the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Feb 2023
MGP: a new 1-hourly 0.25° global precipitation product (2000–2020) based on multi-source precipitation data fusion
Abstract. A high-quality global precipitation product with finer spatiotemporal resolutions and long-term temporal coverage is critical for a variety of science communities (e.g., hydrology, meteorology, climatology, ecology, and agriculture). Here, a novel multi-source precipitation data fusion (MPDF) algorithm, which considers the dependency of precipitation errors on seasonality, was proposed to fully take advantage of the complementary strengths from satellite, reanalysis, and gauge data for generating a higher-quality global precipitation product. Two merging schemes, which used six products (including four satellite precipitation products: IMERG-Late, GSMaP-MVK, TMPA-RT, and PERSIANN-CCS; one reanalysis precipitation product ERA5; one ground-based precipitation product CPCU) and three products (i.e., IMERG-Late, ERA5, and CPCU) as input data sources of the MPDF algorithm respectively, were designed to generate two different high-quality multi-source merged global precipitation products (MGP), i.e., MGP-6P and MGP-3P. The results show that the proposed MPDF algorithm is effective in considering the advantages from satellite, reanalysis, and gauge data. Global comparisons indicate that the MGP suite products with regard to daily mean precipitation share a similar spatial pattern with other global precipitation products (i.e., MSWEP, IMERG-Final, GSMaP-Gauge, ERA5, and CPCU) in most overland regions globally; while large differences between these seven products occur in Australia, southeast China, Europe, near the equator of Africa and South America, and so on. Overall, MGP-3P substantially performs better than the other five research-quality products (i.e., MGP-6P, MSWEP, IMERG-Final, GSMaP-Gauge, and ERA5) in the ground validation on the Chinese mainland, with the highest POD, CC and lowest RMSE of 0.85, 0.71, and 1.21 mm, respectively, at a 3 hourly scale. Especially, the accuracy and detection capability of MGP-3P are the best in most hourly rainfall intensity groups. The MGP-3P product can provide a new precipitation data option for research and applications in the field of hydrology, meteorology, climatology, ecology, and agriculture. MGP-3P (also known as MGP) Version 1.1 is available at the following link: https://www.zenodo.org/record/7386441#.Y8zr4clBxD9 (Chen et al., 2022a).
- Preprint
(15252 KB) - Metadata XML
-
Supplement
(2722 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2023-42', Anonymous Referee #1, 15 Mar 2023
“MGP: a new 1-hourly 0.25° global precipitation product (2000-2020) based on multi-source precipitation data fusion”.
Overall summary:
High-quality global precipitation product with finer spatiotemporal resolutions and long-term temporal coverage is really critical for a variety of science communities. However, after carefully reading this manuscript, there are various aspects confusing me a lot. Most critically, the writing and organization are really too weak to understand the key ideas of this study, as well as lacking scientific innovative contributions for the community, which seems to be just mixing several global precipitation datasets without any clear new thoughts. Some more serious scientific issues could be seen as follows. Considering the high standards of the big journal, ESSD, I think this study have great limitations and too far distances from the standards.
Serious scientific concerns including but are not limited to:
- what’s your basic assumptions for fusing these global dataset? If only consider the CC as the fusion weights, it is too simple and too weak. Beck et al., 2017, 2019 have already investigated such explorations. Please carefully reading such critical references.
- The title tells that this study aims to public a global dataset, however, it only provides precipitation estimates over Land. It is really not rigorous. Please take care of such issues.
- How did the authors consider the negative effects from the different input precipitation estimates, especially in terms of the precipitation events? The weights based on CC could be only achieved at daily scale. So how did you consider the systematic and random errors at hourly scales from the input datasets?
- The resolutions of the MPG is very strange with 1-hourly and 0.25°. most popular satellite and reanalysis precipitation datasets have finer resolutions at 1-hourly and 0.1°, for instance, IMERG, GSMaP, and ERA5-Land. Particularly, PERSIANN-CCS is quiet finer with resolutions half-hourly and 0.04°. So what’s your purpose of the resolutions at 1-hourly and 0.25°?
- The authors seems to have not enough background information on such satellite-based and reanalysis-based precipitation datasets. For instance, the most important aim of PERSIANN-CCS is to capture the first glimpse of the possible precipitation, not the quality. The authors considered the PERSIANN-CCS to provide what information at 1-hourly and 0.25° for developing the qualified research level precipitation product?
- In terms of evaluation and comparison, the results have various weak aspects and do not make me convinced, especially due to the black box merging model: (1) only evaluated at mainland China? Would it be reasonable to represent the global situations? (2) what are the reasons for improving the POD and FAR of MGP-6P and MGP-3P? just because there were merged based on CC only achieved at daily scales? and (3) why not evaluate and compare these precipitation products over CONUS where have enough ground observations for public?
Citation: https://doi.org/10.5194/essd-2023-42-RC1 - AC1: 'Reply on RC1', Hanqing Chen, 21 Mar 2023
-
RC2: 'Comment on essd-2023-42', Anonymous Referee #2, 21 Mar 2023
High-accurate precipitation data are essential for various aspects and estimating precipitation is challenging especially in complex-terrain and ungauged regions. This work produced a global precipitation dataset (0.25°, 1 hourly) by weighted average of multiple precipitation datasets. However, great improvements are needed in terms of the merging method and validation of the dataset. The main concerns are as follows:
1) Validation of the produced dataset is inadequate. The authors produced a global precipitation dataset. However, the validation with gauge observations was only conducted in Chinese Mainland, which does not support its accuracy in other regions of the world. I suggest the authors perform a more convincing validation using worldwide observations.
2) Why CC was used as the weight for merging various datasets rather than other metrics like root mean square error (RMSE), Kling–Gupta efficiency (KGE), Nash‐Sutcliffe efficiency (NSE), et al. In the case where a dataset is highly consistent with the observations in terms of the temporal trend but has a systematic error, this method may lead to a systematic bias in the merging output.
3) The authors repeated the merging procedures three times at different spatial and temporal scales and then spatially and temporally downscale the coarse dataset. What is the underlying logic for applying such a strategy?
4) The authors emphasized that considering the seasonality of errors in the merging procedures is novel. However, no evidence was provided in the manuscript to support the novelty or added value of considering the seasonality of errors in precipitation datasets.
5) The comparison between MGP-6P and MGP-3P makes no sense. If another group of datasets was selected for the merging, the results may differ evidently from those in the manuscript. The key to this problem is not the number of datasets that are applied, but whether the merging method can accurately identify the error characteristics in each dataset.
6) Given that a large number of global datasets have been available with a spatial resolution of 0.1° or finer, the produced dataset with a spatial resolution of 0.25°may not be competitive in the science community.
Some specific comments:
7) Line 74-75, “the quality of gauge observations is extremely dependent on the spatial density of the rain gauges”: the quality of gridded data interpolated from gauge observations, rather than the gauge observation itself, depends on gauge density.
8) Line 143-145, “spatial interpolation was proved to be an effective method in improving the quality of global satellite and reanalysis precipitation estimates”: There is no definitive relationship between spatial interpolation and improving the quality of precipitation datasets.
9) There are many confusing sentences in the manuscript, e.g. Line 177-182, Line 242-243, Line 273-275, Line 467-469, Line 526-527.
10) The used data are not fully introduced, especially for the CPCU dataset.
11) Line 412-414: it is not easy to see the differences from Fig. 3. It is suggested to present the differences between these global datasets and the produced dataset.
12) Section 4.2. Comparing the quality of precipitation datasets on different temporal scales makes no sense, because it is well acknowledged that precipitation product performs worse on a shorter temporal scale than on a longer scale.
Citation: https://doi.org/10.5194/essd-2023-42-RC2 - AC2: 'Reply on RC2', Hanqing Chen, 14 Apr 2023
Status: closed
-
RC1: 'Comment on essd-2023-42', Anonymous Referee #1, 15 Mar 2023
“MGP: a new 1-hourly 0.25° global precipitation product (2000-2020) based on multi-source precipitation data fusion”.
Overall summary:
High-quality global precipitation product with finer spatiotemporal resolutions and long-term temporal coverage is really critical for a variety of science communities. However, after carefully reading this manuscript, there are various aspects confusing me a lot. Most critically, the writing and organization are really too weak to understand the key ideas of this study, as well as lacking scientific innovative contributions for the community, which seems to be just mixing several global precipitation datasets without any clear new thoughts. Some more serious scientific issues could be seen as follows. Considering the high standards of the big journal, ESSD, I think this study have great limitations and too far distances from the standards.
Serious scientific concerns including but are not limited to:
- what’s your basic assumptions for fusing these global dataset? If only consider the CC as the fusion weights, it is too simple and too weak. Beck et al., 2017, 2019 have already investigated such explorations. Please carefully reading such critical references.
- The title tells that this study aims to public a global dataset, however, it only provides precipitation estimates over Land. It is really not rigorous. Please take care of such issues.
- How did the authors consider the negative effects from the different input precipitation estimates, especially in terms of the precipitation events? The weights based on CC could be only achieved at daily scale. So how did you consider the systematic and random errors at hourly scales from the input datasets?
- The resolutions of the MPG is very strange with 1-hourly and 0.25°. most popular satellite and reanalysis precipitation datasets have finer resolutions at 1-hourly and 0.1°, for instance, IMERG, GSMaP, and ERA5-Land. Particularly, PERSIANN-CCS is quiet finer with resolutions half-hourly and 0.04°. So what’s your purpose of the resolutions at 1-hourly and 0.25°?
- The authors seems to have not enough background information on such satellite-based and reanalysis-based precipitation datasets. For instance, the most important aim of PERSIANN-CCS is to capture the first glimpse of the possible precipitation, not the quality. The authors considered the PERSIANN-CCS to provide what information at 1-hourly and 0.25° for developing the qualified research level precipitation product?
- In terms of evaluation and comparison, the results have various weak aspects and do not make me convinced, especially due to the black box merging model: (1) only evaluated at mainland China? Would it be reasonable to represent the global situations? (2) what are the reasons for improving the POD and FAR of MGP-6P and MGP-3P? just because there were merged based on CC only achieved at daily scales? and (3) why not evaluate and compare these precipitation products over CONUS where have enough ground observations for public?
Citation: https://doi.org/10.5194/essd-2023-42-RC1 - AC1: 'Reply on RC1', Hanqing Chen, 21 Mar 2023
-
RC2: 'Comment on essd-2023-42', Anonymous Referee #2, 21 Mar 2023
High-accurate precipitation data are essential for various aspects and estimating precipitation is challenging especially in complex-terrain and ungauged regions. This work produced a global precipitation dataset (0.25°, 1 hourly) by weighted average of multiple precipitation datasets. However, great improvements are needed in terms of the merging method and validation of the dataset. The main concerns are as follows:
1) Validation of the produced dataset is inadequate. The authors produced a global precipitation dataset. However, the validation with gauge observations was only conducted in Chinese Mainland, which does not support its accuracy in other regions of the world. I suggest the authors perform a more convincing validation using worldwide observations.
2) Why CC was used as the weight for merging various datasets rather than other metrics like root mean square error (RMSE), Kling–Gupta efficiency (KGE), Nash‐Sutcliffe efficiency (NSE), et al. In the case where a dataset is highly consistent with the observations in terms of the temporal trend but has a systematic error, this method may lead to a systematic bias in the merging output.
3) The authors repeated the merging procedures three times at different spatial and temporal scales and then spatially and temporally downscale the coarse dataset. What is the underlying logic for applying such a strategy?
4) The authors emphasized that considering the seasonality of errors in the merging procedures is novel. However, no evidence was provided in the manuscript to support the novelty or added value of considering the seasonality of errors in precipitation datasets.
5) The comparison between MGP-6P and MGP-3P makes no sense. If another group of datasets was selected for the merging, the results may differ evidently from those in the manuscript. The key to this problem is not the number of datasets that are applied, but whether the merging method can accurately identify the error characteristics in each dataset.
6) Given that a large number of global datasets have been available with a spatial resolution of 0.1° or finer, the produced dataset with a spatial resolution of 0.25°may not be competitive in the science community.
Some specific comments:
7) Line 74-75, “the quality of gauge observations is extremely dependent on the spatial density of the rain gauges”: the quality of gridded data interpolated from gauge observations, rather than the gauge observation itself, depends on gauge density.
8) Line 143-145, “spatial interpolation was proved to be an effective method in improving the quality of global satellite and reanalysis precipitation estimates”: There is no definitive relationship between spatial interpolation and improving the quality of precipitation datasets.
9) There are many confusing sentences in the manuscript, e.g. Line 177-182, Line 242-243, Line 273-275, Line 467-469, Line 526-527.
10) The used data are not fully introduced, especially for the CPCU dataset.
11) Line 412-414: it is not easy to see the differences from Fig. 3. It is suggested to present the differences between these global datasets and the produced dataset.
12) Section 4.2. Comparing the quality of precipitation datasets on different temporal scales makes no sense, because it is well acknowledged that precipitation product performs worse on a shorter temporal scale than on a longer scale.
Citation: https://doi.org/10.5194/essd-2023-42-RC2 - AC2: 'Reply on RC2', Hanqing Chen, 14 Apr 2023
Data sets
MGP: a new 1-hourly 0.25° global precipitation product (2000-2020) based on multi-source precipitation data fusion Hanqing Chen, Debao Wen, Bin Yong, Jonathan J. Gourley, Leyang Wang, and Yang Hong https://www.zenodo.org/record/7386441#.Y8zr4clBxD9
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 726 | 249 | 37 | 1,012 | 80 | 22 | 30 |
- HTML: 726
- PDF: 249
- XML: 37
- Total: 1,012
- Supplement: 80
- BibTeX: 22
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1