Evaluation of GPM IMERG Performance Using Gauge Data over Indonesian Maritime Continent at Different Time Scales

Round 1

Reviewer 1 Report

The authors have done a fair revision of the manuscript based on the previously submitted comments, and this revision has enhanced its overall presentation and quality. The authors have addressed the major issues pointed in the previous review, and I am quite happy to conclude that the manuscript can be accepted for publication. What is left at the Editor's discretion is my judgment on the basic level of the Discussion section of the MS. The Discussion could be better strauctured and further split formally into several sub-sections, discussing timescale effect on accuracy; influence of rain gauge type; influence of extarnal factors, i.e., altitude; the most important - in-depth discussion on the suitability of satellite-based products with different latency for various applications. Such structuring might seem useful after an extensive description of the results to support the scientific soundness and guide the potential  future use of these products in the IMC.

Also, the authors should verify the use of these abbreviations, MC and IMC, throughout the manuscript, and I suggest that only IMC is retained in the text.

Author Response

We thank the referee for the careful and insightful review of our manuscript. We address all of the concerns of the referee here.

Comment : The authors have done a fair revision of the manuscript based on the previously submitted comments, and this revision has enhanced its overall presentation and quality. The authors have addressed the major issues pointed in the previous review, and I am quite happy to conclude that the manuscript can be accepted for publication. What is left at the Editor's discretion is my judgment on the basic level of the Discussion section of the MS. The Discussion could be better structured and further split formally into several sub-sections, discussing timescale effect on accuracy; influence of rain gauge type; influence of extarnal factors, i.e., altitude; the most important - in-depth discussion on the suitability of satellite-based products with different latency for various applications. Such structuring might seem useful after an extensive description of the results to support the scientific soundness and guide the potential  future use of these products in the IMC.

Response: We appreciate the reviewer's insightful suggestion. However, the current discussion is similar to the structure suggested by the reviewers, although the presentation is given in paragraphs format, not sub-sections. We believe that dividing the discussion into sub-sections, as the reviewer suggested, would be unnecessary. Regarding the use of each application of the three IMERG data types (Early, Late, and Final), it really depends on the needs of the use of the data. Basically, we get better accuracy from data with longer latency, which is related to better data processing. However, if rainfall data is needed for quick application results, then IMERG Early or Late data with shorter latency can be used. This has been described in the discussion section. However, in order for the reader to gain more knowledge about the application, we add more explanation about this application in the discussion section, as follows.

“The accuracy of daily, monthly, and yearly data from IMERG is reasonable and can complement data from traditional rain gauge systems in IMC. As expected, the accuracy of the IMERG is best at the greatest latency (i.e., Final > Late > Early), thus the use of IMERG-F is recommended. The greater the latency, the more information/processing gets into the satellite estimates, the better its performance on average. However, the choice of the product is highly dependent on the requirements of the application. If an application can accept the 18-h latency of IMERG-L, then IMERG-L should be used instead of IMERG-F. Applications that expect fast and near real-time (NRT) results can use the IMERG-E[100], such as flash flood forecasting [101] and NRT hurricane rainfall forecasting [102], NRT erosion, and fires nowcasting. For daily and long-term applications such as agricultural forecasting or drought monitoring[102], we can use the IMERG-L. The IMERG-F offers significant benefits on monthly applications such as hydrological modeling[103] and climate study. However, all of these applications depend on the latency needed because even the same application may have different latency requirements depending on needs.”

Comment: Also, the authors should verify the use of these abbreviations, MC and IMC, throughout the manuscript, and I suggest that only IMC is retained in the text.

Response: Thanks for pointing this out. The MC is no longer used, so every MC word has been changed to "Maritime Continent," and only IMC is preserved in the text.

Reviewer 2 Report

This study performs cross-validation of precipitation from satellite observation and rain-gauge measurement over Indonesia. Although a similar type of analysis has been done before, the focus on Indonesia (with rainy tropical climate and complex terrain) is unique. The cross-validation is performed extensively across many time scales, from hourly to annual. Overall, I find the results substantial and useful. The paper is clearly written.

I do have some comments concerning the technical details:

(1) Since the location of the rain gauge does not coincides with a grid point of the satellite data, some kind of interpolation is needed for the cross-validation. How was the interpolation done, and could this process contribute to the discrepancy between the two data sets?  A particular point that came to mind is that, if a rain gauge is located on a slope, then the interpolation needs to take into account topographical influences.

(2) Related to (1), are the rain gauge measurements just the "raw data" or are they "processed data" after quality control and bias correction?  Although in this study the rain gauge data is regarded as the truth to extract the bias in satellite data, some discrepancies between the two data sets could possibly arise from instrumental biases for the rain gauges themselves. 

(3) The poor agreement between satellite observation and rain gauge data is not surprising. In fact, satellite observation (from a single satellite) rarely has hourly resolution at a particular location. The situation improves slightly when combining multiple satellites, but much of the so-called "hourly data" from satellite archive is still the product of interpolation from the raw observation with a coarser spatial/temporal resolution. Given so, what this paper demonstrates is just that this interpolation does not yield much useful information at the unresolved scales (for satellite).

(4) In this paper, the agreement between satellite and rain-gauge data seems to be judged by an arbitrary threshold of the value of correlation coefficient. To test the statistical significance of the correlation coefficient, one would need to know the degrees of freedom (i.e., number of independent data points), which presumably are quite different for hourly, daily, and monthly data. For example, “CC = 0.7 for daily data” does not imply the same level of statistical significance as “CC = 0.7 for monthly data” since the degrees of freedom are quite different for the two cases. This subtlety seems to have been overlooked in the authors’ interpretation of the results.

Author Response

We thank the referee for the careful and insightful review of our manuscript. We address all of the concerns of the referee here.

Comment: Since the location of the rain gauge does not coincides with a grid point of the satellite data, some kind of interpolation is needed for the cross-validation. How was the interpolation done, and could this process contribute to the discrepancy between the two data sets?  A particular point that came to mind is that, if a rain gauge is located on a slope, then the interpolation needs to take into account topographical influences.

Response: We thank the reviewers for pointing this out. We do not do any interpolation on this data processing. For more details on the process carried out, the explanation of the methodology is revised as follows:

“The comparison between IMERG and rain gauge was carried out using a point-to-pixel grid point approach as done in previous studies [55,56]. One pixel of IMERG data represents precipitation in 0.1° x 0.1° area, so point to pixel matching (rain gauge observation and corresponding satellite estimates) was carried out within ± 0.05° from the grid point of IMERG data. The rain gauge point observations, which are closest to the grid point in the satellite grid (less than 0.05 degrees off the grid points), are evaluated [56]. The pixel from the IMERG Grid used in this study had at least 1 rain gauge observation point in it. If in one pixel there were more than one point of the same type of rain gauge (MRG or AWS), then the amount of rainfall from the rain gauges would be averaged.”

The method used in this study is the same as previous research such as:

Sharifi E, Steinacker R, Saghafian B. Assessment of GPM-IMERG and Other Precipitation Products against Gauge Data under Different Topographic and Climatic Conditions in Iran: Preliminary Results. Remote Sensing. 2016; 8(2):135. https://doi.org/10.3390/rs8020135

Mahmoud MT, Mohammed SA, Hamouda MA, Mohamed MM. Impact of Topography and Rainfall Intensity on the Accuracy of IMERG Precipitation Estimates in an Arid Region. Remote Sensing. 2021; 13(1):13. https://doi.org/10.3390/rs13010013

Comment: Related to (1), are the rain gauge measurements just the "raw data" or are they "processed data" after quality control and bias correction?  Although in this study the rain gauge data is regarded as the truth to extract the bias in satellite data, some discrepancies between the two data sets could possibly arise from instrumental biases for the rain gauges themselves. 

Response: Thank you for pointing this out. The rain gauge is raw data obtained from the BMKG online database. However, as explained in section 2.1, quality control is carried out by BMKG and by ourselves. Differences in the working principle and performance of the rain gauge may affect the results of this validation, as discussed in the discussion section. However, the effect of such difference is not significant because both instruments show almost the same accuracy based on previous research.

Comment: The poor agreement between satellite observation and rain gauge data is not surprising. In fact, satellite observation (from a single satellite) rarely has hourly resolution at a particular location. The situation improves slightly when combining multiple satellites, but much of the so-called "hourly data" from satellite archive is still the product of interpolation from the raw observation with a coarser spatial/temporal resolution. Given so, what this paper demonstrates is just that this interpolation does not yield much useful information at the unresolved scales (for satellite).

Response: Thanks to reviewers for pointing this out. We agree with the reviewer regarding the low accuracy of interpolating satellite data for shorter temporal scales, especially hourly scales. However, the rate of such low accuracy of these satellite observations is different for each validation carried out in each region, such as in Southern China (Fuwan et al., 2020), in Mainland China (Xu et al., 2019), and Canada (Maozami and Najafi, 2021). The validation for the hourly time scale is still very minimal. Most of the validation was carried out for mainland China, and it has not been done for the maritime continent area. For this reason, information on the level of accuracy from different time scales in IMC will be valuable information on the level of accuracy of IMERG data. However, we add information related to the reviewer's suggestions in the discussion section as follows:

“The IMERG data combined ten satellites observation [32] that provide uniformly calibrated precipitation measurements every 2-4 hours around the globe [33]. The poor agreement between IMERG and rain gauge data may demonstrate that interpolation of multiple satellites data with a coarser spatial/temporal resolution for form hourly data or smaller temporal resolution does not yield much useful information at the unresolved scales (for satellite). However, multiple satellite observations must be combined because a single satellite observation rarely has hourly resolution at a particular location. Poor accuracy of hourly data was also found in Canada [68], Brazil [70] and mainland China [71]. However, the accuracy of the IMERG hourly data shows an increase compared to other satellite precipitation datasets [27]”.

Comment: In this paper, the agreement between satellite and rain-gauge data seems to be judged by an arbitrary threshold of the value of correlation coefficient. To test the statistical significance of the correlation coefficient, one would need to know the degrees of freedom (i.e., number of independent data points), which presumably are quite different for hourly, daily, and monthly data. For example, “CC = 0.7 for daily data” does not imply the same level of statistical significance as “CC = 0.7 for monthly data” since the degrees of freedom are quite different for the two cases. This subtlety seems to have been overlooked in the authors’ interpretation of the results.

Response: Thanks to reviewers for pointing this out. We note that the number of independent data points of each time scale is different. This study limits the amount of rain gauge data used to evaluate IMERG data. We use a minimum total of 50 data to test the correlation of the two data. We use at least 7020 (hours) and 86 (daily) total data for hourly and monthly time scales for each station. We get 875 and 196 data lengths for all stations in monthly and yearly data. Thus, the number of independent data points is sufficient for statistical analysis. Thus, qualitative discussions should still be valid even though the number of independent data points differs.

Reviewer 3 Report

Dear Editor,

Please find my comments on a revised manuscript titled "Evaluation of GPM IMERG Early, Late, and Final Rainfall Estimates Using Gauge Data over Indonesian Maritime Continent for Different Time Scales" by Ravidho Ramadhan , Helmi Yusnaini , Marzuki Marzuki , Robi Muharsyah , Wiwit Suryanto, Sholihun Sholihun , Mutya Vonnisa , Harmadi Harmadi , Ayu Putri Ningsih , Alessandro Battaglia , Hiroyuki Hashiguchi , Ali Tokay submitted for consideration for possible publication in MDPI Remote Sensing.

The authors thoroughly revised the manuscript; all my comments have been addressed.

This reviewer recommends accepting the manuscript.

Yours faithfully,

The Reviewer

Author Response

Response: We thank the referee for the careful and insightful review of our manuscript. Because there is no new comment from the reviewer, we do not address any specific response here.