ISFF Project Report: CHATS

NCAR Integrated Surface Flux Facility for CHATS (15 March - 15 June 2007)

The horizontal array in the low-wide position during Phase I.


This document describes the operation and measurements of the Integrated Surface Flux Facility (ISFF) during the Canopy Horizontal Array Turbulence Study field experiment (CHATS), 15 Mar - 15 June 2007.

The objective of CHATS was to make spatial measurements of the velocity and scalar turbulence fields in a uniformly vegetated canopy, using arrays of sonic anemometer/thermometers augmented with fast response water vapor and carbon dioxide sensors. With this spatial information, the three-dimensional fields of velocity and scalar fluctuations can be studied to quantify turbulence transport processes and coherent structures throughout the canopy layer. An ancillary goal was to characterize the turbulent structure of the fields of wind, temperature, humidity, and trace chemical species within and above the orchard canopy.

18 sonic anemometers were deployed in a two-level, horizontal array, which had 9 sonics on each of two horizontal beams with a vertical separation of 1 m. The two horizontal beams were supported by 4 towers which are 9 m tall, and the horizontal length of this array was 15-16 m. The sonics were augmented with 5 Campbell KH2O and 5 Li-Cor LI-7500 fast-response water vapor and carbon dioxide sensors, as well as 5 vertical hot-film anemometers. The dimensions of the array (height and sonic spacing) were changed roughly once per week to obtain data within a range of those dimensions and canopy density.

The main sonic array was supplemented with a 30-m vertical profile of 16 sonic anemometers and 12 hygrothermometers, extending from near the surface, through the canopy and roughness sublayer, to well above canopy top. Also requested are fast-response static pressure sensors and hot-wire/film anemometers to be collocated with the sonic anemometers in the horizontal array, as well as a digital camera system to measure leaf area index (LAI). Finally radiation and soil measurements were made to complete the thermal energy budget of the canopy.

Photographs of the experiment

Field Logbook

A computer-readable field logbook of comments by NCAR and other personnel is copied from the field every 15 minutes during operations and is available in raw text format.

Measurement Sites

The ISFF measurements were made in the Cilker Chandler walnut orchard in the quarter section southwest of the intersection of Sievers and Pitt School Roads near Dixon, CA. ISFF instrumentation was centered on the 59th N/S row of trees west of Pitt School Road. The 30 m tower was located roughly 123 m south of the north edge of the orchard and the horizontal array was located roughly 239 m south of the north edge.

Also located nearby was an NCAR sodar/RASS system and the NCAR REAL scanning lidar. REAL was located north of Sievers Road in order to scan south over the orchard with vertical scans centered between N/S orchard rows 60 and 61 west of Pitt School Road.


The following table shows the instrument heights and spacing for the 6 different configurations of the horizontal array: 3 array heights at 2 different instrument spacings. To view the following table as a separate page for printing, click here.

Horizontal Array Configuration Periods

To view the following table as a separate page for printing, click here.

Sensor Notes

Sonic Anemometers
The CHATS sonic anemometers were deployed in two locations. On the vertical profile tower (vt), they were mounted on the south face of the 30m tower, with the booms pointing to the west, and in the horizontal arrays (ha) the booms were pointed to the south. The horizontal arrays consisted of two levels of 9 sonics each with heights separated by 1m. The horizontal array sonics were numbered 1-9 from west to east and the two levels were designated as (b)ottom and (t)op.

The CSAT sonic anemometers were tested in the EOL wind tunnel both prior to and following the field program. Sonics that had offsets prior to the field project that exceeded the manufacturer's specification of 4 cm/s were sent back to Campbell Scientific for recalibration. Following the field project, several sonics had offsets ranging from 6 to 15 cm/s. For those sonics, the following table lists the offsets to the orthogonal wind components, u, v, w, as well as the offsets to the measured, non-orthogonal components, a, b, c. Note that in most cases, only one of the measured wind components exceeds the manufacturer's specification, but that this can adversely affect more than one of the orthogonal wind components.

Following the project, the sonic tilts were calculated with the planar fit technique, using the 5-minute-averaged wind components. This process was applied to each period for which the mounting of a sonic was unchanged. The sonic tilt angles were then used to rotate the sonic data into a flow-parallel coordinate system with a time-averaged vertical velocity of zero. For the vertical profile tower, the U wind component is positive from west to east and the V component is positive from south to north. For the horizontal array, the U wind component is normal to the array and positive from south to north, and the V component is parallel to the array and positive from east to west.

For CHATS, we restricted the data used for the tilt calculations to nighttime data, since the daytime data appeared to be noisier presumably due to sweeps and ejections from the overlying flow into and out of the canopy. For this purpose, we used the incoming shortwave radiation at 16m to separate daytime from nighttime. For the profile tower we also eliminated data with wind directions within ±45° from the direction of the tower, i.e. into the 'back' of the sonic. For the sonics on the horizontal array, we only accepted wind directions within ±60° from parallel to the sonic U axis, i.e. into the 'front' of the sonic.

For the most part, the sonics on the profile tower were fixed in place for the duration of the project. There were two exceptions. First the original sonic at 6m (S/N 0740) was unable to operate at 60 Hz and was replaced with S/N 1119 on the morning of March 22, apparently around 10 am. The second exception was that the height of the 14m sonic was changed from 14.09 m to 14.52 m around 5 pm March 25 in order to allow raising of the boom to service one of the hot film anemometers. Above 10m, data from the entire project were used to determine the sonic tilt angles. Below 10m, data were used only until leaf-out in mid to late April, and it was assumed that the derived tilt angles applied for the entire project. Among other things, the tilt calculations within and below the canopy after leaf-out exhibited unrealistic vertical velocity offsets. Another difficulty with these data is the lack of data above 1 m/s, a threshold used in the tilt calculation to minimize the occurence of turbulent gusts with large wind elevation angles.

Note that the large vertical velocity offset for the 10m sonic was also detected in the post-project wind tunnel test.

In contrast, the configuration of the horizontal array sonics was changed multiple times during the project. Each configuration is characterized by a sonic spacing and the heights of the two linear arrays. The sonic spacing was changed twice during the project, once on March 31 from 0.5m to 1.72m and again on May 29 from 1.72m back to 0.5m. For each sonic spacing, the array was moved to three different heights by raising and lowering the horizontal ASTER towers supporting the two levels of sonics. With two exceptions, the sonic mountings were not changed during the raising and lowering of the linear arrays, and thus a single tilt fit can be calculated for each of the three periods with a constant sonic spacing.

Both exceptions are associated with the 1.72m sonic spacing. Each of the two levels of these arrays consisted of 4 sonics mounted on each of two horizontal ASTER tower sections, e.g. sonics 1b-4b on the west and 6b-9b on the east, plus a fifth sonic, 5b in the lower array and 5t in the upper array, mounted on a vertical Rohn tower placed between the west and east halves of the wide arrays. Thus the mounting of sonics 5b and 5t was changed every time the height of the array was changed and a separate tilt fit calculation is necessary for each height.

The second exception occurred on April 15, when the EMT tubing used to provide additional support between the individual sonics was stiffened by inserting wooden dowels at the junctions between the tubing. This was done for the upper and lower east arrays, but only for the upper west array. Thus sonic tilts were calculated separately for the periods before and after April 15.

Due to the difficulties after leaf-out mentioned previously, the tilt fits after leaf-out were essentially determined from data at the 9.6m and 10.6m heights and assumed to apply to the lower heights for the same sonic spacing. For the same reasons, tilt fits were not possible after April 15 for the 5b and 5t sonics in the wide array at the lower heights. Further, the data at 9.6m and 10.6m was often scarce and consequently the raw fits gave unrealistic values for the vertical velocity offset. In those cases, the offset was specified from earlier fits of the same sonic calculated with more extensive data.

The sonic orientations or azimuths were measured with a digital sighting compass known as a DataScope. The largest uncertainty in this technique is identifying the features of the sonic that indicate proper alignment of the DataScope with the sonic U axis. Due to poor sight lines through the trees, this was not done for the sonics above 9m on the profile tower. The average of the good readings below 9m, 185.4°, was used for all the unmeasured azimuths.

The sonic azimuths were measured for the horizontal arrays when they were below the canopy at the lowest height, and it was assumed that the azimuths were unchanged as the arrays were raised and lowered. Note that the azimuths for the wide arrays were measured only once, prior to inserting the wooden dowels in the EMT tubing on April 15. Further, the azimuths of sonics 5b and 5t were also only measured this one time, despite moving those sonics to different heights as the wide configuration was changed. In both latter cases, the measured azimuths were assumed to apply during the unmeasured periods.

The absolute values of the sonic temperatures, tc, are inaccurate by perhaps 1-2 °C because of the inability to maintain the sonic path lengths to a sufficient tolerance. Campbell testing and analyses shows that they can hold CSAT path lengths to better than 0.3 mm out of 11.55 cm, or about 0.26 percent. This gives a 0.26 percent error in the speed of sound, which in turn, gives a sonic temperature error of about 0.52 percent. At 300 °K, this is a 1.6 °C error. (Note that the wind and temperature fluctuations in this example are in error by only a fraction of one percent.)

For this reason, we have calibrated the sonic temperature measurements with a linear fit relative to NCAR hygrothermometers mounted nearby,

tccalibrated = offset + gain*tcmeasured

The temperature and humidity sensors in the NCAR hygrothermometers have been calibrated in the laboratory with respect to NIST-traceable standards. The following tables list the offset and gain applied to sonic temperature measured by each individual sonic. The rms deviations from these fits are on the order of 0.14 °C. The general pattern of these deviations follows a 24-hour cycle that suggests they are associated with residual radiation errors in the hygrothermometer mechanically-aspirated shield.

Krypton Hygrometers
Under construction...
CO2 sensors
Hot-film anemometers
With one exception, all radiometers appear to have performed normally. During the second phase of CHATS (6 May -- 10 June), Tcase.out.16m had level shifts and spikes. Field staff attempted unsuccessfully to find a loose connection to fix this problem. Data when this problem occured have been manually determined and were removed from the QC'd data set. Note that without Tcase, Rlw cannot be determined. However, since is always within 1.3 degrees of Tcase.out.16m, it could be used as a surrogate. The resulting calculation of Rlw would be in error by up to 7 W/m^2. This calculation is left to the user. Calculation of Rlw using the pyrgeometer thermopile and thermistor outputs (and Rsw, in the case of the lower Eppley radiometers), along with sensor-specific calibrations, was added to the final data set.

Logger values of radiation reported before the surface radiometers were installed have been removed. Also, data taken during radiometer cleaning events have been removed. The radiometers at 16m were moved to the other side of the tower on 19 March. Data during this move have been removed, but values prior to the move have been retained in the dataset.

The near-surface radiometers were powered by a stand-alone battery, charged by a solar panel. When the orchard canopy closed during the break, charging was weak and the battery drained completely, causing a gap of about 8 days (18--27 April). These data cannot be recovered, but the remaining data are good.

Soil Sensors
The same power problem noted above for the near-surface radiometers affected all soil measurements. Otherwise, the Tsoil and Gsoil measurements were good.

The ECH2O Qsoil measurements using manufacturer's calibration agreed with 4 of 5 gravimetric measurements that were taken (see black circles on this plot). However, manual measurements using our TRIME sensor (red circles) were quite a bit higher. Field staff noted errors using the TRIME sensor due to the soil type and large differences were observed between readings using "permanently" installed TRIME probes and the removable probe. Thus, we assume that the Qsoil measurements are good as reported and that the TRIME readings (normally used for verification of the ECH2O probe) are suspect.

The TP01 thermal properties probe reported good values of thermal conductivity (lambdasoil). However, readings of the decay time constant of the TP01's heat pulse (and thus thermal diffusivity and heat capacity) sometimes were larger (3 or 4 seconds) than the most common value of 2 seconds. To some extent, these were correlated with larger values of lambdasoil, but these values were well within the specifications of this sensor. Assuming that the time constant was always 2 seconds produces heat capacity values that are quite consistent and that track soil moisture. (This plot shows this relation, apparently with a "dry soil" heat capacity of 2.4e6 J/C m^3 when Qsoil<15%, a rise with Qsoil with a slope nearly at the expected slope for water, and a "wet soil" heat capacity of 7.5e6 for Qsoil>28% -- presumably near field capacity.) We produced special code for CHATS that sets the time constant to 2 seconds and added derived values of lambdasoil, asoil, and Cvsoil to the final dataset.

Using all of these measurements, we added derived values of the heat flux at the top of the soil, Gsfc, (that includes the soil heat storage term) to the final data set.

Data Tables and Plots

Click on the above link to view plots.

Plots of wind tunnel hotfilm tests

Data Download

Access to the ISFF CHATS data is initially restricted to the CHATS principle investigators. Contact Ned Patton or Tom Horst for a username and password if you wish to have access to the data.

Data Collection

This is the 5th use of the Viper-based PC104 data systems (NIDAS, from which the HAIPER data system is derived), running the new data acquisition code for primary data acquisition.
Port Assignments
Refer to this table for the port assignments on the data systems.
Estimated Dataset Size
sample size
Licor 750051016+473.10.3
3D Hotfilm3200016+(2+9*2)
(all 9 films)
1D Hotfilm5200016+(2+5*2)
(all 5 films)
(640x480 jpeg)
15 hours/day
Total   208.317.2


This page was prepared by Greg Poulos and is now maintained by Tom Horst NCAR Earth Observing Laboratory, In-situ Sensing Facility