Estimating Latent Heat Flux from a Vineyard using Scintillmnetry

The combination of a near-infrared and a microwave scintillometer were used to estimate the line-averaged, latent heat flux density (A.Es) from a well-irrigated, vineyard valley during summer. The 2 km scintillometer beam passed over the valley floor at a height (z) of 30 m. Eddy covariance towers instrumented at 3 m above the valley surface provided an indepen­ dent check of vineyard evaporation. A formula for free convective A.Es compared favourably to the eddy covariance values, A.Ee at low wind speeds and under clear skies. As the convective boundary layer (CBL) developed and passed through the scintillometer beam it caused additional scintilla­ tions, in which case A.Es >> A.Ee. During unstable conditions the compari­ sons were in close agreement. However A.Es tended to overestimate A.Ee as the wind speed increased. For near-neutral conditions the scintillometer gave unrealistically large values of A.Es. The additional scintillations origi­ nated from the strong entrainment of advected dry and warm air into the newly formed surface boundary layer. During calm overcast days, cold days and during the early morning periods when the surface fluxes were small, the signal strength of the infra­ red scintillometer approached the system noise. (


INTRODUCTION
Over the last ten years several experiments have used a combination of a microwave and near-infrared scintillometer to determine the line-averaged fluxes of the sensible heat H , (W m-2 ) and the latent heat A.Es (W m-2 ), directly above the Earth's surface.Notable is the experi ment of Hill et al. (1988) who used scintillometers operating at millimeter and micrometer 1The Horticultural Research Institute of New Zealand, Kerikeri Research Centre, Kerikeri, Bay of Is lands, New Zealand *Corresponding author address: Dr. Alan E. Green, The Horticultural Research Institute of New Zea land, Kerikeri Research Centre, P. 0. Box 23, Kerikeri, Bay of Islands, New Zealand E-mail: agreen@hort.cri.nzwavelengths.They were the first to outline the methodology for determining H, and A.Es and produce what Hill described as 'a reasonable agreement with those values from micrometeo rological measurements'.The next recorded experiment using multiple scintillometers was Green et al. (1999) who found H. and A.Es compared favorably to eddy covariance measure ments made at the midpoint of the scintillometer transect.This latter experiment confirmed that the recently designed microwave scintillometer was operational and sensitive to inertial sized eddies of heat and water vapor.This research also highlighted the effect of line-aver aged, low frequency humidity fluctuations and their detrimental effect on the scintillometer signals.In both these experiments the surface was flat, expansive and homogeneous in sur face coverage with the scintillometer beam close to the canopy (z < 10 m).
There are provisos for the scintillometer method because it is indirect in its approach, relying upon similarity theory to scale the temperature structure parameter, Ci (K 2 m-213) and the humidity structure parameter, C� ([g m-3]2 m-213 ) as functions of atmospheric stability.In practice, these parameters are first calculated from the scintillometer signals then incorporated into similarity functions to determine H, and A.Es_ To satisfy the prerequisites of similarity theory, scintillometer experiments should be conducted over a flat and homogeneous land scape under a relatively stationary atmosphere.This satisfies the empirical linking (albeit at times tenuous) of the scintillometer beam-intensity perturbations with the surface generated eddies of heat and moisture.
If there is to be an acceptance of the scintillometer method, comprehensive experimental verification and testing over a variety of terrain and atmospheric conditions is mandatory.It is our intent in this paper to contribute an improved understanding of the scintillometer method from initial experience gained with this scintillometer combination (Green et al., 1999).We present here results using a microwave and near-infrared scintilloi:neter propagating across a vineyard valley.Our approach is somewhat different than other close to the canopy scintil lometer experiments.A vineyard has been studied previously using an near-infrared scintil lometer (De Bruin et al., 1995) however the scintillometer beam was very close to the canopy (z = 3.3 m) and the soil-plant conditions were dry.Here we examine the scintillometer capa bility to measure the surface evaporation across an irrigated valley and at a beam height much greater than previously attempted for this two-scintillometer combination.Operating at height is necessary to cover the large distances (up to 10 km).The surrounding farmland was drought stricken.We suspected the contrast in soil-water conditions would cause additional scintilla tions if dry air advected into the valley.The scintillometers were installed at z = 30 m above the valley floor by placing them on opposing and exposed ridges (see Fig. 1).Within the vineyard at two locations eddy covariance instruments monitored H e and A.Ee.These mea surements provided a measure of the vineyard energy balance and a benchmark for the scintil lometer determined fluxes.The valley floor had no significant undulations and the canopy was of a consistent roughness about the center of the scintillometer beam (maximum scintil lometer weighting).Based on past experience (Green et al., 1999) it was believed possible to estimate half hourly A. Es.The preferred wind direction was from the NW, because this was the wind that blew up the valley and across substantial grape planting before intersecting the scintillometer beam.If the wind blew into the valley from across the surrounding hill-sides this could adversely effect the scintillometer performance and A.Es was treated with scepti cism.
2. THEORY Scintillometers provide a wavelength-dependent, line-averaged measurement of the re fractive-index structure parameter, C � (m-213).Embedded within each value of C � reside the weighted spatial effects of temperature and humidity fluctuations that have affected the scin tillometer beam along its entirety.Each received scintillometer signal is time integrated within a bandwidth that encompasses the inertial subrange of turbulent eddies.These conditioned scintillometer signals provide the first step in determining H, and A.E,.

Structure Parameters Ci and C�
Through assuming these structure parameters obey Monin-Obukhov Similarity Theory (MOST) Hill et al. (1988), provide the values of Ci and C� obtained for a dual scintillometer operating at microwave and near-infrared wavelengths (denoted by subscripts m and i respec tively) as, Ci = [ A�.m C� .i+ A�.iC�.m + 2SA Q_mAQ,i �C�,m C�.i] /(QD)2, (2.2) A i and A� are parameters (Hill et al., 1980) that are influenced by the operating wavelength of the scintillometer, absolute temperature T (K), pressure P (N m-2) and humidity Q (g m-3).
The factor D expands to, (2.3) One significant consequence of MOST is that S, the T-Q correlation coefficient takes a value of+ 1 for unstable conditions (Hill, 1989).Such a correlation has not always been observed in scintillation experiments.For instance Kohsiek (1982) found an average S of0.75 while Hill et al. (1988) found S to be much closer to unity (S;;:;: Q_99)_

Sensible and Latent Heat Flux Densities
Our primary interest in scintillometers is determining line-averaged values of H, and A.E, at kilometer scales.To infer these fluxes across large areas scintillation measurements must be made over long horizontal paths.In this experiment the scintillometer averages the scintil lometer signals over a horizontal distance of 2 km.A.E, is our main surface flux of interest because of its potential for irrigation scheduling.Operating over such distances or greater is acceptable provided the scintillometer pathlength is a small fraction of the boundary layer height.For instance Kohsiek (1985) operated a C02 laser scintillometer over 5.9 km and z = 40 m.At such heights slight non-uniformity in the surface maybe from ditches or roads, cloud cover or canopy cover are blended with the surrounding surfaces if there is some mechanical mixing of the boundary layer.Because this scintillometer process is one of line-averaging, a single spatial anomaly at the surface would need to cause a very significant difference in surface flux to produce a measurable signal change at the scintillometer receiver.In such cases it would be unwise to select an experimental site that wasn't fairly uniform.The scintil lation method can be considered a form of low-pass filtering whereby small spatial and tempo ral changes are smoothed by the average signal.Further, as the weighting function of the scintillometer method is symmetrically bell shaped, the scintillometer can be positioned to minimise the effect of any significant change in surface conditions from the normal cover underlying the beam.
Currently MOST provides the only framework for indirectly predicting H, and "-Es from C� and ct.As has been previously discussed the ideal situation for such experiments is a flat and homogeneous site having a stationary atmosphere.Such was the case in the Kansas Plain Experiment from where the original MOST algorithms were derived (Wyngaard, 1973).
In this experiment we assumed MOST was applicable in the vineyard and C� and C� would scale to the scintillometer beam height as functions of atmospheric stability (z/L ).(2.6) The latent heat of vaporisation of water Ly is taken here as 2.45 x 103 kJ kg•1• Obtaining a reliable line-averaged u. for inclusion in (2.6) is difficult to achieve and other options are favoured.Over a homogeneous surface a line-averaged u. is preferred to that obtained from either an eddy covariance measurement or that estimated from the windspeed profile and a surface roughness measurement.Neither of the latter approaches will spatial average u. at the length scale of the scintilloineter transect.It is impractical to measure u. at the scintillometer beam height of 30 m and position of maximum signal weighting (the mid point of the beam).Instead, conventional meteorological instruments placed at z = 3 m esti mate u. which is then taken as being equivalent at z = 60 m because a constant flux layer is assumed.
For near-neutral conditions, typically characterized by overcast skies, strong surface winds and low values for surface fluxes, (z/L0) approaches zero as mechanical turbulence dominates.The wind profile is logarithmic with z and u. can be determined from an average windspeed u0 (m s•1) at z = 3 m, the canopy displacement height d (m) and surface roughness z 0 (m) using, An empirical calculation of z is made assuming z /h = 0.13 and d/h = 0.7 (Sene, 1994).
This requires an estimate of the average canopy height, h0 (m).Deviation from the log-law increases for unstable conditions requiring a modification of (2.7) to become, The full expression for the function 'P(z/L0) can be found in Panofsky and Dutton (1984 ).An iterative solution to (2.8) is obtained by first assuming an initial value for (z/L0) then calculating u*.From (2.5) a new (z/L.) is determined and substituted into (2.8), the process repeating until convergence.In the author's experience this procedure is fairly robust, requiring on average, just three to four iterations to complete.
For a very unstable atmosphere, the free convective situation, when (z/L0) < -1, mechani cal mixing is not significant permitting a simplification of (2.4) using the prior the expressions for L0 and A.Es.This rearrangement expresses C� as, c� = 2 .s(A.Es J 4 13 z-4 '3 ( TQ. J 21 3 Ly gT. (2.9) From (2.9) there is a decrease of C� with height.In reality this is observed as a rapidly diminishing scintillometer signal.Scintillometers are placed at greater heights to reduce the risk of signal saturation and to obtain greater scales of spatial integration.In so doing the resolution of the scintillometer may be compromised.
Continuing with our derivation of A.Es under free convection, from (2.4), By the substitution of (2.10) into (2.9), A.
This is a similar result to the free convective form given by Hill et al. (1988).Note that both C T and C Q are needed to estimate A.Es and so the technique requires at least two different wavelength scintillometers.In contrast the free convective, sensible heat flux only requires a value for CT which can be provided by a single near-infrared scintillometer (Panofsky and Dutton, 1984).

Site Description and Instrumentation
The Brancott Estate situated in the fluvial Fairhall Valley at the foothills of the Southern Ranges is found on the outskirts of Blenheim (Marlborough region, South Island of New Zealand).The Estate was chosen as the experimental site, being one of the largest vineyards in this country encompassing some 240 ha of vines stretching across the valley floor and bounded by low hills (schematic Fig. 1).These hillsides provided a convenient solution to mounting the scintillometers sufficiently high so the beams could see the valley-floor and line-average the vineyard A.Es.Using a laser theodolite the scintillometer transect was established as 1,969 ma t an average height of 30.3 m above the valley floor.The grade was 1:248 from the transmitter end on the hillside above the dry Fairhall riverbed to the receiver end, 'Rob's Knob'.For the purposes of this experiment the scintillometer beam was treated as parallel to the ground surface.The scintillometer beam crossed the central portion of the vineyard in an EW direction from transmitter to receiver.The preferred wind direction was from the NW because it was from this direction that the wind blew up the valley and across the irrigated grapes (a fetch of more than one kilometer) before intersecting the scintillometer beam.Down wind of the vineyard were further grape plantings (Fairhall Estate) but to the NE was the beginning of a dry scrub-land interspersed with a barren surface (local civilian and military airport) and some scattered olive plantings.This surrounding drier farming terrain then ex tended many kilometers to the base of the Nelson foothills (Richmond Range) having more vineyards on either side of the distant Wairau River.Because of a prolonged drought the irrigated Brancott Estate was for all practical purposes an oasis.
'' .� ;• ,., The scintillometers were the same microwave (1.1 cm wavelength) and near-infrared (0.94 µm wavelength) types as described by Green et al. (1999) and were mounted on short 1 m towers, themselves cemented into concrete bases at either end of the scintillometer transect.A plywood cover was built around each scintillometer to protect against wind vibration and the weather.At a later date, clear acrylic covers were added across the front openings as it was suspected the wind shear on the microwave dishes could cause unwanted mechanical scintilla tions.Acrylic is transparent at the scintillometer wavelengths.Following this experiment the scintillometers are now permanently housed in 1 m3 aluminium cubes with portable power systems and telemetry capability.This new construction greatly simplifies installation and is ideal for transportation and weatherproofing.
The Brancott Road running NS separated the Estate into eastern and western sections characterized by the grape-rows orientated NS on the eastern side and EW on the western side.In most cases the vines were planted 1.8 m apart in rows spaced at 3 m but planting age varied from 1 through to 25 years.The vines were trellised and he was estimated as 1.5 m, though some vines at the centre of the scintillometer beam were 2 m in height.The eastern side was planted in predominantly the Sauvignon Blanc variety with the canopy touching the ground.In comparison the vines on the western side were sparser and were pruned below 1 m.The vineyard relied on irrigation to meet the water requirements with all vines receiving water via drip emitters.In contrast the inter-rows were dry and devoid of plant cover due to the enduring drought.
Two, 3 meter high, open-lattice towers were instrumented with rapid response 1-D sonic anemometer-thermometers and krypton hygrometers.Half-hourly values of A. Ee and H0 were calculated using the eddy covariance technique.Other slower sensors included sensitive cup anemometers, a net radiometer R" (W m•2) and a combination temperature-relative humidity sensor.One tower was placed in the mature canopy on the eastern side almost directly under the centre of the scintillometer beams (designated the road site) and the other tower near to the Fairhall river and the much younger vines (the river site).Wind speed, temperature and the relative humidity were also recorded by data-loggers on Rob's Knob.

Weather Conditions and Data Collection
The experiment started in late summer, February 1998, during an extreme drought.There had been no significant rainfall in the Marlborough region during the previous winter-spring period.The surrounding hillsides and pastures were brown and dry with little if any growth.In contrast, the canopy of the vineyard was green and lush.Data collection first began on the afternoon of the 21st February (day of the year, D51) and was continuous for the next three days until the monthly average rainfall ( -35 mm) was experienced over the intervening days, D54 and D55.As the ground was baked hard very little of the rain penetrated the surface soil and most disappeared as runoff.Throughout days D51, D52 and D53 strong winds, mainly N -NW drove up the valley and through the scintillometer beam.Days D52 and D53 were overcast.These days had rel atively low He and A.Ee.By day D56 the rain had stopped and the sky was clear and the early morning winds were a cold SW blowing across the surrounding hills and through the scintillometer beam.In the afternoon the wind direction shifted around to the NE and was accompanied by intermittent cloud cover and an increasing air temperature.Day D57 had a clear sky with low windspeeds in the morning.By the early afternoon ofD57 the windspeed had increased from a northerly direction.D58 started with a brilliantly clear morning which quickly degenerated into overcast conditions with SW wind.The final day of the experiment, 059, began the morning with strong SW winds and overcast skies followed by rain in the late afternoon.
Data from the eddy covariance and the scintillometer instruments was recorded using Campbell Scientific Inc. 21x micro-loggers programmed for 10 minute and half-hour aver ages.The data was retrieved daily by portable computer and analysed using a software spread sheet.Supporting micrometeorological data were recorded using a standard met station lo cated beside the eddy covariance instruments.

Near-neutral and Advective Conditions
Day 051 had a clear sky with strong N-NW winds.In contrast days D52 and 053 were characterized by overcast skies but still with strong N-NW winds (see Fig. 2) that blew up the valley almost perpendicular to the scintillometer beam.For instance at 1200 hrs the windspeed at 30 m was estimated to be uc J om -24 m s 1• The calculation of u c3orn assumed a logarithmic wind profile under near-neutral conditions.Subsequent calculations of u. at z = 3 m were taken as the same value at z = 30 m.   Also contributing to the additional signal variance may be wind turbulence generated by the valley slopes and the ridges where the scintillometers are positioned.Although the scintil lometer path weighting is minimal at the transect extremes (Wang et al., 1978), additional mixing of the atmosphere by these physical barriers could have created unwanted fluctuations in T and Q within the valley.It is expected this effect would be particularly evident at the ridge height rather than near the valley floor as the wind accelerates upwards towards the crest of the hillsides (Kaimal and Finnigan, 1994).For the N -NW wind direction this meant the wind would strike the eastern ridge at an angle of approximately 45 degrees.The scintillom eter data collected at height and under the combination of very strong advective N -NW winds taking place within this valley, remains unusable in its present form.

Unstable Conditions
Constant heavy rain over days D54 and D55 meant all instrumentation had been turned off during this period to protect against failure.The experiment was restarted at 0900 hrs on day D56.As Until 1200 hrs (SW wind) the comparison between A.Es and A.Ee was reasonable.In the afternoon when the wind shifted direction to the EW to blow over the east-ridge and down the scintillometer beam there was a marked underestimation by A.Es.On several occasions during the early afternoon transition in wind shift the structure parameter Ci was observed to vary by more than an order of magnitude over 10 minute periods.Following Hill et aL (1992) we have classified these periods as non-stationary.The atmospheric conditions prevented the applica tion of MOST.By mid afternoon (-1500 hrs) agreement between A.Es and A.Ee had im proved.Drier winds that blew from the NE apparently advected sufficient dry air into the valley so there was a trend for A.Es > A.Ee.The D56 behaviour of A.Esc can be demarcated roughly into three temporal transitions caused by a 180 degree change in wind direction.The results suggest the successful compari son of the scintillometer and eddy covariance methods is dependent on the strength and the direction of the wind, the surrounding topography and their relation to the scintillometer beam orientation.
During the morning of day D57 the atmosphere tended towards highly unstable condi tions to provide a data set with which to test the effectiveness of (2.11).Average wind-speed was low during the first part of the day (uc3m < 1.5 m s-1) and the sky was clear throughout the previous night and extended over the morning of D57 (Fig. 6).By noon the air temperature had climbed to 19°C from an early morning low of 10°C (0800 hrs).From noon the wind-  speed increased from u 3 -2 m s-1 to u 3 -7 m s-1 by 1600 hrs and from a northerly direction.

cm cm
As significant mechanical mixing was now occurring in the boundary layer (2.11) was no longer valid.Figure 7 is a time series comparison between AE s and A.Ee for 057.In the period be tween 0800 and 1330 hrs A.Es was been calculated using (2.11) assuming very unstable con ditions.From 1400 hrs onwards and with the increasing wind-speed, AE s is determined using the iterative method as described for unstable conditions.In the first instance and in contrast to A.Ee, the values of /..E s lying between 0800 and 1030 hrs are larger than believable as Rn < 450 W m•2• During this period the values of C� and C� (Fig. 8) experienced an increase in magni tude peaking at 0900 hrs.Calculated A.Es was then very much greater than expected from a surface latent heat flux.Both the microwave and large aperture infrared scintillometer signals independently showed additional signal variance.In explanation it is proposed these peaks are indicative of the evolution of the CBL (as indicated on Fig. 7) growing in height in re sponse to surface heating.Stull (1988) describes the depth of the CBL increasing slowly, under calm situations, from several tens of metres controlled by the strong nocturnal stable layer capping the CBL.We believe that as the air temperature rose in the mid-morning, the top of the CBL moved upward and passed through the scintillometer beam.The transition in air density as the warmer air moved through the beam caused additional fluctuations to be recorded by the scintillometer, but these went unseen by the eddy covariance instruments.Kohsiek and Herben (1983) infer this behaviour in their paper when describing evaporation made by measurements of radio wave scintillation.Additionally, Wyngaard and LeMone (1980) discuss the behaviour of C� and C� in the CBL and observed that these values can be several orders of magnitude greater near the inversion height than expected through that gen erated by surface fluxes.By late morning (1100 hrs) the height of the CBL had grown beyond both the scintillom eter and eddy covariance instruments and the fluxes recorded by both instruments was con tained within the convective surface layer.Application of (2.11) for free convective condi tions was valid and demonstrated extremely pleasing agreement between A.Es and A.Ee from 1100 to 1330 hrs (Fig. 7).At 1130 hrs an acrylic screen was placed over the scintillometer

539
transmitter.A similar screen was erected at the scintillometer receiver about 1730 hrs.The screens were designed to prevent mechanical oscillations by heavy winds moving the scintil lometers.This modification to the enclosures caused peaks in Ci and C� about 1130 and 1730 hrs (Fig. 8) as the scintillometer beam was broken.There is no evidence there was an improvement in the scintillometers performance by the placement of these screens.From 1400 hrs onwards the free convective formula was discarded in favour of the unstable approach.
Calculations of A.Es constantly overestimate A.Ee in a manner similar to the results of day,

056.
The sudden jump in magnitude of both Ci and C� around 1430 hrs (Fig. 8) from one plateau to another occurs with increased wind-speed (see u03m, Fig. 7).
During the very unstable conditions the signal from the near-infrared scintillometer was very close to the instrument noise.Some discussion on this problem is warranted because the scintillation method relies on accurate measurements of C! i.During very unstable condi tions simple theory states that both Ci and C� are proportional to z•413 within the surface boundary layer.In practical terms this means the magnitude of Ci or C� measured by a scintillometer placed at a height of z = 30 m will be only 5 % of that measured by the same instrument placed at z = 3 m above the valley floor.This reduction in signal strength becomes problematic for low values of evaporation as the scintillometers must be capable of a suitable signal resolution without approaching their own noise limit.During the convective condi tions, Ci was three orders of magnitude less than the values measured over the unstable mid-afternoon situation (Fig. 8).The other two peaks in C� and C� occurring in the early morn ing and the early evening (0700 -0800 hrs and 2000 -2100 hrs) represent the transition be tween the unstable and stable atmospheric states.During these periods C� and C� are inde terminable as will be A.E,.
The final days of the experiment, D58 and D59, are not discussed in detail here as both were overcast, cold and with strong winds blowing NW down the scintillometer transect mak ing data interpretation difficult.During this period the signal output from the infrared scintil lometer was observed to approach its own noise limit ( C� < 10-17 m-213).However the clear early morning of day, D58, provided several half-hour data points which although not pre sented had A.E. and A.Ee in close agreement.
The issue of improving the signal strength of the near-infrared scintillometer is currently being addressed through the use of a more powerful radiation source.It is noted the micro wave scintillometer signal strength was always adequate.

CONCLUSION
The purpose of this research was to identify whether the scintillometer method could be used to estimate evaporation from a vineyard valley to be capable of providing figures for daily irrigation scheduling and the management of water resources.The scintillometer was placed on opposing valley ridges to propagate its beam across and well above the valley floor.
In calculating A.E. it was assumed that similarity theory was workable in the valley and the scintillations were caused by the surface fluxes of heat and moisture_ The results demonstrate the success of these A.E, measurements when compared with eddy covariance data are dependent on prevailing atmospheric conditions and the surrounding terrain.Operating scintillometers at z = 30 m creates additional problems not encountered when the beam is closer to the canopy.These problems are extraordinary scintillations and much weaker signals whose symptoms are an excessive intensity variance and reduced signal to-noise ratio.The cases identified to date include the entrainment of warm dry air under advective conditions and the development of the CBL as it passes through the scintillometer beam.Confirmation of these phenomena is a next experimentalstep.Excessive wind-speed contributed to the overestimation of A.Es relative to A.Ee during unstable conditions.The difference became extreme for near-neutral conditions with a dry to wet transition downwind.
There is no evidence to suggest the scintillometer combination will not perform well when operated at this height and under near-neutral conditions over uniform terrain.The free con vection formula gave good results for the small data set collected.
In summary the results presented indicate the scintillation method can produce credible estimations of A.E. from an irrigated valley.The scintillometer performed best under unstable to very unstable atmospheric conditions.The reasonable comparison between the scintillom eter and eddy covariance time series suggests the application of similarity theory was appro priate.Scintillometers need to be mounted at greater heights than conventional instruments in order to obtain an unobstructed propagation of their beams across kilometer distances and to compensate for curvature of the Earth.As a result they may encounter some of the unwanted effects reported in this experiment.

0
We use the empirical scaling relationship ofHill et al. (1988) for unstable conditions, ,z/Lo<O, (2.4) L0 (m) is the buoyancy length scale, which includes a correction for humidity fluctuations and is defined by, (2.5) K = 0.4 is von Karman's constant, T. (K) is the temperature scaling parameter, g = 9.8 1 m s-2 is gravitational acceleration and p = Hs I A.Es we have called the line-averaged Bowen ratio.Kohsiek (1982) notes that P is proportional to the ratio C/CQ which are two of the structure parameters derived from this experimental scintillometer.The humidity scaling pa rameter Q. (g m•3) and the velocity scale u. (m s•1) combine to give a value for /.,Es as,

Fig
Fig. I.A simplified schematic of the Brancott Valley showing the positions of the scintillometer and eddy covariance instruments.

Fig. 2 .
Fig. 2. Time series of eddy covariance determined fluxes He and A.Ee along with net radiation, Rn and the average windspeed at z = 3m, u c3m for near neutral days 051, 052 and 053.
Fig. 3.A comparison of half hour averages of A.Es and A.Ee obtained under near-neutral and strongly advective conditions.
Fig. 5 using the notation, SW, EW and NE.

Fig. 4 .
Fig. 4. Average wind-speed at z = 3m, u 3 and net radiation R cm n during the unstable conditions of day, D56.

Fig. 5 .
Fig. 5.A time series of A.Es ( •) and A.Ee ( o) for 056, a day of unstable atmo spheric conditions with shifting wind direction.

Fig. 6 .
Fig.6.Air temperature, T, average wind-speed at z = 3m, uc3m and net radiation, Rn for day, 057 which experienced very unstable conditions in the morn ing period.

Fig. 7 .
Fig. 7.A time series for day 057 showing the effect of the CBL passing through the scintillometer beam mid-morning and the reasonable agreement ob tained between very unstable A.E, ( +)and unstable AE8 (•)with val ues of A,Ee (o).

Fig. 8 .
Fig. 8.The behaviour of the structure parameters Ci (o) and C� (•)for day, 057.Each of the spikes in the time traces are periods when reliable data is unavailable.This is due either to obstruction of the beam or the behaviour of the atmosphere as it undergoes various transitions.