97
Thermal Radiation and Energy Closure Assessment
where EB c is the energy balance closure, and the terms have been previously defined.
For many EC measurements, typical closure ratios can range from 0.7 to 0.9, where
higher closure ratios indicate better measurements, whereas lower values suggest not
necessarily bad measurements but the conditions were such that fluxes measured
by the SEB were not well coupled to the boundary layer. There can be many other
reasons for low closure ratios. These range from frequency response issues with the
sonic and IRGA, that is, the instruments are missing important eddies that contain
substantial flux information, sensor separation issues, and misplacement of the SEB
with respect to the surface of interest, that is, placed in a location where a different
surface may be influencing the measurements. The location of the available energy
sensors in a highly variable field will adversely affect the closure ratio, as it may
not be representative of the surface. Equation 5.8 does provide a means of assessing
the repeatability and a quality assurance check of the measurements comprising the
critical components of the SEB.
5.5 DATA PROCESSING
This section provides a brief description of the spectral techniques used to study
turbulence, in particular, the turbulent exchange of heat and water vapor near the
surface. This is the most fundamental technique to begin to gain an understanding through spectral analyses of the fluctuations in the boundary-layer turbulence
exchange of heat and water in a semiarid landscape under advective events. Spectral
analysis is a useful method to assess the reliability of flux measurements (Kaimal
and Finnigan 1972).
5.5.1 calculation of SPectRa
Data were acquired using programmable Campbell Scientific Inc. CR5000 data
loggers. High- and low-frequency data were stored onto high-density compact flash
cards that were part of the CR5000 peripheral package. Compact flash cards were
exchanged weekly from each of the EC SEBSs and stored onto multiple computers at
the CPRL site. We used the MATLAB ® (MathWorks) and Mathematica (Wolfram)
software platforms for developing algorithms for data processing that included turbulent flux calculations and spectral analysis.
Power spectra and cospectra were computed using the complex fast Fourier transform (FFT). The FFT software from the MATLAB and Mathematica platforms do
not require that the number of data be a power of 2, so we were able to use the full
length of our data runs. At 20 Hz, runs of 60 min resulted in high-frequency data
records of 72,000 points for u, v, w, T, CO 2 , and ρ v . Before applying the FFT algorithms, the original 20-Hz data needed to be evaluated using common procedures in
micrometeorology. These included a despiking routine to locate suspect individual
data points that exceeded a threshold value represented by a unique multiple of the
standard deviation of each of the high-frequency parameters. When a parameter
exceeded a threshold limit, it was replaced with an averaged nearest neighbor (10 on
either side) value. Actual data spike replacement for this data set was insignificant.
Thermal Radiation and Energy Closure Assessment
where EB c is the energy balance closure, and the terms have been previously defined.
For many EC measurements, typical closure ratios can range from 0.7 to 0.9, where
higher closure ratios indicate better measurements, whereas lower values suggest not
necessarily bad measurements but the conditions were such that fluxes measured
by the SEB were not well coupled to the boundary layer. There can be many other
reasons for low closure ratios. These range from frequency response issues with the
sonic and IRGA, that is, the instruments are missing important eddies that contain
substantial flux information, sensor separation issues, and misplacement of the SEB
with respect to the surface of interest, that is, placed in a location where a different
surface may be influencing the measurements. The location of the available energy
sensors in a highly variable field will adversely affect the closure ratio, as it may
not be representative of the surface. Equation 5.8 does provide a means of assessing
the repeatability and a quality assurance check of the measurements comprising the
critical components of the SEB.
5.5 DATA PROCESSING
This section provides a brief description of the spectral techniques used to study
turbulence, in particular, the turbulent exchange of heat and water vapor near the
surface. This is the most fundamental technique to begin to gain an understanding through spectral analyses of the fluctuations in the boundary-layer turbulence
exchange of heat and water in a semiarid landscape under advective events. Spectral
analysis is a useful method to assess the reliability of flux measurements (Kaimal
and Finnigan 1972).
5.5.1 calculation of SPectRa
Data were acquired using programmable Campbell Scientific Inc. CR5000 data
loggers. High- and low-frequency data were stored onto high-density compact flash
cards that were part of the CR5000 peripheral package. Compact flash cards were
exchanged weekly from each of the EC SEBSs and stored onto multiple computers at
the CPRL site. We used the MATLAB ® (MathWorks) and Mathematica (Wolfram)
software platforms for developing algorithms for data processing that included turbulent flux calculations and spectral analysis.
Power spectra and cospectra were computed using the complex fast Fourier transform (FFT). The FFT software from the MATLAB and Mathematica platforms do
not require that the number of data be a power of 2, so we were able to use the full
length of our data runs. At 20 Hz, runs of 60 min resulted in high-frequency data
records of 72,000 points for u, v, w, T, CO 2 , and ρ v . Before applying the FFT algorithms, the original 20-Hz data needed to be evaluated using common procedures in
micrometeorology. These included a despiking routine to locate suspect individual
data points that exceeded a threshold value represented by a unique multiple of the
standard deviation of each of the high-frequency parameters. When a parameter
exceeded a threshold limit, it was replaced with an averaged nearest neighbor (10 on
either side) value. Actual data spike replacement for this data set was insignificant.