No additional calibration was applied to the instruments listed below:

CSI CSAT3 serial data are internally flagged by the instrument for bad records (i.e. rain drops). In the post processing bad 20 Hz records are linearly interpolated. If more than 256 CSAT3 records  during 10 minutes are flagged (approx. 12 seconds,  2%), then the 10min-block is completely discarded and no further statistics were computed.
For Gill HS, Gill R2 and the Young 501 no internal checks were used. METEK USA-1 report bad records (MD=10), but the error message format of the USA-1 does not report how many errors in sequence occurred (see 4.2).

4.2 Missing Records

The LabView data aquisition (2) checks for dropped records. If there are missing records in a signal, the integral statistics can still be calculated with the remaining measurements of the block (e.g. with 35900 instead of 36000 values for half an hour, the flux is not substantially altered). Because the number of missing records in sequence is unknown, the gaps cannot be interpolated.
No FFT may be applied to data with missed records, because for FFT a continuous dataset is needed. If METEK USA-1 sonics encounter an internal data check error then this results in dropped records. 10% of all METEK USA-1 data have dropped records, but less than 1% of all data from R2, HS, CSAT3 and Young 81000 Sonics show dropped records (the dropped records also include manual reboot, data transmission problems and FIFO buffer overflows). > individual values for all sonics.

4.3 Despiking

A simple despiking test is run over the 20 Hz data of all sonics including analog inputs. The despiking should avoid effects of single values that are completely off (called "spikes", caused by rain,dust,insects,other errors). The despiking-test checks every single 20Hz value z_i of all parameters  (z=u,v,w,t,q,CO₂) to be within a range:

z_i > mean (z) - s(z) * a
and
z_i  < mean (z) + s (z) * a

a was empirically set a = 6 for u and v, a = 10 for w, t,and q and a = 20 for CO₂.

5. Analog-Inputs

5.1 Krypton hygrometers

A total of 6 CSI KH2O hygrometers were combined with sonics to determine latent heat fluxes during BUBBLE. The KH20s were scaled using the following calibrations:
 

Site / Level	Labeled Photo	Turbulence from Sonic	Sampling	KH2O Serial No.	Calibration date / Calibration type	Provider KH2O
ALLS C 15.8m		Metek USA-1 2002 04003	Metek USA-1  Analog Input Raw Data	1461	12-oct-01 Campbell Scientific	Bulgarian Inst. of Met. and Hydr.
BMES A 2.2m ab. roof		CSI CSAT3 0530	Campbell 23x-Logger (SDM) 10min Cov.	1123	29-Feb-96 Campbell Scientific	TU Dresden
BSPR F 31.7m		Metek USA-1 99 03006	Metek USA-1 Analog Input Raw Data	1094	16-dec-98 Campbell Scientific	Uni Basel
BSPA E 29.9m		Gill HS 000046	Gill HS 000046 Analog Input 1 Raw Data	1448	01-may-01 Campbell Scientific	Uni Basel
GRNZ A 28.0m		Metek USA-1 2002 04004	Metek USA-1  Analog Input Raw Data	1096	6/7-Dec-01 IU Internal	Indiana State University
VLNF A 3.3m		CSI CSAT3 0118	Campbell 21x-Logger (SDM) 10min Cov.	1199	17-Oct-01 Campbell Scientific	Uni Basel

5.1 LICOR 7500

Two Licor 7500 CO2/H2 at BSPR to provide information on Carbon Dioxide Exchange and latent heat flux. Both instruments were sampled with a time offset of 250 msec relatively to the sonic. Due to a software error of the Licor 7500 this offset does not correspond to the real offset.

> Read our technical report on the " Impact of the LICOR 7500 Lag Correction on Field Data sampled 2000-2003 at the University of Basel"

In the postprocessing the offset is shifted back. The signal was transferred using the analog output of the LICOR 7500 and feeding it into the analog inputs of the sonics (Gill HS and Gill R2). The instrument at 14.7m was additionally sampled serially.
 

Site	Operation Period	Labeled Photo	7500 Serial No.	Analog Input at Sonic	Sensor separation	Maximal theoretical input resolution	Lower Range	Upper Range	Provider
BSPR 14.7m	June 14 2002, 15:00 - July 15 2002, 07:30		75H- 0254	11 bit 10 Hz 0 to +5V (Gill R2)	0.24m / 7°	H2O:  1.22(A) / 0.51(B) mmol m-3 CO2:  0.01 mmol m-3	H2O:  0(A) / 200(B) mmol m-3 CO2:  10 mmol m-3	H2O:  2500(A)/1250(2:  30 mmol m-3	NUS Singapore
BSPR 31.7m	June 24 2002, 14:00 - July 13 2002, 15:00		75H- 0332	14 bit 100Hz -5 to +5V (Gill HS)	0.4m/165°or 0.26m / 165°, after July 5., 2002, 09:00	H2O:  0.31 mmol m-3 CO2: 0.005 mmol m-3	H2O:  0 mmol m-3 CO2:  10 mmol m-3	H2O:  2500 mmol m-3 CO2:  50 mmol m-3	Uni Basel

^(A) Settings before 26/06/02, 08:35, ^(B) Settings after 26/06/02, 08:35.

6. Processing of Mean Statistics

6.1 Coordinate System

The coordinate systems of all sonics are rotated so that u+ points towards geographic East and v+ points towards geographic N. w+ always points upward. Note that, this is no rotation into mean wind i.e. no streamline rotation was applied. 

6.2 General Statistics

Statistiscs were calculated from simple blocks over 10 minutes without detrending. Statistical moments are calculated for all parameters (u,v,w,t,q,c) as well as the covariance-matrix for u,v,w,t,q,c

All output parameters are checked to be within a range. Values that are out of this range are removed and set to an error value.

> List of range- and clipping-settings for all parameters.

7. Turbulent Flux Densities

Flux densities labeled "corrected" in the database are:

• Q_H Sensible Heat Flux Density (corrected)

• Q_E Latent Heat Flux Density (corrected)

• Q_S Storage / Soil Heat Flux Density (corrected)

• QCO₂ CO₂ -Mass Flux Density (corrected)

All turbulent fluxes (Q_H,Q_E,from block averages of 20 Hz raw data over one hour without detrending. All fluxes are vertically oriented and no run-to-run streamline rotation was applied. 

7.1 Sensible heat flux density Q_H 

Q_E was calculated from the 60 min block value of the covariance w'a' (vertical wind / absolute humidity) Latent heat of vaporization, air density and heat capacity are calculated for each time step based upon measurements of air temperature, humidity and air pressure at the sites. The following correction were applied in the indicated order:

1. Oxygen-Correction with instrument individual factors (only for KH20 and not the LICOR 7500, after Tanner et al. ,1983). The correction typically enlarges QE  between +1% and +12 %. In relative numbers it is more pronounced at the urban sites (absolute lower QE)

2. Spectral correction and correction for sensor separation (Moore, 1986). The influence on QE is between 2% and 7%.

3. WPL-correction (Webb et al., 1980). The relative influence on QE is between 2% and 25%.