Estimation of Cadmium Content in Lactuca sativa L. Leaves Using Visible–Near-Infrared Spectroscopy Technology
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe main objective at the end of introduction.
Authors need to say what are the advantages of using visible-near infrared spectral technology, what is new or different to Li et al. [17]. Li et al demonstrated that is possible to monitor cadmium content on lettuce. Why to repeat the experiment?
Methodology
To repeat the experiment.
It is not clear how authors ensure the cadmium content on soil?
Spayed layer by layer, how? The soil was put inside the pot in layers? Outside in layers and then put inside the pots?
“Sufficient water supply” It does not indicate quantities. How much, exactly.
Were plants produce under greenhouse or in open field?
Conclusion
Why not to compared the results of these final models with the results obtained by digestion method?
See comments in MS
Comments for author File: 
Comments.pdf
Minor editing of English language required
Author Response
Dear editor and expert:
Thank you for your detailed and important comments on our manuscript. After receiving the review comments, all eight authors of the manuscript have carefully analyzed and discussed the review comments. We believe that these review comments are very constructive and have therefore completed the revisions one by one.
The specific revisions are described below (review comments are in red, responses and revisions are in black)：
Comments 1: The main objective at the end of introduction.
Response 1: Thank you for pointing this out. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We agree with this comment. We have included the research objectives in the final section of the introduction to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 2, paragraph 3, and line 81-89.
Comments 2: Authors need to say what are the advantages of using visible-near infrared spectral technology, what is new or different to Li et al. [17]. Li et al demonstrated that is possible to monitor cadmium content on lettuce. Why to repeat the experiment?
Response 2: Agree. Thank you for raising this issue. We have taken note of this issue. Conducting cadmium pollution research on vegetables such as lettuce using visible-near infrared spectroscopy technology not only enables rapid, accurate, non-destructive detection but also offers advantages such as ease of operation, environmental friendliness, and multifunctionality. It is a highly promising modern analytical technique. The experiment conducted by LI et al. differs from this study in several aspects. Firstly, the variety of test materials used by LI et al. differs from the current experiment; this study utilized "American fast lettuce" while LI et al. employed "European endive," "American endive," and "Qianyu No.1." Secondly, LI et al. utilized a hydroponic experiment, whereas this study employed a soil-based experiment. Additionally, LI et al. used cadmium chloride, while this study used cadmium nitrate. Moreover, in terms of data processing, the spectral preprocessing methods and dimensionality reduction techniques applied in this study differ from those used by LI et al. Consequently, this experiment was conducted based on the research done by LI et al. Mention exactly where in the revised manuscript this change can be found – page 2, paragraph 3, and line 84-89.
Comments 3: It is not clear how authors ensure the cadmium content on soil?
Response 3: Agree. Thank you for pointing this out. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We agree with this comment. The cadmium nitrate solution concentration used in this study was 1000 μg/mL (equivalent to 1 mg/mL). Taking the control of soil cadmium content at 1 mg/kg as an example, for 1.5 kg of soil to achieve a cadmium concentration of 1 mg/kg, a total of 1.5 mg of cadmium needs to be added. With the existing cadmium nitrate solution concentration of 1 mg/mL, to obtain 1.5 mg of cadmium, 1.5 mL of 1000 μg/mL (1 mg/mL) cadmium nitrate solution needs to be added to the soil to ensure that the cadmium concentration in the soil reaches 1 mg/kg. The same applies to other treatments.
Comments 4: Spayed layer by layer, how? The soil was put inside the pot in layers? Outside in layers and then put inside the pots?
Response 4: Agree. Thank you for your feedback. We apologize deeply for our oversight. The spraying was done layer by layer outside the pots: First, prepare the soil by sieving and removing impurities to obtain fine particles for subsequent processing. Then, spray layer by layer. On a large, clean container or surface, lay down a layer of soil (approximately 2.5 cm thick). Next, uniformly spray the pre-prepared cadmium nitrate solution onto this layer of soil using a spray bottle. Then, proceed with mixing and stacking. After spraying, gently mix this layer of soil to ensure even distribution of the solution. Then, lay down the next layer of soil and repeat the steps of spraying and mixing until all the soil has been treated. Finally, transfer the soil to the flower pots. After treating all the soil and allowing it to mature, transfer it to the flower pots.
Comments 5: “Sufficient water supply” It does not indicate quantities. How much, exactly.
Response 5: Agree. Thank you for raising this issue. We have taken note of this issue. We have supplemented the watering amount to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3, and line 112.
Comments 6: Were plants produce under greenhouse or in open field?
Response 6: Agree. Thank you for raising this issue. We have taken note of this issue. We have supplemented the growing conditions to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3, and line 114-116.
Comments 7: Obtained by digestion method??
Response 7: Agree. Thank you for your valuable suggestions, and we sincerely apologize for any confusion in the content of our article. Figure 2 is obtained based on the digestion method.
Comments 8: Why not to compared the results of these final models with the results obtained by digestion method?
Response 8: Agree. Thank you for your valuable suggestions, and we sincerely apologize for any confusion in the content of our article. Using visible-near infrared spectroscopy, this study aims to predict the cadmium content in lettuce leaves. The digestion method is employed to determine the cadmium content in lettuce leaves, which serves as the output variable. The input variables consist of selected wavelength bands. The dataset is divided into a training set and a test set, and partial least squares, BP neural network, and support vector regression models are established for prediction. The final results are compared with the cadmium content obtained by the digestion method, using the error between the predicted values and the actual digestion-based measurements of cadmium content. Error analysis involves calculating the root mean square error (RMSE) between the predicted values and the actual values to evaluate the prediction accuracy of the models. By comparing the errors between the predicted values and the digestion-based measurements, the predictive capability of the models can be assessed, along with their consistency and accuracy in comparison to the digestion-based values.
Comments 9: As keywords use different words than that in title
Response 9: Agree. Thank you for your feedback. We apologize deeply for our oversight. We have amended the keywords to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 1, and line 33-34.
Point 1: Minor editing of English language required
Response 1: Agreed. Thank you for bringing this to our attention. According to your valuable suggestions, we have thoroughly reviewed the article and made the necessary modifications. Mention exactly where in the revised manuscript this change can be found – page 2, and line 95.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsDear Authors,
I have carefully revised your manuscript titled “Estimation of Cadmium Content in Lettuce (Lactuce sativa L.) Leaves Using Visible-Near Infrared Spectroscopy Technology”. The purpose of the article is to evaluate the quantity of cadmium present in lettuce leaves grown through a pot cultivation method, using soil with different levels of cadmium content. The cadmium content in the leaves was determined by using both VIS-NIR spectroscopy and digestion methods. Despite being an interesting subject for Agronomy readers, there are a few issues and inaccuracies that need to be resolved before the article can be published.
1) The issue of cadmium contamination is not limited to China; it's a global problem. Therefore, the introduction needs to be rephrased to appeal to a broader audience.
2) Based on only 135 spectra (because we do not know how many individual lettuces were used to obtain these spectra), the models using spectral data seem to be unreliable. The authors do not include the input spectral data in the Appendix, so it is difficult to verify the correctness of the results. However, fitting a model with r2= 0.99 and 1.00 even in the calibration stage (e.g. table 2 SG+D seems unrealistic to me). It looks like overfitting with the PLS regression method. Maybe too many principal components were used.
The Materials and Methods section needs to be improved, especially the 2.2 and 2.4 sections. Often, references are provided for the pretreatments and the statistical/machine learning methods.
3) It is important to provide information about the number of samples used for model calibration and validation, rather than just the sample ratio. How the dataset was split in training and testing (calibration/validation datasets)? Random? Kennard-Stone algorithm? Or by ranging the Cd content and taking every 3 samples for validation?
4) What was the basis for determining the cadmium content in soil to be 1, 5, 10, and 20 mg/kg? It seems to be a high content. Would the models work with lower cadmium contents?
5) It appears to me that there might be an error in the measurements of the pots and the quantity of soil utilized for the experiments. Can you please confirm if the pots, which measure 480 cm x 230 cm x 160 cm, only contained 1.5 kg of soil?
6) There is no information available about the soil type, experimental conditions (indoor or outdoor), number of irrigation doses, meteorological conditions, etc. Add photos of the experiment.
7) The cadmium content in lettuce leaves was measured using the digestion method. Can you provide relevant literature citations for this method?
8) In the Materials and Methods section, please include information about the ANOVA presented in Figure 1, including the post-hoc test used.
9) Please make sure to use a space to separate values and their respective units throughout the text.
10) To properly format the Latin name of a species, use italics.
11) Could you kindly clarify what the colors represent in Figures 3 and 4? Also, please add suitable legends to the figures for better understanding.
12) In my opinion, the obtained results have not been discussed. E.g. why is PLS working better? Correlation of Cd with leaf constituents having direct spectral response, etc.
Reviewer
L105: 45 days is quite long I suppose for a lettuce? Would have been much better to scan also the leaves a few weeks before.
L108: how many data points (modes)?
L111: not sure if the measurements were outside or inside a greenhouse
L118-119: it is not clear how many single lettuces were analysed to obtain 135 spectra.
In section 2.4, add companies and countries info for each software.
L163: what do you mean by spectral fluctuations?
L167-171: Under different concentrations of cadmium stress, … this is in fact typical for every king of leaves, stress has nothing to do with that.
L176-178: reformulate
Figure 2: Reflectance unit? (also Fig. 3) Is that raw spectra? There is a lot of noise.
In Tables, the number of calibration/validation used could be added. For the PLS, the number of PC
Fig. 4. These graphs are nice but are they necessary? If yes, please explain for the reader how optimal c, g, MSE attributes are found in the planes/curves.
Again, there is no proper discussion were e.g. results are discussed with regard to the literature.
Author Response
Dear editor and expert:
Thank you for your detailed and important comments on our manuscript. After receiving the review comments, all eight authors of the manuscript have carefully analyzed and discussed the review comments. We believe that these review comments are very constructive and have therefore completed the revisions one by one.
The specific revisions are described below (review comments are in red, responses and revisions are in black)：
Comments 1: The issue of cadmium contamination is not limited to China; it's a global problem. Therefore, the introduction needs to be rephrased to appeal to a broader audience.
Response 1: Thank you for pointing this out. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We agree with this comment. We have supplemented the cadmium pollution situation in other countries to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 1, paragraph 2, and line 37-44.
Comments 2: Based on only 135 spectra (because we do not know how many individual lettuces were used to obtain these spectra), the models using spectral data seem to be unreliable. The authors do not include the input spectral data in the Appendix, so it is difficult to verify the correctness of the results. However, fitting a model with r2= 0.99 and 1.00 even in the calibration stage (e.g. table 2 SG+D seems unrealistic to me). It looks like overfitting with the PLS regression method. Maybe too many principal components were used.
Response 2: Agree. Thank you for raising this issue. We have taken note of this issue. The source of the 135 spectra is as follows: a total of 5 concentration gradients were set up, with 3 pots of plants planted at each concentration level. Each pot had 3 lettuce plants planted vertically, resulting in 9 lettuce plants per treatment and a total of 45 lettuce plants. Then, 3 spectral data points were measured for each lettuce plant, leading to 135 spectral data points in total.
In machine learning, overfitting is typically characterized by the model performing well on the training set (R2 close to 1, low RMSE), but poorly on the test set (low Rp2, high RMSEp).
Looking at the results of this study: the R2 on the training set is 1.00, with an RMSE of 0.36, indicating excellent performance. On the test set, Rp2 is 0.89, with an RMSEp of 1.71, showing relatively good performance but slightly lower than the training set.
Overall, while the performance on the test set is slightly lower than on the training set, there are no clear signs of overfitting. Therefore, it can be preliminarily concluded that there is no significant overfitting in these results.
Comments 3: The Materials and Methods section needs to be improved, especially the 2.2 and 2.4 sections. Often, references are provided for the pretreatments and the statistical/machine learning methods.
Response 3: Agree. Thank you for pointing this out. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We agree with this comment. We have already explained the preprocessing and machine learning methods in sections 2.2 and 2.4 of the Materials and Methods, and we have supplemented the references to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3-5, and line 137-143 and 175-195.
Comments 4: It is important to provide information about the number of samples used for model calibration and validation, rather than just the sample ratio. How the dataset was split in training and testing (calibration/validation datasets)? Random? Kennard-Stone algorithm? Or by ranging the Cd content and taking every 3 samples for validation?
Response 4: Agree. Thank you for your feedback. We apologize deeply for our oversight. The training set of this study consists of 90 data samples, while the test set comprises 45 samples, which were randomly partitioned.
Comments 5: What was the basis for determining the cadmium content in soil to be 1, 5, 10, and 20 mg/kg? It seems to be a high content. Would the models work with lower cadmium contents?
Response 5: Agree. Thank you for raising this issue. We have taken note of this issue. The selection of cadmium content in soil is based on the following criteria: According to the "Soil Environmental Quality Standards" (GB15618-2008), in agricultural land, the second-level quality standard value for total cadmium in soil is up to 1.0 mg/kg. Specifically, in vegetable fields with a pH > 7.5, the second-level quality standard value for total cadmium is 0.6 mg/kg. As per the "Pollutant Limits in Food" (GB2762-2012), the cadmium limit in leafy vegetables is 0.2 mg/kg. Additionally, based on recovery tests, the detection limit is generally set at 3-5 times, resulting in the final soil cadmium content being set at 1 mg/kg and 5 mg/kg. Furthermore, according to a survey, the range of cadmium environmental standards in 32 agricultural soils varies from 0.2 to 20.0 mg/kg. Therefore, concentrations are set at 0 (for control), 1, 5, 10, and 20 mg/kg.
This model should also be applicable to lower cadmium concentrations. When it comes to lower cadmium concentrations, the model may have the potential for adaptation. By making appropriate adjustments, the model can be attempted to be applied in studies involving lower cadmium concentrations. However, when applying the model to research on lower cadmium concentrations, validation and optimization work are still required to ensure the accuracy and reliability of the model. This is one of the aspects that need further improvement in future research endeavors.
Comments 6: It appears to me that there might be an error in the measurements of the pots and the quantity of soil utilized for the experiments. Can you please confirm if the pots, which measure 480 cm x 230 cm x 160 cm, only contained 1.5 kg of soil?
Response 6: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. The pot size is 480 mm*230 mm*160 mm. We initially estimated that such a pot could hold approximately 1.5 kg of substrate soil, then accurately measured the aged soil and transferred it into the pot. We have amended this section in the manuscript to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3, and line 106-107.
Comments 7: There is no information available about the soil type, experimental conditions (indoor or outdoor), number of irrigation doses, meteorological conditions, etc. Add photos of the experiment.
Response 7: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We have added content to this section to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3-4, and line 114-116 and 144.
Comments 8: The cadmium content in lettuce leaves was measured using the digestion method. Can you provide relevant literature citations for this method?
Response 8: Agree. Thank you for raising this issue. We have taken note of this issue. We have already supplemented the references to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 4, and line 148.
Comments 9: In the Materials and Methods section, please include information about the ANOVA presented in Figure 1, including the post-hoc test used.
Response 9: Agree. Thank you for raising this issue. We have taken note of this issue. We have already supplemented the content of this section to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 4, and line 170-172.
Comments 10: Please make sure to use a space to separate values and their respective units throughout the text.
Response 10: Agree. Thank you for raising this issue. We have taken note of this issue. We have reviewed the entire document and added spaces to emphasize this point.
Comments 11: To properly format the Latin name of a species, use italics.
Response 11: Agree. Thank you for raising this issue. We have taken note of this issue. We have made modifications to address this issue. Mention exactly where in the revised manuscript this change can be found – page 2, and line 95.
Comments 12: Could you kindly clarify what the colors represent in Figures 3 and 4? Also, please add suitable legends to the figures for better understanding.
Response 12: Agree. Thank you for raising this issue. We have taken note of this issue. The colors in Figures 3 and 4 do not have specific meanings. We have updated the appropriate legend to emphasize this point.
Comments 13: In my opinion, the obtained results have not been discussed. E.g. why is PLS working better? Correlation of Cd with leaf constituents having direct spectral response, etc.
Response 13: Agree. Thank you for raising this issue. We have taken note of this issue. We have added a discussion to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 11, and line 367-381.
Comments 14: L105: 45 days is quite long I suppose for a lettuce? Would have been much better to scan also the leaves a few weeks before.
Response 14: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. The lettuce variety used in this study is the American fast lettuce, which has a growth cycle of 6-8 weeks. Around 45 days marks the harvesting period, during which scanning the lettuce leaves can aid in identifying potential diseases present on the lettuce. This process is crucial for ensuring food safety and quality control.
Comments 15: L108: how many data points (modes)?
Response 15: Agree. Thank you for raising this issue. We have taken note of this issue. We have supplemented the number of data points to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3, and line 125-126.
Comments 16: L111: not sure if the measurements were outside or inside a greenhouse
Response 16: Agree. Thank you for raising this issue. We have taken note of this issue. We have supplemented the growing conditions to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 3, and line 114-116.
Comments 17: L118-119: it is not clear how many single lettuces were analysed to obtain 135 spectra.
Response 17: Agree. Thank you for raising this issue. We have taken note of this issue. The source of the 135 spectra is as follows: a total of 5 concentration gradients were set up, with 3 pots of plants planted at each concentration level. Each pot had 3 lettuce plants planted vertically, resulting in 9 lettuce plants per treatment and a total of 45 lettuce plants. Then, 3 spectral data points were measured for each lettuce plant, leading to 135 spectral data points in total.
Comments 18: In section 2.4, add companies and countries info for each software.
Response 18: Agree. Thank you for raising this issue. We have taken note of this issue. We have added the country of origin and formulas of the software to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 4, and line 163-168.
Comments 19: L163: what do you mean by spectral fluctuations?
Response 19: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. Here, it refers to the spectral curve fluctuations, and we have rephrased this sentence to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 6, and line 221.
Comments 20: L167-171: Under different concentrations of cadmium stress, … this is in fact typical for every king of leaves, stress has nothing to do with that.
Response 20: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We have already omitted this sentence to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 6, and line 225.
Comments 21: L176-178: reformulate
Response 21: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We have rephrased this paragraph to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 6, and line 231-237.
Comments 22: Figure 2: Reflectance unit? (also Fig. 3) Is that raw spectra? There is a lot of noise.
Response 22: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We have already added the units to emphasize this point. Figure 2 represents the raw spectra, which require a series of preprocessing steps to enhance the signal-to-noise ratio due to significant noise. Mention exactly where in the revised manuscript this change can be found – page 6-7, and line 248 and 262.
Comments 23: In Tables, the number of calibration/validation used could be added. For the PLS, the number of PC
Response 23: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We have added this information to the table to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 8-9 and 11, and line 305, 322, 360 and 382.
Comments 24: Fig. 4. These graphs are nice but are they necessary? If yes, please explain for the reader how optimal c, g, MSE attributes are found in the planes/curves.
Response 24: Agree. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. We have added more description to this section to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 9-10, and line 330-339.
Comments 25: Again, there is no proper discussion were e.g. results are discussed with regard to the literature.
Response 25: Agree. Thank you for raising this issue. We have taken note of this issue. We have added a discussion to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 11, and line 367-381.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsDear Authors,
Thank you for your efforts to improve the manuscript. However, some of the comments were left unaddressed. It would be great if you could address them in the next revision.
General comments:
1. Still model results with r2= 0.99 and 1.00 even in the calibration stage seem unrealistic to me.
2. Water lettuce daily to ensure the soil's moisture content remains between 60 % and 70 %.
What was the measurement method used? It appears to be unrealistic to me.
3. Lactuca sativa L. is correct, so please correct it in the manuscript, also in the title of the article.
4. How the dataset was split in training and testing (calibration/validation datasets)? Random? Kennard-Stone algorithm? Or by ranging the Cd content and taking every 3 samples for validation?...
5. What was the basis for determining the cadmium content in soil to be 1, 5, 10, and 20 mg/kg? It seems to be a high content. Would the models work with lower cadmium contents?
6. There is no information available about the soil type, number of irrigation doses, meteorological conditions.
7. Could you kindly clarify what the colors represent in Figures 3 and 4? Also, please add suitable legends to the figures for better understanding.
8. Fig. 4. These graphs are nice but are they necessary? If yes, please explain for the reader how optimal c, g, mse attributes are found in the planes/curves.
Minor comments:
L83: Delete “among”
L185: Delete “exceedingly”
Please make sure to use a space to separate values and their respective units throughout the text.
Author Response
Dear editor and expert:
Thank you for your detailed and important comments on our manuscript. After receiving the review comments, all eight authors of the manuscript have carefully analyzed and discussed the review comments. We believe that these review comments are very constructive and have therefore completed the revisions one by one.
The specific revisions are described below (review comments are in red, responses and revisions are in black)：
Comments 1: Still model results with r2= 0.99 and 1.00 even in the calibration stage seem unrealistic to me.
Response 1: Thank you for pointing this out. We deeply apologise for this, due to our negligence, your understanding of the article has brought you a bad experience. To elucidate the issue with the PLSR model as an example, firstly, it's imperative to recognise that the PLSR model integrates the characteristics of multiple linear regression analysis, principal component analysis, and canonical correlation analysis. It has the capability to eliminate multicollinearity among independent variables, thereby averting the occurrence of overfitting within the model. During the modelling phase, the data utilised are actual measurements, ensuring the data's authenticity and dependability. By way of illustration, employing the SG+D preprocessing method, the outcomes of modelling via PLSR are as follows.
Comments 2: Water lettuce daily to ensure the soil's moisture content remains between 60 % and 70 %. What was the measurement method used? It appears to be unrealistic to me.
Response 2: Agree. Thank you for raising this issue. We have taken note of this issue. The narrative provided might lead to a certain misunderstanding; it involves watering lettuce daily to ensure that the soil moisture remains between 60% and 70%. The measurement of this moisture level is conducted using the Y315 soil temperature, moisture, and pH triple-function meter. The equipment used is depicted in the image below. Mention exactly where in the revised manuscript this change can be found – page 3, and line 114-117.
Comments 3: Lactuca sativa L. is correct, so please correct it in the manuscript, also in the title of the article.
Response 3： Agree. Thank you for your feedback. We apologize deeply for our oversight. We have made the necessary adjustments in the manuscript to emphasise this point. Mention exactly where in the revised manuscript this change can be found – page 1-2, and line 2 and 95.
Comments 4: How the dataset was split in training and testing (calibration/validation datasets)? Random? Kennard-Stone algorithm? Or by ranging the Cd content and taking every 3 samples for validation?...
Response 4：Agree. Thank you for raising this issue. We have taken note of this issue. The training and test sets were divided in a 2:1 ratio, wherein the division was based on the range of cadmium content, with one sample being randomly selected as a validation sample from every three samples. Mention exactly where in the revised manuscript this change can be found – page 8, and line 293-295.
Comments 5: What was the basis for determining the cadmium content in soil to be 1, 5, 10, and 20 mg/kg? It seems to be a high content. Would the models work with lower cadmium contents?
Response 5：Agree. Thank you for raising this issue. We have taken note of this issue. The concentration levels were set at 1, 5, 10, and 20 mg/kg, based on the variation range of 0.2 to 20.0 mg/kg for 32 cadmium environmental standard values in agricultural land soil obtained from a survey, following a principle of incremental increase.
The model was established based on the data measured in this experiment, resulting in a model that is real, accurate, and reliable. The gradient of cadmium concentrations used in the experiment was set to 1, 5, 10, and 20 mg/kg. As to whether the model is applicable to lower cadmium concentrations, further development and utilisation of the model are required. This also represents one of the issues we need to address in future research. Mention exactly where in the revised manuscript this change can be found – page 3, and line 104-108.
Comments 6: There is no information available about the soil type, number of irrigation doses, meteorological conditions.
Response 6：Agree. Thank you for your feedback. We apologize deeply for our oversight. The soil type is a full-value seedling substrate, consisting of pure natural black nutrient peat. Irrigation is carried out daily to ensure that the soil moisture remains between 60% and 70%. The meteorological conditions consist of daytime and nighttime temperatures at 25°C/18°C ±2°C, with a relative humidity of 60% to 70%. The environmental temperature and humidity are measured using the DL9010 thermo-hygrometer (Deli, China). Mention exactly where in the revised manuscript this change can be found – page 2-3, and line 96-98 and 114-122.
Comments 7: Could you kindly clarify what the colors represent in Figures 3 and 4? Also, please add suitable legends to the figures for better understanding.
Response 7：Agree. Thank you for raising this issue. We have taken note of this issue. The different colored lines in Figure 3 represent different cadmium stress concentrations, as explained in the legend. In Figure 4, the colors do not have specific meanings. The figure consists of 135 spectral curves under different preprocessing methods, mainly to observe the variations in the spectral curves due to different preprocessing methods; hence, there is no need to add a legend.
Comments 8: Fig. 4. These graphs are nice but are they necessary? If yes, please explain for the reader how optimal c, g, mse attributes are found in the planes/curves.
Response 8： Agree. Thank you for your feedback. We apologize deeply for our oversight. We have replaced the figure with a table to emphasize this point. Mention exactly where in the revised manuscript this change can be found – page 10, and line 343.
Comments 9: L83: Delete “among”
Response 9：Agree. Thank you for raising this issue. We have taken note of this issue. We have deleted 'among' to emphasise this point. Mention exactly where in the revised manuscript this change can be found – page 2, and line 85.
Comments 10: L185: Delete “exceedingly”
Response 10：Agree. Thank you for raising this issue. We have taken note of this issue. We have deleted 'exceedingly' to emphasise this point. Mention exactly where in the revised manuscript this change can be found – page 5, and line 194.
Comments 11: Please make sure to use a space to separate values and their respective units throughout the text.
Response 11: Agree. Thank you for your feedback. We apologize deeply for our oversight. We have reviewed the entire text and made the necessary adjustments to emphasise this point.
Author Response File: Author Response.pdf
