Relationships between CO2 evolution, moisture content and temperature for a range of soil types

Soils were collected from two locations at six sites on each of eight soil types. After sieving, soil CO2 evolution was measured at three moisture contents and at 5, 10, 15 and 20 degrees C. Our objectives were to (i) establish a regression model for the relationships between CO2 evolution, temperature and soil moisture content, and (ii) to investigate which aspects of this regression apply universally and which are specific to individual soil types. The same soil samples were used at the different temperatures, giving CO2 data which were repeated or longitudinal measurements. This experimental structure has implications for the covariance structure of the data. A preliminary analysis of the data showed that a natural logarithmic transformation of CO2 evolution rates was effective both in linearizing the mean response and in stabilizing its variance. When CO2 evolution was expressed as ln(ml CO2 kg(-1) loss-on-ignition h(-1)) and the soil moisture content as a proportion of the water-holding capacity, the shapes of the curves of CO2 evolution against moisture content and the positions of their maxima became more similar than when CO2 evolution and soil moisture content were expressed on an oven-dry weight basis. The average responses of ln(ml CO2 kg(-1) LOI h(-1)) of the soil types to temperature over the range 5-20 degrees C were remarkably close to a set of parallel straight lines. This resulted in the formulation of a model in which: (a) the mean value of ln(ml CO2 kg(-1) LOI h(-1)) at each site within a given soil type can be described by a regression which is linear on temperature and quadratic on soil moisture content; (b) the mean value of the linear coefficient of temperature is common to all the soil types studied; and (c) the mean values of the intercept, and of both the linear and quadratic coefficients of moisture, vary stochastically between soil types. A classification of soil types on the basis of their mean regression coefficients is: (i) brown calcareous earths, typical brown earths, (ii) humic alluvial gley soils, (iii) typical brown podzolic soils, podzols, stagnohumic gley soils, stagnopodzols, (iv) raw peat soils.