PROCESSING AND QUALITY CONTROL OF EDDY COVARIANCE DATA

PROCESSING AND QUALITY CONTROL OF EDDY COVARIANCE DATA DURING LITFASS-2003 1 1 2 2 Thomas Foken , Matthias Mauder , Mathias Göckede , Claudia Liebethal , Frank Beyrich , Jens-Peter Leps 1 University of Bayreuth, Department of Micrometeorology, Germany 2 German Meteorological Service, Meteorological Observatory Lindenberg, Germany Furthermore, all sites where investigated for their footprint characteristics and the existence of 1. INTRODUCTION 4. EDDY COVARIANCE DATA ANALYSIS internal boundary layers. Measurements were excluded if the sensor was not located within the During the experiment LITFASS-2003, which took place in a 20 x 20 km 2 area near the The eddy covariance data of all LITFASS-2003 sites were post-processed in a uniform way. new equilibrium layer of an internal boundary layer δ after a sudden change of the surface Meteorological Observatory Lindenberg, Germany, turbulent fluxes of momentum, sensible, Therefore, we used the comprehensive turbulence software package TK 2, which was developed characteristics. To determine the land use composition within the source area of each and latent heat were measured at nine agricultural sites, two grassland sites, one forest site, at the University of Bayreuth. It includes quality tests of the raw data and necessary corrections measurement position, the three dimensional Lagrangian stochastic trajectory model of two lake sites, and at two levels of a 100 m tower. For more information, see presentations of the covariances, as well as quality tests for the resulting turbulent fluxes (see Figure 1). Langevin type (Thomson, 1987) was used. The parameterization of the flow statistics and the 9. 1 by Beyrich et al. as well as 9. 2 through 9. 5, all from this conference. effect of stability on the profiles were in line with those used in (Rannik et al. , 2000; Rannik et High frequency data [10 – 20 Hz] For this presentation, only the six stations operated by the University of Bayreuth (UBT) and al. , 2003; Göckede, 2004). The principle aim of the footprint study was to determine the flux the German Meteorological Service, Meteorological Observatory Lindenberg (MOL), were contribution of the target land use area for different sets of wind direction and stability classes in Calculation of averages, variances and covariances for 30 min intervals excluding physically included in the study, because both groups were responsible for all data processing and order to check whether the measurements are representative for the specified land use type not possible values and spikes (Vickers and Mahrt, 1997) quality control issues. Other stations were operated by the GKSS Research Centre under different conditions (see Table 3). Geesthacht, the Royal Meteorological Institute of the Netherlands, the University of Hamburg, • coordinate rotation after Planar Fit method (Wilczak et al. , 2001) and the University of Technology Dresden. An overview of the micrometeorological 5. SOIL HEAT FLUX • correction of oxygen cross sensitivity for Krypton hygrometers (Tanner, 1993) measurements during the LITFASS-2003 experiment is given on the poster next to this one. Regarding the soil heat flux, we tested different approaches to determine this component of the • correction of spectral loss due to path length averaging, spatial separation of sensors and frequency energy balance from in-situ measurements such as soil temperature, soil moisture, and heat dynamic effect of signals (Moore, 1986) 2. EXPERIMENTAL SETUP flux plate records. We concentrated on two methods: first, a combination of the gradient • conversion of sonic temperature fluctuations into fluctuations of actual temperature for calculation of At each site, all components of the energy balance were measured. For this poster approach and calorimetry and second, a combination of heat flux plate measurements and the sensible heat flux (Schotanus et al. , 1983; Liu et al. , 2001) presentation six micrometeorological sites of LITFASS-2003 operated by MOL and UBT were calorimetry. For each approach, we tested several reference depths (depths where the • correction for density fluctuations to determine fluxes of scalar quantities H 2 O und CO 2 (Webb et al. , selected: cereal(A 5), maize (A 6), grassland (NV 2, NV 4), pine forest (HV, canopy height 14 m), temperature gradient or the heat flux plate measurement is made). To get an idea of the 1980; Liebethal and Foken, 2003) and a lake (FS). correctness of the results, we conducted a sensitivity analysis, testing the sensitivity of the two Iteration of the corrections until error < 0, 01% approaches to different reference depths and variations in the input data file. From these analyses, we draw the following conclusions: 3. INTERCOMPARISON PRE-EXPERIMENTS Post-field quality control (Foken et al. , 2004) • Test for steady state conditions (Foken und Wichura, 1996) - It is safer to use the gradient approach rather than heat flux plates The intercomparison of the sensors had already started one year before the experiment. All • Test for integral turbulence characteristics (Foken und Wichura, 1996) radiation sensors and the anemometers of the participating institutes were compared, most of - The deeper the reference depth, the better. them during a pre-experiment in May/June 2002 at the Boundary-layer field site Falkenberg of - Measurements in shallow depths should receive the most attention and effort. Corrected and quality assured results of turbulent fluxes the German Meteorological Service. - It is critical to measure temperatures correctly. Figure 1: Flow chart of the comprehensive software package developed we at the University of Bayreuth (http: //www. bayceer. uni. Of specific interest was the intercomparison of the gas analyzers done in the field and in the bayreuth. de). It performs all post processing of turbulence measurements and produces quality assured turbulent fluxes. For further investigations, we decided to use the soil heat flux calculated from the laboratory. The results of the comparison between the variance of the absolute humidity and The application of the quality control procedure, after Foken and Wichura (1996), on the data gradient/calorimetry approach with a reference depth of 0. 20 m, because this approach exposes the latent heat flux, measured at about 3. 25 meters above grasslands, are given in Tables 1 from the micrometeorological stations during the LITFASS-2003 campaign allows us to asses the minimal sensitivity to most of the input parameters. and 2. quality of turbulent fluxes. Figure 2 shows the proportion of half hour values of latent heat flux between 06: 00 and 20: 00 UTC, which were classified as the highest quality, indicating data 6. CONCLUSIONS Table 1: Comparison of the variance of the humidity Table 2: Comparison of the latent heat flux during the which can be used for fundamental research according to Foken et al. (2004). fluctuations during the comparison experiment 2002, reference UBT#1, CSAT 3 / LI The processing of the eddy covariance data of all these sites with one software tool, including 100% 19. 5. 20. 5. 21. 5. 22. 5. 23. 5. 24. 5. 25. 5. 26. 5. 27. 5. 28. 5. 29. 5. 30. 5. 31. 5. 1. 6. 2. 6. 3. 6. 4. 6. 5. 6. 6. 6. 7. 6. 8. 6. 9. 6. 10. 6. 11. 6. 12. 6. 13. 6. 14. 6. 15. 6. 16. 6. 17. 6. reference UBT#1, LI 7500 90 - 99% trans-formations and corrections like planar-fit rotation, Schotanus-correction, Moore-correction, A 5 80 - 89% A 6 and WPL-correction, was very beneficial in comparing all data produced by different groups. 70 - 79% N 2 Furthermore, all fluxes were quality checked on the basis of the tools pro-posed by Foken and N 4 60 - 69% Wichura (1996) and flagged as high quality data, moderate quality data, and low quality data, 50 - 59% HV combined with the height of possible internal boundary layers and the footprint sector. FS < 50% Using the calculated ‘surface’ soil heat flux, we also get a much better closure of the energy Figure 2: Availability of high quality latent heat flux data in % between 0600 and 2000 UTC. Micrometeorological sites LITFASS-2003 A 5 A 6, N 2, N 4, HV, FS in Brandenburg/Germany. Measurement period from May 19 th to June 17 th 2003 balance than using pure plates measurements. Considering these findings for flux calculations, corrections, and data quality, it was possible to Table 3: Fetch x, height of the new equilibrium layer δ, and percentage of the flux from the target land use area dependent on the wind direction and stability for the micrometeorological site LITFASS-2003 A 6 (corn field) determine composite time series for different surface types to provide validation data for models and to determine area-averaged fluxes. sector in° 30° 60° 90° 120° 150° 180° 210° 240° 270° 300° 330° 360° All Krypton (KH-20) and infrared (LI 7500) hygrometers used at the different x in m 29 41 125 360 265 203 211 159 122 81 36 28 micrometeorological stations during LITFASS-2003 were calibrated both before and after the δ in m 1. 6 1. 9 3. 4 5. 7 4. 9 4. 3 4. 4 3. 8 3. 3 2. 7 1. 8 1. 6 7. ACKNOWLEDGMENTS flux contribution from the target land use area in % field campaign in the laboratory using a dew point generator (LI-610, Li. Cor Inc. ) to generate a The project is funded by the Federal Ministry of Education, Science, Research and Technology stable 36 49 81 99 96 92 93 88 81 70 44 35 sequence of five pre-defined dew point values. The difference in the slope of the calibration neutral 51 63 90 100 98 98 95 90 82 59 50 (DEKLIM, project EVA-GRIPS). line between the two calibrations was typically less than 2 %. unstable 62 74 98 100 100 100 98 91 70 61 * Corresponding author's address: Thomas Foken, University of Bayreuth, Dept. of Micrometeorology, D-95440 Bayreuth; email: thomas. foken@uni-bayreuth. de