It is not clear if you have multiple sites of data collection,  But assuming that you DO have multiple sites, I then assume that you are interested in fitting a growth model to the data and looking for predictors of the trajectories .  .  and this can be done in several packages inducing HLM; SAS Proc Mixed and SPSS linear mixed etc. 

I think that Days  is a fixed effect as you are NOT randomly sampling s set of days from a populations of days.  But IF you are collecting data orm a set of sites then the sties might be random effects.

Thanks William,

To make clear,  my experiment was a lab experiment. So no multiple site effect. I have taken the gas samples in specific Day, hence as your suggestion, that could be a fixed effect. My experiment designed was quite complex. After taking  sample in 1, 3 and 7 days from an incubated ring-soil,  I destruct the sample for mineral N analysis.For days 10 and 14 I took different sets of ring-soil and  destruct again for mineral N . Same process continued for 1 month. Now, whether I should use a multiple linear regression approach or mix effect modelling for the analyses or repeated measurement ? Could you suggest me.

Data are dependent in time in terms of the process. Without know the specific questions and goals, and the data structure, I suggest you might to look at least three types of approach. The linear mixed model approach (e.g. Pinheiro & Bates) have several forms to fight with the correlated structure based on time; It appears as most soundy for your inquiry. But also the time series approach and the survival approach may provide some conceptual insight to decision-making, and both experimental and statistical design.

Mariano, thank you very much  for the suggestion.