« Previous | Contents | Next »
Listen
Appendix 1 Additional information for Module 1: Regression equation to predict dry bulk density of organic surface horizons in Scotland.
Organic carbon values are held within the Scottish soils database as a percentage of dry weight, however, in order to determine the overall carbon store, it is necessary to convert this value to a volumetric basis. This is achieved by multiplying the gravimetric measurement by the dry bulk density. However, there is insufficient data on the dry bulk density of Scottish soils in order to give meaningful estimates of mean values or any other 'average' values for individual soil types. Thus, regression equation-based pedotransfer functions were developed to predict bulk density for soil horizons grouped according to their mode of deposition and subsequent pedogenesis. In order to develop these equations all data on organic soil horizons (those where the organic carbon content exceeds 18%) were grouped.
There are dry bulk density measurements for 39 soil horizons. The soil horizons come from 15 profiles that were sampled for various research projects and 3 profiles sampled specifically to derive dry bulk density for organic horizons. The profiles were located in Galloway, Loch Bradan catchment, Mourne Mountains (Northern Ireland), Mharcaidh catchment and Shetland.
The organic carbon content of each horizon was determined by C and N elemental analyser and bulk densities were determined by oven drying a known volume of peat for 48 hrs at a temperature of 105 o C and expressing the results as grams dry weight per cubic centimetre.
A number of regression equations were tested: simple linear regression between organic carbon content (org_C) and dry bulk density (Db), between log transformed organic carbon (logn_org_C) and dry bulk density and a multiple linear regression with organic carbon content and depth as independent variables.
Results: Initial simple linear regression between dry bulk density and organic carbon content explained 63.5% of the variation in dry bulk density.
Db= 0.6654 - 0.00999*org_C (1)
A log normal transformation of the organic carbon value improved this simple linear relationship by explaining 70.4% of the variation.
Db= 1.772 - 0.4127*logn org_C (2)
As the degree of peat decomposition generally increases with depth, the uppermost depth that the sample was taken from was used as a predictor along with org_C and logn_org_C in a multiple linear regression. There was no improvement in the percentage variation explained (63.4% and 70.1% respectively).
Plots of dry bulk density against depth showed a sharp decline up to approximately 15 cm depth but then leveled out until depths of around 100cm where there was a slight increase in dry bulk density. The inclusion of depth as a predictor means that this slight increase can be modeled though this has to be balanced by the decline in the percentage variance explained. As this increase in bulk density with depth is only evident in a few samples, it is perhaps best that further data be collected to substantiate this before applying such a model to the data in this project.
There is a little to be gained by using the log transformed organic carbon contents as a predictor of dry bulk density though this should be matched against the additional effort involved in converting all organic carbon values and the small increase in percentage variance explained. It is more than likely that the bulk density values of fibrous 'litter' layers will be over predicted using these transfer functions.
The recommended equation to predict dry bulk density is equation 2. However, it would be prudent to test this equation against an independent dataset when available. This has significant implications for soil carbon stock estimates for Scotland.
« Previous | Contents | Next »