Response: we added an explanation of nighttime selection. Because high altitude in western China,
sunlight time may from 05:00 to 21:00 in summer, therefore, we selected the time from 22:00 to 03:00 in the night to acquire the relationship of NEE/T.
12. Page 11, Section 3.3., Authors need to provide much more detailed information on how they calculate LUE.
Response: information was added in section 2.5.
13. Page 12, section 3.4.1, while vegetation indices are correlated with LUE, it does not mean that one
can use vegetation indices to replace LUE. One must realize that vegetation indices and LUE are two
different terms and biological processes. In that same paragraph, authors also found good correlation
between vegetation indices and fapar.
Response: we agree with the suggestions by the reviewer, perhaps because we authors are not native
speakers, and we made mistakes in the use of “replace”. These VIs just could be indicators or proxies
of the variables. The revision will be proofread by a native speaker and we are trying to make it clear and consistent.
14. Page 15, it stats \change
by biome? Or does that relationship change by pixel?
Response: the statement of “it seems unlikely there is a universal relationship of fapar or LUE to a VI”
is about the specific regression coefficients. Our mean is that a regression model (specific coefficients)
derived from maize ecosystems may not valid for other ecosystems. As many researches (Gitelson et al., 2005, Chl estimation different for maize and soybean, GPL) indicated that different
ecosystems may act differently. Therefore, in the revision, although all reviewers did not suggested
to apply our method in other ecosystems, we decided to add a further validation part in the manuscript
to see if this method works in forest and grassland. As we did not get enough auxiliary data (LAI),
we just apply the model derived from maize for forest and grassland systems. We got clear difference
in regression and this result implies, first, our method may also work for forest and grassland, and second, different species may affect the regression.
15. This manuscript only uses 1-month long CO2 flux data, and it does not cover the entire plant growing season of maize. It is not sure how well the GPP equation would work in early and late part
of the maize growing season.
Response: We agree with this concern. Due to earthquake in China 2008, we could not get part of the data, especially in the later mature season. Our experiments started from the beginning (3-4 leaves, 5cm in height) to a middle stage (about 1.8 m in height) of maize. Thus, the method may work in the early stage. For the late, especially the senesces, more work will be needed. We will continue this method in future research. We added some clarifications in the discussion part.
16. The linear correlation or regression analysis in this manuscript does show the usefulness of vegetation indices, however, if a model does not include climate constraint, it is of no use to estimate
the effect of weather variation (e.g., temperature, water).
Response: Yes, we consider it’s right to concern more about the climate variables in the estimation of GPP. The linear correlation may be the most important findings in this study and is explained by Monteith logic. The main aim of the paper is trying to find a model that can use all inputs from remote sensing observations for GPP estimation as it will be helpful for crop growth evaluation as large areas, quickly and iteratively. However, this method has some limitations, for example, maybe can not track the seasonal changes. We added a new discussion part to better evaluate our method, especially for the limitations.
17. Figure 5, LUE values are very small. It does not make sense, when compared with other publications.
Response: we changed the unit of LUE (mol CO2mol-1PPFD) in the revised version so that this result
could be better compared with other literatures.
Reviewer #3 (Highlight):
I think that the theme is important because it can take advantage for other agronomical cultures using
MODIS data or other remote sensor. Reviewer #3 (Comments): Comments to manuscript:
1. The authors mention that R2 is the correlation coefficient, which is false because R2 talks about the determination coefficient.
Response: we followed this suggestion and made changes throughout the paper.
2. Adapt the introduction to the work that they developed and not to mix introduction with methodology.
Response: we removed some part of method in section 2.1 as a new part of methodology.
3. They present acronyms that are not defined in the manuscript, and I suggest reviewing the writing
of this because in some paragraphs is confused.
Response: we checked throughout the paper and made corrections where needed.
4. In the variables that used they do not show the units in the manuscript and either in the title of the figures.
Response: we added the units of variables both in the text and in figures.
5. Change the color of some symbols of figures 6 to 8, because are not appraised well. Response: we changed the color in figures.
6. There are orthographic errors and of writing in the text corrected them. Response: we checked through the manuscript and made changes accordingly.
7. Some literature were not in the references and three references are not mentioned in the text. Response: all references were checked for consistency.
8. If the authors wrote 22:00 p.m., then to correct 3:00 a.m. instead of 3. Response: we followed this suggestion.
9. I consider that the period of study is very short and it is not possible to speak of seasonal changes
Response: Yes, we agree with the concern and provide more information about the stages of maize.
The limitations were also stated in the new discussion part.
10. They do not present discussions sufficient to support the results, and the figures little are taken advantage of, is necessary to discuss more on the matter.
Response: we agree with this suggestion and we added a new part of further validation of our method
with additional discussions.
11. I think that the simple objective to estimate GPP derived from data MODIS is not sufficient and
makes lack as GPP modeled related to in situ data, as well as its variation.
Response: yes, we consider the in situ data (e.g., temperature, water content) would be more related
with the GPP and GPP models incorporating these variables may more logically be accepted by scientists.
Remote observations, such as MODIS and other sensor data, would be offering a method of GPP evaluation quickly, nondestructively and timely in those cases where in situ data were impossible. Recently, less dependent on input parameters becomes a momentum for development of new models
in GPP estimation. Much work has been done; for example, using VIs (Gitelson, IEEEL, Landsat;
Sims, 2008, RSE, MODIS). We are trying to find a model with less input variables. For example, Sims et al., 2008 showed a model of GPP estimation with land surface temperature. On the other hand, if we can find some indicators of LST from satellite observations, we may avoid using the in situ measurements. Here is an explanation of the objective of the study, we are trying to use all remote sensing data for GPP estimation. With Monteith equation, we used VIs for estimation of both LUE and fAPAR. Estimation of PAR from satellite observations would be very helpful for our method and this work is undergoing. However, we also know that this method do have limitations
in the operational application. These limitations were added in the discussion part of manuscript.