Respirometric tests on a soil contaminated by crude oil were performed. Continuous measurements of oxygen and carbon dioxide concentrations and temperature in the soil atmosphere resulted in a large volume of data. Time series and system identification theories were used to analyze data as a biological signal, allowing us to detect some particularities related to daily cycles of the studied variables as well as its time relationships through autocorrelation and cross-correlation functions. Using system identification techniques, it was possible to build black box models, namely autoregressive moving average models which enable to predict oxygen concentration at the outlet in a good agreement with measured data.