An evaluation of image analysis for measuring changes in void space and excremental features on soil thin sections in an upland grassland soil

This paper evaluates the extent to which image analysis can be applied to determining faunal impacts on soils. It is based on experience gained from a 2-year experiment using an upland grassland soil to investigate impacts on soil micromorphology. Eighty undisturbed samples were collected in June of 1999, 2000 and 2001 from limed and control plots on the Macaulay Institute's Sourhope Research Station in the Scottish Borders. Soil thin sections were prepared from these samples and composite images (10.7 x 6.4 mm) were acquired from each component horizon. Protocols for image analysis were developed to segment void space and excremental features. For the latter, areas complementary to void space above two size thresholds were defined. Areas of mineral and plant fragments were subtracted from these and the resultant areas, after further image processing, corresponded to excrement attributable to enchytraeids and earthworms. Quantification by point counting on a 2-mm, grid was used to validate the image analysis protocols. In the H horizons of the control plots, the total void space increased between the start and the end of the experiment with the bulk of this increase in large voids (>1000 mum). More marked increases were evident for the H horizons in void space for the limed plots. Correlation was also found between excremental features as measured by image analysis and point counting (r(2) value of 0.44 for earthworm-derived features). Though the image analysis protocols for segmenting excremental features were validated, variability in the results precluded identification of change. This is explained in terms of the small representative elementary areas which were the basis for image analysis; larger areas are required when features such as earthworm excrements are present. Overall the advantages of using image analysis are highlighted, especially with regard to speed of data extraction and the ability to analyse void space using different void size thresholds. However, it is essential for image analysis results to be validated by conventional micromorphology and point counting; furthermore, careful consideration needs to be given to selecting the size of representative elementary areas. (C) 2003 Elsevier B.V. All rights reserved.