Linking Remote Sensing with APSIM through Emulation and Bayesian Optimization to Improve Yield Prediction
Round 1
Reviewer 1 Report
after a minor correction in the results section and improved the English language after that the paper is ok for publication. well done and really nice research work.
Author Response
"Please see the attachment."
Author Response File: Author Response.pdf
Reviewer 2 Report
The manuscript provides a framework to apply advanced satellite observations to constrain and improve the performance of crop model simulations. The work addresses the model calibration and improvement at both single-site and global levels. Thus, the results are of great merit to both science and society. However, there are a lot of unexplained abbreviations and redundant expressions making parts of the manuscript hard to follow. Descriptions of figure captions can be largely improved as well. Details of these issues and concerns regarding the usage of the LAI and NDVI datasets are listed below:
1. There are a few abbreviations not fully explained in the manuscript. For example, APSIM, DSSAT, NDVI, IA, PROSAIL, ANOVA, etc. This would make readers hard to understand and follow the content. Thus, providing the full terminology will be helpful.
2. Line 72: What does ‘the G x E x M inference space’ refer to?
3. As the latest LAI and NDVI satellite observations have been used in the study, more detailed information regarding the datasets (spatial and temporal information) and improvements in the datasets that might benefit model calibration should be included.
4. Regarding the measurements of LAI and NDVI, they are both affected by certain observational conditions. As the study focuses on the growing season from July-October, would the cloud affect the usage of the data? If so, is there any cloud mask or preprocessing for LAI and NDVI data used?
5. Since the goal of adopting LAI and NDVI is to improve maize prediction, do the authors have any idea whether LAI or NDVI is more important to maize yield (i.e., a simple explanation regarding connections between the yield and LAI and NDVI would be appreciated)?
6. The captions of the majority of the figure were too simple to miss critical information:
Figure 1: Is this the flow chart regarding the simulations and calibration procedures? The usage of this figure in the text was not marked.
Figure 2: Since there are 2 subplots in the figure, subtitles, and description (regarding the meaning of the colored dots and contours) of both plots are necessary. (The same issue for the rest of the figures in this manuscript)
Author Response
"Please see the attachment."
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors have analyzed the APSIM-Maize model to understand changes in yield predictions along with the help of satellite-based indices such as LAI and NDVI. The study has used three different Bayesian multi-criteria optimization frameworks across 13 sites across the U.S Midwest. The concept is well formulated but there are some issues that need to address. The specific comments are given below. Accordingly, a major revision of the manuscript has been recommended.
Major Comments:
1) Abstract: Model descriptions part needs to be shortened and the Novelty of the study needs to highlight.
2) Introduction is not properly contextualized. For instance, there are so many process-based models available but why authors only picked up APSIM and DSSAT (see L36). In L74 to L86, so many state variables, such as LAI, plant N concentration, and plant biomass..so on.. are mentioned but among thewhy only NDVI and LAI picked up?
3) Methodology: Section 2.3.2: The heading is only LAI…why not LAI and NDVI ? In this section, it is not clear why LAI taken from Sentinel-2 and NDVI from Landsat-8. Authors could have taken both LAI and NDVI from a single sensor data rather than varying resolutions from two sensors.
Many of the abbreviations used in the equations are not defined which makes unclear about the steps under the methodology
4) Results: Fig 5: Use 2 as superscript for R2. What are these colors indicate? I think legend is missing here
5) L369: Explain d-index under methodology
6) Show Fig 9a and 9b. In 9a, show the legend
7) Discussion: Global optimization scheme delivered lower RMSE. It is not clear why
8) Limitation of the study must be included under discussion (e.g., two sensors with varying resolutions….why only two remote sensing-based variables are used in the APSIM, …so on)
Minor Comments
1) L6: Something wrong here How LAI is derived from NDVI?
2) L93: What is IA?
3) L99: what is nRMSE?
4) L121: write to compare
5) NN, NDVI, S2, ME, CV, GEX so on....
6) L422/428: HPDA and HDPA used. Which one is correct ?
7) Landsat8 should be written as Landsat-8, Sentinel2 as Sentinel-2 throughout the ms
8) Eq. 4-8, many of the variables are not defined (e.g. Cab, NP, Cm , so on...)
Author Response
"Please see the attachment."
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
All concerns have been clearly addressed.
The full name of NDVI can be added in the abstract as for LAI.
Regarding the explanation of using LAI and NDVI variables. 1-2 sentences can be added to the manuscript to state the reason for using them (as good and well-used proxies for crop yield).
Author Response
"Please see the attachment."
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors have improved the overall quality of the manuscript and addressed all queries.
Author Response
"Please see the attachment."
Author Response File: Author Response.pdf