Accuracy and bias of methods used for root length measurements in functional root research

Research output: Journal contributionsJournal articlesResearchpeer-review

Standard

Accuracy and bias of methods used for root length measurements in functional root research. / Delory, Benjamin; Weidlich, Emanuela W. A. ; Meder, Leonie et al.
In: Methods in Ecology and Evolution, Vol. 8, No. 11, 11.2017, p. 1594-1606.

Research output: Journal contributionsJournal articlesResearchpeer-review

Harvard

APA

Vancouver

Bibtex

@article{d5b2ce44749f45ef89551be36ad4c7fb,
title = "Accuracy and bias of methods used for root length measurements in functional root research",
abstract = "Functional root traits are becoming a key measure in plant ecology, and root length measurements are needed for the calculation of root traits. Several methods are used to estimate the total root length (TRL) of washed root samples [e.g. modified line intersect (LI) method, WinRHIZOTM and IJ_Rhizo], but no standardized comparison of methods exists. We used a set of digital images of unstained root samples to compare measurements given by the LI method and automated methods provided by WinRHIZOTM and IJ_Rhizo. Linear regression models were used to detect bias. Both linear regression models and the Bland-Altmans` method of differences were used to evaluate the accuracy of eight methods (1 manual, 2 semi-automated and 5 automated) in comparison with a reference method that avoided root detection errors. Length measurements were highly correlated, but did not exactly agree with each other in 11 of 12 method comparisons. All tested methods tended to underestimate the TRL of unstained root samples. The accuracy of WinRHIZOTM was influenced by the thresholding method and the root length density (RLD) in the pictures. For the other methods, no linear relationship was found between accuracy and RLD. With WinRHIZOTM (global thresholding + pixel reclassification; RLD = 1 cm cm-2), the Regent's method and the Tennant's method underestimated the TRL by 7·0 ± 6·2% and 4·7 ± 7·9%, respectively. The LI method gave satisfactory results on average (underestimation: 4·2 ± 6·0%), but our results suggest that it can lead to inaccurate estimations for single images. In IJ_Rhizo, the Kimura method was the best and underestimated the TRL by 5·4 ± 6·1%. Our results showed that care must be taken when comparing measurements acquired with different methods because they can lead to different results. When acquiring root images, we advise to (i) increase the contrast between fine roots and background by staining the roots, and (ii) avoid overlapping roots by not exceeding a RLD of 1 cm cm-2. Under these conditions, good length estimates can be obtained with WinRHIZOTM (global thresholding + pixel reclassification). The Kimura method in IJ_Rhizo can be an alternative to WinRHIZOTM.",
keywords = "Ecosystems Research, IJ_Rhizo, Winrhizo TM, Functional ecology, ImageJ, Line intersect method, Root length, Washed root samples",
author = "Benjamin Delory and Weidlich, {Emanuela W. A.} and Leonie Meder and Anna L{\"u}tje and Duijnen, {Richard van} and Rafael Weidlich and Temperton, {Victoria Martine}",
year = "2017",
month = nov,
doi = "10.1111/2041-210X.12771",
language = "English",
volume = "8",
pages = "1594--1606",
journal = "Methods in Ecology and Evolution",
issn = "2041-210X",
publisher = "British Ecological Society",
number = "11",

}

RIS

TY - JOUR

T1 - Accuracy and bias of methods used for root length measurements in functional root research

AU - Delory, Benjamin

AU - Weidlich, Emanuela W. A.

AU - Meder, Leonie

AU - Lütje, Anna

AU - Duijnen, Richard van

AU - Weidlich, Rafael

AU - Temperton, Victoria Martine

PY - 2017/11

Y1 - 2017/11

N2 - Functional root traits are becoming a key measure in plant ecology, and root length measurements are needed for the calculation of root traits. Several methods are used to estimate the total root length (TRL) of washed root samples [e.g. modified line intersect (LI) method, WinRHIZOTM and IJ_Rhizo], but no standardized comparison of methods exists. We used a set of digital images of unstained root samples to compare measurements given by the LI method and automated methods provided by WinRHIZOTM and IJ_Rhizo. Linear regression models were used to detect bias. Both linear regression models and the Bland-Altmans` method of differences were used to evaluate the accuracy of eight methods (1 manual, 2 semi-automated and 5 automated) in comparison with a reference method that avoided root detection errors. Length measurements were highly correlated, but did not exactly agree with each other in 11 of 12 method comparisons. All tested methods tended to underestimate the TRL of unstained root samples. The accuracy of WinRHIZOTM was influenced by the thresholding method and the root length density (RLD) in the pictures. For the other methods, no linear relationship was found between accuracy and RLD. With WinRHIZOTM (global thresholding + pixel reclassification; RLD = 1 cm cm-2), the Regent's method and the Tennant's method underestimated the TRL by 7·0 ± 6·2% and 4·7 ± 7·9%, respectively. The LI method gave satisfactory results on average (underestimation: 4·2 ± 6·0%), but our results suggest that it can lead to inaccurate estimations for single images. In IJ_Rhizo, the Kimura method was the best and underestimated the TRL by 5·4 ± 6·1%. Our results showed that care must be taken when comparing measurements acquired with different methods because they can lead to different results. When acquiring root images, we advise to (i) increase the contrast between fine roots and background by staining the roots, and (ii) avoid overlapping roots by not exceeding a RLD of 1 cm cm-2. Under these conditions, good length estimates can be obtained with WinRHIZOTM (global thresholding + pixel reclassification). The Kimura method in IJ_Rhizo can be an alternative to WinRHIZOTM.

AB - Functional root traits are becoming a key measure in plant ecology, and root length measurements are needed for the calculation of root traits. Several methods are used to estimate the total root length (TRL) of washed root samples [e.g. modified line intersect (LI) method, WinRHIZOTM and IJ_Rhizo], but no standardized comparison of methods exists. We used a set of digital images of unstained root samples to compare measurements given by the LI method and automated methods provided by WinRHIZOTM and IJ_Rhizo. Linear regression models were used to detect bias. Both linear regression models and the Bland-Altmans` method of differences were used to evaluate the accuracy of eight methods (1 manual, 2 semi-automated and 5 automated) in comparison with a reference method that avoided root detection errors. Length measurements were highly correlated, but did not exactly agree with each other in 11 of 12 method comparisons. All tested methods tended to underestimate the TRL of unstained root samples. The accuracy of WinRHIZOTM was influenced by the thresholding method and the root length density (RLD) in the pictures. For the other methods, no linear relationship was found between accuracy and RLD. With WinRHIZOTM (global thresholding + pixel reclassification; RLD = 1 cm cm-2), the Regent's method and the Tennant's method underestimated the TRL by 7·0 ± 6·2% and 4·7 ± 7·9%, respectively. The LI method gave satisfactory results on average (underestimation: 4·2 ± 6·0%), but our results suggest that it can lead to inaccurate estimations for single images. In IJ_Rhizo, the Kimura method was the best and underestimated the TRL by 5·4 ± 6·1%. Our results showed that care must be taken when comparing measurements acquired with different methods because they can lead to different results. When acquiring root images, we advise to (i) increase the contrast between fine roots and background by staining the roots, and (ii) avoid overlapping roots by not exceeding a RLD of 1 cm cm-2. Under these conditions, good length estimates can be obtained with WinRHIZOTM (global thresholding + pixel reclassification). The Kimura method in IJ_Rhizo can be an alternative to WinRHIZOTM.

KW - Ecosystems Research

KW - IJ_Rhizo

KW - Winrhizo TM

KW - Functional ecology

KW - ImageJ

KW - Line intersect method

KW - Root length

KW - Washed root samples

UR - http://www.scopus.com/inward/record.url?scp=85018578504&partnerID=8YFLogxK

U2 - 10.1111/2041-210X.12771

DO - 10.1111/2041-210X.12771

M3 - Journal articles

VL - 8

SP - 1594

EP - 1606

JO - Methods in Ecology and Evolution

JF - Methods in Ecology and Evolution

SN - 2041-210X

IS - 11

ER -

DOI

Recently viewed

Publications

  1. Modeling of microstructural pattern formation in crystal plasticity
  2. Special Issue: A variational pragmatic approach to regional variation in language: Celebrating the work of Klaus P. Schneider
  3. Managing information in the case of opinion spamming
  4. Leading Knowledge Exploration and Exploitation in Schools
  5. Random year intercepts in mixed models help to assess uncertainties in insect population trends
  6. A Study on the Impact of Intradomain Finetuning of Deep Language Models for Legal Named Entity Recognition in Portuguese
  7. Categorizing urban tasks
  8. From theory to experience
  9. Managing Multiple Logics: The Role of Performance Measurement Systems in Social Enterprises
  10. Dynamische Mathematik
  11. Performance of the Chemcatcher ® passive sampler when used to monitor 10 polar and semi-polar pesticides in 16 Central European streams, and comparison with two other sampling methods
  12. Functions of Constitutions
  13. Comparison of Panel Cointegration Tests
  14. External State-Building and Why Norms Matter
  15. rSOESGOPE Method Applied to Four-Tank System Modeling
  16. Do Linguistic Features Influence Item Difficulty in Physics Assessments?
  17. History of Embryology: Visualizations Through Series and Animation
  18. Responsibility and environment
  19. Using the learner-generated drawing strategy
  20. Development and characterisation of a new interface for coupling capillary LC with collision-cell ICPMS and its application for phosphorylation profiling of tryptic protein digests
  21. Clashing Values
  22. Stabilisierung durch Implementieren, Sichern und Evaluieren
  23. Choice and quantity demand for improved and unimproved public water sources in rural areas
  24. Negotiation complexity
  25. Learning-related emotions in multimedia learning
  26. Nonautonomous control of stable and unstable manifolds in two-dimensional flows