Gabriela, I assume your mean sensible heat FLUX right? Simply stated, this flux is estimated in the following way: You determine an appropriate averaging interval, then take the time average of the vertical component of the wind (w) and of temperature (T), estimate the deviations from the mean for every data record in this interval (w' and T') and then calculated the covariance between w' and T' : . This is the sensible heat flux valid over the time interval. Repeat over multiple time intervals to get the temporal evolution of the sensible heat flux. This is the simple story. However a lot of considerations usually go toward getting the best possible estimates, that are not influenced by sampling errors. There is quite a body of literature on the subject. Don't go at it blindly. I would suggest reading publications by Larry Mahrt & others to educate yourself on the subject of estimating fluxes by eddy covariance measurements.

Gabriela, I assume your mean sensible heat FLUX right? Simply stated, this flux is estimated in the following way: You determine an appropriate averaging interval, then take the time average of the vertical component of the wind (w) and of temperature (T), estimate the deviations from the mean for every data record in this interval (w' and T') and then calculated the covariance between w' and T' : . This is the sensible heat flux valid over the time interval. Repeat over multiple time intervals to get the temporal evolution of the sensible heat flux. This is the simple story. However a lot of considerations usually go toward getting the best possible estimates, that are not influenced by sampling errors. There is quite a body of literature on the subject. Don't go at it blindly. I would suggest reading publications by Larry Mahrt & others to educate yourself on the subject of estimating fluxes by eddy covariance measurements.

Thanks Robert. I would like to run a model to estimate the footprint of the estimated flux (CO2, H2O), and one of the inputs is sensible heat.

I remain unclear about what exactly you need for your footprint calculations. Is it a heat flux or an average measure of temperature?

Dear Gabriela,

see my "picture"? That's the logo of EddyPro, a software intended exactly to calculate fluxes (including sensible heat and momentum fluxes), starting from sonic anemometer measurements, according to the Eddy Covariance method mentioned by others. EddyPro will apply all necessary data screening and corrections procedures (Vickers and Mahrt, 1997; Schotanus et al., 1981; Moncrieff et al., 1997; etc.). If you also have gas measurements collected at the same frequency, you also get CO2, CH4 and H2O (and latent heat) fluxes, again with all appropriate corrections.

On top of that, EddyPro will also give you an estimate of the flux footprint.

EddyPro, released by LI-COR Biosciences, is one of the several Eddy Covariance software available. To my knowledge, it is the only one that is free, open source, fully documented and fully supported. Its documentation can also serve as an introduction to the Eddy Covariance method, but you may rather want to consider the (free) book mentioned below.

Please visit:

www.licor.com/eddypro, to download the software;

www.licor.com/env/forum/forum.php?id=8, the forum where most common problems or doubts with the software are discussed;

envsupport.licor.com/help/EddyPro4/Default.htm, EddyPro's web help;

www.licor.com/env/products/eddy_covariance/ec_book.html, an introduction to Eddy Covariance, by G. Burba.

Hope this helps.

Thanks Robert and Fulvio,

I am trying to use a code that runs in Matlab. In the input variables it comes the latent heat, I think, the units are in W/m2. As we recorded wind in three directions (u,v,w) and CO2, and H2O density, I wondered how is that I can get the latent heat. Probably is not latent heat and is sensible heat. So, if that´s the case I could follow what Georg says.

I have used before Eddy Pro and I am not keen on using it , as we prefer to use a code.

Thanks all for your comments.

Gabriela, I understand the wish to use scripts. But please allow me to give a gentle warning: depending on how accurate you want the sensible heat flux (H), it can be as easy as calculating the covariance of w and sonic T (w'T'), and then multiply by air density (rho) and by specific heat at constant pressure (Cp):

H = rho * Cp * w'T' (1)

or it can get as complicated as any Eddy Covariance software developer will tell you. (I guess that's why they exist "Eddy Covariance software developers").

The correct calculation of H (W/m2) actually requires the knowledge of latent heat flux (LE, W/m2), that's why you get it as an input to the code. Calculating LE from anemometric and gas data is much more complicated than calculating H itself, and the way one calculates it depends strongly on whether the gas analyzer is an "open path" or a "closed path".

Everything considered - if you decide to implement everything from scratch (highly not recommended) - I believe you'll make a smaller error if you use Eq. (1), rather than trying to first calculate LE correctly and then correct H for LE.

I do believe you need an Eddy Covariance software. If EddyPro does not fit your needs, you may want to try "EddyUH", released by the University of Helsinki. It's a Matlab software, but I am not sure they share the source code. Try to google it.

Good luck!

Dear Gerardo,

Thanks very much for your advice I am already having a look into EddyUH. I will give it a try, I guess what are you saying is that developing a software to do Eddy Covariance calculations can be very complicated and in the end I do want to analyze data and not to spend tons of time developing things that have been already developed.

On the other hand though, I want to understand the details and the implications of the data analysis....Going back into the point of sensible heat calculation, if I have used an open path analyzer shall I have to do calculations in a different way?

I think it is wise to consider already existing software. At first glance, estimating a flux seems simple enough. However getting fluxes that are not plagued with important errors is more involved. The "devil is in the details" really.

Going back to your question: If you have high-frequency measurements of water vapor, you have access to latent heat flux (LE). Say your measurements are water vapor density (q in kg m-3), the the same general "recipe" as I outlined above applies, meaning that you average over a given interval, derive deviations from the mean (i.e. q') and calculate the covariance (w'q'). The latent heat flux is just L(w'q') (in W m-2), where L is the latent heat of vaporization.

Yes, that's exactly what I meant: if you're an average micro-meteorologist and programmer, you would spend a considerable amount of time for doing that code correctly. If you're not either, you may never do that correctly.. :)

I do understand the wish to "understand the details and implications", but as a matter of fact, it takes to be a micro-meteorologist *and* a programmer, to understand details and implications. I guess a good analogy is this: do you require to understand all the mathematical and implementation details of an FFT code before you use it to calculate a spectrum? Or do you assume that someone - probably more knowledgeable than us in the field - did the job in a way that we won't be able to evaluate critically (at least in this life), because we specialized in something else?

With Eddy Covariance it's quite the same: the topic is complex and rich enough to be practically a field of investigation in itself. So, there are experts who can do things good enough, and non-expert who are allowed to avoid lots of headaches by exploiting the work of the experts. Well, that's how I see it.

Back to the open path: this type of analyzer provides gas densities, which means that you'll have to deal with "air density fluctuations" when you calculate LE. This is done by adding the so called WPL term (from the names of the authors of a paper, Webb et al. 1980) to the fluxes calculated using the gas density (not concentration) and wind measurements.

You probably don't want to hear that H and LE must be calculated iteratively because to calculate H you need LE, but to calculate the WPL term for LE you need H...

Before that, and exactly in the "right" place, you'll have to introduce spectral corrections to avoid flux underestimations.

But before anything else, you'll first have to apply some quality screening procedure, and in particular a despiking procedure (e.g. Vickers and Mahrt 1997; Mauder et al. 2013), to eliminate implausible wind or concentration values that would mess up the covariances and hence the fluxes. So, you see.. a lot to do.

I am guessing EddyUH would guide you through all this, but if you need a reference, try with the links I sent you in the previous post, notably the web help of EddyPro and the EC book by G. Burba. Ciao!

My group has matlab code for computed eddy correlation fluxes from sonic anemometers combined with fast CO2/H2O sensors like the Licor 7200 or 7500. There are a number of issues with Webb (dilution) corrects, units, and cross talk.

Sonic anemometers have velocity-temperature cross talk that depends on the sonic geometry. Some vendor correct internally and some dont. Also, recall sonic temperature has a water vapor contribution (approximately but not exactly virtual temperature). There are also coordinate rotation methods, flow distortion, and spectral vs time-series processing.

Lots of fun.

Contact Ludovic Bariteau if you want some matlab code.

[email protected]

Cheers Fairall,

Thanks all of you guys for your great ideas and comments

I think the attached presentation will be useful to you. It explains how to compute "sensible heat flux" from sonic data and also some other necessary concepts of Eddy covariance system.

I noticed people discussing EddyUH software, which was developed in my group. I can recommend coming to our intensive training course, which, I believe, will be organized again in Jan 2014 in Helsinki. We teach the use EddyUH on realistic example data; the course gives a very solid understanding of the whole EC process starting from high-frequency measurements and up to the most advanced corrections.

Thanks for the presentation Piyush! Cheers

Hola Gabriela,

Una alternativa a tu pregunta sobre el postprocesamiento de flujos por covarianza turbulenta, es contactar algun experto cercano a tu lugar de trabajo para que te ayude. Entiendo que el CICESE (posiblemente este mal escrito el acronimo) tiene una torre de flujos en Baja California.

Seguramente algun colega de MexFlux podria ayudarte. Encontraras los nombres y afiliaciones de los que participamos en este consorcio en el siguiente articulo:

Vargas et al. (2013) Progress and opportunities for monitoring greenhouse gases fluxes in Mexican ecosystems: the MexFlux network. Atmosfera 26(3):325-336.

En caso de que no encuentres respuesta en Mexico, yo te puedo ayudar desde Singapur donde radico.

Saludos

The above discussion helped me a lot thanks to the contributors

Very useful discussion. While calculating Sensible heat flux from sonic anemometer care to be taken with wind speed ranges as per the some field studies.please refer this paper.

Atmos. Meas. Tech., 5, 2095–2111, 2012

Piyush you suggested a very good presentation, highly recommended!