Uh, it seems you have a problem, possibily a big one. First of all: is it really ONE farmer? Then I'd say forget it. You will not be able to unearth anything that is publishable. Second, you do not seem to have any replication at the local level (only one field each in plan, hill, etc.). Third, you have too many treatments (all non-pasture fields are different at the different locations, so you would not be able to separate crop effect from location effect). Thus you have no real variation at all, and sorry to be this brutal, but  this project is not viable.

Hello Gàbor, first of all thanks for beeing brutal, I do need brutal and honest opinions. I'll try to clarify the situation in better way.

1) yes, it is one farm (total of 6000 sheep). This farmer uses different locations depending on the season. For example now he has flocks in the plain area. Then he will move on the hills and eventually in the mountain area (summer).

2) yes, I dont have any field replication but the filed I have are very representative of the local situation. I've seen a published paper  where they "focused on characterizing the major controls on soil respiration within representative sites rather than comparing CO2 efflux rates among different land uses ... we chose to intensively study one site in each land use type rather than attempting to replicate sites across the ecosystem" (Almagro et al., 2009). Now the problem is that they had 30 pseudo-replication per field, I have 3. 

3) I would like to have separted system analysys: Plain vs Hills vs Mountain. Therefore I have 2/3 treatments per area. 

4) last but not least, the thing I'm gonna compare (sorry I didn't specify it before) the transumant effect on the system. I don't care about the actual crops, I care of which level co2 emission I have after 10 years of "normal cropping system" compared with the 10 years transumant presence (alfa alfa grazed fields).

I hope i've explained myspef in a better way.

Please, feel free to be brutal again :-)

Article Temperature dependence of soil CO2 efflux is strongly modula...

Ciao Matteo

It seems I had understood the system correctly. Still, if it is only one farmer, it is nearly impossible to reach results that are representative. You mention several flocks (6000 heads of sheep are many, and they rarely go together, I believe) - maybe you can make those (or areas that these different flocks use) as your units of treatment/replication?

Transhumance (which is seasonally different grazing) areas should then be rather large. Is there no way you can subdivide these e.g. by flocks? The no. of other 8non-grazed) fields can be sampled in a representative way. Here I suggest that ou measure rates in as many fields as you can - so that you do not pseudoreplicate this "treatment" at least (which is the non-grazed, "modern" (?) cropping system.

There is also space & time variation - so that when grazing happens in one area, it is a different season and fo course, a different place. So I would go and measure CO2 efflux at various times during the yea. What limits your sampling locations to 3 only? Can you not increase that, even at the expense of measuring less frequently?

Well, indeed the farmer owns 6000 sheep which means that he has 10/12 flocks composed by 4/5/600 sheep each. I'm gonna follow one of them.

The major constrain is time. Litterature suggests to get data in a time span of 4 h. Each sampling lasts 120 sec. Therefore I cannot increase them too much.

Consider also that I have to insert PVC collars in the groud and convince the land owner to give me permission it was not the easiest thing in the world. 

Concerning the time of course iI will consider space and time variation but I remark that my intention its not to investigate what happens now, but what 10 years of grazing vs. classic land management produced in terms of CO2 emission. 

Thank you again for the advices 

Matteo, you will not gain too much by this setup. Now you write me that you are going to follow only one of the flocks - it seems that this project is beyond your possibilities. I understand that you are keen to find the effect of grazing out but while it may be of interest, for lack of any statistical replication, you will not be able to publish the results in a scientific journal. It might be a good piece of narrative, for a conference, or for a book, - but not for a scientific journal.

I understand the constraints, and i have sympathy for you for the problems - but these things are no excuse for not following scientific practice. If I understand your intent, you will use these measurements to calculate what the CO2 emission has been over a 10year period.  In order to do that, you need reliable measurements now - and with this setup, you cannot achieve that. The estimation of the amount over 10 years brigns its own, new set of uncertainties - but the basis for the 10y effect is the reliability of CO2 emission now - under the different grazing regimes. In order to measure that, you need replication. And you do not have it: one flock of one farmer in one field/pasture? The measurement precisin on that one field gives you no confidence that the same will happen at any other situation. You do need replication, real replication.

I understand. So your advice is to find at least 4/5 grazed alfa-alfa fields 4/5 wheat fileds, 4/5 sunflower fileds (for one area). Therefore I will have real replication. I can try to manage this. Again, thank you very much.

Yes, Matteo, precisely. If you need to, you may reduce the within-field replication. For example, if you have capacity to do 60 measurements in total, it is much better to do 4 repeat measurements/field, and increase the no. of fields measured (like you write: 5 each of pasture, wheat & sunflower) than 20 measurements per field, but only one field each of pasture, wheat and sunflower. This will make your life more difficult, but I am afraid there is no other way to get resutls in which you can be confident.