Validation and Modification of the Van Genuchten Model for Eroded Black Soil in Northeastern China

Round 1

Reviewer 1 Report

Thank you for incorporating my suggestions. You may still want consider reducing the length of the manuscript since there are parts that discuss minor matters in great detail. But I will leave that up to you. 

You may also wish to consider what parts of the Soil Water curve is most important to your research. Is the wet/low suction more important than the dry/high suction? Either way, you may want to (perhaps in another article) consider statistically weighing your measured data so that the more important portion of the SWCC is fit more accurately.

You may also want to consider other pedotransfer models in your next study (see your reference 54: Estimation of the van Genuchten Soil Water Retention Properties from Soil Textural Data by Ghanbarian-Alavijeh, Liaghat, Huang and Van Genuchten).

I also would have liked to see a more thoughtful discussion of the erosion process/change in soil fabric-grain size/change in SWWC. You have these samples and analyzed them but I did not get the impression you were thinking about whether you were testing similar horizons, similar fabrics, or were trying to see where each of your samples fit in the erosion "time line" . 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper "Validation and Modification of the Van Genuchten Model for Eroded Black Soil in Northeastern China" is interesting for the topic addressed, but it presents some critical issues, which should be resolved to improve the qualitative and scientific level of the manuscript.
In particular, the structuring of the work starts from the classification of soils based on the level of incidence of erosion (table 1), but there is no introduction that explains what are the causes of soil erosion and what they are determined.
It would also be important to document with photographs the different types of black soil eroded in the 15 sites studied.
Finally, in the conclusions it would be appropriate to include considerations on the practical applications of the research results and comparison with other studies made in similar contexts.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors responded to the comments made in the review, integrating text and improving it, so the work is now publishable.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This paper presents the author's efforts to apply Rosetta correlations to a soil that is important to their region (Black Soil). There was an extensive sampling and testing program as well as attempts to use the Van-Genuchten model to fit SWWC's. In order to generalize their test data for lightly and heavily-eroded Black soil, they compared data generated from laboratory tests on lightly and heavily eroded black soil to Van-Genuchten curve models, and then applied Rosetta soil fabric, index test data to better generalize the test data. 

The paper will require revision, however, since there are some important points not clearly explained in the paper. 

What were the specific SWCC tests , and how many were performed? A table showing this would be helpful.  What were the classification/fabric/index tests performed (how many and on what samples) in order to use as input to the Rosetta model? The method to evaluate goodness of fit of the predictions is confusing or misleading. The authors fit the data with VG formula, which is fine, but how many data points are they using to make the prediction? There is no indication or graph of a VG curve and measured data together. Graphs of predicted vs measured water content seem misleading, such as figure 2a. Is the R² value based on the data points in that graph? If so, it does not look like a typical R²=0.88 fit. The data is too dispersed, just looking at the graph. If the fit is based on the computed regression line and the 1:1 line, then it makes little sense since you have ignored the true variability of your data.  Computing the variability/error should be explained more clearly, see the file titled "Statistical measures" that I have attached.  I made a large number of grammatical corrections and clarifications on the first half of the manuscript, but decided not to continue since there is still so much more revision required in the second half of the paper.

So, I would suggest including a better discussion on the actual test data with some measured data points on graphs, a simpler, more direct method to compute goodness of fit, a better treatment of the index properties, showing their variability as well (better than box plots), number of tests on which samples etc...   and then the final discussion about making the modifications to your original predictions. 

Comments for author File: Comments.zip

Author Response

Response to Reviewer1 Comments

This paper presents the author's efforts to apply Rosetta correlations to a soil that is important to their region (Black Soil). There was an extensive sampling and testing program as well as attempts to use the Van-Genuchten model to fit SWWC's. In order to generalize their test data for lightly and heavily-eroded Black soil, they compared data generated from laboratory tests on lightly and heavily eroded black soil to Van-Genuchten curve models, and then applied Rosetta soil fabric, index test data to better generalize the test data.

The paper will require revision, however, since there are some important points not clearly explained in the paper.

Specific comments:

 

Point 1: What were the specific SWCC tests, and how many were performed? A table showing this would be helpful.

Response 1: Sorry, we did not describe this clearly. We have revised the part of data collection in the manuscript as follows,

“The soil water characteristic curve, mechanical composition, and bulk density of the soil samples collected from all three layers of each field site were measured. (1) The soil water characteristic curve for a soil is defined as the relationship between water content and suction for the soil. Soil water contents under 8 suctions (0, 33, 50, 100, 300, 500, 1000, and 1500 kPa) were measured and produced a data set of 360 (15 soil sample sites × 3 layers × 8 soil water suctions) pairs of measured suction (h) versus water content. Soil water contents for unsaturated soil were measured using pressure membrane apparatus, Soil saturated water content (0 kPa) were measured with several 100 cm³-cutting rings. (2) The mechanical composition were measured with the pipette method and produced a data set with 45 (15 soil sample sites × 3 layers) groups. (3) Bulk density were measured with cutting rings, and 45 (15 soil sample sites × 3 layers) data were collected.” (line 134 to line 144 on page 4).

 

Point 2: What were the classification/fabric/index tests performed (how many and on what samples) in order to use as input to the Rosetta model? The authors fit the data with VG formula, which is fine, but how many data points are they using to make the prediction?

Response 2: We have added a table in the revised manuscript, and rewrote revised some unclearly sentences. Some changes to the sentences are made as follows,

“According to classification of soil erosion in Table 1, 7 field sites (4 lightly eroded and 3 severely eroded soil field sites) of soil water contents were selected for fit, and the remaining 8 field sites of soil water contents (5 lightly eroded and 3 severely eroded soil field sites) were used for verification (Table 2). Firstly, 168 pairs of measured suction (h) versus water content from the 7 fitted field sites were used to fit a set of parameters in VG model using the program under different eroded soil. Then, the derived set was used as input in the VG model equation and received another data set of soil water content. Finally, 192 pairs of measured suction (h) versus water content from the 8 validated field sites were used to compare with the results derived from the model.” (line 153 to line 161 on page 5)

“The measured data set from 7 field sites (4 lightly eroded and 3 severely eroded soil field sites) were classified according to the Table 2. 21groups hierarchical sequence data were obtained as the (c) and (e) model forms of the Rosetta model according to the Table 3. The 21 groups fitted parameters as input of the VG model, including ,,  and , can obtain the 168 pairs of fitted suction (h) versus water contents by calculation. Model accuracy were verified by comparing with the measured values. If the model doesn’t perform well according to the verification results, the relation between m and n in the VG model need modification. Finally, 192 pairs of measured water contents of the 8 field sites (5 lightly eroded and 3 severely eroded soil field sites) in Table 2 were used to validate the performance of revised VG model.” (line 179 to line 188 on page 5)

Table 2. The number date were used in this study.

	Eroded degree	Sample sites	Layers	water suctions	Samples
Fitting	Lightly	4	3	8	96a
	Severely	3	3	8	72b
Verification	Lightly	5	3	8	120c
	Severely	3	3	8	72d
Total					360

^a4 (Lightly soil sample sites) × 3 (layers) × 8 (soil water suctions) = 96.

^b3 (Severely soil sample sites) × 3 (layers) × 8 (soil water suctions)=72.

^c5 (Lightly soil sample sites) × 3 (layers) × 8 (soil water suctions) = 120.

^d3 (Severely soil sample sites) × 3 (layers) × 8 (soil water suctions)=72.

 

Point 3: The method to evaluate goodness of fit of the predictions is confusing or misleading. Computing the variability/error should be explained more clearly, see the file titled "Statistical measures" that I have attached.

Response 3: Thank you for your suggestions. We have referenced your suggestions and added a method analyzes of the mean of residual in the manuscript. Added a part of 2.4 Accuracy Assessment Methods in the revised manuscript.

“The measured soil water contents were compared with the predicted soil water contents to assess the model performance by calculating the mean of residuals (MR) [56], Root Mean Square Error (RMSE), and using a 1:1 line regression method (one of the T test method, by estimating whether the confidence interval of the slope and intercept of the regressed equation include the number of 1 and 0, respectively. If included, it indicates that there is no difference between the regressed curve and the 1:1 line. It further indicates that there is no difference between the simulated values and observed values) [57, 58].

The residuals (R), mean of residuals (MR), and root mean square error (RMSE) were obtained using the following equations (3), (4) and (5),

                                                                                   (3)

                                                                                (4)

                                                                       (5)

where N is the total number of events,  is the measured value of soil water content, and  is the estimated value of soil water content.” (line 186 to 200 on page 6).

 

Point 4: There is no indication or graph of a VG curve and measured data together.

Response 4: Thank you for your suggestions. The fitted soil water characteristic curves for different eroded soils by the VG model have been added to the manuscript (Figure 2).

Figure 2. The soil water characteristic curve for different eroded soils

 

Point 5: Graphs of predicted vs measured water content seem misleading, such as figure 2a. Is the R² value based on the data points in that graph? If so, it does not look like a typical R²=0.88 fit. The data is too dispersed, just looking at the graph. If the fit is based on the computed regression line and the 1:1 line, then it makes little sense since you have ignored the true variability of your data. 

Response 5: We are sorry for making you feel confused. In the revised figure, we used the average measured values from 8 field sites (5 lightly eroded and 3 severely eroded soil field sites) and the simulated value derived from the VG model at each suction for the 1:1 line test. Figure 3 has been revised as follows,

Figure 3. Comparison of the measured and simulated soil water contents for different eroded soils.

 

Point 6: How did you measure theta residual. A value of 0.0 is very unusual in table 2.

Response 6: Theta residual was computed by the VG model program. Actually, it is not 0, just close to 0. Sorry for making you feel confused. We have attached the results of VG model as follows:

Figure a. The parameters output by the VG model for lightly eroded soil

Figure b. The parameters output by the VG model for severely eroded soil

 

Point 7: I made a large number of grammatical corrections and clarifications on the first half of the manuscript, but decided not to continue since there is still so much more revision required in the second half of the paper.

Response 7: Thank you for your suggestions. We have checked the manuscript for language carefully. 

 

Point 8: I would suggest including a better discussion on the actual test data with some measured data points on graphs, a simpler, more direct method to compute goodness of fit, a better treatment of the index properties, showing their variability as well (better than box plots), number of tests on which samples etc...   and then the final discussion about making the modifications to your original predictions.

Point 8: Thank you for your suggestions. We have revised them in the manuscript.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Generally, I would strongly encourage you to spellcheck your document - there are numerous grammatical errors and quirky sentences.

l. 120 ff: There are several things that seem unclear to the reader: How did you define "lightly eroded" and "severly eroded" - maybe add some description and/or pictures of typical sites.

Apart from that - please try to refer to your soils as FAO/WRB soil types with a proper description of your soil horizons.

This leads to another question: If you're investigating different sites, horizon depth surely differed - why did you decide to sample at certain depths (20/40/60) and not certain horizons? Especially when you want to focus on effects caused by soil erosion, there are certainly cut profiles with differing horizon-structure.

l. 165: "R" is missing at RMSE

l.193: It's just my personal opinion, but I would split result description and discussion. As I know this would mean a lot of work on the manuscript, I would delegate this issue to the editor.

l.198: Can you deliver the results of the T-test in your manuscript?

Figure 2 and others: Cut "the" from the axis labels

l. 235 / Figure 2: I have to admit, this might be a too easy question: I can't exactly understand your dataset distribution: Why are there always different measured soil moistures for the same simulated soil moisture? Is this caused by analyzing the different suctions? If this is the case - please explain, why it is reasonable the calculate your regression for all suctions and not for each suction exclusively.

l.408: Your discussion and the results are quite interesting - alas, I have a fundamental question regarding your setup: I know that you've collected a lot of individual samples - but you're trying to validate empirical models with an n of only 15 sites, which are, on top of that, split into two groups (lightly/severely eroded). My personal feeling is, that such an investigation may need more sites to deliver some profound results. I intepret your dataset somehow as some kind of small case study (and yes - I know that you've colleted *a lot* of samples).

Author Response

Response to Reviewer2 Comments

Generally, I would strongly encourage you to spell check your document - there are numerous grammatical errors and quirky sentences.

Specific comments:

 

Point 1: 120 ff: There are several things that seem unclear to the reader: How did you define "lightly eroded" and "severly eroded" - maybe add some description and/or pictures of typical sites.

Response 1: Thanks for your suggestion. We did not describe this clearly. We have added sentences to explain this: “Erosion phases were classified based on the structure of soil profiles, which was placed in Table 1.”(Line 126 to 130 on page 3)

Table 1. Classification of soil erosion.

Eroded soil degree	Indication
Lightly	A significant portion of horizon A remaining
Severely	Exposure of horizon C or evidence of incorporation of horizon C material into exposed horizon B through cultivation

A, mineral horizon; B, illuvial horizon; C, parent material

 

Point 2: Apart from that - please try to refer to your soils as FAO/WRB soil types with a proper description of your soil horizons.

Response 2: We have added sentences to clarify this: “Soil type is classified as Udic Isohumisols in the Chinese Soil Taxonomy, and Udic Argiboroll in the USDA Soil Taxonomy, or a Luvic Phaeozem in the FAO/UNESCO system.” (Line 112 to 113 on page 3)

 

Point 3: This leads to another question: If you're investigating different sites, horizon depth surely differed - why did you decide to sample at certain depths (20/40/60) and not certain horizons? Especially when you want to focus on effects caused by soil erosion, there are certainly cut profiles with differing horizon-structure.

Response 3: The structure of soil profile in the study area has been greatly changed by soil erosion, and changes in horizon thicknesses were great at different positions. To get comparable information at different sites, we sampled at certain depths.

 

Point 4: 165: "R" is missing at RMSE

Response 4: Thanks for your suggestions. We have revised it in the manuscript.

 

Point 5: 193: It's just my personal opinion, but I would split result description and discussion. As I know this would mean a lot of work on the manuscript, I would delegate this issue to the editor.

Response 5: Thanks for your suggestions.

 

Point 6: 198: Can you deliver the results of the T-test in your manuscript? Response 6: Yes, of course.

T-test in Table 4 were listed as follows.

Independent Samples Test

		Levene’s Test for Equality of Variances		t-test for Equality of Means
		F	Sig.	t	df	Sig.(2-tailed)	Mean Difference	Std. Error Difference	95% Confidence Interval of the Difference
									Lower	Upper
θs	Equal variances assumed Equal variances no assumed	3.686	.070	2.605 2.428	19 12.308	.017 .031	.07250 .07250	.02783 .02986	.01425 .00762	.13075 .13738
θr	Equal variances assumed Equal variances no assumed	0.021	.886	-.021 -.021	19 17.035	.983 .983	-.00056 -.00056	.02624 .02638	-.05547 -.05619	.05436 .05508
α	Equal variances assumed Equal variances no assumed	0.062	.806	.197 .212	19 18.045	.846 .834	.04222 .04222	.21453 .19900	-.40679 -.37580	.49123 .46024
n	Equal variances assumed Equal variances no assumed	7.185	.015	-1.508 -1.337	19 9.403	.148 .213	-.1114 -.1114	.0739 .0833	-.2660 -.2987	.0433 .0759

 

T-test in Figure 4 for lightly eroded soil using two inputs were listed as follows.

Independent Samples Test

		Levene’s Test for Equality of Variances		t-test for Equality of Means
		F	Sig.	t	df	Sig.(2-tailed)	Mean Difference	Std. Error Difference	95% Confidence Interval of the Difference
									Lower	Upper
θs	Equal variances assumed Equal variances no assumed	3.911	.061	-1.278 -1.278	22 12.707	.215 .224	-.00917 -.00917	.00717 .00717	-.02404 -.02470	.00571 .00637
θr	Equal variances assumed Equal variances no assumed	.539	.470	-.566 -.566	22 21.544	.577 .577	-.01000 -.01000	.01768 .01768	-.04666 -.04671	.02666 .02671
α	Equal variances assumed Equal variances no assumed	23.114	.000	-4.451 -4.451	22 11.000	.000 .001	-.02083 -.02083	.00468 .00468	-.03054 -.03113	-.01113 -.01053
n	Equal variances assumed Equal variances no assumed	.422	.523	7.602 7.602	22 18.674	.000 .000	.17417 .17417	.02291 .02291	.12665 .12616	.22168 .22218

 

T-test in Figure 4 for severely eroded soil using two inputs were listed as follows.

 

Independent Samples Test

		Levene’s Test for Equality of Variances		t-test for Equality of Means
		F	Sig.	t	df	Sig.(2-tailed)	Mean Difference	Std. Error Difference	95% Confidence Interval of the Difference
									Lower	Upper
θs	Equal variances assumed Equal variances no assumed	.203	.658	.816 .816	16 14.400	.426 .428	.00667 .00667	.00816 .00816	-.01064 -.01080	.02398 .02413
θr	Equal variances assumed Equal variances no assumed	2.825	.112	-.789 -.789	16 9.758	.442 .449	-.03111 -.03111	.03942 .03942	-.11468 -.11924	.05246 .05702
α	Equal variances assumed Equal variances no assumed	21.689	.000	-1.693 -1.693	16 9.052	.110 .124	-.01000 -.01000	.00591 .00591	-.02252 -.02335	.00252 .00335
n	Equal variances assumed Equal variances no assumed	5.041	.039	-.986 -.986	16 8.002	.339 .353	-1.26111 -1.26111	1.27861 1.27861	-3.97164 -4.20944	1.44942 1.68722

 

Point 7: Figure 2 and others: Cut "the" from the axis labels

Response 7: Thanks for your suggestions. We have revised it in the manuscript.

 

Point 8: 235 / Figure 2: I have to admit, this might be a too easy question: I can't exactly understand your dataset distribution: Why are there always different measured soil moistures for the same simulated soil moisture? Is this caused by analyzing the different suctions? If this is the case - please explain, why it is reasonable the calculate your regression for all suctions and not for each suction exclusively.

Response 8: We are sorry for making you feel confused. In the revised figure, we used the average measured values from 8 field sites (5 lightly eroded and 3 severely eroded soil field sites) and the simulated value derived from the VG model at each suction for the 1:1 line test. Figure 3 has been revised as follows,

Figure 3. Comparison of the measured and simulated soil water contents for different eroded soils.

 

Point 9: Your discussion and the results are quite interesting - alas, I have a fundamental question regarding your setup: I know that you've collected a lot of individual samples - but you're trying to validate empirical models with an n of only 15 sites, which are, on top of that, split into two groups (lightly/severely eroded). My personal feeling is, that such an investigation may need more sites to deliver some profound results. I intepret your dataset somehow as some kind of small case study (and yes - I know that you've colleted *a lot* of samples).

Response 9: Very good questions. Our study area was located in the Black Soil Region of Northeastern China. The soil type was the same across the area (Udic Isohumisols in the Chinese Soil Taxonomy, and Udic Argiboroll in the USDA Soil Taxonomy, or a Luvic Phaeozem in the FAO/UNESCO system), and soil properties were similar. The only difference from place to place was the degree of soil erosion caused by different slope gradient. The study sites we selected included different slope gradients and slope positions. Therefore, this study was representative, the results could be used in the Black Soil Region. We have collected soil samples in the other three field sites from south to north in the Black Soil Region (Figure 1) and compared the soil properties with the 15 sites in our study (Table 1). The results showed that the soil properties of the three selected reference sites were similar with the sites in the manuscript.

Table 1 Soil physical and chemical properties in black soil region of Northeast China

Indicators	Values at the 15 sites used in our study			Values at reference sites
	Max	Min		R1	R2	R3
Clay content	0.39	0.04		0.21	0.23	0.22
Silt content	0.47	0.06		0.14	0.39	0.46
Sand content	0.88	0.20		0.66	0.38	0.32
Organic carbon (%)	3.45	0.06		0.73	0.58	1.68
PH	6.38	5.55		6.07	6.03	5.99

 

 

Author Response File: Author Response.pdf