A Novel Hybrid Spatiotemporal Missing Value Imputation Approach for Rainfall Data: An Application to the Ratnapura Area, Sri Lanka

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an interesting paper which proposes a good review of the state of the art of the statistical toolboxes applied to increasing the use and significance of an uncomplete database. The authors show that a model mixing several tools can improve the rainfall forecasting skills in the particular case of a poorly covered area, i.e where the number of available recording stations is limited.

The paper is obviously addressed to readers specialists of the field but could be, in my view, of interest to a wider public provided special care is taken to shortly describe each acronym cited for the first time in the text (see some suggestions in my comments of yellow highlightings)

The results are correctly documented, please check that all labels are understandable (e.g. Fig.7)

Comments for author File: Comments.pdf

Comments on the Quality of English Language

The English is of very good quality. Only one sentence (highlighted in red) should be checked.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

General comments

 

Nivithigala (St7_Niv) and Ulinduwawa (St8_Uli) have the least data available. When applying the hybrid model to interpolate available data to missing data, how the percentage of missing data impact the model performance?

 

When considering spatial interpolation, it’s understandable to interpolate surrounding data to the missing data points based on the spatial distance. While considering temporal interpolation, it interpolates available data at a certain time to the missing value at a time. Since the rainfall is not continuous, the interpolation may not make sense. How did the authors deal with this issue?

 

If the correlation of rainfall values from different stations is non-linear, then how did the authors treat their relationship?

 

The authors mentioned the artificially generated missing data. Does it mean that the original data set has no missing data? After implementing the hybrid model, does the interpolated data match with the original data well?

 

Figures and Tables

 

Figure 3: explain what the black and white blocks are.

 

Combine Figure 3 and Table, since both describe the data sets.

 

Table 2: Did the authors tune these common parameter settings in MLP models? Which one gives better performance? Or how did the authors select these parameters listed in Table 2?

 

Figure 7: The explanation of Figure 7 is not clear. It’s hard to follow. Please explain it more.

 

Specific comments

 

P2L87: “uses” to “use”

P6L230: it’s the first time to use “MSE”. Add its full name, mean squared errors.

P8L273: “The smaller the values of MAE and RMSE, and larger the 𝑅² value better the forecasting is.” to “The smaller the values of MAE and RMSE, and the larger the 𝑅² value, the better the forecasting is.”

 

Important references

[18]: Cuenca, J.; Correa-Flórez, C.; Patino, D.; Vuelvas, J. (2020) Spatio-Temporal Kriging Based Economic Dispatch Problem Including Wind Uncertainty. Energies. 13. 6419. 10.3390/en13236419.

Comments on the Quality of English Language

Minor change needed

Author Response

Please see the attachment.

Author Response File: Author Response.pdf