Estimating length, average diameter and surface area of roots using two different Image analyses systems

Image analyses systems provide a quick determination of various root morphological parameters. Generally, a specific testing procedure should be conducted at the beginning of every measurement process. In this study, the performance of two image analyses programs using different measuring algorithms was compared: a commercial package WinRHIZO and a freeware ROOTEDGE. Roots of field grown cereal crops, wheat (Triticum durum Desf.) and barley (Hordeum vulgare L.), were used. Several types of tests were executed: 1. Comparison of image analyses and manually conducted measurements of root length; 2. Comparison between root length, average diameter and surface area measurements performed with ROOTEDGE and WinRHIZO; 3. Tests of root arrangement to assess the importance of random orientation of the scanned roots for accurate measurements; 4. Evaluation of the maximum acceptable scanning density (cm scanned root length per cm(2) scanning area). The results suggest that ROOTEDGE and WinRHIZO provide fairly correct measurements of root morphological parameters. There were small differences between manually and image analyses measurements of root length, in particular using a transparent light unit for scanning. Ratios of image analyses to manual estimations ranged from 0.95 to 1.03 for different root samples of winter barley. Comparisons of the programs generated almost equal root estimates. Discrepancies between diameter and surface area were slightly higher than between length measurements. The average root diameter was a little overestimated by ROOTEDGE compared to WinRHIZO. The most significant source for these discrepancies presumably was the difference between the fixed threshold for ROOTEDGE and the flexible threshold, automatically optimized for every single image by WinRHIZO. ROOTEDGE and WinRHIZO image analyses showed small sensitivity to root sample orientation. Estimations of root length, average diameter and surface area were well reproducible. For the scanning density to 3 em cm(-2) CV values for the replicated measurements varied between 0.3% and 3.4% by both programs. High scanning density of roots resulted in increasing underestimation of root length and overestimation of root average diameter. For the common scanning density range in root research between 1 and 3 cm cm(-2) discrepancies did not exceed 5%. Higher scanning densities than 3 cm cm(-2) are not recommended.