Fertilization and Soil Nutrients Impact Differentially Cranberry Yield and Quality in Eastern Canada

Round 1

Reviewer 1 Report

Dear Authors

Review of the manuscript
„ Nutrient Management of Sand-grown Cranberry in Eastern Canada”
Manuscript ID: horticulturae-1272985

The title of the manuscript is too general and needs to be corrected. It should reflect the whole content of the paper. The abstract in the section describing the research methodology should be modified after completing / correcting the M&M chapter. The keywords seem right. The introduction presents the research problem well. The aim of the research and research hypotheses were presented. The research methodology is described vaguely and requires some additions. The introductory sentence for the M&M chapter, section 2.1, line 74 is unnecessary - a truism. The doses of P, Mg, B and Cu should be given, and the yield parameters were to depend on them (see research hypothesis). There is also a lack of the times and methods of their application. Why were different nitrogen fertilizers applied to the organic cropping system each year? My other doubts: a) representativeness of the very small harvested area of ​​berries (0.37 m2); b) too early harvesting of fruits, i.e. 2-3 weeks before commercial picking, it could have significantly influenced the values ​​of the parameters tested - these issues require careful clarification.

Figure one should be more detailed, now it is a fragment of a general, hardly readable map. Soil characteristics should be divided into individual research areas - the scatter measures indicate significant differentiation of the obtained values. Soil pH is missing. In this table, please indicate which forms are total content and which are available for plants (according to Mehlich III method).

Chapter "Results"

lines 175, 177: Were the doses of P differentiated? (see M&M chapter doubts);

lines 177-178: I suggest to present the influence of Mg, B and Cu regimes (in any form);

lines 184-185: in what combination?

In the discussion, the authors attempted to explain the obtained dependencies. They also confronted the results of their observations with the data contained in the literature.

The conclusions are right.

Reviewer

Author Response

REVIEWER #1

Review of the manuscript
„ Nutrient Management of Sand-grown Cranberry in Eastern Canada”
Manuscript ID: horticulturae-1272985

1. The title of the manuscript is too general and needs to be corrected. It should reflect the whole content of the paper.
2. We changed title to:

Fertilization and Soil Nutrients Impact Differentially Cranberry Yield and Quality in Eastern Canada

The abstract in the section describing the research methodology should be modified after completing / correcting the M&M chapter.

Abstract rewritten.

R: The M&M section was summarized as follows: “A 3-yr trial was conducted on permanent plots at four production sites in Quebec, Canada. We analyzed yield parameters, marketable yield and fruit quality in response to fertilization and soil properties.”

The keywords seem right.

Thank you

The introduction presents the research problem well. The aim of the research and research hypotheses were presented.

Thank you

The research methodology is described vaguely and requires some additions.

The introductory sentence for the M&M chapter, section 2.1, line 74 is unnecessary - a truism.

Removed

The doses of P, Mg, B and Cu should be given, and the yield parameters were to depend on them (see research hypothesis). There is also a lack of the times and methods of their application.

We detailed dosage in Table 2 and Appendix A2. Fertilizer sources are also provided in section 2.3. We provide times and methods for N, P, K, Mg, Cu, and B fertilization on conventional and organic site (line 114 to 120). The effect of Mg, Cu and B regimes on dependent variables is presented in Appendix A3 and A4.

Why were different nitrogen fertilizers applied to the organic cropping system each year?

We had to adjust to changes in product certification. Amino acids product was denied in 2016 and was replaced by fish emulsion. The SCU, amino acids and certified fish emulsions were used to test effects of slow-release fertilizers compared to fast-release ammonium sulfate on cranberry yield and quality.

My other doubts:

a) representativeness of the very small harvested area of ​​berries (0.37 m2);

Such areas are normal in cranberry research. Cranberry harvest surface per plot was close compared to Parent et al. (2006) (0.45 m2) and twice as large compared to Davenport (1996) (0.18 m2).

b) too early harvesting of fruits, i.e. 2-3 weeks before commercial picking, it could have significantly influenced the values ​​of the parameters tested - these issues require careful clarification.

Harvesting experimental plots 2-3 wk before commercial harvest is unavoidable and normal because commercial cranberry beds are flooded before commercial harvesting (Parent et al., 2006). Indeed, farm managers may decide to harvest the bed early, even before date planned for plot harvest (Davenport, 1996). Hence, TAcy index for anthocyanin content (Fulecki and Francis, 1968) is lower at hand picking time (Parent et al., 2006). Nevertheless, trends in berry yield and quality related to N and K fertilization remained significant.

Figure 1 should be more detailed, now it is a fragment of a general, hardly readable map.

The zoom map (Figure 1) shows study sites. Therefore, we had several versions of the map. The large map allows international readers to capture the proximity of Quebec farms to Wisconsin and Massachusetts, the leading US states in cranberry production.

Soil characteristics should be divided into individual research areas - the scatter measures indicate significant differentiation of the obtained values. Soil pH is missing. In this table, please indicate which forms are total content and which are available for plants (according to Mehlich III method).

Table 2 was reconfigured to include pH and Mehlich-3 elements.

Chapter "Results"

lines 175, 177: Were the doses of P differentiated? (see M&M chapter doubts);

We mention that in line 110 phosphorus was applied as triple superphosphate (46% P₂O₅) or bone meal (13% P₂O₅) in three P doses (0, 15, 30 kg P ha^-1) (Table 2 and Appendix A2). The P treatment were applied at four occasions same as N and K treatment. Please see the P result in 13 dependent variable in Figures 3 or 4 and Appendix A3 or A4.

lines 177-178: I suggest to present the influence of Mg, B and Cu regimes (in any form);

OK. Figure in Appendix (A3 and A4).

lines 184-185: in what combination?

Nutrient combinations detailed in Table 2.

In the discussion, the authors attempted to explain the obtained dependencies. They also confronted the results of their observations with the data contained in the literature.

Thank you.

The conclusions are right.

Thank you.

Reviewer 2 Report

In my opinion the paper with title: “ Nutrient Management of Sand-Grown Cranberry in Eastern Canada” is suitable to be published in the journal Horticulturae with  minor revision

In chapter 2.5. please specific the devices you have used in order to determine the Brix (refractometer type)

Please introduce in the text at 149 row what represent the clr transformations. and adding the reference for more information.

Please introduce in the text at 140 row what represent the nlme model and adding the reference for more information

In article is used like code number, ammonium sulfate - (21-0-0) for example. Please explain the procedure to develop this code . First number represent    , second number         , third   number   represent

Author Response

REVIEWER #2

In my opinion the paper with title: “Nutrient Management of Sand-Grown Cranberry in Eastern Canada” is suitable to be published in the journal Horticulturae with minor revision

In chapter 2.5. Please specific the devices you have used in order to determine the Brix (refractometer type)

The refractometer was used to measure total soluble solid concentration (Brix) in fruits. Ocean Spray is used the Atago RX-5000 refractometer models, which is an automatic digital refractometer (Available at: https://www.atago.net/product/?l=en&f=products-rx-alpha-top.php) to measure Brix. In addition, the Ocean Spray quality department evaluated fruit quality by using the USDA (2007) shipping point and market inspection instructions for fresh cranberries.

Please introduce in the text at 149 row what represent the clr transformations, and adding the reference for more information.

The clr transformation is used to account for the special properties of compositional data. We referred to Aitchison (1986) and Pawlowsky-Glahn et al. (2011). We added Greenacre (2021): Greenacre M. Compositional data analysis. Ann. Rev. Stat. Appl. 2021. 8, 271–99. https://doi.org/10.1146/annurev-statistics-042720-124436

Please introduce in the text at 140 row what represent the nlme model and adding the reference for more information

The combination of performance-treatment has been analyzed by a mixed model. The nonlinear mixed-effects (nlme) model using the variance-covariance matrix for random effects and residuals (Pinheiro, 2019) is common in biology, environment and agriculture research. We analyzed the fertilization experiment as linear mixed-effect models by using the R nlme package with fertilization regimes and plant functional groups as fixed effects and site and year of survey as a random effect. Three combination of performance treatment were tested to fit 1) a linear model, 2) a quadratic model or 3) no model at all. See the result in Figures 3-4, and Appendix 3-4.

In article is used like code number, ammonium sulfate - (21-0-0) for example. Please explain the procedure to develop this code.

First number represent %N, second number represents %P2O5, third number represents %K2O. Fertilizer compositions was expressed as commercial units (%N, %P2O5, %K2O). Nutrient additions were expressed as elemental composition in Table 2.

Reviewer 3 Report

In general, this seems to be a well-conducted study, but I do have a number of questions and comments below.

61-65  This section on redundancy analysis and log-ratio transformation comes out of nowhere and lacks context. Provide a brief introductory statement of what problem is being addressed with these analyses.

139  What were the dependent variables analyzed by the mixed model?

146  What are the “cranberry performance indexes”? The aforementioned yield and quality variables? As a minor point, “indexes” and “indices” (Fig. 6) seems to be used interchangeably in the manuscript.

147-148  Add one sentence with the goal of this analysis.

Table 1 should be a supplementary table (Appendix).

Fig. 1: The zoom level of the map needs to be increased to show more detail of the study area. It seems to me that all that needs to be shown is latitude 45 to 50, longitude -65 to -80.

Fig. 2: I do not understand Fig. 2; there needs to be more information explaining it in the legend. What are the scales on the various figures? Absolute differences in relation to the 21-0-0 standard? Based on per-plot averages? What is the boxed value in the middle? The mean or the median? What are the ranges? 95% confidence intervals? Why is it that not all graphs show all three treatments (amino acids, SCU, fish emulsions)? Where in the graph does it show which effect is statistically significant and which one is not?

Fig. 3 and 4: Again, more information is needed in the legend, e.g., that this is averaged across fertilizer types, locations, and years. Also indicate what the gray area surrounding the regression line is – the prediction interval? The legend to Fig. 4 mentions response to N and K, but P is also shown in the figure. It is not clear to me why some of the individual figure panels for P have so few data points – shouldn’t there be a value for each field plot?

I also do not understand Fig. 5. Take the first row and the first column as an example, which is flower counts crossed with flower counts. Shouldn’t the vertical and horizontal axes have the same scale in this case? Instead, the vertical axis is from 0 to 0.0020 and horizontal axis from 300 to 900. How can that be? Also, explain the graphs shown in the diagonal of the matrix and add units for each of the variables.

Fig. 6: It looks like the location (left vs. right) of the two plots got switched in the legend. Also explain what is meant by the “relative abundance (>0.5%) of cranberry performance indices”.

Appendix A1: It looks like the mean temperature value for December 2015 is erroneous. For monthly precipitation, use a bar graph, not a line graph.

Appendix 2: Indicate sample size (n) used in the calculation of mean and SD.

The table shown as Appendix A3 should go into the main text. This information is critical toward understanding what was done.

Author Response

REVIEWER #3

In general, this seems to be a well-conducted study, but I do have a number of questions and comments below.

l. 61-65  This section on redundancy analysis and log-ratio transformation comes out of nowhere and lacks context. Provide a brief introductory statement of what problem is being addressed with these analyses.

Redundancy Analysis (RDA) is an extension of “simpler” algorithms such as correspondence analysis and principal component analysis. RDA is a commonly used to analyze the impact of "explanatory variables" on "response variables" (l. 177-180). Due to closure to the bounded sum of measurement unit resulting in spurious correlations and possibly to confidence intervals scanning beyond the compositional space (e.g. below 0 or above 100%), running statistical with raw compositional data leads to distorted, sometimes physically absurd, results if not log-ratio transformed beforehand (Aitchison, 1986; Filzmoser et al., 2009). We added Greenacre (2021): Greenacre M. Compositional data analysis. Ann. Rev. Stat. Appl. 2021. 8, 271–99. https://doi.org/10.1146/annurev-statistics-042720-124436. The clr enabled us to use RDA for data reduction. RDA algorithm allowed to rank the importance of variables in the model by using the vegan package (Oksanen et al. 2013). In our case, a multiple linear regression does guarantee that the model fits the data well. Every RDA has two matrices, a response matrix and an explanatory matrix. Orthogonal axes are defined in the space of the explanatory variables. See the Numerical ecology book of Legendre & Legendre 2012 for details on the implementation and interpretation of RDA (Available at: https://www.elsevier.com/books/numerical-ecology/legendre/978-0-444-53868-0).

l. 139 what were the dependent variables analyzed by the mixed model?

We analyzed fertilization regimes (N, P, K, Mg, B and Cu) and plant functional groups as fixed effects in mix model (l. 168-169). We have 13 dependent variables. There were seven yield parameters as follows: flower counts, number of reproductive uprights, number of flowers per reproductive up right, berry counts, number of fruiting uprights, berry counts per fruiting upright, and fruit set. Other dependent variables were quality indices (TAcy, Brix, firmness and berry moisture), marketable yield and berry weight. The outputs of dependent variables are presented in Figure 3 or 4, and appendix 3 or 4. We analyzed the fertilization experiment as nonlinear mixed-effects (nlme) model (Pinheiro, 2019), with fertilization regimes and plant functional groups as fixed effects and site and year of survey as a random effect.

l. 146  What are the “cranberry performance indexes”? The aforementioned yield and quality variables? As a minor point, “indexes” and “indices” (Fig. 6) seems to be used interchangeably in the manuscript.

They are indeed yield and quality variables (sorry for the confusion), l. 163-168 : yield, yield parameters and quality indices. We used ‘indices’ across the text.

l. 147-148  Add one sentence with the goal of this analysis.

l. 147-148: commercial criteria.

Table 1 should be a supplementary table (Appendix).

Done.

Fig. 1: The zoom level of the map needs to be increased to show more detail of the study area. It seems to me that all that needs to be shown is latitude 45 to 50, longitude -65 to -80.

We zoomed the study area but kept the larger map to indicate the location compared to WI and MA. If we go in too much detail, international readers may not capture the proximity of Quebec farms to Wisconsin and Massachusetts, the leading states in cranberry production. Therefore,

Fig. 2: I do not understand Fig. 2; there needs to be more information explaining it in the legend. What are the scales on the various figures? Absolute differences in relation to the 21-0-0 standard? Based on per-plot averages? What is the boxed value in the middle? The mean or the median? What are the ranges? 95% confidence intervals? Why is it that not all graphs show all three treatments (amino acids, SCU, fish emulsions)? Where in the graph does it show which effect is statistically significant and which one is not?

Coefficients of a linear mixed model is a core characteristic for understanding the effect of N source on cranberry performance (Figure 2). We compared four N source as amino acids, S-coated urea and fish emulsions to ammonium sulfate. The lme base R function calculates the intervals and proximate 95% confidence intervals. The caption was made more explicit.

Fig. 3 and 4: Again, more information is needed in the legend, e.g., that this is averaged across fertilizer types, locations, and years. Also indicate what the gray area surrounding the regression line is – the prediction interval? The legend to Fig. 4 mentions response to N and K, but P is also shown in the figure. It is not clear to me why some of the individual figure panels for P have so few data points – shouldn’t there be a value for each field plot?

There were no significant site and year effects on dependent variables. We ran linear or quadratic model using nonlinear mixed-effects (nlme) model (Pinheiro, 2019). Tests of significance (p = 0.05) were used to reject the null hypothesis and not to accept it as true; non-significant results do not mean that there is no difference between groups or no effect of a treatment (Amrhein et al. 2019, Available at: https://www.nature.com/articles/d41586-019-00857-9?fbclid=IwAR1jzbGpWu9wsHIwBdOu3byOielCLEQxPZMvHJ-3X4GW2gvy4eD98a7a9EU). Therefore, the shaded area is by default the asymptotic 95% confidence envelope. We used geom smooth (Wickham, 2016, Available at: https://link.springer.com/chapter/10.1007/978-3-319-24277-4_2) based on a linear or quadratic model (Fig. 2.6). The results is shown in Figure 3 or 4, and Appendix 3 or 4.

I also do not understand Fig. 5. Take the first row and the first column as an example, which is flower counts crossed with flower counts. Shouldn’t the vertical and horizontal axes have the same scale in this case? Instead, the vertical axis is from 0 to 0.0020 and horizontal axis from 300 to 900. How can that be? Also, explain the graphs shown in the diagonal of the matrix and add units for each of the variables.

The vertical axis of flower counts has now same scale as flower counts in horizontal axes in Figure 5. We added units for each variable in the vertical and horizontal axes. The scatterplots above the diagonal show the original data after adjusting for all other variables. A matrix information was added in the legend subsequently. A correlation (Pearson correlation, Benesty, et al. 2009) matrix is a key figure for understanding correlation coefficients between flower counts, number of reproductive uprights, number of flowers per reproductive up right, berry counts, number of fruiting uprights, berry counts per fruiting upright, and fruit set in our study. We used ggpairs (available at: https://www.rdocumentation.org/packages/GGally/versions/1.5.0/topics/ggpairs).function in R to make a correlation matrix.

Fig. 6: It looks like the location (left vs. right) of the two plots got switched in the legend. Also explain what is meant by the “relative abundance (>0.5%) of cranberry performance indices”.

In the legend, the location of the plot (left vs. right) was corrected (thanks for pointing that out). We determined soil‐nutrient intercorrelation with cranberry indices by using redundancy analysis (RDA). We used a very basic scenario based on, when one of the data sets (Y) is to be explained by the other (X), redundancy analysis (RDA) was used to understanding to ranking and ordinate relative importance or abundance of variables. Therefore, the RDA was applied to both response matrix Y (cranberry performance indices) and an explanatory matrix X (pH and N, P, K, Mg, Cu, Ca, Zn, Mn, Fe, Al, and C) and the results (Figure 6) were displayed using two triplot. The redundancy calculations were provided by multiple linear regression relating TAcy, Brix, Firmness, yield and berry moisture or weight in plot on the right to multiple explanatory variables (soil properties and pH) in left plot. See the Numerical ecology book of Legendre & Legendre 2012 for more details on the implementation and interpretation of RDA (Available at: https://www.elsevier.com/books/numerical-ecology/legendre/978-0-444-53868-0).

In order to make measurements comparable across samples, we computed the relative abundance of data sets (the soil nutrients and cranberry performance indices). We use the anova.cca function (https://www.rdocumentation.org/packages/vegan/versions/2.4-2/topics/anova.cca) and the log-ratio transformation (Aitchison, 1986; Filzmoser et al., 2009) in vegan package to evaluate the relative importance of soil nutrients expressed as log ratios of nutrients. We removed any reference to a minimum percentage. The relative abundance is the proportion or fraction of one component relative to total. For soil nutrient as interactive compositional system, relative abundance was expressed as log ratio allowing to compute Euclidean distance between two compositions, a requirement for RDA.

Appendix A1: It looks like the mean temperature value for December 2015 is erroneous. For monthly precipitation, use a bar graph, not a line graph.

There is no mistake in Appendix. Bar and line graphs are indeed readable options. Meteorological data (Supplementary Material S1) were obtained directly from the closest weather stations, which use the Meteorological Service of Canada's, Available online at: (https://climate.weather.gc.ca/historical_data/search_historic_data_e.html) archived temperature and precipitation. We used weathercan package (Available at: https://cran.r-project.org/web/packages/weathercan/weathercan.pdf) for historical weather data such as monthly precipitation or temperature.

Appendix 2: Indicate sample size (n) used in the calculation of mean and SD.

Sample size specifically refers to the number of plots (144) across four production cranberry field in each year.

The table shown as Appendix A3 should go into the main text. This information is critical toward understanding what was done.

We made a new table (Table 2). Fertilizer sources are provided in Appendix A2. Effects of Mg, Cu and B are presented in Appendix A3-A4.