MangiSpectra: A Multivariate Phenological Analysis Framework Leveraging UAV Imagery and LSTM for Tree Health and Yield Estimation in Mango Orchards
, Muhammad Shahid Iqbal 2, Asim Dilawar Bakhshi 3, Ejaz Hussain 1 and Javed Iqbal 1
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis manuscript presents the workflow of MangiSpectra and multi-temporal UAV data for monitoring and mapping the health and yield of Mango orchards. The results were high accuracy for monitoring health and low accuracy for estimated mango yield. However, several parts need improvement.
1. In the introduction section, Table 1 lacks essential information such as data used, study area and accuracy results. Also, the authors should summarise the main details of these previous studies. In addition, the last paragraph, as the objectives, is still unclear to me.
2. The parameter setting of UAV observation must improve and provide more details. The workflow of data preprocessing is very hard to understand when implemented in Pix4D software. You need to add the information.
3. Why have you provided Figure 6 in the method part? Is it the result?
4. For your predictive models, why do you not use the CHM data for mapping and monitoring?
5. Figure 8 should be moved to the result part.
6. As the essential thing of this study is estimating mango yield. However, the authors did not provide details about the observed yields. This dataset is one of the significant pieces of information needed to construct the predictive model. Moreover, how do you spit the data for training and validating?
7. The statistical RMSE values are generally missing for model and result validation.
8. The estimated yield in this study was very low accuracy (R2 and RMSE). What is the way to improve it? Is it beneficial for the next studies?
9. The discussion should be improved, and details of the talks should be provided. The related studies did not compare and mention.
10. Is it possible to combine the UAV and satellite data for this implemented workflow?
11. Does crop management, such as cutting down on the mango leaf, have effects on your predictive models?
12. The constructed models in this study can transfer to other fields and other crops. Why?
13. In conclusion, you need to revise and improve. I recommend you provide the main findings and further recommendations for future studies. Your main study is missing.
14. The references are missing in several parts of this paper. It would help if you improved it.
Comments on the Quality of English LanguageThe English Language is required to improve.
Author Response
Thank you for your suggestions and observations which have improved our manuscript. We have addressed all the observations and updated the manuscript accordingly. The detailed response is attached as a separate file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript presents an innovative approach leveraging UAV imagery and LSTM for health classification and yield estimation of mango orchards. The research methodology is robust, and the findings are well-validated with experimental data. The manuscript is generally well-structured; however, some grammatical errors, formatting inconsistencies, and clarity issues need to be addressed. Once these issues are addressed, the paper will be in a strong position for publication.
English rating 7/10. English is alright, but certain sections contain awkward phrasing and redundancy, some are complex and can be simplified for better readability, certain terminology used may need further explanations to add more clarity. Minor language editing is required to improve clarity and readability.
Introduction – overall need improvement to be more concise about the interest and meet the study objective.
Detail comments as below;
Line 1-4- "However, the large canopy size, complex tree morphology, and unique phenological properties of mango trees limit the effectiveness of low-resolution satellite imagery for health and yield estimation." Not sure the used of word properties here is suitable in this aspect. May be change to characteristics
Methods and data – need to present the collected data in the summary table including the no of images, samples and range of the yield data
Line 10- grammar on have been change to were developed to quantify the degree of canopy yellowness…
Line 17- abstract. RMSE 0.18 or 50.18? which one?
Line 44-47- These "This variability complicates the monitoring regime due to spectral mixing on one hand and distinct responses in phenological stages on the other. Finally, there are environmental strains such as temperature fluctuations, wind, drought, and pest infestations impacting tree health and yield." May be authors can consider to add the factor of the planting distance used, the canopy covered area and the other such as pruning activities may effect on the canopy structure shape.
Table 1- CNN-LSTM abbreviations were not firstly introduce in the abstract section. Please cross check to all the short name used.
Figure 1: The study area map needs a more descriptive caption. Are these 4 different areas? What are their names? Location, total area? Total no of trees?
Figure 5: the captions can be improved by putting the full name of the MTYI etc.
Figure 9- missing the ref for Fig 9a-d in the images.
Figure 10- captions used can be improved. Legend font size can be improved vs Figure 12
Figure 11- recommended to standardize the range of the Y- scale to be similar for all four graphs. Missing the unit for the yield
Equation 2,3,4 ,5 (section 4.11 and so on) – need better explanation for the symbols used and the reference no for the equations used.
Comments on the Quality of English LanguageEnglish rating 7/10. English is alright, but certain sections contain awkward phrasing and redundancy, some are complex and can be simplified for better readability, certain terminology used may need further explanations to add more clarity. Minor language editing is required to improve clarity and readability.
Author Response
Thank you for your suggestions and observations which have improved our manuscript. We have addressed all the observations and updated the manuscript accordingly. The detailed response is attached as a separate file.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe manuscript presents a comprehensive study on the use of the MangiSpectra framework for monitoring mango orchard health and yield using UAV imagery combined with advanced machine learning techniques, such as LSTM models. The authors successfully introduce innovative indices (MTYI, WYI, and NAFDI) tailored for mango canopy analysis and demonstrate their applicability in phenological health assessments and yield predictions. The integration of high-resolution imagery with temporal and spatial data, as well as the effort to address phenological complexities, is commendable. The study contributes to the field of precision agriculture by offering a near-real-time solution to orchard management challenges.
However, the manuscript would benefit from addressing the following points to improve clarity and rigor:
1.Challenges in Capturing Underlying Tree Conditions
The manuscript notes that traditional remote sensing methods struggle with capturing the complexity of mango canopies due to overlapping leaves, varying phenological responses, and environmental factors. While UAV imagery offers higher spatial and temporal resolution, it is not immune to similar issues, such as canopy occlusion and limited penetration to underlying layers. It would strengthen the manuscript if the authors explicitly discussed how MangiSpectra mitigates these limitations or provided a comparison of the framework’s performance with traditional methods in addressing these challenges.
2.Insufficient Details on LSTM Model Training and Decision Trees
The manuscript lacks sufficient details regarding the training process of the LSTM model and the decision tree methodology. Key aspects, such as the rationale for selecting specific hyperparameters, the configuration of LSTM layers, and the criteria for model evaluation, remain unclear. Additionally, the decision tree component would benefit from a more detailed explanation of its integration with health classification outputs and its role in yield estimation. Providing this information is critical for reproducibility and for readers to fully assess the robustness of the proposed framework.
3.Relocation of Formulas and Methodological Details to Supplementary Files
The manuscript includes numerous formulas and methodological descriptions that, while important, detract from the readability of the main text. These technical details could be relocated to supplementary materials, allowing the main manuscript to focus more on the interpretation of results and the broader implications of the study. This adjustment would make the manuscript more accessible to a wider audience without compromising technical rigor.
4.Lack of Clear Contextualization in the Literature Review
While the manuscript references a significant number of studies, it does not provide a clear synthesis of the current state of the field or how this study builds upon and addresses gaps in existing research. For instance, while the use of UAV imagery and LSTM models has been explored in precision agriculture, the manuscript could better articulate its novelty, such as the development of custom indices tailored for mango canopies or the integration of phenological data for yield prediction. A more structured discussion in the introduction would help to position this work more effectively within the existing literature and highlight its unique contributions.
Author Response
Thank you for your suggestions and observations which have improved our manuscript. We have addressed all the observations and updated the manuscript accordingly. The detailed response is attached as a separate file.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsAll comments have been addressed. However, units and all figure descriptions should be rechecked and revised.
Author Response
All table and figure captions are rechecked and revised as per the reviewer's suggestion. Minor editing has further improved the manuscript. Thank you for taking the time to review the paper thoroughly.
Reviewer 3 Report
Comments and Suggestions for Authorsgood
Author Response
Thank you for taking the time to review the paper thoroughly.
