Determination of an averaged human olfactory response as odour concentration to odours by instrumental means has been sought to simplify on-site measurement and reduce cost. Different concentrations of hydrogen sulphide, 4-methyl phenol, ammonia and acetic acid were used to simulate livestock waste odour. A second-order uniform precision rotatable central composite statistical design identified 31 odour mixtures to establish a human olfactory response model. An odour panel containing eight people was selected according to recognised criteria. They had an average age of 43.3 years. A multiple linear regression model accounted for 74.1% of the variance, with a decreasing order of influence of hydrogen sulphide > 4-methyl phenol>>acetic acid within the bounds of the model described. Surprisingly, 4-methyl phenol concentration produced a negative effect on the olfactory response. To enable better fitting of the complex olfactory response radial basis function, neural network modelling was used to account for 88% of the variance. With pig manure odours outside the studied concentration range the network model did not function. Predictions using the linear regression model were on average 20% less than observed values. These findings demonstrate that this approach is appropriate to determine the odour concentration from pig manure using the main odorants.
|Number of pages||6|
|Journal||Journal of the Science of Food and Agriculture|
|Publication status||Published - 15 Jan 2001|
- odour mixtures