Acquisition and Processing of Atmospheric Boundary Layer Data

Much of what we know about the structure of the boundary layer is empirical, the result of painstaking analysis of observational data. As our understanding of the boundary layer evolved, so did our ability to define more clearly the requirements for sensing atmospheric variables and for processing that information. Decisions regarding choice of sampling rates, averaging time, detrending, ways to minimize aliasing, and so on, became easier to make. We find we can even standardize most procedures for real-time processing. The smaller, faster computers, now within the reach of most boundary layer scientists, offer virtually unlimited possibilities for processing and displaying results even as an experiment is progressing. The information we seek, for the most part, falls into two groups: (1) time-averaged statistics such as the mean, variance, covariance, skewness, and kurtosis and (2) spectra and cospectra of velocity components and scalars such as temperature and humidity. We discuss them separately because of different sampling and processing requirements for the two. A proper understanding of these requirements is essential for the successful planning of any experiment. In this chapter we discuss these considerations in some detail with examples of methods used in earlier applications. We will assume that sensors collecting the data have adequate frequency response, precision, and long-term stability and that the sampling is performed digitally at equally spaced intervals. We also assume that the observation heights are chosen with due regard to sensor response and terrain roughness. For calculations of means and higher order moments we need time series that are long enough to include all the relevant low-frequency contributions to the process, sampled at rates fast enough to capture all the high-frequency contributions the sensors are able to measure. Improper choices of averaging times and sampling rates can indeed compromise our statistics. We need to understand how those two factors affect our measurements in order to make sensible decisions on how long and how fast to sample.

Sensors and Techniques for Observing the Boundary Layer

Sensors used for boundary layer measurements fall into two broad categories: in situ sensors that can be mounted on the ground, on masts, towers, tethered balloons, free balloons, or aircraft; and remote sensors, ground-based or aircraft-mounted, that infer atmospheric properties through their effects on acoustic, microwave, and optical signals propagating through the air. In situ sensors are the traditional instruments of choice for surface and lower boundary layer studies, being the only ones capable of the accuracy and resolution needed for quantitative work. A major portion of this chapter will therefore be devoted to discussions of their characteristics. Remote sensors have the advantage of increased range and spatial scanning capability, but the constraints on minimum range and spatial resolution limit their usefulness for surface layer measurements. Used in combination, however, the two types of sensors provide a more complete description of the flow field being studied than either of the two can provide separately. New remote sensors with shorter minimum ranges and finer range resolutions are now becoming available for boundary layer applications. A brief discussion of such devices is also included in this chapter. The variables of greatest interest to boundary layer meteorologists are wind speed, temperature, humidity, and the fluxes of momentum, heat, mass, and radiant energy. Given suitable fast-response measurements of wind velocity and scalar fluctuations, we can calculate the eddy fluxes directly from the products of their fluctuating components as explained in Chapter 1. If only the gradients of their means are available, however, then over a flat homogeneous surface the fluxes may be inferred from the Monin-Obukhov relationships of Chapters 1 and 3. Practical methods for doing that are described in many texts; see, for example, Monteith (1975, 1976). (Those simple relationships do not hold, as we know, under advective conditions, in plant canopies, and over hills.) There are also sensors in use that measure surface and near-surface fluxes directly, such as the drag plate (surface stress), the lysimeter (latent heat flux), flux plates (soil heat flux), and radiometers (radiant heat flux). We will discuss these and a few other types as well because of their application to studies of plant canopies.

Flow Over Hills

We now move on to the next obstacle to understanding how the boundary layer behaves in general through the study of flow over ridges and hills. In Chapter 4 we examined simple changes in surface conditions and showed how their effects extend upwards with increasing downwind distance. The distinguishing features of the flow over those changes were a small perturbation in the pressure field and an internal boundary layer, the depth of which was controlled by turbulent diffusion from the new surface. Here, we confront not a change in surface properties but a change in surface elevation that forces large-scale changes in the pressure field. The response to this forcing is more complicated than any we have tackled so far, but the work of many scientists over the past 25 years gives us a measure of understanding of the processes involved. In addition to extending to hillsides the kind of analyses of wind and turbulence we have already presented, there are new questions that only arise in the context of hill flows. One, with ramifications for large-scale prediction of the weather and climate, is how much drag hills exert on the atmosphere flowing over them. For large hills and mountains this problem is dominated by the behavior of the internal gravity waves initiated by hills; over lower topography, however, it involves a subtle balance between changes in the surface stress distribution and the pressure field. In questions of wind turbine siting, understanding the position and magnitude of accelerations in the mean wind becomes crucial, whereas changes to both the mean wind and turbulence are important when predicting the fate of atmospheric pollutants in hilly terrain or estimating wind loads on buildings. The pattern of airflow around a hill is determined not only by the hill shape but also by its size. A characteristic feature of the atmosphere as a whole is its static stability, extending all the way to the ground at night-time and down to zi during the day. As a result, the vertical movement of air parcels that must occur as the wind flows over a hill is accompanied by a gravitational restoring force.

Flow Over Changing Terrain

The micrometeorologist setting out to find a field site that satisfies the requirements of horizontal homogeneity will soon be reminded that most of the earth’s surface is not flat and that most of the flat bits are inconveniently heterogeneous. This is what forced the location of early pioneering experiments to remote sites such as Kansas, Minnesota, or Hay (Chapter 1), where the elusive conditions could be realized. Vital as these experiments were to the development of our understanding, they are merely the point of departure for applications to arbitrary terrain. The components of arbitrariness are two: changes in the land surface and hills. In this chapter we discuss the first of these, flow over changing surface conditions; in Chapter 5 we look at flow over hills. In the real world, the two conditions often occur together — in farmland it is the hills too steep to plow that are left covered with trees — but we separate them here to clarify the explication of phenomena and because treating them in combination would exceed the state of the art. We simplify the problem of horizontal heterogeneity still further and discuss mainly single changes in surface conditions from one extensive uniform surface to another. Furthermore, the change will typically be at right angles to the wind direction so the resulting flow field is two-dimensional. Although multiple changes are now receiving theoretical attention (Belcher et al., 1990; Claussen, 1991), there exist as yet no experimental data for comparison. Two types of surface change may be distinguished at the outset: change in surface roughness, which produces a change in surface momentum flux with a direct effect upon the wind field, and change in the surface availability of some scalar. Those of most interest are the active scalars, heat and moisture. (These are called active because their fluxes and concentrations affect stability and thereby turbulent mixing and momentum transfer, as we saw in Chapters 1 and 3.) We shall discover significant differences in flow behavior according to whether the wind blows from a smooth to a rough surface or a rough to a smooth surface.

Flow Over Flat Uniform Terrain

We start with the simplest of boundary layers, that over an infinite flat surface. Here we can assume the flow to be horizontally homogeneous. Its statistical properties are independent of horizontal position; they vary only with height and time. This assumption of horizontal homogeneity is essential in a first approach to understanding a process already complicated by such factors as the earth's rotation, diurnal and spatial variations in surface heating, changing weather conditions, and the coexistence of convective and shear-generated turbulence. It allows us to ignore partial derivatives of mean quantities along the horizontal axes (the advection terms) in the governing equations. Only ocean surfaces come close to the idealized infinite surface. Over land we settle for surfaces that are locally homogeneous, flat plains with short uniform vegetation, where the advection terms are small enough to be negligible. If, in addition to horizontal homogeneity, we can assume stationarity, that the statistical properties of the flow do not change with time, the time derivatives in the governing equations vanish as well. This condition cannot be realized in its strict sense because of the long-term variabilities in the atmosphere. But for most applications we can treat the process as a sequence of steady states. The major simplification it permits is the introduction of time averages that represent the properties of the process and not those of the averaging time. These two conditions clear the way for us to apply fluid dynamical theories and empirical laws developed from wind tunnel studies to the atmosphere's boundary layer. We can see why micrometeorologists in the 1950s and 1960s scoured the countryside for flat uniform sites. The experiments over the plains of Nebraska, Kansas, and Minnesota (USA), Kerang and Hay (Australia), and Tsimliansk (USSR) gave us the first inklings of universal behavior in boundary layer turbulence. Our concept of the atmospheric boundary layer (ABL) and its vertical extent has changed significantly over the last few decades.

Flow Over Plant Canopies

Any land surface that receives regular rainfall is almost certain to be covered by vegetation. Most of the inhabitable regions of the globe fall into this category. Often the vegetation is tall enough to call into question the assumption, implicit in the discussion of the first two chapters, that the roughness elements on the ground surface are much lower than any observation height of interest to us. In fact, if we venture to make measurements too close to tall vegetation, we discover significant departures from many of the scaling laws and formulas that seem to work in the surface layer above the canopy. To take one example, momentum is absorbed from the wind not just at the ground surface but through the whole depth of the canopy as aerodynamic drag on the plants. Consequently, although we still observe a logarithmic velocity profile well above the canopy, its apparent origin has moved to a level z = d near the top of the plants. The precise position of this “displacement height,” d, depends on the way the drag force is distributed through the foliage and this in turn depends on the structure of the mean wind and turbulence within the canopy. Our interest in the nature of within-canopy turbulence, however, is not motivated solely by its influence on the surface layer above. The understanding of turbulent transfer within foliage canopies provides the intellectual underpinning for the physical aspects of agricultural meteorology. As such it has a history almost as venerable as investigations of the surface layer itself. The landmark study of Weather in Wheat by Penman and Long (1960) was the first of a series of seminal papers to establish the quantitative link between the turbulent fluxes in a canopy and the physiological sources and sinks of heat, water vapor, and carbon dioxide (CO2). Prominent and influential among these early publications were those by Uchijima (1962), Denmead (1964), Brown and Covey (1966), and Lemon and Wright (1969). Whereas these authors were motivated by curiosity about plant physiology and the transfer of water and other scalars through the soil-plant-air continuum, other workers forged the link between the classical surface layer studies detailed in Chapter 1 and the structure of within-canopy turbulence.

Spectra and Cospectra Over Flat Uniform Terrain

Turbulent flows like those in the atmospheric boundary layer can be thought of as a superposition of eddies—coherent patterns of velocity, vorticity, and pressure— spread over a wide range of sizes. These eddies interact continuously with the mean flow, from which they derive their energy, and also with each other. The large “energy-containing” eddies, which contain most of the kinetic energy and are responsible for most of the transport in the turbulence, arise through instabilities in the background flow. The random forcing that provokes these instabilities is provided by the existing turbulence. This is the process represented in the production terms of the turbulent kinetic energy equation (1.59) in Chapter 1. The energy-containing eddies themselves are also subject to instabilities, which in their case are provoked by other eddies. This imposes upon them a finite lifetime before they too break up into yet smaller eddies. This process is repeated at all scales until the eddies become sufficiently small that viscosity can affect them directly and convert their kinetic energy to internal energy (heat). The action of viscosity is captured in the dissipation term of the turbulent kinetic energy equation. The second-moment budget equations presented in Chapter 1, of which (1.59) is one example, describe the summed behavior of all the eddies in the turbulent flow. To understand the conversion of mean kinetic energy into turbulent kinetic energy in the large eddies, the handing down of this energy to eddies of smaller and smaller scale in an “eddy cascade” process, and its ultimate conversion to heat by viscosity, we must isolate the different scales of turbulent motion and separately observe their behavior. Taking Fourier spectra and cospectra of the turbulence offers a convenient way of doing this. The spectral representation associates with each scale of motion the amount of kinetic energy, variance, or eddy flux it contributes to the whole and provides a new and invaluable perspective on boundary layer structure. The spectrum of boundary layer fluctuations covers a range of more than five decades: millimeters to kilometers in spatial scales and fractions of a second to hours in temporal scales.