Abstract
The presence of shallow cold flows in the atmospheric boundary layer (ABL) instigates changes in the turbulent regime of the atmospheric surface layer (ASL). This small scale flow circulation introduces radiative cooling controls over large areas in polar latitudes during winter. In this study, microscale dynamic and turbulent variables have been obtained in the framework of the Winter Boundary Layer Experiment in Fairbanks, Alaska, developed during the winters of 2009/2010 and 2010/2011. Multiscale surface turbulence observations based on Eddy covariance and laser scintillometry were combined with Doppler acoustic sounding to document simultaneous changes in the ABL flow and ASL turbulence. We computed changes in momentum and heat fluxes characterizing intermittent and persistent modes of the drainage flow over three study cases. On the basis of laser scintillometry observations, we argue that a significant source of turbulence aiming at the surface fluxes has origins in the upper level shear-induced thermal turbulence at the top of the ABL.
Keywords
- surface turbulence
- optical scintillation
- radiative cooling
- small scale drainage flows
- stable boundary layer
1. Introduction
The absence of daylight combined with snow-covered surfaces during the extreme winters in continental Alaska sets up a unique meteorological condition in the lower troposphere where surface radiative cooling becomes the dominant forcing mechanism initiating the formation of stably-stratified ABL [1, 2]. This meteorological feature is present under synoptic anticyclone conditions (i.e., surface high pressure, clear skies) and is particularly prominent and episodic throughout the winter. In general, when this meteorological configuration sets up, several days of a persistent stably-stratified ABL occur [3, 4, 5, 6]. This local ABL feature, also known as surface-based temperature inversion (SBI), occurs statistically over 85% of time in the presence of multiple elevated temperature inversion (EI) layers from synoptic origin [1]. In such anti-cyclone conditions and when weak horizontal synoptic flow forcing prevails, the topographic configuration and the orientation of mountains in the Interior of Alaska constrain the low-level tropospheric circulation. At the regional scale, the air flow in the ABL becomes locally quasi-laminar and regionally stagnant under strong surface radiative cooling and therefore prone to disruptions from local-scale circulation mechanisms. The investigation of this process was the objective of the Winter Boundary-Layer Experiment [7].
Mountains and hills in polar regions shelter large cold pools of air between them that are exposed to a higher radiative cooling rate than adjacent low-laying areas that may be more opened and oriented toward south (i.e., the south-sunny side). These topographic and radiative conditions set the stage for the development of density flows that takes the form of drainage flows [7]. Such small-scale flows channeled through the varying terrain morphology connect basins on the north-facing to those on the south-facing sides of the mountains and hills and profoundly affect the stability of the relatively lower and warmer basins, as well as the surface turbulent flux and momentum regime. The occurrence of these flows has been identified in other experiments on lower latitudes [8, 9] as a cold down-slope winds initiated by the temperature contrast between basins influencing therefore the strongly stratified ABL that builds up locally at the basin scale.
On the other hand, extreme winter conditions in Fairbanks, Alaska, have been known for several decades to represent a very complex air pollution problem [3, 10, 11]. Besides local anthropogenic emissions, what exacerbates the air pollution problem is precisely the naturally occurring strong surface cooling rate and, consequently, the formation of extremely low-level SBI, often complicated with multiple stratified layers [1, 2] under a stagnant flow condition. As indicated previously, these episodes are mostly present under specific synoptic conditions (e.g., calm winds and stagnant quiescent synoptic anticyclonic flows) [3, 5] over periods lasting several days [1]. Moreover, the meteorological setup, the flow configuration, and the complexity of the developed ABL structure impose stringent modeling restrictions when an accurate representation of the ABL is needed in particular for air pollution assessments [12].
The first observational study of the winter ABL in Fairbanks was carried out in mid-1970s using a bi-static sodar instrument from the National Oceanic and Atmospheric Administration, Wave Propagation Laboratory. The set of observations illustrated several dynamic aspects of the stable ABL such as propagation of waves, inertial oscillations within the ABL structure, and formation and destruction of temperature inversion and stratified layers [5, 6]. Despite of the importance of these observations in terms of improving the understanding of physical processes in the ABL and the outstanding mesoscale modeling challenges that winter high latitudes imposes no additional efforts have been dedicated to continuing these investigations until the development of Wi-BLEx [1, 2, 7]. In fact, previous studies [10, 13, 14, 15] have indicated that during the extreme winter, a heavy air mass lies close to the ground, where SBI forms as the result of a strong surface radiative cooling rate pronounced by the absence of sunlight and under a stagnant airflow often with wind speed less than ~1 ms−1. In such conditions, katabatic flows have been found above the local SBI as they come down from the mountains surrounding the area. In these previous studies, and perhaps with limited observational capabilities at the time, it was also indicated that the extreme cold pool dominating the Tanana Valley air mass prevented the local SBI airflow to mix with the upper level katabatic flow. This feature is self-evident in several photographs of Fairbanks winter scenery, where a clear differentiation can be visually established when comparing tall power plant emission stacks and home heating emission systems. In this study, we provide observational evidence that during the winter period, the valley ABL is penetrated by a small-scale cold flow from the north-northwest sector of the observational site (see Figure 1) close to the hill slopes around Fairbanks. This shallow cold flow was observed to interact with the local ABL under specific prescribed stagnant anticyclone conditions and to dominate the thermodynamic structure and circulation of the ABL close to the foothills for specific periods of time [7]. In this case, penetrative drainage flows were observed to introduce small scale mixing, increasing thermal turbulence near the top of the ABL and re-stratification at the surface layer when the flow ceased. This flow originates upstream in sheltered cold pools of air (i.e., northern slopes of the Cranberry Hills), where the combination of geographic orientation and very low solar elevation angle during the winter allows for a more efficient radiative cooling than in the central valley region, i.e., the south facing slopes of the Tanana Valley (see Figure 1).
Section 2 describes the experimental setup during Wi-BLEx, and Section 3 describes the large scale meteorological conditions during the observations. Section 4 introduces the observations of drainage flows and ABL structure, and Section 5 analyzes and discusses the cases and the determination of structural and turbulent parameters of the drainage flow that influence the state of surface turbulent fluxes. This chapter concludes in Section 6 where a summary conclusion is provided based on the experimental evidence that supports the hypothesis that shear-induced thermal turbulence in the upper level depth of the drainage flow propagates downward to induce large-eddy turbulent exchanges in the ASL fluxes.
2. Experiment and methodology
Wi-BLEx took place at the University of Alaska Fairbanks (UAF) experimental farm located in the south-facing basin in the foothills of the Cranberry Hills (see Figure 1) on the UAF campus. The specific sets of observations discussed in this paper were carried out from November 1, 2010, to March 30, 2011, and three cases were selected for further analysis.
The instrumentation consisted of two 3D-sonic anemometers installed at 2- and 4-m height on a meteorological mast around the center of the farm (see Figure 2).
The sonic anemometers acquired at 10 Hz, the three components of the velocity field, as well as the sonic virtual temperature. This allows calculating turbulent quantities u’, v’, w’, and
Based on the retrieved turbulent magnitudes, the combined auto-covariance of wind speed components and their cross-covariance allowed calculation of turbulent kinetic energy (tke) Eq. (1), surface friction velocity (u*) Eq. (2), surface momentum (τ0) Eq. (3), and the flux Richardson number (Rif) Eq. (4).
In Eqs. (1)–(4), θsfc is the surface potential temperature and U(z) is the wind profile and in Eq. (5), D is the diameter of the receiver’s optical lens, d is the distance between emitter and receiver, while σln(I)2 is the variance of the natural logarithm of the recorded optical intensity over a given time period. The LAS emitter was operated at 125 Hz and 1 min averaging in the receiver at 520 m distance across the basin (see Figure 2). The acquired data were processed to calculate
3. Meteorological and mean flow conditions during observations
Three cases were selected for analysis from the set of experiments carried out during Wi-BLEx. In this section, the synoptic meteorology framework and the mean flow conditions are described.
4. Drainage flow events and changes in the surface turbulence
Significant changes are observed in the surface turbulent regime obtained by auto-covariance of turbulent velocities and cross-covariances to calculate tke, u*, and τ0. These variables are shown in Figure 7 where in the right panel is the tke (m2 s−2), central panel the increase of u*(ms−1) and right panel is the momentum (m2s2). At the beginning, previous to the initiation of the drainage flow, the surface layer static stability was maintained to 1°C/m of stratification. During drainage flow penetration (15:00 to 21:00 UTC), the tke increased to 0.15 m2 s−2, u* increased to 0.15 ms−1, and τ0 increased to 0.02 m2s2. The surface wind speed increased to 3.5 ms−1, and the temperature dropped down by ~4°C. At 00:00 UTC, the temperature steadily decreased from −27.5°C to −30°C. When the drainage flow ceased, the surface layer re-stabilize as indicated by the calculation of the Rif number.
The vertical structure of the ABL from the 00:00 UTC 18 January shows a surface-based inversion at 108 m from the radiosonde measurements at the NWS-NOAA nearby station with winds from the south in the stable ABL and the FA [25]. This height can also be retrieved in the backscatter signature of the sodar profile by searching the point of maximum gradient indicating the decaying of thermal turbulence structure in the ABL by the
The
The
During this period of time, the surface layer was in a transitional period until the vertical cooling initiated the formation of a stable ABL. As the initial drainage flow from the northwest direction penetrated the surface layer, the tke increased from ~0 to 0.2 m2 s−2, u* increased from ~0.05 to 0.2 ms−1, and τ0 increased from ~0 to 0.03 m2 s−2, while the temperature decreased from −3°C to −18°C, a drop in about 15°C in ~15 h.
The Rif calculated was highly variable taking negative and positive values before the onset of the drainage flow. Rif decreased and reached a steady value above 1 corresponding to a quasi-laminar flow. The end of the disruptive flow was indicated by the change of wind direction to the southeast and a reduction in the wind speed and the surface micrometeorological variables tke, u*, and τ0 at 20:00 UTC. The sun rose at 16:21 UTC, and the surface-layer temperature began to rise at 17:00 UTC. The most active turbulent activity was verified to occur between 09:00 and 15:00 UTC, when the surface parameters were at their maximum dynamic swing and wind speed was sustained overtime, while the Rif number indicates a dynamic unstable flow. The vertical ABL structure observed by Doppler sodar indicates increasing
The vertical
Summarizing, records of the ABL vertical structure obtained from Doppler sodar observations in Figures 8 and 11
5. Discussion
The turbulent state of the ASL was observed to significantly change in the presence of the shallow cold drainage flow into the basin. After scrutinizing all recorded cases during Wi-BLEx, it was found that this flow develops based on two modes: persistent flow represented by Case I and III and intermittent flow represented by Case II. Changes in the turbulence regime of the ASL, instigated by the penetration of the drainage flow into the basin’s ABL, have been determined by means of two instruments sonic anemometer and LAS. These two instruments sample basically the same turbulence spectrum with the difference that sonic anemometers are in-situ sensors, while LASs are large-scale area-average turbulence sensing devices. Both instruments respond to the turbulence developed by the flow upstream on overlapping footprints. However, in this experiment, the LAS was installed across the basin (see Figure 2) to continuously evaluate the turbulent state of the drainage flow in space and time and thus fully record the turbulent structures developing from microscale to basin scale.
Based on Monin-Obukhov similarity hypothesis, the sensible heat flux was calculated for both instruments LAS (HLAS) and sonic anemometer as indicated previously using 5-min intervals eddy-covariance integration (HEC) [23, 24]. The HLAS was calculated using friction velocity (u*) and the Obukhov length (L) obtained based on sonic anemometers measurements. The LAS covered an optical path length of 520 m across the basin. In this case, HEC was proven to capture in the range of 70–80% up to 100% of the surface turbulent fluxes developing at the basin scale when compared to HLAS [7]. This result, indicating a divergence between the calculated heat fluxes over certain time periods, pointed to the idea that large-eddy turbulence could be present in the ASL [31, 32]. On the other hand, it can be argued that given the variability of the flow in the ASL, a time variable integration is needed to account for the entire eddy flux for the case of HEC. Nevertheless, the nature and source of the large eddies present in the basin can be independently investigated. These eddies can be either part of the natural mode by which the drainage flow develops breaking up the stratification in the basin (i.e., by carrying large eddy momentum) or they can be eventually part of the turbulence generated at the ABL level that would breakdown and dissipate at the surface. To further speculate on the nature of the large-eddy inducing optical turbulence in the LAS system, Figure 15 conceptualizes the turbulent transfer mechanism supporting the argument that shear-induced turbulence along the side of the drainage flow enhances
Therefore, in what follows we analyze, each study case based on LAS optical scintillation times series and the calculated spectrogram. This analysis is guided by the flow dynamic and turbulence (i.e., wind direction, speed, and the
In the analysis of Case I, it was noted that between 15:00 and 00:00 UTC of the following day, at the initiation of the drainage flow, HEC on average is ~−4 Wm−2, while HLAS is ~−8 Wm−2 without major flux differences. However, in the period from 00:00 to 09:00 UTC, the time averaged turbulent flux values are HLAS −18 Wm−2 and HEC −14 Wm−2, showing an increasing divergence between the measurements. This discrepancy is consistent with the increasing probability of wind direction change thus affecting the HEC measurements. Because of the dynamic changes in the flow, the HEC and HLAS measurements are therefore not convergent during this period (see Figures 3 and 6).
The time-series of
For Case II on 7 February 2011, LAS was not working for about 30 min. after 21:00 UTC. Thus, in order to provide a continuous spectral analysis, the signal was cut-off at 21:00 UTC as seen in Figure 17 both panels. In this case, the average HLAS and HEC reached values of ~ −20 Wm−2 with the first sustained drainage flow pulse, while the second intermittent flow pulse developed an HLAS and HEC on average ~−16 Wm−2. The wind speed in the two intermittent pulses changes rapidly over a short period of time, and eddy covariance was calculated over 5 min. integration period as indicated previously. LAS and EC flux measurements compared well in time. Similarly, to the analysis of Case I, the time series of
Finally, for Case III, computed fluxes based on HLAS and HEC diverge in the presence of the drainage flow. For this period, the average values of HEC and HLAS are −20 and −40 Wm−2, respectively. HLAS is systematically higher than HEC throughout the period (03:00 to 18:00 UTC), and this difference is larger than the statistical fluctuation and random errors in EC methodology. The drainage flow enters the basin in the period of time from 9:00 to 15:00, and both HLAS and HEC measurements peak to similar values. The time series of
6. Conclusions
As a summary, this chapter presents an analysis of atmospheric turbulence based on field experiments including instruments for the ASL and the ABL. The objective of Wi-BLEx was to provide a new data set illustrating the dynamic and turbulence regime of a small scale flow penetrating a high latitude basin during extreme winter conditions. The study focuses on quantifying the temporal and spatial aspects of the developing turbulent structures in the ABL and in the ASL and their possible interaction in the conditions high-latitude polar atmospheres. Data sets from Wi-BLEx observations clearly demonstrate the influence of shallow cold drainage flows in the surface turbulent fluxes and the occurrence of large-eddy spectral structures of turbulence during winters in polar regions. Of particular importance in this experiment is the use of high-frequency optical scintillometer to determine the presence of large-eddy turbulence in the ASL that would have been difficult to resolve by sonic anemometers alone in particular when nonstationary flows are under analysis [32].
In this work, three cases were selected for analysis and discussion. The selected cases develop different temporal behavior, surface signatures, and turbulence patterns throughout the vertical structure of the ABL. Of special interest here is the notion that a small scale shallow cold flow entering the basin introduces surface mixing and localized areas of turbulence within the vertical structure of the ABL. Moreover, based on tower observations, it was verified that after drainage flow ceased, the surface layer rapidly re-stratifies in response to the outstanding radiation cooling rate of winter polar atmospheres. In all cases, the drainage flow observed in the vertical by Doppler sodar is represented by wind speed profile that has been observed previously under similar flow dimensions [9, 30, 33].
It is important to note that during winter, the absence of shortwave incoming radiation sets the surface radiation budget to low levels in absolute terms and mostly depending on longwave net radiation. Therefore, heat fluxes resulting from small scale dynamic processes are limited in absolute values up to 20–30 Wm−2 with episodic events of 40 Wm−2.
The cooling effectiveness of the drainage flow was the stronger in the intermittent Case II from all cases but only for short periods of time. A temperature drop of 11° C was verified with an average heat flux of −40 Wm−2 during the first intermittent pulse, while the rest of the intermittent events exhibited a lower negative average heat flux in the range of −10 and −5 Wm−2.
In conclusion, during winter, the interaction between polar atmospheres and landscapes combined with the presence of specific synoptic meteorological configurations clearly evidence the possibility of an inter-valley density flow impacting the surface energy balance at a regional scale. This mechanism was evident in the three analyzed cases but also throughout the Wi-BLEx data sets.
Large scale synoptic flows play an important role at regional level to onset the occurrence of drainage flow. Based on the analyzed information, the mode in which the drainage flows develop sustained or intermittent has a combined synoptic and topographic dependence. However, the spatial and temporal scales of the resulting drainage flow depend upon the actual flow-basin dynamic, turbulent, and radiative conditions.
All cases occurred under a surface high pressure forcing in the region. This synoptic meteorological feature is important because it normally comes with a weak pressure gradient force characterized therefore by weak horizontal winds and clear skies, strongly driving the radiative cooling in the basin. During the late winter, Case III, because of the diurnal effect of solar radiation, the study fundamentally differs from the central-winter cases I and II. The drainage event under analysis developed during the night of March 5, 2011 and the day of March 6, 2011. The flow dynamic setting for this case resembles the cases collected during Wi-BLEx during the late winter, in particular for the period March 1 to 10, 2011.
Finally, the three cases described in this study can be summarized as: Case I characterized by a sustained flow lasting ~18 h in the central part of the winter; Case II is characterized by intermittent flow pulses occurred lasting less than ~3 h in the central part of the winter; and Case III resulted a sustained flow lasting for about 9.5 h in late winter. Altogether surface and vertical observations demonstrated that localized turbulence in the ABL depth, as measured in the terms of
Acknowledgments
Wi-BLEx was supported by the Air Quality Office of the Fairbanks North Star Borough and by funding from the Department of Environmental Conservation of Alaska. Instrumental support is also recognized from the US Eielson Air Force Base in Alaska. During the Wi-BLEx observational field campaign, both authors were supported by the Geophysical Institute and the College of Natural Science and Mathematics of the University of Alaska Fairbanks. Support for the Chapter Book publication was granted form the Office of the Science of the Vice-Chancellor for Research Prof. Larry Hinzman, University of Alaska Fairbanks.
References
- 1.
Mayfield JA, Fochesatto GJ. The layered structure of the winter atmospheric boundary layer in the interior of Alaska. Journal of Applied Meteorology and Climatology. 2013; 52 :953-973 - 2.
Malingowski J, Atkinson D, Fochesatto GJ, Cherry J, Stevens E. An observational study of radiation temperature inversions in Fairbanks, Alaska. Polar Science. 2014; 8 (1):24-39 - 3.
Bowling SA, Ohtake T, Benson CS. Winter pressure systems and ice fog in Fairbanks, Alaska. Journal of Applied Meteorology. 1968; 7 :961-968 - 4.
Beran D, Hooke W, Clifford S. Acoustic echo-sounding techniques and their application to gravity-wave, turbulence, and stability studies. Boundary-Layer Meteorology. 1973; 4 :133-153 - 5.
Holmgren B, Spears L, Wilson C, Benson CS. Acoustic soundings of the Fairbanks temperature inversions. In: Weller G, Bowling SA, editors. Climate of the Arctic: Proceedings of the AAAS-AMS Conference, Fairbanks, Alaska. Geophysical Institute, University of Alaska; 1975, 1973. pp. 293-306 - 6.
Brown EH, Hall FF Jr. Advances in atmospheric acoustics. Reviews of Geophysics and Space Physics. 1978; 16 :47-110 - 7.
Fochesatto GJ, Mayfield JA, Gruber MA, Starkenburg D, Conner J. Occurrence of shallow cold flows in the winter atmospheric boundary layer of interior of Alaska. Meteorology and Atmospheric Physics. 2013. DOI: 10.1007/s00703-013-0274-4 - 8.
Clements WE, Archuleta JA, Hoard D. Mean structure of nocturnal drainage flow in a deep valley. Journal of Applied Meteorology. 1989; 28 :457-462 - 9.
Doran JC, Horst TW. Observations and models of simple nocturnal slope flows. Journal of the Atmospheric Sciences. 1983; 40 :708-717 - 10.
Benson CS. Ice Fog: Low Temperature Air Pollution, Defined with Fairbanks, Alaska as Type Locality. College, Alaska: University of Alaska Fairbanks, Geophysical Institute; 1965. p. 134 - 11.
Wendler G, Jayaweera K. Some measurements of the development of the surface inversion in Central Alaska during winter. Pure and Applied Geophysics. 1972; 99 :209-221 - 12.
Mölders N, Kramm G. A case study on wintertime inversions in interior Alaska with WRF. Atmospheric Research. 2010; 95 :314-332 - 13.
Benson CS. Ice fog. Weather. 1970; 25 :11-18 - 14.
Benson CS. Ice Fog: Low Temperature Air Pollution. CRREL Research Report 121; 1970 - 15.
Benson CS, Weller G. A Study of Low-Level Winds in the Vicinity of Fairbanks, Alaska. Report to Earth Resources Co. Geophysical Institute, University of Alaska; 1970 - 16.
Starkenburg D, Metzger S, Fochesatto GJ, Alfieri J, Gens R, Prakash A, et al. Assessment of de-spiking methods for turbulent flux computations in high latitude forest canopies using sonic anemometers. Journal of Atmospheric and Oceanic Technology. 2016; 33 :2001-2013. DOI: 10.1175/JTECH-D-15-0154.1 - 17.
Kaimal JC, Finnigan JJ. Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press; 1994 - 18.
Wilczak J, Oncley S, Stage S. Sonic anemometer tilt correction algorithms. Boundary-Layer Meteorology. 2001; 99 :127-150 - 19.
Vickers D, Mahrt L. Quality control and flux sampling problems for tower and aircraft data. Journal of Atmospheric and Oceanic Technology. 1997; 14 :512-526 - 20.
Lee X, Massman W, Law B. Editors Handbook of Micrometeorology. A Guide for Surface Flux Measurements and Analysis. Dordrecht, Netherlands: Kluwer Academic Publishers; 2004 - 21.
Kleissl J, Gomez JSH, Hong SH, Hendrickx JMH, Rahn T, Defoor WL. Large aperture scintillometer intercomparison study. Boundary-Layer Meteorology. 2008; 128 :133-150 - 22.
De Bruin HAR, Meijninger WML, Smedman AS, Magnusson M. Displaced-beam small aperture scintillometer test. Part I: The Wintex data-set. Boundary-Layer Meteorology. 2002; 105 :129-148 - 23.
Gruber MA, Fochesatto GJ. A new sensitivity analysis and solution method for scintillometer measurements of area-average turbulent fluxes. Boundary-Layer Meteorology. 2013; 149 :65-83. DOI: 10.1007/s10546-013-9835-9 - 24.
Gruber MA, Fochesatto GJ, Hartogensis OK, Lysy M. Functional derivatives applied to error propagation of uncertainties in topography to large-aperture scintillometer-derived heat fluxes. Atmospheric Measurement Techniques. 2014; 7 :2361-2371. DOI: 10.5194/amt-7-2361-2014 - 25.
Fochesatto GJ. Methodology for determining multilayered temperature inversions. Atmospheric Measurement Techniques. 2015; 8 :2051-2060. DOI: 10.5194/amt-8-2051-2015 - 26.
Neff WD. An observational and numerical study of the atmospheric boundary layer overlying the east antarctic ice sheet. Ph.D. Thesis. Boulder, Colorado: University of Colorado; 1980 - 27.
Neff WD. Observations of complex terrain flows using acoustic sounders: Echo interpretation. Boundary-Layer Meteorology. 1988; 1988 (40):363-392 - 28.
Neff WD, King CW. Observations of complex-terrain flows using acoustic sounders–Experiments, topography, and winds. Boundary-Layer Meteorology. 1987; 40 :363-392 - 29.
Hootman BW, Blumen W. Analysis of nighttime drainage winds in boulder, Colorado during 1980. Monthly Weather Review. 1983; 111 :1052-1061 - 30.
Sakiyama SK. Drainage flow characteristics and inversion breakup in two Alberta Mountain valleys. Journal of Applied Meteorology. 1990; 29 :1015-1030 - 31.
Drobinski P, Brown RA, Flamant PH, Pelon J. Evidence of organized large Eddies by ground-based Doppler Lidar, sonic anemometer and sodar. Boundary-Layer Meteorology. 1998; 88 :343-361 - 32.
Mahrt L. The near-calm stable boundary layer. Boundary-Layer Meteorology. 2011; 140 :343-360 - 33.
Horst TW, Doran JC. Nocturnal drainage flow on simple slopes. Boundary-Layer Meteorology. 1986; 34 :263-286