Forage Biomass Estimated with a Pre-Calibrated Equation of a Rising Platemeter in Pastures Grown in Tropical Conditions

Round 1

Reviewer 1 Report

line 91 In the discussion, I recommend commenting on the relative differences obtained, between the 3 pasture types, with studies that highlight the parameters on which the accuracy of the pre-calibrated methods depends.

 

line 110 If this assumption is correct, then it would be reliable (for the biomasses estimated the ascending to meter as well as for the comparisons you made) to use pasture-specific prediction equations, i.e., equations developed from each type of grass species sampled; given that there are <<different equations provided by the ascending to meter manufacturing company>>

 

line 261 If "the goodness of fit of the ascending tome meter is greater for ryegrass compared to kikuyu and stargrass as your hypothesis suggests because of the prediction equations developed with ryegrass", then this predisposes your results to a basis of comparison that is inconsistent; as it is more accurate for ryegrass and less accurate for kikuyu and stargrass.

 

line 285 Please give other reasons why you chose the equation you used. Although <<other equations are specific for months or seasons in temperate New Zealand conditions>> or <<added an additional difficulty of choosing a calibration equation>>, how appropriate is the one you used for the pastures being evaluated?

 

line 329 To eliminate the effect of this type of "variation" in your analysis, comparison indicators (such as r2) should be studied on the results obtained per evaluated station, i.e., these comparisons will be made per pasture and not between pastures. In addition, it would be important to indicate the geographic distance between the 3 evaluated pastures (from each other) in the study area, to get an idea of these variation factors and any other influencing indices you propose in the discussion.

Comments for author File: Comments.pdf

Author Response

Dear reviewer:

We adjusted the manuscript following your suggestions included in the attached document. Your comments in the system and our responses are the following:

1. line 91 In the discussion, I recommend commenting on the relative differences obtained, between the 3 pasture types, with studies that highlight the parameters on which the accuracy of the pre-calibrated methods depends. We used the same references cited in the introduction to discuss how different indicators affect the biomass estimates (lines 335–348)
2. line 110 If this assumption is correct, then it would be reliable (for the biomasses estimated the ascending to meter as well as for the comparisons you made) to use pasture-specific prediction equations, i.e., equations developed from each type of grass species sampled; given that there are <<different equations provided by the ascending to meter manufacturing company>>. You are correct. Part of our findings is to highlight the importance to develop specific equations to have more accurate data. That is why we don’t recommend using the equations developed in this study in a platemeter as one of our conclusions (lines 429–431) but instead developing specific calibration equations (lines 432–435), which is the second part of the study (not published).
3. line 261 If "the goodness of fit of the ascending tome meter is greater for ryegrass compared to kikuyu and stargrass as your hypothesis suggests because of the prediction equations developed with ryegrass", then this predisposes your results to a basis of comparison that is inconsistent; as it is more accurate for ryegrass and less accurate for kikuyu and stargrass. In this case we refer to the equations developed in the study based on the average intercepts and slopes from the linear regressions of Botanal (Table 4).
4. line 285 Please give other reasons why you chose the equation you used. Although <<other equations are specific for months or seasons in temperate New Zealand conditions>> or <<added an additional difficulty of choosing a calibration equation>>, how appropriate is the one you used for the pastures being evaluated? We appreciate this comment. We totally acknowledge that the choice of the equation is debatable. We used the “factory default” equation as it is the one that comprises a set of six months (April through September) of pasture growth. We have included this detail in the discussion (lines 297–300).
5. line 329 To eliminate the effect of this type of "variation" in your analysis, comparison indicators (such as r2) should be studied on the results obtained per evaluated station, i.e., these comparisons will be made per pasture and not between pastures. In addition, it would be important to indicate the geographic distance between the 3 evaluated pastures (from each other) in the study area, to get an idea of these variation factors and any other influencing indices you propose in the discussion. We have edited this section to explain that even though the three farms were in the same region, the elevation gradients influenced the response as it was evident with the difference in regrowth periods (lines 346–351).

 

The comments suggested in the PDF are the following:

• There are some references highlighted in the document without any comment that we did not know how to address.
• We changed a little bit of the wording in the abstract 
• Vartha & Matches (1977) we acknowledge this was an old reference to start. We included two more recent references.
• ’t Mannetje this is how the last name of this author is spelled.
• You asked for the map of the study environment. We don't understand what kind of map you are referring to. Also, for the conditions of the study is that we provided the data included in table 1 as part of the methods to characterize each species.
• In the results, the first paragraph in Italics must have been an error from the system as the original version don't include any section in Italics.
• Table 3 highlighted. The journal is the one in charge of checking that tables are not separated in two parts.
• Figures 2, 3, and 4 were formatted according to the editor's requests and they don't need a legend.
• In the discussion, the first two paragraphs about Pasture Management and Forage biomass don't pretend to be extensive but a brief characterization of the pastures as our goal is to compare the three methods. The forage biomass was within the ranges estimated in previous studies under these conditions.
• The references were adjusted to the IIEE system as the journal requested.

Reviewer 2 Report

Comments to the article

"Forage biomass estimated with a universal equation of a rising platemeter in pastures grown in tropical conditions."

 

1.    Then  the "universal equation" is far from universal, am I right? The title must be changed following your objectives and results.

2.    This study was not designed to compare the three standing aerial biomass estimation methods. Table 1 shows slope and management are confounded with pasture grass species. The sampling design is not mentioned, so I assume the data sets came from measurements whose purpose was not to compare the sampling methods.

3.    The phenological stage should be expressed as kg of DM of leaves per kg of DM shoot. You used the number of leaves per shoot, a count variable, for which the median and range are better statistics, given the non-parametrical nature of counts.

4.    The disk meter had an area-to-weight ratio of 0.5564 g/cm², equivalent to 5.6 kg/m²; how does this compare to other disk designs mentioned in other studies?

5.    You used 30 height readings if the pasture was <0.5 ha and 40 records for paddocks larger than half a hectare. What was the reason for this difference? Were those numbers (n = 30; n = 40) calculated with a formula? Do you have experimental evidence that the number of height readings was enough to minimize the variance? (See Castillo et al., 2009).

6.    To estimate the standing aerial biomass (SAB, kg/ha), you used three reference plots instead of the five used in the original Australian paper (Haydock and Shaw, 1975). Why? One would expect more points in the comparison plots, say seven or nine, but not less than five. You limited the prediction power of your calibration regression, as you ended up with only one error degree of freedom for the calibration scale of each pasture sampled.

7.    Erroneously you used the three quadrats harvested in the botanal to obtain a mean SAB. Then, there is no wonder that the results of the BOTANAL and your "direct" sampling resulted in very high coefficients of determination (R² > 0.95). The fact is that both methods result in a high correlation because one variable is a component of the other. Thus, comparing your BOTANAL and clipping methods is invalid, as both estimates are artificially highly correlated. I suggest you eliminate the "direct" sampling from your article.

8.   For the standard linear regression model, the coefficient of determination, R-squared (R²), is a widely used goodness-of-fit measure whose usefulness and limitations are more or less known to the applied researcher. Applying this measure to nonlinear models generally leads to values that can lie outside the [0,1] interval and decrease as you add more regressors. Kvalseth (1985) indicates that the R² must be used as a measure of goodness of fit only when:

8.1 R² must possess a utility as a measure of goodness of fit and have an intuitively reasonable interpretation.

8.2 R² ought to be independent of the units of measurement of the model variables; that is, R² ought to be dimensionless.

8.3. The potential range of values of R² should be well defined with endpoints corresponding to perfect fit and complete lack of fit, such as 0 ≤ R² ≤ 1, where R² = 1 corresponds to perfect fit and R² ≥ 0 for any reasonable model specification.

8.4. R² should be sufficiently general to apply: (a) to any model, (b) whether the x_j are random or nonrandom (mathematical) variables, and (c) regardless of the statistical properties of the model variables (including residual ε).

8.5. R² should not be confined to any specific model-fitting technique; R² should only reflect the model's goodness of fit per se, irrespective of how the model has been derived.

8.6. R² should be such that its values for different models fitted to the same data set are directly comparable.

8.7. Relative values of R² ought to be generally compatible with those derived from other acceptable measures of fit (e.g., standard error of prediction and root mean squared residual).

8.8. Positive and negative residuals () should be weighted equally by R².

 

9.  Therefore the coefficient of determination (R²) is not the best statistic to help you decide which pasture sampling method is the most adequate. The best thing to do is to "correct" the R-square by the number of parameters in the model or to do an F-test if a simpler model, a 2nd-order polynomial, is contained or nested within a more complicated one, say, 4th-order polynomial. The unadjusted coefficient of determination is:
R²_Unadj. = 1 – (SS _residual/SS _total); the adjusted version is:
R²_Adj. = 1 – ((SS _residual/(n – k)))/((SS _total/(n – 1))) where n is the number of 'x-y' pairs and K the number of parameters of the model; here, the lower the value of the adjusted R², the less adequate the fit. Thus, R²_Unadj. > R²_Adj. (Motulsky and Christopolous, 2004).

10. This reviewer considers the submitted article unsuitable for publication due to its statistical flaws.

 

References

Castillo-Gallegos, E., Valles-de la Mora, B. and Jarillo-Rodríguez, J. (2009). Relación entre materia seca presente y altura en gramas nativas del trópico mexicano. Técnica pecuaria en México, 47(1), 79-92.

Haydock, K.P., and Shaw, N.H. (1975). The comparative yield method for estimating dry matter yield of pasture. Australian Journal of Experimental Agriculture, 15(76), 663-670.

Kvalseth, T.O. (1985) Cautionary note about R². American Statistician, 39, 279-285.

 

Motulsky, H., & Christopoulos, A. (2004). Fitting models to biological data using linear and nonlinear regression: a practical guide to curve fitting. Oxford University Press.

Comments for author File: Comments.pdf

Author Response

Dear reviewer, we appreciate your comments.

Below we explain how those were addressed and give details for some of your questions using the same numbers of your list:

1. We acknowledge the universal equation is not commonly used in the titles of manuscripts and we replaced it with pre-calibrated equation as it was originally explained in the methods.
2. While we understand that you could have thought that this study was not designed to compare the three methods, we would like to point out that this study was in fact designed to evaluate and compare the three methods and it was run during the same year for the three grass species simultaneously. We acknowledge that, based on the assumption made, the sampling design may have been confusing, which is why we edited and extended the section 2.2 Data collection to provide more details of each sampling procedure.
3. The phenological stage was an indicator collected as complementary information to characterize the grazing management but this was not a variable comprised in the study. That is why we included these indicators in Table 1 as part of the methods.
4. The disk meter used in this study is not different  than other disks used in other studies. It actually has the same size and weight as the one used by Sanderson et al. (2001). In their methods they indicate as it follows: "The rising plate meter has a disk with a diameter of 362 mm (0.1 m²) and mass of 0.315 kg". This implies that the area is 0.11341 m² which is equal to 8.81 times to cover a square meter or 8.81 x 0.315 kg = 2.77 kg/m². 
5. We followed the sampling procedure given by the manufacturing company of the platemeter. As it was mentioned in number 2, we have edited the section 2.2 Data collection. We have explained how this only happened in two paddocks of Stargrass.
6. The biomass was estimated using three levels or reference points that are included in the Botanal spreadsheet and weighed against the 50 visual samples. We have used this approach in previous studies (http://www.mag.go.cr/rev_agr/v38n01_133.pdf, http://www.mag.go.cr/rev_agr/v37n02_091.pdf, http://www.mag.go.cr/rev_agr/v37n01_091.pdf) to reduce the subjectivity when recording the 50 visual samples. For example, if we were to use 5 levels, it has been our experience that find the difference between a number 3 and 4 can be challenging. We included further details in the methods section.
7. We included further details in the methods section explaining how in a previous study we made this comparison of methods with minor differences. We also included more details in the discussion. We acknowledge that both methods are correlated, however, we think it is worth including this method as it reduces the time spent when collecting samples for producers or technicians lacking experience in grass sampling techniques.
8. We acknowledge that it we want to decide the best method (among the three evaluated) another type of analysis should be performed. Our interest was never to pick the best method but to compare between methods. Even when we mentioned in the data analysis section that “At each comparison, one method was considered the most accurate (hand-clipping>Botanal>platemeter) to represent the predictability of the indirect methods as it has been previously studied R² is an indicator of goodness of fit”, we know that each methods has particular reasons to have more or less differences with respect to other method. That is why we used the residual standard deviation to compare two methods at the time and this indicator has been used in previous studies such as Harmoney et al., (1997) when comparing four indirect methods.
9. Like the previous point, we included more details in the closing paragraph of the introduction explaining our aim in this study as well as in the data analysis section to explain that it was never the intention to pool the three methods to pick the “best one”.
10. After explaining each suggestion and adding more details, we hope that the manuscript is suitable.

Reviewer 3 Report

I think it's an interesting paper on productivity estimation of forage. However, it is judged that more investigations on various grass species and environments are needed.

 

 

Review report

Manuscript ID: Grasses-2244630

Manuscript title:  The review about ‘Forage biomass estimated with a universal equation of a rising platemeter in pastures grown in tropical conditions’

 

Specific comments to Authors:

 From the perspective of the article as a whole, the written expression is accurate and fluent, with few grammatical errors.

1、About the introduction part:

(1)   The paragraphs provide information on different methods of estimating forage biomass, including direct and indirect methods, and their advantages and disadvantages.

(2)   They introduce the concept of remote sensing methods for estimating forage biomass, which is a promising area of research.

(3)   The paragraphs describe a specific study that compares the accuracy of different methods for estimating forage biomass in tropical conditions, which may be helpful for researchers and producers in those regions.

(4)   They do not provide a comprehensive review of all available methods for estimating forage biomass but rather focus on a few specific methods and their limitations.

(5)   They do not discuss the potential environmental impacts of different methods for estimating forage biomass, which may be an important consideration for sustainable agriculture.

 

2、About the Method and Results parts:

(1)   The study evaluates multiple methods to estimate forage biomass, providing a comprehensive comparison of their accuracy and suitability.

(2)   The study compares three different grass species, providing a more diverse and representative sample.

(3)   The study considers multiple variables such as paddock size, stocking density, days of regrowth, number of leaves, nitrogen rates, slope, and senescence, providing a more detailed analysis of the factors that affect forage biomass.

(4)   The study uses statistical analyses such as the coefficient of determination and linear regression equations, providing a more objective and quantitative evaluation of the methods.

(5)   The study does not consider other potential methods to estimate forage biomass, such as near-infrared spectroscopy or remote sensing techniques.

(6)   The study only evaluates three grass species, which may not be representative of all possible forage species.

(7)   The study only considers a limited number of variables, and other factors such as soil quality, weather conditions, or management practices may also affect forage biomass.

 

3、About the Discussion and Conclusions parts:

(1)   The article discusses the importance of accurate estimation of forage biomass for livestock operations.

(2)   The study evaluates the accuracy of the rising platemeter for estimating forage biomass compared to the Botanal and hand-clipping methods.

(3)   The article discusses the limitations of indirect methods and the need for species-specific calibration equations.

(4)   The study highlights the interference of senescent material on the accuracy of platemeter readings.

(5)   The article suggests the potential benefits of using the rising platemeter with specific calibration equations for estimating forage biomass in tropical conditions.

(6)   The discussion and conclusion section is somewhat repetitive and could be more concise.

(7)   The study only evaluates the rising platemeter and does not compare it to other indirect methods for estimating forage biomass.

 

(8)   The limitations of the study are not fully discussed, including the sample size and the specific environmental conditions of the study.

Author Response

Dear reviewer, we appreciate your comments.

Below we explain how those were addressed and give details for some of your questions using the same numbers of your list:

Specific comments to Authors:

From the perspective of the article as a whole, the written expression is accurate and fluent, with few grammatical errors.

1.Introduction:

1)   The paragraphs provide information on different methods of estimating forage biomass, including direct and indirect methods, and their advantages and disadvantages.

2)   They introduce the concept of remote sensing methods for estimating forage biomass, which is a promising area of research.

3)   The paragraphs describe a specific study that compares the accuracy of different methods for estimating forage biomass in tropical conditions, which may be helpful for researchers and producers in those regions.

4)   They do not provide a comprehensive review of all available methods for estimating forage biomass but rather focus on a few specific methods and their limitations. We appreciate this observation and it is totally correct. It was not our intention to provide an extensive review of all the methods used to estimate biomass in pastures but the ones we considered most relevant for this manuscript.

5)   They do not discuss the potential environmental impacts of different methods for estimating forage biomass, which may be an important consideration for sustainable agriculture. We briefly mentioned how some agronomic traits may influence the accuracy of pre-calibrated methods (lines 86–92) but we extended the implications of environmental factors in the discussion section. In the discussion we included details of how the farms were in the same region but with differences in elevation that influence their management (lines 346–355)

 

2.Methods and Results:

1)   The study evaluates multiple methods to estimate forage biomass, providing a comprehensive comparison of their accuracy and suitability. We added a brief description of our aim in the last paragraph of the introduction (lines 103–105) to provide more details of how the comparisons were made between two methods at the time and not among the three methods (pooled).

2)   The study compares three different grass species, providing a more diverse and representative sample.

3)   The study considers multiple variables such as paddock size, stocking density, days of regrowth, number of leaves, nitrogen rates, slope, and senescence, providing a more detailed analysis of the factors that affect forage biomass.

4)   The study uses statistical analyses such as the coefficient of determination and linear regression equations, providing a more objective and quantitative evaluation of the methods.

5)   The study does not consider other potential methods to estimate forage biomass, such as near-infrared spectroscopy or remote sensing techniques.

6)   The study only evaluates three grass species, which may not be representative of all possible forage species. We acknowledge that there are more grass species used in dairy farms in the tropical regions and in Costa Rica (Brachiaria/Urochloa and Panicum/Megathyrsus), but we chose the three main species used in specialized dairy farms in Costa Rica based on a previous study. Also, these other grass species have a different plant architecture (bunchgrasses) as compared to the grass species recommended for platemeters.

7)   The study only considers a limited number of variables, and other factors such as soil quality, weather conditions, or management practices may also affect forage biomass. We added more details in some paragraphs of the methods (lines 123–125, 135–142, 151–154, 166–168, 178–180, 192–193)

 

3.Discussion and Conclusions:

1)   The article discusses the importance of accurate estimation of forage biomass for livestock operations.

2)   The study evaluates the accuracy of the rising platemeter for estimating forage biomass compared to the Botanal and hand-clipping methods. We explained the reason to use the universal equation pre-loaded in the platemeter (lines 298–301)

3)   The article discusses the limitations of indirect methods and the need for species-specific calibration equations.

4)   The study highlights the interference of senescent material on the accuracy of platemeter readings.

5)   The article suggests the potential benefits of using the rising platemeter with specific calibration equations for estimating forage biomass in tropical conditions.

6)   The discussion and conclusion section is somewhat repetitive and could be more concise.

7)   The study only evaluates the rising platemeter and does not compare it to other indirect methods for estimating forage biomass. You are correct, it was our intention to compare the platemeter against the other two methods. That is why we included the description of our aim in the last paragraph of the introduction (lines 103–105).

8)   The limitations of the study are not fully discussed, including the sample size and the specific environmental conditions of the study. Because we followed each method as it is referenced and has been used in previous studies, each paddock represented an experimental unit. We thought that it would not have been accurate to increase the sample size from each paddock without having the corresponding number of replicates at each method. In other words, we would have ended up with an unbalanced study. We included two explanations of how the environmental conditions (lines 347–351) and comparisons between Botanal and hand-clipping were related (lines 414–421). Finally, in the conclusions we explained that our results are specific for the conditions of this study (lines 431–433) and the need to develop specific equations for these conditions (lines 440–443)

Round 2

Reviewer 2 Report

Is it OK to use three or two points instead of the 5-point original scale in the comparative yield method to estimate pasture’s aerial biomass?

Consider the following example. We sampled two 5000 m² divisions before grazing on a ten divisions 30-day rotation (3 days grazing/27 days recovery). We followed the comparative yield method as closely as possible, as described in the original paper. The data (g/0.25 m² of DM) we obtained for the calibration line were as follows:

Visual Rating	Paddock 1 5 points	Paddock 1 3 points	Paddock 1 1 points	Paddock 3 5 points	Paddock 3 3 points	Paddock 3 1 points
1	80.4	80.4	80.4	45.1	45.1	45.1
2	107.1			56.2
3	135.2	135.2		68.2	68.2
4	143.1			75.4
5	179.3	179.3	179.3	94.6	94.6	94.6

 

The result of fitting a simple linear regression to the data was as follows:

	Paddock 1 5 points	Paddock 1 3 points	Paddock 1 1 points	Paddock 3 5 points	Paddock 3 3 points	Paddock 3 1 points
Best-fit values
B0	58.82	57.41	55.62	32.45	32.20	32.76
B1	23.40	24.75	24.75	11.81	12.36	12.36
Std. Error
B0	7.263	5.315		3.257	1.657
B1	2.190	1.556		0.9821	0.4850
Goodness of Fit
Error DOF	3	1	0	3	1	0
Adj. R2	0.9659	0.9921	1.000	0.9729	0.9969	1.000
Sy.x	6.925	4.401		3.106	1.372
RMSE	5.997	3.112		2.690	0.9699
AICC	46.80			38.78
F-value	114.16	252.92	N. C.	144.54	649.53	N. C.
P > F	0.0018	0.0400	N. C.	0.0012	0.0250	N. C.

 

Therefore, if you have well-trained operators, the risk of using three points and not having a significant linear response is very small. So the researcher can get away with allowing the loss of 40% of his(her) calibration points.

I do not criticize using three instead of 5 points. What is not correct, from the statistics standpoint, is to use the 3-point average aerial biomass because it is one component of the equation used to estimate the amount of aerial biomass in a quadrat. According to the table above, the calibration equation for paddock 1 with three points is Y = 57.41 + 24.75X. The average of comparative evaluations can substitute for the X. For this sampling, the average was 1.19, so Y = 54.71 + 24.75(1.19) = 86.86 g/0.25 m² of DM. If you average the three points corresponding to visual ratings 1, 3, and 5, you’ll get 131.63 g/0.25 m² of DM. The difference between 86.86 and 131.63 is vast, so I wonder why you obtained such significant correlations when you related BOTANAL with the clipped biomass. Besides, comparing clipped vs. BOTANAL violates the assumption of independence (Both are highly correlated according to your plots).

 

You quoted on page 4, lines 179-181, “This procedure was previously evaluated and compared in Villalobos et al. (2013) against the Botanal method with minor differences reported and reducing the time spent during the sampling.” However, I found nothing related to your claim in such a paper. I consulted the article on Agronomía Costarricense, 37(2):91-103, 2013. Please verify your information.

Author Response

Dear reviewer, we appreciate your comments and we have addressed your concerns as it follows (bolded text is our explanation for each comment):

• Therefore, if you have well-trained operators, the risk of using three points and not having a significant linear response is very small. So the researcher can get away with allowing the loss of 40% of his(her) calibration points. This is a good point, and we totally agree that well-trained operators would be necessary (recommended) to use a reduced number of subsamples when using the hand-clipping method. We have included this consideration in the discussion section (lines 420-423). Also, we suggested that producers and technicians lacking experience in pasture sampling may use the hand-clipping method only when the number of subsamples are increased to obtain a more robust estimate of biomass (lines 423-426 and 454-456).
• I wonder why you obtained such significant correlations when you related BOTANAL with the clipped biomass. Besides, comparing clipped vs. BOTANAL violates the assumption of independence (Both are highly correlated according to your plots). We agree that the real samples from Botanal and the subsamples used in the hand-clipping method are the same and we acknowledged that the high r² values were expected due to the high correlation between samples (lines 419-420).
• You quoted on page 4, lines 179-181, “This procedure was previously evaluated and compared in Villalobos et al. (2013) against the Botanal method with minor differences reported and reducing the time spent during the sampling.” However, I found nothing related to your claim in such a paper. I consulted the article on Agronomía Costarricense, 37(2):91-103, 2013. Please verify your information. We have edited this sentence indicating that the comparison of methods comprised data not published in the manuscript cited (lines 179-181).