Hi,
I wish to know the ways to compute cospectra.
I found one of resource stated that the cospectra were calculated for each 30-minute data set using smoothed and spliced fast Fourier transforms (Kaimal & Finnigan 1994) in EdiRe.
Could someone explain to me the ways to compute the cospectral?
Below is my understanding for the cospectral computed by EdiRe,
I have 48 half hourly CO2 data sets measured at 10 Hz.
Split each half hourly data set into two sets i.e. high frequency range and low frequency range.
High frequency range: a set of r short records each with N/r data points.
Low frequency range: a series made up of N/r nonoverlapping block averages (r points averages) of the original time series.
My question:
How to determine suitable r?
How to merge the resulting two spectral curves into a single spectral curve?
Please correct me if Im wong for the procedure below,
Let say for half hourly CO2 flux:
High frequecncy range:
Split N=18000 data points into 35 short records each with 512 data points + last record (80 data points + 432 zero padding).
Compute cospectra
Short record 1:
512 cospectra
Short record 2:
512 cospectra
.
.
.
.
Short record 36:
80 cospectra
Tahe the composite average form 36 records, thus it produce 512 cospectra.
Low frequency range:
Take the bolck average of 36 block. Thus, we 36 points average.
Using the method above the cospectra were computed for 48 half hourly CO2.
The cospectra plot contained 48 data points which generated by averaging the cospectra for each half hour.
Thank you....