Abstract
Abstract Root traits get increasing attention as functional equivalents of above‐ground traits. Image analysis software such as WinRhizo ™ and IJ_Rhizo facilitate root trait analyses. Delory et al. ( ) presented a comparison between the accuracy of WinRhizo ™ and IJ_Rhizo for measuring root length. We complement their analyses with a comparison of diameter and volume estimates and a comparison of different image resolutions with manual and automatic threshold. We analysed 100 images of fibrous and taproot systems, which were obtained using the root model ArchiSimple. As a result, for each image, diameter, length and volume were known. The images were analysed with WinRhizo ™ and IJ_Rhizo and we compared the estimates of diameter, length and volume to ground‐truth values. We further computed relative errors and their magnitude and analysed their dependency on image characteristics and root system properties. At 1,200 and 800 dpi, diameter and length estimates provided by WinRhizo ™ and IJ_Rhizo were of comparable accuracy. Diameter errors were balanced. Volume estimates were subjected to a systematic error caused by the assumption of constant diameter. WinRhizo ™ , however, provides the opportunity to calculate correctly computed volumes from diameter classes. At 1,200 dpi, IJ_Rhizo failed to automatically find an appropriate threshold for pixel classification, which fundamentally decreased accuracy. The magnitude of diameter errors increased with root overlap for IJ_Rhizo. The length errors increased with increasing root length, overlap and root length density for WinRhizo ™ . The magnitude of underestimation of the volume (WinRhizo ™ ) decreased with volume. It was higher for taproot than for fibrous root systems. All errors increased with lower resolution. Our results confirm the results of Delory et al. ( ) regarding the accuracy for length. They further confirm that estimates derived from different software packages or at different resolution should not be compared directly. The characteristics of root systems should be standardized for image analysis. The dependency of errors on the response variable of interest can influence the effect size and increase the probability of errors. Validation of methods should be conducted for each analysed dataset. New image analysis tools should be validated against a real ground‐truth.
Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 16
Claim · model-written summary
WinRhizo and IJ_Rhizo gave comparable accuracy for diameter and length estimates at 1,200 and 800 dpi.
Not reportedTaxon
Verbatim
e0“At 1,200 and 800 dpi, diameter and length estimates provided by WinRhizo ™ and IJ_Rhizo were of comparable accuracy”
Claim · model-written summary
Diameter errors were balanced when estimated against ground-truth values.
Not reportedTaxon
Verbatim
e0“Diameter errors were balanced”
e1“we compared the estimates of diameter, length and volume to ground‐truth values”
Claim · model-written summary
Volume estimates had a systematic error caused by the assumption of constant diameter.
- Outcome
- Volume estimatese0
- Qualifier
- systematic error caused by the assumption of constant diametere0
- Comparator
- ground‐truth valuese1
Not reportedTaxon
Verbatim
e0“Volume estimates were subjected to a systematic error caused by the assumption of constant diameter”
e1“we compared the estimates of diameter, length and volume to ground‐truth values”
Claim · model-written summary
WinRhizo provides the opportunity to calculate correctly computed volumes from diameter classes.
- Outcome
- correctly computed volumese0
Not reportedTaxon
Verbatim
e0“WinRhizo ™ , however, provides the opportunity to calculate correctly computed volumes from diameter classes”
Claim · model-written summary
At 1,200 dpi, IJ_Rhizo failed to find an appropriate threshold, which decreased accuracy.
Not reportedTaxon
Verbatim
e0“At 1,200 dpi, IJ_Rhizo failed to automatically find an appropriate threshold for pixel classification, which fundamentally decreased accuracy”
Claim · model-written summary
The magnitude of diameter errors increased with root overlap for IJ_Rhizo.
Not reportedTaxon
Verbatim
e0“The magnitude of diameter errors increased with root overlap for IJ_Rhizo”
Claim · model-written summary
Length errors of WinRhizo increased with root length, overlap and root length density.
- Treatment
- increasing root length, overlap and root length densitye0
- Outcome
- length errorse0
- Direction
- increasede0
Not reportedTaxon
Verbatim
e0“The length errors increased with increasing root length, overlap and root length density for WinRhizo ™”
Claim · model-written summary
The magnitude of WinRhizo volume underestimation decreased with volume.
Not reportedTaxon
Verbatim
e0“The magnitude of underestimation of the volume (WinRhizo ™) decreased with volume”
Claim · model-written summary
Volume underestimation was higher for taproot than for fibrous root systems.
Not reportedTaxon
Verbatim
e0“The magnitude of underestimation of the volume (WinRhizo ™) decreased with volume. It was higher for taproot than for fibrous root systems”
Claim · model-written summary
All errors increased with lower resolution.
Not reportedTaxon
Verbatim
e0“All errors increased with lower resolution”
Claim · model-written summary
The results confirm the earlier accuracy results of Delory et al. for length.
Not reportedTaxon
Verbatim
e0“Our results confirm the results of Delory et al. ( ) regarding the accuracy for length”
Claim · model-written summary
Estimates from different software packages or resolutions should not be compared directly.
- Outcome
- estimates derived from different software packages or at different resolutione0
Not reportedTaxon
Verbatim
e0“They further confirm that estimates derived from different software packages or at different resolution should not be compared directly”
Claim · model-written summary
Characteristics of root systems should be standardised for image analysis.
- Outcome
- characteristics of root systemse0
Not reportedTaxon
Verbatim
e0“The characteristics of root systems should be standardized for image analysis”
Claim · model-written summary
Methods should be validated for each analysed dataset.
- Outcome
- Validation of methodse0
Not reportedTaxon
Verbatim
e0“Validation of methods should be conducted for each analysed dataset”
Claim · model-written summary
New image analysis tools should be validated against a real ground-truth.
- Outcome
- New image analysis toolse0
Not reportedTaxon
Verbatim
e0“New image analysis tools should be validated against a real ground‐truth”
Claim · model-written summary
The dependency of errors on the response variable of interest can increase the probability of errors.
- Treatment
- dependency of errors on the response variable of intereste0
- Outcome
- probability of errorse0
- Direction
- increasee0
Not reportedTaxon
Verbatim
e0“The dependency of errors on the response variable of interest can influence the effect size and increase the probability of errors”