Can a Combination of UAV-Derived Vegetation Indices with Biophysical Variables Improve Yield Variability Assessment in Smallholder Farms?

Round 1

Reviewer 1 Report

General comments:

The manuscript by Adewopo et al. deals with the use of remotely sensed vegetation indices from drones to predict maize yield variability in controlled plots and farmers’ field in Nigeria. The answer to the question in the title, I guess, should be “no”. The Authors found a statistically significant correlation between vegetation indices and yield variability in experimental fields, whereas they demonstrated that very weak relationships emerge when the same methodology is applied to farmers fields.  The topic could be of interest for the reader of Agronomy, and the Authors carried out a huge work which deserves consideration. However, the paper is hard to read, the English form can be much improved, as well as the presentation of the results. My main concern is that the comparison between the two conditions tested should be central in the paper, and the reasons of the failure of the method on farmers’ fields should be more investigated. It is fine to have “bad” results and to report them, but then the reasons should be discussed. Also, the statistical analysis could be improved.

The Abstract clearly frames the study objectives, methodologies and results. I suggest the Authors to avoid to use acronyms, or to limit them at least in this section. My main concern here is about the specific results: it is good to report that yield prediction in controlled experimental conditions is more accurate than in actual farmer fields, but the Authors could provide some perspectives and potential improvements in their methodology to foster its use in yield forecasting activities in the conditions where no dedicated experimental trials are available.

The Introduction deals with relevant topics: it starts characterizing the maize growing area, then gives an extensive background of vegetation indices and finally presents current limits and potentialities of using Unmanned aerial/air (please choose one) vehicles for yield estimation. The dissertation is well supported by references, despite many sentences are hard to read and need rephrasing. Please see specific comments. I suggest the Authors to rewrite the paragraph where the objectives are outlined, by giving a clear description of the research question and of its implications for the users/community.

Materials and Methods section reports the necessary information on the study area, although the section dealing with the setting up of the field experiments both in NOTs and in farmer fields could be improved. The statistical analysis can be questioned: did the Authors used some measure of complexity (AIC, BIC) to compare the different models? This is not trivial, because they also take into account the number of parameters/variables used in the model formulation. Another concern is the significance of the “location” random factor in the mixed effect model. The significance of location as a factor explaining the yield variability strongly limits the scalability of the results to other areas. I suggest the Authors to consider this limitation.

The presentation of the results lacks structure: the Authors should try to systematically presents the comparison between NOTs and actual fields throughout this section, given that this is the central topic of the paper. I personally do not like listing Figures and Tables without describing them, please consider to rewrite completely this section trying to give essential information to follow a logical workflow.

The Discussion section requires a deep revision of the English form, and a more articulated comparison of the results of the study with similar studies. The reasons of the failure of the methodology on farmers’ field should be addressed here.

Title:

I suggest to change the “(s)”  with “s” UAV is quite extensively used in literature as an acronym, but I suggest to report the extended form.

Abstract:

Line 20 five instead of 5

Line 25,26 check the apex in the unit

Line 25: this sentence is too vague: how many genotypes? Were they analysed in the paper? Not clear from the Abstract.

Line 33 RMSEP is used and not explained.

Keywords: please avoid to use keywords already present in the title.

Materials and Methods

Lines 126-127: scientific names in italics.

Line 149: the acronym NOT was already explained, but I would suggest to avoid it in the title of the subsection

Line 151: 100 NOTs, but only very few are shown in the map. Could you explain please?

Line 154: ok, these are the 12 ones shown in Figure 1.

Lines 160-166: the description of the agronomic practices in NOTs lacks basic information: sowing dates, fertilization amounts, cultivar names (or type).

Line 163: WAS week after sowing is not explained. How these two dates were chosen? Which is the standard period of the maize growing season in Nigeria? Please report average climatic conditions, such as  average temperature and precipitation in the maize season in the study area.

Line 165-166: ok, but please provide at least these basic info.

Line 168-174: which is the extension of these farmers fields as compared to NOTs? Total and average

Line 218: farmers’ fields throughout the manuscript.

Line 261: accrue à belong

Results

Line 288: t/ha please be consistent in writing the units of measure.

Figure 2 is misleading: This chart seems a frequency chart. Please check it out.

Figure 3 Boxplots are nice, but this Figure is superficially discussed in the Results. If the Authors would like to keep it, please provide some description, and even better statistical test on the distributions.

Line 298: this is no correlation, not poor.

Lines 301-302: I did not understand why these analysis were performed only on NOTs. Also, what is the reproducibility of these results on a different areas, if the location is significant? Please consider to avoid using location as a random factor in the linear mixed effect models.

Table 2: UNDVI? Yld? All the acronyms are not explained in the caption.

Figure 4: these are linear regressions, better than “univariate relationship”

Table 1: “multivariate assessment”à these are the results of the linear mixed effect model.

Figure 7: reading the text in the plots is very hard.

Discussion

Line 382: very yield?

Line 383-385: this is quite obvious indeed.

Line 392: why Maize in capital letter?

Line 411: “with intent to” à aiming at improving

Line 424: without disparity à with no significant differences

Author Response

Please see the attachment.

We would also like to thank the reviewers for the thorough review and candid suggestions which have helped to improve the manuscript.

Author Response File: Author Response.docx

Reviewer 2 Report

This research uses UAV-derived vegetation indices to predict and assess yield variability. The research is in-general well-organized and presented.  Below are my comments:

Abstract: Please explain or use the full name for some terms when first using, e.g., tha-1, Ht.

8WAS: If r< 0.3 and p < 0.001, usually the variables are correlated, but the model is not well-designed for less explanatory.

Section 2.5.2: R² and RMSEP are well-recognized statistical parameters for measuring correlation level, just citations would be good enough to explain the calculation,

For the georeferencing, it is recommended to add the precise level of the handhold GPS, since the UAV with high-resolution might need a more rigorous registration protocol.

For Figure 7 & 8, better to combine figures or use tables to summerize the features for more informative

More proofreading is needed. Such as double space, some parts of the ms are left align while some are justify. Check some term as well, e.g., use images, or imagery, instead of imageries.

More literature review can be included for updated research of UAV derived NDVI and other Vis:

Yang, B., Hawthorne, T. L., Torres, H., & Feinman, M. (2019). Using Object-Oriented Classification for Coastal Management in the East Central Coast of Florida: A Quantitative Comparison between UAV, Satellite, and Aerial Data. Drones.

Cao, J., Leng, W., Liu, K., Liu, L., He, Z., & Zhu, Y. (2018). Object-Based mangrove species classification using unmanned aerial vehicle hyperspectral images and digital surface models. Remote Sensing, 10(1).

Author Response

Please see the attachment

We would like to thank the reviewer for the detailed review and the very helpful suggestions on the overall content of the article.

Author Response File: Author Response.docx