Estimation of Biomass and N Uptake in Different Winter Cover Crops from UAV-Based Multispectral Canopy Reflectance Data
Round 1
Reviewer 1 Report
Brief summary:
The authors discusses the linear relationship between growth parameters (GAI, N uptake, and DM) sampled from cover crops and VIs derived from their canopy reflectance using UAV-based multispectral sensors. The prediction by a universal model based on all four cover crop species was better for GAI than for DM and N uptake and the best score was obtained by SRred. The species-individual models for SV and SO provided better results than the universal model. For the cover crop mixtures, there were slight differences between prediction performances by the mixture-individual and universal models.
General comments:
Novelty of this study is to estimate canopy parameters of cover crops on higher resolution using UAV‐based multispectral sensors. However, the analytical method is not sophisticated and the results cannot be interpreted clearly due to several sources of errors.
Aside from the scientific content of the study, there are many file conversion errors displaying "Error! Reference source not found." in this manuscript. These errors make the manuscript difficult to be reviewed correctly. I recommend the authors to resubmit the manuscript after revising completely.
Specific comments:
Line 145-146: The authors should explain the reason why they used the median of the extracted pixels as a representative value instead of the mean.
Line 179-181: The authors used adjusted r2 instead of r2 for performance comparison of the models. However, adjusted r2 is normally used for comparison of multiple regression models with different number of independent variables. The authors should explain the reason why they used adjusted r2 for their analysis.
Line 190-192: The original sentence in the template has to be removed.
Line 208-209: The authors should explain the reason why the data for cover crop mixture was restricted to DM and N uptake in 2018.
Line 279-281: The authors presented only differences of rMAE between the mixture-individual and universal models. The authors should show the original values of the rMAE and adjusted r2.
Line 349-358: The authors discussed the comparison of the species-individual model and the universal model only for SV. They should discuss the reasons why the species-individual models for SV and SO provided better results than the universal model and why the species-individual models for OR and WR were less accurate than the universal model.
Author Response
Dear reviewer,
Thank you very much for your critical feedback to my manuscript. I am sorry for the error messages. I renewed all references for tables and figures. Additionally, I changed the error value from adj.r² to R² in all figures and tables. In the following I address your comments:
Line 145-146: The authors should explain the reason why they used the median of the extracted pixels as a representative value instead of the mean.
Added the explanation to the aforementioned line.
Line 179-181: The authors used adjusted r2 instead of r2 for performance comparison of the models. However, adjusted r2 is normally used for comparison of multiple regression models with different number of independent variables. The authors should explain the reason why they used adjusted r2 for their analysis.
Revised the results and changed adj. r² to R², as it is the more suitable performance value for linear regressions.
Line 190-192: The original sentence in the template has to be removed.
Removed the sentence from the aforementioned line.
Line 208-209: The authors should explain the reason why the data for cover crop mixture was restricted to DM and N uptake in 2018.
The experimental design was not designed for this form of evaluation.
Line 279-281: The authors presented only differences of rMAE between the mixture-individual and universal models. The authors should show the original values of the rMAE and adjusted r2.
rMAE and R² for the mixture-individual models for DM and N uptake were added by Table 5.
Line 349-358: The authors discussed the comparison of the species-individual model and the universal model only for SV. They should discuss the reasons why the species-individual models for SV and SO provided better results than the universal model and why the species-individual models for OR and WR were less accurate than the universal model.
Figure 8 was subsequently included to compare the estimations of species-individual models and universal models for the sole cover crops canopy parameters. Discussion was adapted accordingly.
Reviewer 2 Report
Manuscript for revision has a lot of Error! Reference source not found it must be properly completed, these errors make reading manuscript very difficult without proper references to tables and figures manuscript can not be accepted. I have marked it in Overall Recommendation !
Line 117 2.2. Trial designs
what plant was a forecrop and how was the nitrogen dose, how far from experiment date manure (or organic fertilizer) was added?
2.4. Plant sampling
line 151 When plant material was dug out did the authors check plant density of sown plants species on 0,5m2 fields real and theoretical ?
Line 159 and 169 Does it mean that senescent material DM, GAI N uptake was not taken into modelling?, If yes Author need to take into account that part of nitrogen was accumulated in that part of plants and can be a storage of N and its source for a succeeding corp - Line 36.
Line 195 at frost events how was the temprature?
line 201 senescent percentage of WR was 45,7% after frost events rye is one of the most frost resistant plants how the author explain that comparing to the rest used plants
Figure 3 comparing winter rye to the spring plants (non - frost resistant) in my opinion is not good idea due to significantly different dry mass accumulation rate
4. Discussion
Spring vetch can cause some errors due to biologically nitrogen fixation which amount (concentration, contents) can differ significantely year to year it depends on microbioligical activity. Highest r2 values measured vs predicted N for SV can suggest that part of accumulated nitrogen could be atmosphere origin not from fertilizers or soil.
Author Response
Dear reviewer,
Thank you very much for your feedback to my manuscript. I am very sorry for the error messages in the pdf document. I renewed all the references of tables and figures. In the following I address each of your comments:
Manuscript for revision has a lot of Error! Reference source not found it must be properly completed, these errors make reading manuscript very difficult without proper references to tables and figures manuscript cannot be accepted. I have marked it in Overall Recommendation!
References to tables and figures were corrected.
Line 117 2.2. Trial designs
what plant was a forecrop and how was the nitrogen dose, how far from experiment date manure (or organic fertilizer) was added?
Forecrop and nitrogen dose is mentioned in the named chapter. Further fertiliser information was not necessary for the objective of the study.
2.4. Plant sampling
line 151 When plant material was dug out did the authors check plant density of sown plants species on 0,5m2 fields real and theoretical?
No, there was no measurement of the plant density.
Line 159 and 169 Does it mean that senescent material DM, GAI N uptake was not taken into modelling. If yes Author need to take into account that part of nitrogen was accumulated in that part of plants and can be a storage of N and its source for a succeeding crop - Line 36.
Senescent material was measured. As the sensor can only see the green vegetation, only green vegetation was included into the calibration data set.
Line 195 at frost events how was the temperature?
Temperature of the frost events were added to the aforementioned line.
line 201 senescent percentage of WR was 45,7% after frost events rye is one of the most frost resistant plants how the author explains that comparing to the rest used plants
WR-individual model performed the lowest, compared to the other species individual models. This may be traced back to the high amount of senescent material, which disturbs the spectral reflectance signal.
Figure 3 comparing winter rye to the spring plants (non - frost resistant) in my opinion is not good idea due to significantly different dry mass accumulation rate
The experimental design of the project was not developed for the current study.
- Discussion
Spring vetch can cause some errors due to biologically nitrogen fixation which amount (concentration, contents) can differ significantly year to year it depends on microbiological activity. Highest r2 values measured vs predicted N for SV can suggest that part of accumulated nitrogen could be atmosphere origin not from fertilizers or soil.
Actually, SV was strongly suppressed by voluntary oilseed rape. The result was the best, but the least trustworthy. This was also mentioned in the discussion.
Reviewer 3 Report
In this research, the authors tried to build some simple empirical models to estimate GAI, N, and DM from UAV multispectral data in Kiel, Northern Germany. In general, this is solid research. The authors did some interesting exploration on universal model, species individual models, and mixture individual models. The results are informative and encouraging. The manuscript is well written and organized. I think it just need a minor revision. Here are some detailed problems.
It seems that a lot of figure reference links are failed in the current version, like those in lines 114, 120, 123... Please correct them.
Line 35. “1,5-8”. Should it be “[1,5-8]”?
Line 129. In section 2.3, please also provide more detailed spectral information such as band width, SNR, quantitative levels of each band, which is important for remote sensing estimation missions.
Line 240. All the estimation models are linearly in table 4. But I notice that the relationships between the NDred predicted GAI and measured GAI is clearly nonlinear in figure 5. Considering the distribution of the samples, I suggest to try some simple transforms (such as logarithmic transformation) on the dependent variable and then do the linear fitting. I believe the results will be much better than the current form.
Line 243. The authors used full names of different crops in this figure. I suggest to use abbreviations that been defined in the text.
Line 248. Figure 6 is very interesting. I suggest to do more discussion about the changing of model performance in different time/growing period. For instance, the authors can also calculate the difference rMAE that been illustrated in figure 8 in different period. The results will provide more specific suggestions of model selection and better support the potential of universal model, which is much more useful in remote sensing monitoring missions. I think a general suggestion about the latest time of using general model is interesting.
Author Response
Dear reviewer,
Thank you very much for you positive feedback and comments on my manuscript. I am very sorry for the error messages in the pdf-document. I renewed all references of tables and figures. In the following, I address each of your comments:
It seems that a lot of figure reference links are failed in the current version, like those in lines 114, 120, 123... Please correct them.
References to tables and figures were corrected.
Line 35. “1,5-8”. Should it be “[1,5-8]”?
Corrected
Line 129. In section 2.3, please also provide more detailed spectral information such as band width, SNR, quantitative levels of each band, which is important for remote sensing estimation missions.
Band width and mid-wavelength of the four spectral bands were mentioned in the named section.
Line 240. All the estimation models are linearly in table 4. But I notice that the relationships between the NDred predicted GAI and measured GAI is clearly nonlinear in figure 5. Considering the distribution of the samples, I suggest to try some simple transforms (such as logarithmic transformation) on the dependent variable and then do the linear fitting. I believe the results will be much better than the current form.
NDred or commonly known as NDVI was often stated, as not sensitive to higher values of GAI. Growth of cover crops reaches the critical point of NDred, which leads to exclusion of the VI for the further validations.
Line 243. The authors used full names of different crops in this figure. I suggest to use abbreviations that been defined in the text.
Sole cover crops are named by their full name in all figures, to make it easier for the reader to keep an overview about the variety of the species.
Line 248. Figure 6 is very interesting. I suggest to do more discussion about the changing of model performance in different time/growing period. For instance, the authors can also calculate the difference rMAE that been illustrated in figure 8 in different period. The results will provide more specific suggestions of model selection and better support the potential of universal model, which is much more useful in remote sensing monitoring missions. I think a general suggestion about the latest time of using general model is interesting.
Agreed. Discussion was further elaborated.
Round 2
Reviewer 1 Report
The manuscript has been revised well. I think this manuscript is acceptable now.
Reviewer 2 Report
Accepted in present form
