Mapping Dragon Fruit Croplands from Space Using Remote Sensing of Artificial Light at Night

Reviewer Comments

Response

This paper attempted to relate the area and the production of dragon fruit crop with the nightlight radiance values obtained from VIIRS over the Binh Thuan Province in Vietnam. It is known that in some instances artificial light at night is provided to dragon fruit crop when the natural light is deficient. The purpose of this practice is to maintain crop growth at optimum level. The assumption that the nightlight radiance represents the artificial light for fruit crop is correct. However, to relate the nightlight radiance with the fruit production is not appropriate since the fruit production is influenced by the cumulative effect of both daylight and nightlight.

Thanks for your comment.

The monthly values of nightlight radiance were extracted for the cropland in the study area. Apparently, cropland is not exclusively dragon fruit area. Since cropland covers other major crops of the region as well, nightlight of cropland is not representative of dragon fruit crop area. It is therefore not appropriate to relate cropland nightlight radiance with dragon fruit crop area.

In this study, we assumed that NTL from cropland was caused by the artificial lighting in dragon fruit plantations. And we chose one dragon fruit field shown in discussion, and verified our idea.

In order to contrast the statistical data based on administrative district, we sum the NTL brightness by the corresponding regions.

Therefore, it seems to be able to reflect the dragon fruit with NTL data.

In equations 1 and 2 on page 6, all notations have not been defined. Apparently, the subscript ‘i’ to DNB refers to month. Therefore, in both equations notation i is expected to range between 1 and 12, and not between 0 and 12.

We revised the mistake in the equations and defined the notations.

In table 1 on page 12, the last two columns include ‘Sum of Mean’ and ‘Sum of STD’ of DNB. It is not clear how these two measures were derived since no detail has been provided. It is to be notated that equations 1 and 2 are to provide mean values not ‘the sum of means’.

We defined the “Sum of Mean” and “Sum of STD” in the methodological part and explained the calculation method.

Sample size of 10 is too small to draw any meaningful conclusion.

The sample size is a little bit small, but it can still reflect the positive correlation in some degree.

We are considering to extend the time series of NTL dataset (from 1 year to 5 years or more), thus can increase the sample size.

Text of this paper needs to be carefully checked for clarity of expression and grammar.

We refined the text in this paper.

Reviewer Comments

Response

The authors analyse in the paper connection between NTL and agricultural production under condition of artificial light.

The relation between such production and NTL should be strong because it is a simple cause-effect relation with almost no extra noise. The results are as expected. The paper is written clearly but can be improved.

Thank you for your approval.

1) Pictures 11-16 are scatterplots with linear regression. The data shows a nonlinearity that the authors does not interpret. The titles of the images state that this is the correlation, but in the text, the authors consider correlations as Pearson (I guess) correlation. The authors should correct the titles of the images, and the nonlinearity should be analyzed. The authors should explain the way of correlation calculation.

We corrected the titles of the images (11-16 in original manuscript, 8-13 in revised manuscript) and analyzed the nonlinearity.

The way of correlation calculation was Pearson correlation, and it was also explained in revised version.

2) Why the standard deviation (eq. 2) is a biased version of the estimator?

There was a mistake in equation 2, and was corrected. It was just a normal standard deviation.

3) The sentence in the abstract "The research in this paper demonstrates the possibility of applying NTL remote sensing data in agricultural production monitoring and opens up a new application field of NTL remote sensing data." sounds great but is not valid. Such analysis can be fruitful for agriculture with artificial light regime only.

We revised the sentence in Line 21.

Reviewer Comments

Response

nice job and an interesting new application. I feel that there are still a few miss citations. Please cite a paper about the VIIRS composites calibration.

Thank you for your approval.

We cited these two papers as 37 and 38.

Reviewer Comments

Response

The study presents an investigation of the dynamics of nighttime lights in agricultural areas in a province in Vietnam.

Authors find correlation between NTL emission and dragon fruit cropland.

I think that this is an interesting aspect and as far as I can see it is unique.

Thank you for your approval.

Furthermore, there are some issues with the data analsis.

Not all details are given, some data could be better communicated in tables and there could be more analysis.

We revised the paper on your comments.

the discussion is missing and conclusion is too brief

We added the discussion and reorganized the conclusion.

An important paper that has been overseen by the authors deals with NTL and CO2 emissions in the area studied, also mentioning dragon fruit plantages as a problem for NTL studies in terms of population.

We cited the paper as 40.

Abstract needs to be more focused

We revised the abstract.

please cite

Kumari et al. (2016). Induced flowering with artificial light for year round production of dragon fruit in Sri Lanka Ann. Sri Lanka Dep. Agric. 18, 128–114

We cited the paper as 29.

the study site is at 11°N so close to the equator - daylength does not change dramatically throughout the year

please explain

Although the study area is tropical and nearby the equator, the change of daylength is less than 2 hours throughout the whole year (shown in Figure 2), it is quite a big influence for long-day plants, such as dragon fruit. It is generally agreed that dragon fruit need over 12 hours daylength a day to keep blooming. Without artificial lighting, dragon fruit in this area only blossom and bear fruit in a short period just from May to September every year, when is called in-season, and other month is called off-season for dragon fruit.

there are two Methods sections (Materials & Methods, Methods), please merge

We merged the two sections.

"the rainy season is from May to October, and the dry season is from November to April"

how does this affect the NTL data? could it be that the presence of clouds mask the NTL?

this is important because this would create artefacts in the data and create false correlation etc.

please find stable bright light source (e.g. airport) and check stability

In the raining season, the clouds will affect the NTL data, but occasionally for the monthly composite data. The situation will be a little worse in July or August. The MEAN and STD value of the whole year can eliminate the influence by clouds masking.

We selected an urban area of Phan Thiet City and a dragon fruit cropland to compare the brightness. It is seemed that the brightness in urban area is comparatively stable remaining at about 50.

It is important to give more details about the selection of urban vs. cropland area

What is the size?

Is the size the same?

How many pixels per site?

Why were only 2 sites chosen, one each?

The size of both selection area is 2km by 2km, which covers about 25 pixels per site.

These two sites, one stands for artificial lighting in urban, the other stands for artificial lighting in cropland, are representative areas for two different human activities.

In my view, this section needs to be written with much more detail on the statistical part and the methodological part to guide the reader through what was done and what the pitfalls are.

We rebuilt the structure of paper, transmitting the land cover analysis to the chapter of discussion, and refining the methodological part.

The MS is lacking a proper discussion.

We rebuilt the structure of the paper and added the chapter of discussion.

Conclusion is very brief and just repeating what was stated already

the last sentence should be extended to give an outlook to the future where and how this method can be really applied

We revised the conclusion and extended the outlook.

Reviewer Comments

Response

1.     Authors did not address this reviewer’s original comment that “to relate the nightlight radiance with the fruit production is not appropriate since the fruit production is influenced by the cumulative effect of both daylight and nightlight.” Authors are advised that an appropriate response to this comment is included in the discussion.

It is true that fruit production is controlled by many factors, including daylight and artificial nightlight. Our results show that DNB values have positive correlations with dragon fruit production in the study area. Such correlations do not necessarily imply causation. We have added comments in lines 217-220:

“Although dragon fruit production is controlled by daylight, artificial nightlight, and other factors, the above results suggest that the sum of Mean and STD values of DNB in the statistical areas is an excellent measure of the cultivation, harvest and production of dragon fruit in Binh Thuan Province, Vietnam.”

2.     Authors should acknowledge in the discussion that ‘the total cropland area’ has been taken as a surrogate for ‘dragon fruit cropland’.

We discussed this in line 242-246:

“It should be noted that the croplands in the land cover dataset include not only dragon fruit croplands but also other croplands. However, by comparing the DNB values from croplands and other land cover types in Figure 14, it is reasonable to believe that artificial lighting in dragon fruit croplands may be a major factor contributing to the high DNB values of the croplands.”

3.     The terms ‘average’ and ‘mean’ have been used in the text interchangeably. Please specify how ‘Average value of DNB’ was calculated as shown in Figure 16, and how the ‘Mean of DBN’ was calculated as shown in Figure 14.

We have replaced “average” with “mean”, and revised Lines 239 and 256 to explain how the mean DNB values are calculated.

4.     The text related to Figures 7 – 12 indicates that ‘correlation coefficient’ was calculated to show the relationship whereas all the six figures indicate that it was the ‘coefficient of determination’ (R-squared) which was used. Authors are advised to correct the discrepancy.

We corrected the discrepancy, in line 191.

5.     Considering the small sample size and the skewed dataset, a nonparametric method should have been used instead of parametric linear regression method. A comment on this point should be included in the discussion.

We included the comment in the discussion in line 292.

Reviewer Comments

Response

Please be careful with significant digits. For example in table 1 for the sum of lights and STD in my opinion it does not make sense to give any value behind the comma.

We changed the expression of digits by using comma for number which are more than 9,999, and removing redundant digit behind decimal. Such as numbers in Table 1, Fig 7 – 12.

Furthermore, the STD as factor appears odd - please give some more information about the motivation why it was used. Pearson R values are almost identical for MEAN and STD - STD is very high.

We believed that the Mean value can express the occurrence of agricultural lighting, while STD can express the seasonal changes (dispersion). The Persons R values of both are very high, which just verified our vision.

Please improve fig. 16 - for me it is relatively clear but it might not be obvious for all readers and gain check digits. 4 digits behind comma seem excessive in a small table in a figure.

We keep 1 digit behind comma in Fig 16, which is clearer for all readers.