • Zhang, M. H., , S. Xie, , R. T. Cederwall, , and J. J. Yio, 2001: Description of the ARM operational objective analysis system. Tech. Note ARM TR-005, 19 pp.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Time series of temperature: observed, TOBS (dot); ECMWF model forecast, TEC (solid line); and a simple time integration of the ECMWF total tendencies, TINT (dotted line), at approximately 860 m

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 2.
Fig. 2.

Histogram distribution of 1-h changes in the horizontal advection tendency for three model levels between 600 and 700 mb from successive runs of the operational ECMWF model for the month of Jul. Distributions are for all times (dashed) and end-of-the-day times (i.e., 35-h forecast valid at 2300 UTC minus 12-h forecast valid at 0000 UTC on the following day) (solid)

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 3.
Fig. 3.

Comparison of observed TOA downwelling SW flux with the observed column radiative flux divergence, RAD: (a) Apr, (b) May, (c) Jun, and (d) Jul. The solid line is a least squares fit to the data

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 4.
Fig. 4.

The correction term in the (a) temperature advection and (b) water vapor mixing ratio advection in Jul

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 5.
Fig. 5.

Scatterplot of the corrected and unmodified ECMWF (a) temperature and (b) water vapor mixing ratio advection for Jul

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 6.
Fig. 6.

Time series of the observed (left) temperature and (right) water vapor mixing ratio for Apr–Jul

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 7.
Fig. 7.

Time series of errors (predicted minus observed) in the modeled temperature profiles for the (left) baseline and (right) corrected simulations

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 8.
Fig. 8.

Time series of errors (predicted minus observed) in the modeled water vapor mixing ratio profiles for the (left) baseline and (right) corrected simulations.

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 9.
Fig. 9.

Rms error in the modeled (top) temperature and (bottom) water vapor mixing ratio profiles for the baseline (solid) and corrected (dotted) simulations.

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 10.
Fig. 10.

Time series of modeled baseline (dotted), corrected (dashed), and observed (solid) surface downwelling (a) shortwave and (b) longwave radiative flux

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Fig. 11.
Fig. 11.

Temperature differences between simulations forced with advections calculated using the baseline radiative flux divergence and the (a) baseline radiative flux divergence minus 28 W m−2 and (b) baseline radiative flux divergence plus 28 W m−2

Citation: Monthly Weather Review 132, 3; 10.1175/1520-0493(2004)132<0687:ANAFOA>2.0.CO;2

Table 1.

(a) Values for terms in Eq. (7) derived from SHEBA observations. Units are 10−3 g m−2 s−1. (b) Values for terms in Eq. (9) derived from SHEBA observations. Units are W m−2

Table 1.
Table 2.

(a) Values for terms in Eqs. (14) and (16). Units are 10−3 g m−2 s−1. (b) Values for terms in Eqs. (15) and (17). Units are W m−2

Table 2.
Table 3.

A comparison of modeled and retrieved mean liquid and ice water paths. Units are g m−2

Table 3.
Table 4.

A comparison of modeled and observed total monthly precipitation. Units are cm (liquid equivalent)

Table 4.
Table 5.

A comparison of modeled and observed/retrieved mean LWP and IWP (g m−2), cloud fraction (%), and time-averaged column radiative flux divergence (RAD; W m−2) for Jul

Table 5.
Save