• Anderson, D., A. Hollingsworth, S. Uppala, and P. Woiceshyn, 1991: A study of the use of scatterometer data in the European Centre for Medium-Range Weather Forecasts operational analysis-forecast model. 2: Data Impact. J. Geophys. Res.,96, 2635–2648.

  • Atlas, R., P. M. Woiceshyn, S. Peteherych, and M. G. Wurtele, 1982: Analysis of satellite scatterometer data and its impact on weather forecasting. Oceans,82, 415–420.

  • Brown, R. A., 1970: A secondary flow model for the planetary boundary layer. J. Atmos. Sci.,27, 742–757.

  • ——, 1972: The infection point instability problem for stratified rotating boundary layers. J. Atmos. Sci.,29, 850–859.

  • ——, 1978: Similarity parameters from first-order closure and data. Bound.-Layer Meteor.,14, 381–396.

  • ——, 1981: Modelling the geostrophic drag coefficient for AIDJEX. J. Geophys. Res.,86, 1989–1994.

  • ——, and T. Liu, 1982: An operational large-scale marine PBL model. J. Appl. Meteor.,21, 261–269.

  • ——, and G. Levy, 1986: Ocean surface pressure fields from satellite sensed winds. Mon. Wea. Rev.,114, 2197–2206.

  • ——, and L. Zeng, 1994: Estimating central pressures of oceanic midlatitude cyclones. J. Appl. Meteor.,33, 1088–1095.

  • ——, and Coauthors, 1982:Surface wind analysis for Seasat. J. Geophys. Res.,87, 3355–3364.

  • Duffy D. G., and R. Atlas, 1986: The impact of SEASAT-A scatterometer data on the numerical prediction of the Queen Elizabeth II storm. J. Geophys. Res.,91, 2241–2248.

  • Endlich, R. M., 1967: An iterative method for altering the kinematic properties of wind fields. J. Appl. Meteor.,6, 837–844.

  • ——, D. E. Wolf, C. T. Carlson, and J. W. Maresca Jr., 1981: Oceanic wind and balanced pressure-height fields derived from satellite measurements. Mon. Wea. Rev.,109, 2009–2016.

  • Harlan, J., Jr., and J. J. O’Brien, 1986: Assimilation of scatterometer winds into pressure fields using a variational method. J. Geophys. Res.,91, 7816–7836.

  • Holton, J. R., 1979: An Introduction to Dynamic Meteorology. 2d ed. Academic Press, 391 pp.

  • Hsu, S. C., and W. T. Liu, 1996: Wind and pressure fields near tropical cyclone Oliver derived from scatterometer observations. J. Geophys. Res.,101, 17 021–17 027.

  • ——, ——, and M. G. Wurtele, 1997: Impact of scatterometer winds on hydrological forcing and convective heating. Mon. Wea. Rev.,125, 1556–1576.

  • Jones, W. L., L. C. Schroeder, D. H. Boggs, E. M. Bracelente, R. A. Brown, G. J. Dome, W. J. Pieson, and F. J. Wentz, 1982: The Seasat-A satellite scatterometer: The geophysical evaluation of remotely sensed wind vectors over the ocean. J. Geophys. Res.,87, 3297–3317.

  • Kondo, J., 1975: Air-sea bulk transfer coeffiecients in diabatic conditions. Bound.-Layer Meteor.,9, 91–112.

  • Lenzen, A. J., D. R. Johnson, and R. Atlas, 1993: Analysis of the impact of Seasat scatterometer data and horizontal resolution on GLA model simulations of the QE II storm. Mon. Wea. Rev.,121, 499–521.

  • Levy, G., and F. S. Tiu, 1990: Thermal advection and stratification effects on surface winds and the low level meridional mass transport. J. Geophys. Res.,95, 20 247–20 257.

  • ——, and R. A. Brown, 1991: Southern Hemisphere synoptic weather from a satellite scatterometer. Mon. Wea. Rev.,119, 2803–2813.

  • Liu, W. T., K. B. Katsaros, and J. A. Businger, 1979: Bulk parameterization of air-sea exchanges of heat and water vapor including the molecular constraints at the interface. J. Atmos. Sci.,36, 1722–1735.

  • McMurdie, L. A., and K. B. Katsaros, 1985: Atmospheric water distribution in a midlatitude cyclone observed by the Seasat Scanning Multichannel Microwave Radiometer. Mon. Wea. Rev.,113, 584–598.

  • Overland, J. E., P. M. Woiceshyn, and M. G. Wurtele, 1980: SEASAT observations of cyclones. Tropical Ocean–Atmos. Newslett.,3, 7.

  • Paulson, C. A., 1970: The mathmatical representation of wind speed and temperature profiles in the unstable atmospheric surface layer. J. Appl. Meteor.,9, 857–886.

  • Peteherych, S., M. G. Wurtele, P. M. Woiceshyn, D. H. Boggs, and R. Atlas, 1984: First global analysis of SEASAT scatterometer winds and potential for meteorological research. Proc. URSI Commission F Symp. and Workshop, Shoresh, Israel, NASA, 575–585.

  • Stoffelen, A. C. M., and G. J. Cats, 1991: The impact ofSeasat-A scatterometer data on high-resolution analyses and forecasts: The development of the QE II storm. Mon. Wea. Rev.,119, 2794–2802.

  • Webb, E. K., 1970: Profile relationships: The log-linear range, and extention to strong stability. Quart. J. Roy. Meteor. Soc.,96, 67–90.

  • Woiceshyn, P. M., Ed., 1979: SEASAT Gulf of Alaska Experiment Workshop, Vol. II, Comparison data base: Conventional marine meteorological and sea surface temperature analyses, Appendices A and B. Jet Propulsion Laboratory Document 622–101. [Available from JPL, 4800 Oak Grove Drive, Pasadena, CA 91109.].

  • ——, M. G. Wurtele, and G. F. Cunningham, 1989: Wave hindcasts forced by scatterometer and other wind fields. Second Int. Workshop on Wave Hindcasting and Forecasting, Vancouver, BC, Canada, 268–277.

  • Wunsch, C., 1978: The North Atlantic general circulation west of 50°W determined by inverse methods. Rev. Geophys. Space Phys.,16, 583–620.

  • View in gallery

    Dealiased, subjectively analyzed synoptic scatterometer wind fields (m s−1) and model-derived surface pressure field (mb) for 1800 UTC 11 September 1978. Winds are reduced to one-third of the original resolution (1° × 1°).

  • View in gallery

    PBL-model-derived surface pressure field from SASS synoptic scatterometer wind vectors with 1°C air–sea temperature difference for 1200 UTC 11 September 1978 (upper), 1800 UTC 11 September 1978 (middle), and 1800 UTC 14 September 1978 (lower).

  • View in gallery

    The corresponding NMC analysis as in Fig. 2.

  • View in gallery

    The corresponding ECMWF analysis as in Fig. 2.

  • View in gallery

    Reanalyzed surface pressure field for the region near Gulf of Alaska at 1800 UTC 11 September 1978 (after McMurdie et al. 1985).

  • View in gallery

    Reanalysis of surface pressure field using GOASEX observed data at 1800 UTC 14 September 1978.

  • View in gallery

    Visible satellite cloud image for the northern Pacific at the time of Fig. 6.

  • View in gallery

    PBL-model-derived surface pressure fields as in Fig. 2 except for neutral stability.

  • View in gallery

    Relationship of the magnitude of surface wind vs geostrophic wind under different stratification.

  • View in gallery

    As in Fig. 2 except for balanced pressure fields.

  • View in gallery

    Pressure difference charts derived by substracting the balanced pressure fields (Fig. 10) from the PBL-model-derived pressure fields (Fig. 2).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 109 109 1
PDF Downloads 4 4 0

Construction of Marine Surface Pressure Fields from Scatterometer Winds Alone

View More View Less
  • a Department of Atmospheric Sciences, University of California–Los Angeles, Los Angeles, California
  • | b Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California
© Get Permissions
Full access

Abstract

A series of 6-h, synoptic, gridded, global surface wind fields with a resolution of 100 km was generated using the dataset of dealiased Seasat satellite scatterometer (SASS) winds produced as described by Peteherych et al. This paper is an account of the construction of surface pressure fields from these SASS synoptic wind fields only, as carried out by different methods, and the comparison of these pressure fields with National Centers for Environmental Prediction (NCEP) analyses, with the pressure fields of the European Centre for Medium-Range Weather Forecasts (ECMWF), and with the special analyses of the Gulf of Alaska Experiment.

One of the methods we use to derive the pressure fields utilizes a two-layer planetary boundary layer (PBL) model iterative scheme that relates the geostrophic wind vector to the surface wind vector, surface roughness, humidity, diabatic and baroclinic effects, and secondary flow. A second method involves the assumption of zero two-dimensional divergence, leading to a Poisson equation (the “balance equation”) in pressure, with the wind field serving as a forcing function.

The pressure fields computed from the SASS winds using a two-layer PBL model closely approximate the NCEP and ECMWF fields. In some cases, the PBL-model-derived pressure fields can detect mesoscale features not resolved in either the NCEP or ECMWF analyses. Balanced pressure fields are much smoother and less well resolved than the PBL-model-derived or NCEP fields. Systematic differences between balanced pressure fields and the PBL-model-derived fields are attributed to the neglect of horizontal divergence in the balance equation. The effect of stratification is found to produce a larger impact than secondary flow or thermal wind effects on the derived pressure fields. Inclusion of secondary flow tends to weaken both low and high pressure centers, whereas inclusion of stratification intensifies low centers and weakens high centers.

* Current affiliation: Jet Propulsion Lab, California Institute of Technology, Pasadena, California.

Corresponding author address: Dr. Carol S. Hsu, JetPropulsion Laboratory, California Institute of Technology, Mail Stop 300-320, 4800 Oak Grove Drive, Pasadena, CA 91109.

Abstract

A series of 6-h, synoptic, gridded, global surface wind fields with a resolution of 100 km was generated using the dataset of dealiased Seasat satellite scatterometer (SASS) winds produced as described by Peteherych et al. This paper is an account of the construction of surface pressure fields from these SASS synoptic wind fields only, as carried out by different methods, and the comparison of these pressure fields with National Centers for Environmental Prediction (NCEP) analyses, with the pressure fields of the European Centre for Medium-Range Weather Forecasts (ECMWF), and with the special analyses of the Gulf of Alaska Experiment.

One of the methods we use to derive the pressure fields utilizes a two-layer planetary boundary layer (PBL) model iterative scheme that relates the geostrophic wind vector to the surface wind vector, surface roughness, humidity, diabatic and baroclinic effects, and secondary flow. A second method involves the assumption of zero two-dimensional divergence, leading to a Poisson equation (the “balance equation”) in pressure, with the wind field serving as a forcing function.

The pressure fields computed from the SASS winds using a two-layer PBL model closely approximate the NCEP and ECMWF fields. In some cases, the PBL-model-derived pressure fields can detect mesoscale features not resolved in either the NCEP or ECMWF analyses. Balanced pressure fields are much smoother and less well resolved than the PBL-model-derived or NCEP fields. Systematic differences between balanced pressure fields and the PBL-model-derived fields are attributed to the neglect of horizontal divergence in the balance equation. The effect of stratification is found to produce a larger impact than secondary flow or thermal wind effects on the derived pressure fields. Inclusion of secondary flow tends to weaken both low and high pressure centers, whereas inclusion of stratification intensifies low centers and weakens high centers.

* Current affiliation: Jet Propulsion Lab, California Institute of Technology, Pasadena, California.

Corresponding author address: Dr. Carol S. Hsu, JetPropulsion Laboratory, California Institute of Technology, Mail Stop 300-320, 4800 Oak Grove Drive, Pasadena, CA 91109.

Introduction

The data gathered by the Seasat scatterometer system (SASS) represent a successful first effort to measure marine surface wind globally. Before the launch of the satellite, significant improvements in surface weather analysis and forecast skill over open ocean and coastal regions had been limited by the sparsity of observations. The high resolution marine wind data produced from information gathered by the ocean-observing Seasat satellite for 96 days during 1978 greatly increased the observational coverage, most particularly the wind field over the oceans (see Fig. 1). The work done by Overland et al. (1980) and Atlas et al. (1982) clearly showed that the scatterometer system was highly successful in delineating surface weatherpatterns with significantly greater resolution than can possibly be achieved by in situ observational systems. A positive impact on forecast skill has been found in some cases when SASS winds are incorporated into numerical weather prediction models (see, e.g., Duffy and Atlas 1986; Stoffelen and Cats 1991; Lenzen et al. 1993). In the Southern Hemisphere, the differences in the forecasts with and without SASS winds can be large (about 20–30 mb within one day), according to Anderson et al. (1991). However, a method for the optimal assimilation of the relatively finescale scatterometer wind data has yet to be developed (Anderson et al. 1991). Positive impact of ERS-1 scatterometer winds has recently been demonstrated with the U.K. Meteorological Office (UKMO), Goddard Space Flight Center (GSFC), and National Centers for Environmental Prediction [NCEP, previously known as the National Meteorological Center (NMC)] models, and these data are currently being used operationally to improve NCEP and UKMO analyses and forecasts (P. Woiceshyn 1996, personal communication).

There are different approaches to the use of scatterometer data. NCEP’s approach incorporates the nonsynoptic swaths of wind observations into its weather prediction models according to its assimilation schemes. Another approach is to interpolate the swath wind observations in time and space and use the resulting synoptic wind field as a valid dataset in itself. This has been done, for example, in wave hindcasting (Woiceshyn et al. 1989) and in precipitation estimate during the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) (Hsu et al. 1997).

Surface pressure may also be computed directly from wind data by various methods. Some work has already been carried out on this problem. Pressure fields were derived by Endlich et al. (1981) by balancing the rotational part of the wind and the pressure field. Stratification, humidity, secondary flow, and baroclinity are not accommodated in the balance equation, and this method requires a knowledge of the pressure at all boundary points to solve the resulting Poisson equation. Furthermore, pressure fields derived from the balance equation will not be accurate in tropical areas or within weather systems where significant convergence exists.

Another approach is to use a variational method to assimilate scatterometer winds into surface pressure fields by a reduction–rotation method. This was done by Harlan and O’Brien (1986) with the aid of two constraints: 1) minimize the difference between relative vorticities calculated from the data and those calculated from the variational solution and 2) minimize the average kinetic energy. This method requires a first-guess pressure field and assumes a constant turning angle between surface wind and geostrophic wind. Harlan and O’Brien recognized that more sophisticated models may be required, especially during explosive cyclogenesis, as in the Queen Elizabeth II storm case.

The pressure fields deduced by the methods mentioned above both exhibit substantial errors (up to 10 mb) in the vicinity of a front or storm. In this paper, we use a two-layer planetary boundary layer (PBL) model (Brown and Liu 1982) to derive surface pressure fields from synoptic SASS winds. The output of this methodology is a geostrophic wind or pressure gradient, so that at least one reliable observation (e.g., buoy or coastal station) is required to produce the pressure field.

Cases in which an existing PBL model is used to obtain surface pressure fields from SASS winds are presented; these fields are then validated by NCEP surface pressureanalyses, by detailed Gulf of Alaska Experiment (GOASEX) analyses, and by 1000-mb data from the European Centre for Medium-Range Weather Forecasts (ECMWF).1

Data

The scatterometer radar onboard the Seasat-A satellite provided global measurements of oceanic surface wind speed and direction accurate to ±2 m s−1 and ±20°, respectively (Jones et al. 1982). A 15-day, subjectively dealiased, asynoptic gridded dataset from asynchronous orbital data were produced with a resolution of 100 km × 100 km (Peteherych et al. 1984). Satellite cloud images are taken into account in determining the directions of wind vectors. For meteorological studies, we have converted this record into a synoptic dataset by interpolating these data in time, using a third-order polynomial scheme.

The two-layer PBL model

Two-layer similarity model

To obtain pressure fields, we use the two-layer PBL model of Brown and Liu (1982) that relates the geostrophic wind vector to the surface wind vector, surface roughness, humidity, stratification, and thermal wind. The model includes a surface layer and an outer layer. The velocity in the surface layer increases logarithmically with height and is corrected for variable stratification using the Businger–Dyer model (Paulson 1970). The outer layer extends from the top of the surface layer to the top of the boundary layer where the flow is assumed to be in geostrophic balance. The classical Ekman spiral is unstable to infinitesimal disturbances, so it is modified by the addition of secondary flow as proposed by Brown (1970, 1972). The flows at the interface of the surface layer and the outer layer are patched by matching velocities and the vertical shear of velocity there to relate Ekman and surface layer flow to the geostrophic wind. Kondo’s (1975) empirical method, which considers the flow to be laminar at very low wind speeds or small roughness, is used to obtain the roughness length from the winds through the friction velocity. Molecular effects in the interfacial layer mentioned by Liu et al. (1979) are also considered.

The basic equations are as follows.

In the Ekman–Taylor layer, U and V can be expressed as
i1520-0450-36-9-1249-e1
where ut and υt are the zonal and meridional components of the thermal winds, U2 and V2 are secondary flow, α is the geostrophic departure angle, and the lower boundary condition is the wind at the top of the surface layer. The direction of U is defined as the direction of the surface wind. Here, V is the wind perpendicular to the surface wind at any level, ζz/L is the nondimensional stability function, and L is the Monin–Obukhov length. The effect of moisture fluctuations on buoyancy is included in L.
In the surface layer, the logarithmic profile is used:
i1520-0450-36-9-1249-e2
Here, Us is the surface wind (measured by the scatterometer in our case), u∗ is the surface wind stress, z is the height of wind speed measurement, z0 is the surface roughness length, k is von Kármán’s constant, and ψ is a function of ζz/L.
Then we match the solutions of the two layers at the top of the surface layer (which is assumedtobe the height where scatterometer winds are measured, which is 19.5 m) to obtain the following conditions (Brown 1981).
i1520-0450-36-9-1249-e3

Stratification, thermal wind, and humidity effects can be added to the neutrally stratified profile if air–sea temperature difference and humidity are known. For our problem, only US is known and ECMWF air–sea temperatures are used. We can first employ Kondo’s (1975) method for obtaining CD’s and z0 in different ranges of US’s. Here, u∗ can be obtained by an iterative method considering stratification given in the term z/L. Since L and z0 are also functions of u∗, further iteration is required.

At this point, we can use matching conditions to solve for the turning angle α and G (geostrophic wind) since the similarity parameters B and A′ can be obtained if we know λ, ut, υt, U2, and V2. Here, λ is the ratio of the height of the surface to the outer layer characteristic scale δ. It is found to be approximately 0.15 (Brown 1978).

After G and α are determined, horizontal pressure gradients can be obtained from the geostrophic relations. Thus, as long as we know boundary conditions, the pressure field can be computed through surface winds using this boundary layer model.

In summary, we first obtain roughness length z0 and frictional velocity u∗ by iteration from the observed surface velocity, using the method of Kondo (1975). Then we determine the turning angle and magnitude of the geostrophic wind from the matching conditions. Finally, we use the geostrophic wind vectors to construct the surface pressure field through the geostrophic relations. An inversion method (Wunsch 1978) is used to reduce errors during integration.

It has been known for many years (e.g., Webb 1970) that PBL models based on the log-linear profile are deficient in its representation of the flow under conditions of strong stability. An alternative procedure is the one suggested by Webb (1970), who uses a modified log-profile under the conditions of large stability. We found little difference in the results using the two procedures. In our cases, the wind speeds are too small to produce any significant change in geostrophic winds.

This model has been used in several major field experiments to derive surface wind fields from surface pressure fields to verify observed winds. Some examples are the Arctic Ice Dynamics Joint Experiment (AIDJEX) (Brown 1981), GOASEX (Woiceshyn 1979), and the Joint Air–Sea Interaction Experiment (JASIN; Brown et al. 1982). Here, we invert the model to obtain pressure fields from surface winds.

Various forms of this model have also been used rather successfully in deriving surface pressure fields from the data gathered by scatterometer winds. Scatterometer winds were used to reconstruct surface pressures over limited regions during different synoptic storm situations (Brown and Levy 1986; Brown and Zeng 1994), in the Southern Hemisphere (Levy and Brown 1991), and for tropical cyclones by incorporating gradient wind dynamics (Hsu and Liu 1996).

Air temperature was not a parameter measured bySeasat,but the PBL model is sensitive to air–sea temperature difference as a measure of the near-surface atmospheric stratification. Thus, in keeping with the effort to restrict our data input to satellite observations, we have explored the effect of arbitrarily prescribing an air–sea temperature difference. Where the winds have a poleward component, we consider the air to be 1°C warmer than the sea surface, and where the winds have an equatorward component, we consider the air to be 1°C colder than the sea surface. The contribution of atmospheric stratification to the derived pressure field is discussed in detail in section 4c.

Inverse method

We derive pressure values from geostrophic winds and observations using an inverse method described by Wunsch (1978).
Axb
where x is a column vector with unknown pressure values. Here, Aij is an m × n matrix with elements equal to the points with latitude index i and longitude index j, and n is equivalent to the total grid points considered. The rows of Aij contain zeros everywhere except for the elements corresponding to the points that are used in the finite-difference scheme used to represent the pressure gradient at the point j in the b matrix. Here, b is an n-dimensional column vector that contains the pressure gradients. Pressure values at each point can be determined by adding a constraint on the mean pressure level from the observations. Both A and b are inexact. Any solution will be an estimate of the true x. The matrix A can be inverted by singular value decomposition (SVD), where
AVΛUT
where the superscript T denotes the transpose of a matrix A and Λ is a diagonal matrix with elements equal to eigenvalues. The system can be solved by
VΛ−1UTb
to obtain the pressure field with least squares errors.

PBL-model-derived pressure fields and comparison with NCEP analyses and ECMWF analyses

The PBL-model-derived pressure field (hereafter PSASS) over the North Pacific, extending from 20° to 55°N and 155°E to 135°W at 1200 UTC 11 September 1978 is shown in Fig. 2a. There are two primary low centers in the northern part of the domain and one high center in the eastern subtropical Pacific. For comparison, we show in Figs. 3a and 4a the operational surface pressure analyses for 1200 UTC 11 September 1978 produced at NCEP (Fig. 3a) and ECMWF (Fig. 4a). The patterns in these three fields—PSASS, NCEP, and ECMWF—are broadly similar.

Looking at the fields in more detail, we see that in the PSASS field, the low in the northwestern Pacific is located at 47°N, 166°E and has a central pressure of 990 mb. A second cyclonic center is located in the Gulf of Alaska near 49°N, 151°Wand has a central pressure of 987 mb. The subtropical high in the eastern Pacific has a central pressure of 1024 mb and is located slightly east of an anticyclonic center in the wind field.

The NCEP analysis also shows two low centers: one near 47°N, 167°E the other near 50°N, 152°W. The low center in the derived pressure field in the northwestern part of the domain is located about 1° west of the corresponding NCEP analyzed low and has a central pressure 2 mb higher than that in the NCEP analysis (988 mb). The other low center is 5 mb lower than that in the NCEP analysis and is located about 2° west and 1° south of the low in the NCEP analysis. The position of the derived subtropical high center in the eastern Pacific is nearly identical to that in the ECMWF analysis and is 2 mb lower than that in the NCEP analysis (1026 mb).

The two lows in the ECMWF analysis are located near 47°N, 169°E and 50°N, 151°W. The central pressures are 989 mb and 991 mb. The position of the subtropical high is close to that in both the NCEP and PSASS fields, while it has a central pressure of 1024 mb, which is exactly the same as that in PSASS. The pattern of the derived low in the eastern part of the domain bears a closer resemblence to that in the ECMWF analysis.

Six hours later, at 1800 UTC, the PSASS primary low center in the northwestern Pacific has moved about 6° east to 46°N, 172°E (Fig. 2b), while the NCEP analyzed low has moved only slightly east to 47°N, 170°E (Fig. 3b). In the ECMWF analysis (Fig. 4b), the position of this low is closer to that in the PSASS field. Its central pressure is now 991 mb in the PSASS field, which is 5 mb higher than that in the NCEP analysis and 4 mb higher than that in the ECMWF analysis. Again, we can see that our PSASS field resembles the ECMWF pressure field slightly more. The principal low in the Gulf of Alaska, according to the NCEP analysis, is near 50°N, 152°W with a central pressure of 984 mb, while in our derived pressure field is located at 51.5°N, 151°W with a central pressure of 979 mb. When we examine the cyclonic centers of the high density wind barbs, we find that the low centers are very close to these cyclonic centers. For the NCEP analyses, there are no observations near the center of the lows plotted in their surface analyses. Therefore, we strongly favor our derived pressure fields for locating low centers. The position and central pressure of the subtropical high in the eastern Pacific is close to that in the NCEP analysis. However, the subtropical high in the ECMWF analysis is about 3° east and the central pressure is 1 mb lower than both NCEP analysis and PSASS field.

A second case, that of 1800 UTC 14 September 1978, showing a belt of high pressure over the northern Pacific Ocean, is shown in Figs. 2c, 3c, and 4c. In the NCEP analysis (Fig. 3c), the trailing end of a cold front in the northeastern part of the domain separates two high centers with the same central pressure of 1030 mb. A third high center is located in the western Pacific at 44°N, 165°E with a central pressure of 1029 mb.

The PBL-derived pressure field shows this belt of highs (Fig. 2c). There is general agreement between the derived pressure field and the NCEP and the ECMWF analysis. The major discrepancy is a low that appears at 36°N, 139°W. Our analysis shows a trough between the high centers, whereas the NCEP analysis exhibits a continuous belt of high pressure. Unlike the NCEP analysis, the ECMWF-analyzed pressure field also shows a local trough near this region, except the position is further north of that in the derived field.

Even though the highsare similar both in magnitudes and features in most regions for PBL-derived, NCEP, and ECMWF fields, there is one exception—the PSASS-derived high in the western Pacific is stronger than that in both the NCEP analysis and the ECMWF analysis. There is no observational evidence to support pressures in this region as high as 1036 mb. ECMWF has 1031 mb (Fig. 4c) and NCEP has 1029 mb (Fig. 3c). This PSASS central pressure is presumably due to the strong winds south of 40°N and north of 45°N, with speeds of 15 to 20 m s−1 (not shown). As an alternative hypothesis, we considered that anomalous high pressure gradient might be due to a thermal wind effect, but introducing this (from ECMWF surface temperature fields) into the PBL model produced no appreciable difference in the pressure field. This absence of thermal wind influence is consistent with the results reported by Brown and Levy (1986), who remark that in the presence of strong winds and stable flow, the boundary layer is too shallow to allow baroclinic influence. [See also Levy and Tiu (1990), who emphasize the greater effect of stratification in comparison with the thermal wind and note that the effect of thermal wind is mainly on the turning angle rather than the magnitude of the geostrophic wind.] A summary of the positions and central pressures is given in Table 1.

Comparison with GOASEX, ship and buoy data, and stratification effect

Comparison with GOASEX

Another dataset available for comparison with our derived pressure field in a region limited to the northeast Pacific are the special analyses of the GOASEX experiment in the Gulf of Alaska. The pressure field at 1800 UTC 11 September 1978 was reanalyzed for a region near the low center in the Gulf of Alaska using GOASEX data (Fig. 5). The reanalyzed low center is 980 mb. This is closer to the central pressure of that in our derived pressure field, which is 979 mb, than the 984 mb value in the NCEP and ECMWF analyses. The location of this low is 52°N, 151°W in the GOASEX reanalysis, which is very close to that in our SASS wind-derived pressure field.

A reanalysis of the pressure field by Woiceshyn (1979) at 1800 UTC 14 September (Fig. 6) using the GOASEX observed data shows a low near 38°N, 140°W, which is consistent with our derived pressure pattern. The central pressure of the GOASEX analyzed low is 1026 mb, which is precisely what we obtain in the PBL-model-derived pressure field. A commalike cloud is near the low center in the model-derived pressure field on a visible satellite image of the eastern Pacific at that time (Fig. 7).

Comparison with surface buoy and ship observations

During GOASEX, special spot observations were made by the NOAA buoys, ocean station PAPA, and the NOAA ship Oceanographer. Data were recorded coincident with the satellite’s passage and represent the most accurate estimates of the true surface conditions at the time of Seasat measurements. Locations of NDBO (National Data Buoy Office) buoys and ships in the Gulf of Alaska are given in Table 2.

Observed surface pressure measurements from these buoy and ship observations are compared with the values at the corresponding locations and time (nearest available) obtained from the analyses from the PBL model, NCEP analysis, and ECMWF analysis. From Table 3, there is broad agreement between the values obtained from the different sources. The differences vary from 0 to 4 mb for single points. The meandifferences for all the cases are less than or equal to 2 mb. The NCEP analyses show the smallest differences from the buoy and ship observations. For the PBL-derived fields, only one surface observation (20°N, 151°W) is included and it is far away from this region. Moreover, some uncertainties may be caused by the interpolation of the wind fields. Considering these factors, the performance of the PBL-derived analyses is quite satisfactory. Though the inclusion of more surface observations into the PBL analyses will improve the results, our main objective is to use as few other resources as possible to speed up and simplify the calculation.

When we compare the analyses with weather station observations near the low center at 1800 UTC 11 September 1978 (see Table 4), we find that NCEP analyses still perform the best and ECMWF analyses are slightly inferior to the others. The errors vary from 0 to 5 mb and the mean errors have increased slightly and vary from 2.0 mb (NCEP analyses) to 2.6 mb (ECMWF analyses). Even though the PBL-model-generated pressure fields seem to capture the minimum pressure of the low (980 mb), the position of the minimum is slightly west of that in Fig. 5, which included the observations from the weather stations listed in Table 4.

Stratification effect

Without question the stratification of the atmosphere is quantitatively significant in the derivation of pressure fields from wind fields. We have made control runs for each of our cases using neutral stratification instead of the arbitrary 1° air–sea temperature difference as described above. The effect is not difficult to anticipate. For unstable stratification there is an increase in the downward momentum transfer and therefore in the sea surface roughness. Thus, inverse reasoning from a given 19.5-m wind speed would infer a weaker pressure gradient when the atmosphere is unstable and a stronger gradient when it is stable.

The question we posed for ourselves was whether it was possible to formulate a simple rule for assessing a stability that would present an improvement in results over an assumption of neutrality everywhere. The answer appears to be affirmative. The neutral runs referred to are uniformly less satisfactory in their agreement with the weather service analyses than the counterparts with the estimated temperature difference assigned. At 1800 UTC 11 September, the Aleutian low center is insufficiently developed in the neutral run (Fig. 8b), with a central pressure of 987 mb. The 1° assigned temperature difference in the strong stable southerly current strengthens the gradient and results in a low of 979 mb, about 5 mb lower than the weather service values and 1 mb lower than the GOASEX analysis. It is entirely possible that the temperature difference is greater than this, but if a 2° difference is assigned to southerly winds, the effect is to deepen the Aleutian center only slightly more to 978 mb.

The careful GOASEX reanalysis revealed a maximum of 3°C air–sea temperature difference in the Gulf of Alaska, but if this difference is applied to all southerly winds, the accumulated effect of the southerly flow north of 35°N is so great as to overdeepen the low center to 977 mb.

Thus, although stratification is important in the deviation of pressure fields from wind fields, and air–sea temperature difference is not a quantity available directly from remote sensing, it does appear that even a very simple-minded algorithm can provide an assessment of its effect that is practically useful.

Figure 9 shows the magnitudes of geostrophic wind versus surface winds under different stratification. Under unstable conditions, the geostrophic wind is not sensitive to the increase of surface wind speed. It tends todecrease the geostrophic wind slightly through mixing with lower level winds that carry lower momemtum. On the other hand, under stable conditions, the geostrophic wind is smaller than the neutral case when winds are smaller than 5 m s−1. However, when wind speeds exceed 5 m s−1, the stable stratification starts to increase the geostrophic wind tremendously. This effect can be seen very clearly near cyclones. Near the stable side (east part in the Northern Hemisphere) of a cyclone, isobars can be much denser than the other side of the cyclone. As a result, isobars around cyclones are not symmetric. This phenomena can also be seen in the NMC or ECMWF analyses.

When a secondary flow effect is included (not shown), the pressure gradients always decrease slightly. Central pressures are within 1 or 2 mb of the neutral case with this effect considered.

Balanced pressure

With this high-resolution dataset, one might be curious whether we can do just as well using some other scheme with simpler physics to derive pressure fields. One method of constructing pressure fields entirely from surface winds is by balancing the rotational part of the wind and pressure as done by Endlich et al. (1981).

Any velocity field can be partitioned into a nondivergent part plus an irrotational part such that
VVψVe
where
VψVe
For midlatitude synoptic-scale motions V is quasi-nondivergent according to scale analysis of the vorticity equation (Holton 1979). The remaining terms in the vorticity equation, when divergent winds are neglected, imply a relationship of nondivergent winds to the height (pressure) fields, which is the well-known balance equation. It can be written as
g2ZβuJu, υ
where Z is the height of a pressure surface, f is the Coriolis parameter, ζ is the relative vorticity, β = ∂f/∂y, and J(u, υ) = (∂u/∂x)(∂υ/∂y) − (∂u/∂y)(∂υ/∂x).

To obtain height (pressure) fields from the balance equation, pressures at all boundary points are required when solving the Poisson equation. The computed height fields are converted to pressure fields through the hydrostatic relation. A successive relaxation scheme was applied to solve the equation after the boundary values were prescribed. Pressure values at boundaries are taken from the NCEP analyses. For all the cases calculated, the pressure values were found to converge very rapidly.

To divide the winds into the rotational and divergent parts, the method of direct vector alterations (Endlich 1967) are used. The desired wind fields are obtained by a point iterative method applied to the two simultaneous linear partial differential equations that define horizontal divergence and relative vorticity. The divergent vectors are believed to be much smaller than the nondivergent vectors. For the cases we studied, however, the divergent winds are usually of comparable magnitudes with the nondivergent winds and are not negligible near the low centers.

We have applied this technique to derive the balanced pressure fields for the three cases examined above. Figure 10a shows the balanced pressure field for 1200 UTC 11 September 1978. We can see that the balanced pressurefield is much smoother and has less structure than either the NCEP (Fig. 3a), ECMWF (Fig. 4a), or PSASS fields (Fig. 2a). Even very near the northern boundary, where we have the same boundary values as those in the NCEP field, the primary low center near the northwestern Pacific has a central pressure of 995 mb. This is 9 mb higher than that in the NCEP analyis and 5 mb higher than that in our PBL-derived pressure field with a 1°C air–sea temperature difference imposed. The primary low in the eastern Pacific has a central pressure of about 994 mb, which is 7 mb higher than that in the PSASS analysis. The position of the low center in the northwestern Pacific is close to that in the NCEP analysis (note that the domain extends only to 50°N in the balanced pressure field), while the low center in the Gulf of Alaska is 2° west of that in the NCEP analysis. Even worse is the position of the subtropical high, which is about 10° east of the high in the NCEP analysis and is also much weaker.

The pressure gradients are underestimated in the balanced pressure fields; therefore, the central pressures are usually higher in the lows and lower in the highs (cf. Figs. 2). This can be seen most clearly in the pressure difference chart (Fig. 11), obtained by substracting balanced pressures from PBL-model-derived pressures. The most prominent differences appear at high and low centers where divergence is largest. These differences can be as large as 10 mb. A region of low values in the difference chart near 42°N, 165°W is mainly due to the failure of the balanced pressure field to capture the position and strength of the low center in that region.

We have made extensive computations of the divergence and vorticity values from the SASS winds and found that although divergence is generally smaller than vorticity, the values are of the same order of magnitude. Endlich et al. (1981) attribute the discrepancies in their derived pressure fields to the asynoptic nature of their wind data. From our synoptic analysis, we can see that the difference is systematic. Therefore, we conclude that the errors in the balanced pressure fields are chiefly due to the neglect of divergence.

Conclusions

The construction of detailed synoptic marine surface pressure fields from surface winds alone, measured by satellite, has been achieved. The pressure fields are obtained from SASS winds through a two-layer marine PBL model that includes ageostrophic winds produced by secondary flow, stratification, thermal wind, and humidity. It should be noted that our results are subject to limitations by the boundary layer model used, notably, the assumption of the geostrophic balance of the wind and pressure fields at the top of the boundary layer fields. These PBL-model-derived pressures are compared with NCEP and ECMWF analyses and special analyses based on data gathered by GOASEX. Balanced pressure fields are also obtained as references for comparison. We can make following conclusions.

  1. Pressure fields derived from SASS winds using a two-layer PBL model are of a quality comparable to those of the NCEP and ECMWF analyses in the North Pacific that were produced in 1978.
  2. Balanced pressure fields are systematically inferior to those derived using a two-layer PBL model.
  3. PBL-model-derived pressure fields can detect mesoscale features not resolved in the weather service analyses.
  4. The SASS-derived pressure fields are as close to the NCEP and ECMWF fields in pattern and central pressures as these two are close to each other.
  5. Central pressures, especially those in low centers, are highly sensitive to stratification due to the strong northerly and southerlycurrents surrounding them. Even without observational evidence of the air–sea temperature difference, the assumption of a ±1° difference (according to whether the flow is toward the north or south) produces a marked improvement over the assumption of a neutral atmosphere, as judged by agreement with conventional analyses.
  6. Atmospheric stratification tends to intensify low centers and suppress high centers. The effect is most prominent near the stable side of cyclones where wind speeds are strong. Inclusion of atmospheric stratification in the calculation yields more accurate results.

The successful derivation of pressure fields from scatterometer wind fields is significant in two respects. First, the accuracy of these fields will be increased in areas of sparse data. Second, fields derived from scatterometer data alone can be available without the delay occasioned by the data processing and assimilation in forecasting centers. This should be of great value in short-range forecasting or nowcasting. Even with the launch of ERS-1, the data are single swath, separations between swaths are large, and there is not enough coverage to derive synoptic pressure fields, especially near fast-developing storms as the cases we demonstrated. The recent launch of NASA scatterometer (NSCAT) on the Advanced Earth Observing Satellite (ADEOS1) in August 1996 and the scheduled launch of SeaWinds scatterometer on ADEOS2 in 1999 with better coverage give us the promise of advancing our investigations.

Acknowledgments

The authors greatly acknowledge Prof. R A. Brown of the University of Washington for his insightful discussions on the PBL model. We are also grateful to Dr. Ichiro Fukumori for his stimulating discussions on the inverse methods. Comments from three anonymous reviewers are very much appreciated. I am indebted to Dr. Don Collins at the PO.DAAC for his encouragement and support. The research described in this paper was carried out in part by the Jet Propulsion Laboratory, California Institutes of Technology, under a contract with the National Aeronautics and Space Administration.

REFERENCES

  • Anderson, D., A. Hollingsworth, S. Uppala, and P. Woiceshyn, 1991: A study of the use of scatterometer data in the European Centre for Medium-Range Weather Forecasts operational analysis-forecast model. 2: Data Impact. J. Geophys. Res.,96, 2635–2648.

  • Atlas, R., P. M. Woiceshyn, S. Peteherych, and M. G. Wurtele, 1982: Analysis of satellite scatterometer data and its impact on weather forecasting. Oceans,82, 415–420.

  • Brown, R. A., 1970: A secondary flow model for the planetary boundary layer. J. Atmos. Sci.,27, 742–757.

  • ——, 1972: The infection point instability problem for stratified rotating boundary layers. J. Atmos. Sci.,29, 850–859.

  • ——, 1978: Similarity parameters from first-order closure and data. Bound.-Layer Meteor.,14, 381–396.

  • ——, 1981: Modelling the geostrophic drag coefficient for AIDJEX. J. Geophys. Res.,86, 1989–1994.

  • ——, and T. Liu, 1982: An operational large-scale marine PBL model. J. Appl. Meteor.,21, 261–269.

  • ——, and G. Levy, 1986: Ocean surface pressure fields from satellite sensed winds. Mon. Wea. Rev.,114, 2197–2206.

  • ——, and L. Zeng, 1994: Estimating central pressures of oceanic midlatitude cyclones. J. Appl. Meteor.,33, 1088–1095.

  • ——, and Coauthors, 1982:Surface wind analysis for Seasat. J. Geophys. Res.,87, 3355–3364.

  • Duffy D. G., and R. Atlas, 1986: The impact of SEASAT-A scatterometer data on the numerical prediction of the Queen Elizabeth II storm. J. Geophys. Res.,91, 2241–2248.

  • Endlich, R. M., 1967: An iterative method for altering the kinematic properties of wind fields. J. Appl. Meteor.,6, 837–844.

  • ——, D. E. Wolf, C. T. Carlson, and J. W. Maresca Jr., 1981: Oceanic wind and balanced pressure-height fields derived from satellite measurements. Mon. Wea. Rev.,109, 2009–2016.

  • Harlan, J., Jr., and J. J. O’Brien, 1986: Assimilation of scatterometer winds into pressure fields using a variational method. J. Geophys. Res.,91, 7816–7836.

  • Holton, J. R., 1979: An Introduction to Dynamic Meteorology. 2d ed. Academic Press, 391 pp.

  • Hsu, S. C., and W. T. Liu, 1996: Wind and pressure fields near tropical cyclone Oliver derived from scatterometer observations. J. Geophys. Res.,101, 17 021–17 027.

  • ——, ——, and M. G. Wurtele, 1997: Impact of scatterometer winds on hydrological forcing and convective heating. Mon. Wea. Rev.,125, 1556–1576.

  • Jones, W. L., L. C. Schroeder, D. H. Boggs, E. M. Bracelente, R. A. Brown, G. J. Dome, W. J. Pieson, and F. J. Wentz, 1982: The Seasat-A satellite scatterometer: The geophysical evaluation of remotely sensed wind vectors over the ocean. J. Geophys. Res.,87, 3297–3317.

  • Kondo, J., 1975: Air-sea bulk transfer coeffiecients in diabatic conditions. Bound.-Layer Meteor.,9, 91–112.

  • Lenzen, A. J., D. R. Johnson, and R. Atlas, 1993: Analysis of the impact of Seasat scatterometer data and horizontal resolution on GLA model simulations of the QE II storm. Mon. Wea. Rev.,121, 499–521.

  • Levy, G., and F. S. Tiu, 1990: Thermal advection and stratification effects on surface winds and the low level meridional mass transport. J. Geophys. Res.,95, 20 247–20 257.

  • ——, and R. A. Brown, 1991: Southern Hemisphere synoptic weather from a satellite scatterometer. Mon. Wea. Rev.,119, 2803–2813.

  • Liu, W. T., K. B. Katsaros, and J. A. Businger, 1979: Bulk parameterization of air-sea exchanges of heat and water vapor including the molecular constraints at the interface. J. Atmos. Sci.,36, 1722–1735.

  • McMurdie, L. A., and K. B. Katsaros, 1985: Atmospheric water distribution in a midlatitude cyclone observed by the Seasat Scanning Multichannel Microwave Radiometer. Mon. Wea. Rev.,113, 584–598.

  • Overland, J. E., P. M. Woiceshyn, and M. G. Wurtele, 1980: SEASAT observations of cyclones. Tropical Ocean–Atmos. Newslett.,3, 7.

  • Paulson, C. A., 1970: The mathmatical representation of wind speed and temperature profiles in the unstable atmospheric surface layer. J. Appl. Meteor.,9, 857–886.

  • Peteherych, S., M. G. Wurtele, P. M. Woiceshyn, D. H. Boggs, and R. Atlas, 1984: First global analysis of SEASAT scatterometer winds and potential for meteorological research. Proc. URSI Commission F Symp. and Workshop, Shoresh, Israel, NASA, 575–585.

  • Stoffelen, A. C. M., and G. J. Cats, 1991: The impact ofSeasat-A scatterometer data on high-resolution analyses and forecasts: The development of the QE II storm. Mon. Wea. Rev.,119, 2794–2802.

  • Webb, E. K., 1970: Profile relationships: The log-linear range, and extention to strong stability. Quart. J. Roy. Meteor. Soc.,96, 67–90.

  • Woiceshyn, P. M., Ed., 1979: SEASAT Gulf of Alaska Experiment Workshop, Vol. II, Comparison data base: Conventional marine meteorological and sea surface temperature analyses, Appendices A and B. Jet Propulsion Laboratory Document 622–101. [Available from JPL, 4800 Oak Grove Drive, Pasadena, CA 91109.].

  • ——, M. G. Wurtele, and G. F. Cunningham, 1989: Wave hindcasts forced by scatterometer and other wind fields. Second Int. Workshop on Wave Hindcasting and Forecasting, Vancouver, BC, Canada, 268–277.

  • Wunsch, C., 1978: The North Atlantic general circulation west of 50°W determined by inverse methods. Rev. Geophys. Space Phys.,16, 583–620.

Fig. 1.
Fig. 1.

Dealiased, subjectively analyzed synoptic scatterometer wind fields (m s−1) and model-derived surface pressure field (mb) for 1800 UTC 11 September 1978. Winds are reduced to one-third of the original resolution (1° × 1°).

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 2.
Fig. 2.

PBL-model-derived surface pressure field from SASS synoptic scatterometer wind vectors with 1°C air–sea temperature difference for 1200 UTC 11 September 1978 (upper), 1800 UTC 11 September 1978 (middle), and 1800 UTC 14 September 1978 (lower).

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 3.
Fig. 3.

The corresponding NMC analysis as in Fig. 2.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 4.
Fig. 4.

The corresponding ECMWF analysis as in Fig. 2.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 5.
Fig. 5.

Reanalyzed surface pressure field for the region near Gulf of Alaska at 1800 UTC 11 September 1978 (after McMurdie et al. 1985).

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 6.
Fig. 6.

Reanalysis of surface pressure field using GOASEX observed data at 1800 UTC 14 September 1978.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 7.
Fig. 7.

Visible satellite cloud image for the northern Pacific at the time of Fig. 6.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 8.
Fig. 8.

PBL-model-derived surface pressure fields as in Fig. 2 except for neutral stability.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 9.
Fig. 9.

Relationship of the magnitude of surface wind vs geostrophic wind under different stratification.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 10.
Fig. 10.

As in Fig. 2 except for balanced pressure fields.

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Fig. 11.
Fig. 11.

Pressure difference charts derived by substracting the balanced pressure fields (Fig. 10) from the PBL-model-derived pressure fields (Fig. 2).

Citation: Journal of Applied Meteorology 36, 9; 10.1175/1520-0450(1997)036<1249:COMSPF>2.0.CO;2

Table 1.

Comparison of position and central pressure for different analyses.

Table 1.
Table 2.

Position of buoys and ships used for Table 3.

Table 2.
Table 3.

Comparison of sea level pressure (mb) from different analyses with buoy observations for the three cases studied. Numbers inside the parentheses are absolute values of the differences from buoy/ship observations. Mean error is obtained by averaging the accumulated differences.

Table 3.
Table 4.

Comparison of sea level pressure (mb) from different analyses with buoy observations for 1800 UTC 18 September 1978.

Table 4.

1

The 1° × 1° ECMWF data for 12 days in September 1978 were kindly provided by Dr. Anthony Hollingsworth, using their 1986 forecast and assimilation code.

Save