Calibration of Near-Infrared Spectra for Phosphorus Fractions in Grassland Soils on the Tibetan Plateau
Round 1
Reviewer 1 Report
As a General comment this is more a manuscript on agronomic results that it is about NIR. I’m not entitle to comment on the agronomic part, but if this paper focus on NIR it must be profoundly revised
I strongly suggest the Author to read
Williams P, Dardenne P, Flinn P. Tutorial: Items to be included in a report on a near infrared spectroscopy project. Journal of Near Infrared Spectroscopy. 2017;25(2):85-90. doi:10.1177/0967033517702395
But also
TI - Critical review of chemometric indicators commonly used for assessing the quality of the prediction of soil attributes by NIR spectroscopy, JF - Trends in Analytical Chemistry, Bellon-Maurel, Véronique; Fernandez-Ahumada, Elvira; Palagos, Bernard; Roger, Jean-Michel; McBratney, Alex. PY - 2010
Particularly the first one, indicates what is necessary for an NIR paper
Comments
2.1 Soil samples
192 (-1 missing) means 6 Treatments x 4 depth x 4 repeats x 2 duplicates. This is highly redundant samples as treatments are only 6 x 4 depth and the other are basically replicates. Keeping duplicates for the same treatment x depth x repeats is not acceptable, as it doesn’t provide any further information, but it improved accuracy ONLY if they are averaged. Please explain why you want to keep duplicates.
The analytical methods MUST (see Williams) indicate reference analytical error also known as Standard Error of Laboratory (SEL)
2.3 Near-Infrared spectroscopy
Ln151: 11500-to 3800nm? Nanometer? Must be wrong
Ln160-162: dividing in 80% calibration and 20% validation a higly redundant data set DOESN’T make and independent validation. In your case you can block your data by repeat (you have 4 of them) and take one entire repeat out for validation. Validation and calibration blocking is repeated unecessarly at Ln191-194
Ln176-185. I don’t understand the authors just set their own RPD limits at their own convenience. Must read Bellon-Maurel paper and their discussion about RPD.
2.4 Data Analysis
Beside mentioning the software there is nothing about data analysis , they mention PCA, while infact PLS should be the most appropriate, no information about math pretreatments, number of latent variables used, how outliers were treated. This must be rewritten.
It must have descriptive analysis of each P fraction as total, not by depth as in supplemental material, also must report cross correlation among P fractions
3.1 Hedley P fraction
If this is a paper about NIR the entire section and relative table and figure must be removed, doesn’t bring any information understanding the performance of NIR
3.2 NIRS Models
This is also poor. The only table about NIR performance shows RSQ and RPD only which are the expression of the same thing (see Williams), no SECV, SEP, Bias, slope.
4.1 Influence……
Remove agronomic discussion not relevant for NIR
4.2 NIRS model
It seems that Authors desperately try to find positive results on their calibrations, and yet they don’t show any prediction error (SEP) so we don’t know if such errors are acceptable from an agronomic point of view. I think they have to try much harder with better convincing data before stating “promising “ results. As the Authors stated the bibliography is against their final statement. I just want to point out as they want to quantify by NIR an inorganic element at a very low concentration and in this study mostly below what is normally regarded as lower limit of detection of 0.1%
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
In their manuscript “Calibration of near-infrared spectra (NIRS) for phosphorus fractions in grassland soils on the Tibetan Plateau” the authors present a study with various objectives (1) characterization of fertilizer effects on grassland soils of the Tibetan plateau on inorganic and organic soil P fractions (2) application of the Hedley sequential extraction to the grassland soils and (3) the calibration of NIRS to predict soil P fractions. The focus of the work lies, as the title also suggests, on the calibration of the NIRS model.
Overall, the manuscript is already very well-written and well-balanced, contains recent literature in the extensive introduction and discussion section.
The conclusions they reached based on their finding from model calibration that NIRS might be a promising tool to predict stable P pools in combination with the Hedley fractionation for labile P pool characterization for soils in alpine grasslands on the Tibetan Plateau appears plausible.
Although the role/effect of some of the treatments, especially and grazing – non-grazing for the P fractions has remained a bit obscure to me, I acknowledge that the primary focus lies on the NIRS calibration.
All figures and tables were prepared with care and are of good quality. This is equally true for the literature review and figures and tables in supporting information. So, atypically, I do not have a lot of criticism to offer.
I am sure that due to the diligent work of the authors in preparing this manuscript with nicely written and data-wise sound presentation of the NIRS model fitting this manuscript will find its readership and might be used for future reference in other studies. I therefore recommend minor revisions.
Point-by-Point comments
L46 and iron oxides?
L86 Although widely used, it is disputable of that the Hedley method provides information on P bioavailability in soil (after a quck literature search I found a study that stated that it is not suitable for mechanistic studies – ecosystem research and operational fractions might be fine. I would advise to include literature that studied the underlying mechanisms of the Hedley fractionation for a well-balanced and informed introduction. – ok, I noted you included some of this in the discussion L314ff.
L104-107 “In this study we showed…” move to conclusion section or abstract and focus on objectives, hypothesis and work to be presented in the following: present tense.
L122 “three” equal parts? Not “six” for all repeats?
L125 So there were 3 x 6m sub-plots for grazing? Which animal was grazing? So you actually have 12 “treatments”? With and without grazing? Also, which crops were grown and thereby amounts P removed with harvesting
L127 So you took 4 samples per each 3x6 m plot, right?
L172 ff. defined all variables of the equation (e.g. y and Differ is missing)
L360 What is the “fission model”? Please explain somewhere its meaning in the manuscript above
L361 How is one fraction more easily irradiated than another – must be by physical protection- pls explain?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf