the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Dynamic Rainfall Erosivity Estimates Derived from GPM IMERG data
Robert Alexander Emberson
Abstract. Soil degradation is a critical threat to agriculture and food security around the world. Understanding the processes that drive soil erosion is necessary to support sustainable management practices and to reduce eutrophication of water systems from fertilizer runoff. The erosivity of precipitation is a primary control on the rate of soil erosion, but to calculate erosivity high frequency precipitation data is required. Prior global scale analysis has almost exclusively used ground-based rainfall gauges to calculate erosivity, but the advent of high frequency satellite rainfall data provides an opportunity to estimate erosivity using globally consistent gridded satellite rainfall. In this study, I have tested the use of GPM IMERG rainfall data to calculate global rainfall erosivity. I have tested three different approaches to assess whether simplification of IMERG data allows for robust calculation of erosivity, finding that the highest frequency 30-minute data is needed to best replicate gauge-based estimates. I also find that in areas where ground-based gauges are sparse, there is more disparity between the IMERG derived estimates and the ground-based results, suggesting that IMERG may allow for improved erosivity estimates in data-poor areas. The global extent and accessibility of IMERG data allows for regular calculation of erosivity on a month-to-month timeframe, permitting improved dynamic characterisation of rainfall erosivity across the world in near-real time. These results demonstrate the value of satellite data to assess the impact of rainfall on soil erosion and may benefit practitioners of sustainable land management planning.
- Preprint
(2266 KB) - Metadata XML
-
Supplement
(35759 KB) - BibTeX
- EndNote
Robert Alexander Emberson
Status: closed
-
RC1: 'Comment on egusphere-2022-1315', Anonymous Referee #1, 10 Feb 2023
In the submitted paper author investigates the performance of the GPM IMERG rainfall data to calculate global rainfall erosivity. Three approaches are tested for the estimation of the global rainfall erosivity patterns. A detailed comparison with the GloREDa dataset and derived global erosivity map is conducted. The submitted paper is in the scope of the HESS journal. It is very well written and it is easy to follow the conducted methodological steps and discussion of the results. The presented results can be regarded as an important step towards determination of the so-called dynamic global rainfall erosivity maps that could be used as input data for the soil erosion models. Therefore, I only have a few moderate comments/suggestions:
Can author add some additional (besides what is written in sections 4.2 and 4.1) discussion about the uncertainty related to the satellite-based rainfall dataset and how does this transforms into the rainfall erosivity estimation and does uncertainty perhaps has some seasonal patterns. Would it be perhaps possible to compare calculated IMERG monthly rainfall erosivity to REDES dataset (subset of GloREDa data; monthly rainfall erosivity maps for Europe).
At the end of the manuscript, author wrote that more detailed comparison is needed with ground-based data in order to verify the IMERG dataset. Could perhaps some additional discussion be added regarding this possible further step (e.g., how this could be done). It should be also noted that also GloREDa dataset has its own limitations (e.g., different data periods were used for different stations, data temporal resolution was not uniform, etc.). Even if assume that GloREDa represents the “true” rainfall erosivity patterns it is question, how accurate this actually is. Is there any alternative way, without using GloREDa dataset. Perhaps satellite-based drop-size-distribution data and calculation of the rainfall erosivity directly from the DSD?
Figure 2: It would be perhaps good to include the 1:1 line to the first (A) figure as well.
Figure 4: Units should be added for x-axis. Is it meaningful to include MFI data? I see that you included Figure S1 but at least it should be noted in the Figure 4 caption that that MFI has different units. Perhaps the same could be done for Figure S7 and Figure S8.
Figure 5: Perhaps the y-axis caption is not the most clear, it should be total rainfall erosivity in specific month (units mo-1), right? Also it would be perhaps better to plot the absolute values (not log) since this can be easier to put in context (compared to annual erosivity values). Additionally, figure caption should specify what do black and red lines represent (mean and median?).
Figure 6: Similarly, as for Figure 5.
Figure 6: It is interesting why the monthly variations for S. America (and also for some other continents) are much larger as compared to Figure 5 (this is noted also in the text), any explanation?Citation: https://doi.org/10.5194/egusphere-2022-1315-RC1 - AC2: 'Response to reviewers', Robert Emberson, 28 Apr 2023
-
RC2: 'Comment on egusphere-2022-1315', Anonymous Referee #2, 28 Feb 2023
The manuscript “Dynamic Rainfall Erosivity Estimates Derived from GPM IMERG data” evaluated the performance of IMERG-Final rainfall product in estimating global rainfall erosivity estimates. Three methods were tested in this manuscript.
The main drawback of the manuscript is that it lacks high quality Introduction, and explanations that describe the reasons for getting those results. The author also uses many subjective words, which is not recommended for scientific writing. Specific values are encouraged.
Specific comments:
- Lines 16-18: I think the differences between the IMERG-derived and gauge-derived estimates cannot suggest that the IMERG data may allow for improved erosivity estimates in ungauged areas unless the author can provide corresponding evidence. The findings should be supported by study results. I noted that the IMERG-derived erosivity estimates are more similar to ground-based estimates in areas with a high spatial density of gauges, such as Europe. This means that the quality of erosivity estimates based on IMERG depends on the spatial density of gauges if the ground-based estimates are regarded as “true”. So, the erosivity estimates based on IMERG might have large errors and uncertainties in ungauged areas due to error propagation.
- Line 33: please give specific factors, and add citations.
- I believe that the Introduction section needs to be rewritten because of the lack of many previous studies, although the author introduced two papers that focused on estimating soil erosivity using CMORPH and TMPA, respectively. In fact, there are many papers investigating soil erosivity using multiple gridded precipitation datasets. Given that this study aims to evaluate the potentials of the IMERG data and different methods in estimating erosivity, by analyzing the advantages and limitations of using different rainfall datasets and different approaches in estimating erosivity, it can draw out the reason why you did this study and what scientific questions this study will solve. So, please review related papers and improve the Introduction section.
- Lines 49-50: please give the specific methods directly, rather than vague descriptions.
- Lines 65-66: IMERG-Final is derived from the error correction of IMERG-Late by using GPCC as a reference.
- Lines 70-74: how to separate or remove solid precipitation from precipitation? I suggest the author describe this point as much as possible because errors may be introduced in erosivity estimates when the solid precipitation cannot be removed accurately.
- Lines 120：please provide citations for this sentence.
- 7: what is and k, please provide specific meaning.
- Figure 1 A: no colour bar to represent the 0 value.
- Line 179: Figure 1C?
- Lines 182-183: which papers supported this point?
- Lines 188-190: why do the differences increase with increasing erosivity, and why does the 30-minute model generally produce higher values? I think the readers will be interested in the reasons. In addition, underlying reasons might be helpful for us to find a more reliable method to accurately estimate erosivity.
- Lines 200-207: lack of explanations for the results.
- Lines 229-230: how did you judge the absolute estimates from the 30-minute model are closer to those of GloREDa? Based on the results in Table 2?
- Captions of Figures 5 and 6: 2020 or 2021?
- It is an interesting phenomenon that the IMERG estimates are much larger than those of GloREDa in several coastal areas, why does this phenomenon appear? Is it caused by large errors in IMERG precipitation estimates or other reasons?
- I suggest the author discuss potential reasons for large differences between IMERG-derived estimates and GloREDa estimates.
- Line 334: delete “is”.
- The author claimed that “These estimates provide informative comparisons to the study of Bezak and coauthors [2022]” (in line 49). However, I did not see the relevant comparison test in this study. In addition, the comparison is possible to demonstrate that IMERG performed better than CMORPH in estimating erosivity. So, a comparison may be added to demonstrate that IMERG is more suitable for use in estimating erosivity compared to CMORPH. More importantly, I think the erosivity obtained from this work is not limited to comparison with Bezak and coauthors [2022], please use more appropriate sentences to sublimate your work.
- AC2: 'Response to reviewers', Robert Emberson, 28 Apr 2023
-
RC3: 'Comment on egusphere-2022-1315', Anonymous Referee #3, 05 Mar 2023
In this study, the author investigates the potential of the “Integrated Multi-satellitE Retrievals for GPM (IMERG) data” to estimate rainfall erosivity at almost global scale. Multiple approaches were tested, and the results were compared against a global map of rainfall erosivity obtained through a Gaussian regression model.
I believe that the subject of the study is interesting and within the scope of Hydrology and Earth System Sciences. I do not have a large number of comments to make. The methodology to estimate rainfall erosivity is per se quite straightforward. In addition, several studies applied and discussed rainfall erosivity estimates, including satellite data as rainfall input data. And last but not least, the author made a good job in the application and presentation of the results obtained.
I have the feeling that the author aimed at keeping the study concise. Which is fine, despite some more detail in the Introduction and in the Discussion sectors could make the study more comprehensive. The description of the results is good but their discussion and implications could be improved. Also the abstract could be enriched by presenting some results of the statistical analysis carried out to support the statement made.
However, in my amble opinion, the conditio sine qua non to suggest publication rest on the data evaluation. In its current form, this study do not provide a valid evaluation of its results. The author limits the validation exercise by comparing the multiple results obtained against the GloREDa map. However, GloREDa is fruit of an interpolation, and therefore an estimate as well. To meaningfully evaluate the performance of the approach described in this study, the rainfall erosivity estimates should be evaluated against a set of true data, or, better said, rainfall erosivity values (annual average or single storms) estimated in several meteorological stations. Adopting a statistical significant number of meteorological stations around the globe and across different climate zones. As it was done to evaluate GloREDa and other studies using satellite data at global (Bezak et al. 2022; Liu et al. 2020) or regional scale (Kim et al. 2020).
Without a proper evaluation/validation of the results against true data, in my opinion the study cannot be published. The comparison against GloREDa is a good exercise but cannot assess the performance of the estimates provided in this study. Therefore, I would recommend to the author to evaluation/validation his results against a dataset of measured rainfall erosivity data. Once this is done, and the results fully evaluated and described in the paper, the paper will make a fine contribution to current literature.
Citation: https://doi.org/10.5194/egusphere-2022-1315-RC3 - AC2: 'Response to reviewers', Robert Emberson, 28 Apr 2023
-
CC1: 'Comment on egusphere-2022-1315', Panos Panagos, 06 Mar 2023
A very interesting article which focus on a topic getting a lot of attention. Obviously, the erosivity Is better estimated with high temporal rainfall data.
According to our experience, the satellite products are not yet “mature” enough to capture the variability of rainfall erosivity. According to Bezak (2021), 11% of the erosive events contribute to around 50% of the total erosivity. This was done based on the detailed rainfall erosivity records in GloREDa.
Similar observations have been done by Matthews et al. (2022) based on 300,000 erosive events.
Therefore, based on measured erosivity data, very few events are the ones who contribute to major part of total erosivity. Unfortunately, the satellite products are not yet mature enough to capture the high erosive events.
The satellite products are tending to smooth the high erosive values.
Therefore, there will be differences between your results and the ones of GloREDa in high erosive areas (continents). That is the case for Table 2.
I would also to propose a comparison per climatic zone where you will find big discrepancies in the tropical zones.
The global erosivity map (Panagos et al., 2017) has been tested and evaluated against the 3,625 measured R-factor values. Please see the figure 4 in the Panagos et al. (2017) and the excellent performance of Global assessment.
As you state in the manuscript, seems that you have estimated better than GloREDa which is not the case. If you insist your statement, you should prove that you perform better than GloREDa in the measured 3,625 stations. The GloREDa measured stations data will be available soon with a data paper.
Another remark: The MFI is much problematic and this has been shown in a recent review of Chen et al (2023) .
In my opinion, a mixture of satellite products with measured GloREDa would be an ideal and operational solution. That is why the EU Soil Observatory launched also a data collection campaign to get more measured stations data for GloREDa. More info about this call for data in the European Soil Data Centre 2.0 newsletter (February 2023).
Lines 30-35: the cost of soil erosion at global scale, at least for agricultural productivity losses has been estimated to about 8 billion dollars per year. You can find more information in Sartori et al. (2019).
In Europe, “Ballabio et al (2017) – Mapping Monthly erosivity in Europe” have developed monthly erosivity maps and datasets based on REDES. Similar has done also for GloREDa at global scale and an article is under preparation to present the monthly erosivity maps. Therefore, measured R-factor data on GloREDa can derive monthly erosivity maps at global scale.
References:
Bezak, N., Mikoš, M., Borrelli, P., Liakos, L. and Panagos, P., 2021. An in-depth statistical analysis of the rainstorms erosivity in Europe. Catena, 206, p.105577.
Matthews, F., Panagos, P. and Verstraeten, G., 2022. Simulating event-scale rainfall erosivity across European climatic regions. Catena, 213, p.106157.
Chen, W., Huang, Y.C., Lebar, K. and Bezak, N., 2023. A systematic review of the incorrect use of an empirical equation for the estimation of the rainfall erosivity around the globe. Earth-Science Reviews, p.104339.
Citation: https://doi.org/10.5194/egusphere-2022-1315-CC1 -
AC1: 'Reply on CC1', Robert Emberson, 27 Apr 2023
Note: in replying to this comment, I have added my response in Italics.
A very interesting article which focus on a topic getting a lot of attention. Obviously, the erosivity Is better estimated with high temporal rainfall data.
I appreciate the time taken by the reader to comment on the article – this is additional input that is certainly welcome.
According to our experience, the satellite products are not yet “mature” enough to capture the variability of rainfall erosivity. According to Bezak (2021), 11% of the erosive events contribute to around 50% of the total erosivity. This was done based on the detailed rainfall erosivity records in GloREDa.
Similar observations have been done by Matthews et al. (2022) based on 300,000 erosive events.
Therefore, based on measured erosivity data, very few events are the ones who contribute to major part of total erosivity. Unfortunately, the satellite products are not yet mature enough to capture the high erosive events.
The satellite products are tending to smooth the high erosive values.
This is overall an excellent point. I want to provide a full answer, but first I want to stress that I think the value of satellite-based and gauge-based data is very different, and any discussion of the two side-by-side does need to acknowledge their relative merits.
First – I am not sure I completely agree that satellite rainfall is guaranteed to underperform gauges when it comes to estimating very large rainfall events. For areas where gauge density is very high, this is likely to be true, but the local rainfall intensity in extreme rainfall events can vary widely and associated wind can also interrupt effective gauge measurement (e.g. Medlin et al. 2007). Satellite rainfall can underestimate extreme rainfall if the overpass of the Microwave satellite does not coincide with peak rainfall intensity, but studies have also shown that the uncertainty of satellite rainfall products is lower at higher rainfall intensities (e.g. Tian and Peters-Lidard (2010)).
Using Bezak et al. 2021 as inspiration, I have explored the storm histories of 4 locations from around the world; two in areas of concern for soil erosion (Near Wichita, USA, and Lucknow, North India) one in a critical region of degradation where the IMERG estimate exceeds GloREDa (Central Sierra Leone) and also near San Pedro de Atacama, in the Atacama desert of Northern Chile. Bezak et al. (2021) clearly show that the 11% of erosive events are responsible for the bulk of erosivity, but I wanted to test what maximum rainfall values would correspond to the storms most responsible for erosivity. In Lucknow and Wichita, more than 80% of the total rainfall kinetic energy calculated from the satellite data comes from storms with a peak rainfall intensity of <30mm/hr. In Sierra Leone, this value is much higher (<70mm/hr), but the same overall pattern emerges. In the Atacama, only three rainfall storm events were found over 20 years of data.
Although one could argue that the satellite products may just be missing or underestimating the large storms, the fact that the Sierra Leone case shows the same patterns as in Lucknow and Wichita suggests that the satellite is capable of capturing rainfall events far in excess of the 30mm/hr events (or lower) that are relevant in those locations (since it does so in Sierra Leone). I suggest that although the larger rainfall events do contribute the bulk of the erosivity, those rainfall events are in many locations are not reaching a maximum intensity that would lead them to be significantly underestimated; the key types of rainfall event that are underestimated by satellite products are major tropical storms (Marc et al. 2022), occurring in areas of extreme topography, where rainfall may exceed 100-200mm/hr – somewhat less relevant to agricultural soil degradation. Since the underestimation of extreme rainfall by satellites is generally because peak rainfall is missed by the microwave satellite overpass, we would still expect to sample some of the largest events - i.e., I would not expect the IMERG based estimates to be systematically lower for every storm. Curves for all 4 locations are shown below.
Overall – I agree that there are challenges to using satellite precipitation data, and in revising this study I want to avoid giving the sense that one product is ‘better’ than the other. It is clear that GloREDa is highly valuable in large parts of the globe. In revising this study I hope to provide a more nuanced perspective about the value of satellite vs gauge based products, but not to suggest that one or the other is superior.
Therefore, there will be differences between your results and the ones of GloREDa in high erosive areas (continents). That is the case for Table 2.
I would also to propose a comparison per climatic zone where you will find big discrepancies in the tropical zones.
This is certainly interesting, and something I will be pursuing in future study.
The global erosivity map (Panagos et al., 2017) has been tested and evaluated against the 3,625 measured R-factor values. Please see the figure 4 in the Panagos et al. (2017) and the excellent performance of Global assessment.
As you state in the manuscript, seems that you have estimated better than GloREDa which is not the case. If you insist your statement, you should prove that you perform better than GloREDa in the measured 3,625 stations. The GloREDa measured stations data will be available soon with a data paper.
I look forward to seeing the data! Thank you for letting me know. In revision, as mentioned above, I want to rephrase to ensure it is clear I am not stating that the satellite product is ‘better’ that GloREDa – this is more a comparison of the two for illustrative purposes.
Another remark: The MFI is much problematic and this has been shown in a recent review of Chen et al (2023) .
In this case, I entirely concur. In revising the study, I want to ensure that it is clear to the reader that although I am testing MFI here I agree that it should not be widely used. Since prior authors have used TRMM era satellite precipitation products to estimate MFI, it seems important to at least mention it.
In my opinion, a mixture of satellite products with measured GloREDa would be an ideal and operational solution. That is why the EU Soil Observatory launched also a data collection campaign to get more measured stations data for GloREDa. More info about this call for data in the European Soil Data Centre 2.0 newsletter (February 2023).
I think this is an excellent overall point. I address it more fully above, but I should stress that a combined model is likely to be the best performing output. In this case, the intent of the study is to discuss the value and limitations of a solely IMERG-derived model, but a more general assessment would include diverse input datasets.
Lines 30-35: the cost of soil erosion at global scale, at least for agricultural productivity losses has been estimated to about 8 billion dollars per year. You can find more information in Sartori et al. (2019).
Thank you – an excellent point – I will add this in revision.
In Europe, “Ballabio et al (2017) – Mapping Monthly erosivity in Europe” have developed monthly erosivity maps and datasets based on REDES. Similar has done also for GloREDa at global scale and an article is under preparation to present the monthly erosivity maps. Therefore, measured R-factor data on GloREDa can derive monthly erosivity maps at global scale.
Again – I am very much looking forward to reading the upcoming study and comparing results.
References:
Bezak, N., Mikoš, M., Borrelli, P., Liakos, L. and Panagos, P., 2021. An in-depth statistical analysis of the rainstorms erosivity in Europe. Catena, 206, p.105577.
Matthews, F., Panagos, P. and Verstraeten, G., 2022. Simulating event-scale rainfall erosivity across European climatic regions. Catena, 213, p.106157.
Chen, W., Huang, Y.C., Lebar, K. and Bezak, N., 2023. A systematic review of the incorrect use of an empirical equation for the estimation of the rainfall erosivity around the globe. Earth-Science Reviews, p.104339.
Citation: https://doi.org/10.5194/egusphere-2022-1315-AC1
-
AC1: 'Reply on CC1', Robert Emberson, 27 Apr 2023
Status: closed
-
RC1: 'Comment on egusphere-2022-1315', Anonymous Referee #1, 10 Feb 2023
In the submitted paper author investigates the performance of the GPM IMERG rainfall data to calculate global rainfall erosivity. Three approaches are tested for the estimation of the global rainfall erosivity patterns. A detailed comparison with the GloREDa dataset and derived global erosivity map is conducted. The submitted paper is in the scope of the HESS journal. It is very well written and it is easy to follow the conducted methodological steps and discussion of the results. The presented results can be regarded as an important step towards determination of the so-called dynamic global rainfall erosivity maps that could be used as input data for the soil erosion models. Therefore, I only have a few moderate comments/suggestions:
Can author add some additional (besides what is written in sections 4.2 and 4.1) discussion about the uncertainty related to the satellite-based rainfall dataset and how does this transforms into the rainfall erosivity estimation and does uncertainty perhaps has some seasonal patterns. Would it be perhaps possible to compare calculated IMERG monthly rainfall erosivity to REDES dataset (subset of GloREDa data; monthly rainfall erosivity maps for Europe).
At the end of the manuscript, author wrote that more detailed comparison is needed with ground-based data in order to verify the IMERG dataset. Could perhaps some additional discussion be added regarding this possible further step (e.g., how this could be done). It should be also noted that also GloREDa dataset has its own limitations (e.g., different data periods were used for different stations, data temporal resolution was not uniform, etc.). Even if assume that GloREDa represents the “true” rainfall erosivity patterns it is question, how accurate this actually is. Is there any alternative way, without using GloREDa dataset. Perhaps satellite-based drop-size-distribution data and calculation of the rainfall erosivity directly from the DSD?
Figure 2: It would be perhaps good to include the 1:1 line to the first (A) figure as well.
Figure 4: Units should be added for x-axis. Is it meaningful to include MFI data? I see that you included Figure S1 but at least it should be noted in the Figure 4 caption that that MFI has different units. Perhaps the same could be done for Figure S7 and Figure S8.
Figure 5: Perhaps the y-axis caption is not the most clear, it should be total rainfall erosivity in specific month (units mo-1), right? Also it would be perhaps better to plot the absolute values (not log) since this can be easier to put in context (compared to annual erosivity values). Additionally, figure caption should specify what do black and red lines represent (mean and median?).
Figure 6: Similarly, as for Figure 5.
Figure 6: It is interesting why the monthly variations for S. America (and also for some other continents) are much larger as compared to Figure 5 (this is noted also in the text), any explanation?Citation: https://doi.org/10.5194/egusphere-2022-1315-RC1 - AC2: 'Response to reviewers', Robert Emberson, 28 Apr 2023
-
RC2: 'Comment on egusphere-2022-1315', Anonymous Referee #2, 28 Feb 2023
The manuscript “Dynamic Rainfall Erosivity Estimates Derived from GPM IMERG data” evaluated the performance of IMERG-Final rainfall product in estimating global rainfall erosivity estimates. Three methods were tested in this manuscript.
The main drawback of the manuscript is that it lacks high quality Introduction, and explanations that describe the reasons for getting those results. The author also uses many subjective words, which is not recommended for scientific writing. Specific values are encouraged.
Specific comments:
- Lines 16-18: I think the differences between the IMERG-derived and gauge-derived estimates cannot suggest that the IMERG data may allow for improved erosivity estimates in ungauged areas unless the author can provide corresponding evidence. The findings should be supported by study results. I noted that the IMERG-derived erosivity estimates are more similar to ground-based estimates in areas with a high spatial density of gauges, such as Europe. This means that the quality of erosivity estimates based on IMERG depends on the spatial density of gauges if the ground-based estimates are regarded as “true”. So, the erosivity estimates based on IMERG might have large errors and uncertainties in ungauged areas due to error propagation.
- Line 33: please give specific factors, and add citations.
- I believe that the Introduction section needs to be rewritten because of the lack of many previous studies, although the author introduced two papers that focused on estimating soil erosivity using CMORPH and TMPA, respectively. In fact, there are many papers investigating soil erosivity using multiple gridded precipitation datasets. Given that this study aims to evaluate the potentials of the IMERG data and different methods in estimating erosivity, by analyzing the advantages and limitations of using different rainfall datasets and different approaches in estimating erosivity, it can draw out the reason why you did this study and what scientific questions this study will solve. So, please review related papers and improve the Introduction section.
- Lines 49-50: please give the specific methods directly, rather than vague descriptions.
- Lines 65-66: IMERG-Final is derived from the error correction of IMERG-Late by using GPCC as a reference.
- Lines 70-74: how to separate or remove solid precipitation from precipitation? I suggest the author describe this point as much as possible because errors may be introduced in erosivity estimates when the solid precipitation cannot be removed accurately.
- Lines 120：please provide citations for this sentence.
- 7: what is and k, please provide specific meaning.
- Figure 1 A: no colour bar to represent the 0 value.
- Line 179: Figure 1C?
- Lines 182-183: which papers supported this point?
- Lines 188-190: why do the differences increase with increasing erosivity, and why does the 30-minute model generally produce higher values? I think the readers will be interested in the reasons. In addition, underlying reasons might be helpful for us to find a more reliable method to accurately estimate erosivity.
- Lines 200-207: lack of explanations for the results.
- Lines 229-230: how did you judge the absolute estimates from the 30-minute model are closer to those of GloREDa? Based on the results in Table 2?
- Captions of Figures 5 and 6: 2020 or 2021?
- It is an interesting phenomenon that the IMERG estimates are much larger than those of GloREDa in several coastal areas, why does this phenomenon appear? Is it caused by large errors in IMERG precipitation estimates or other reasons?
- I suggest the author discuss potential reasons for large differences between IMERG-derived estimates and GloREDa estimates.
- Line 334: delete “is”.
- The author claimed that “These estimates provide informative comparisons to the study of Bezak and coauthors [2022]” (in line 49). However, I did not see the relevant comparison test in this study. In addition, the comparison is possible to demonstrate that IMERG performed better than CMORPH in estimating erosivity. So, a comparison may be added to demonstrate that IMERG is more suitable for use in estimating erosivity compared to CMORPH. More importantly, I think the erosivity obtained from this work is not limited to comparison with Bezak and coauthors [2022], please use more appropriate sentences to sublimate your work.
- AC2: 'Response to reviewers', Robert Emberson, 28 Apr 2023
-
RC3: 'Comment on egusphere-2022-1315', Anonymous Referee #3, 05 Mar 2023
In this study, the author investigates the potential of the “Integrated Multi-satellitE Retrievals for GPM (IMERG) data” to estimate rainfall erosivity at almost global scale. Multiple approaches were tested, and the results were compared against a global map of rainfall erosivity obtained through a Gaussian regression model.
I believe that the subject of the study is interesting and within the scope of Hydrology and Earth System Sciences. I do not have a large number of comments to make. The methodology to estimate rainfall erosivity is per se quite straightforward. In addition, several studies applied and discussed rainfall erosivity estimates, including satellite data as rainfall input data. And last but not least, the author made a good job in the application and presentation of the results obtained.
I have the feeling that the author aimed at keeping the study concise. Which is fine, despite some more detail in the Introduction and in the Discussion sectors could make the study more comprehensive. The description of the results is good but their discussion and implications could be improved. Also the abstract could be enriched by presenting some results of the statistical analysis carried out to support the statement made.
However, in my amble opinion, the conditio sine qua non to suggest publication rest on the data evaluation. In its current form, this study do not provide a valid evaluation of its results. The author limits the validation exercise by comparing the multiple results obtained against the GloREDa map. However, GloREDa is fruit of an interpolation, and therefore an estimate as well. To meaningfully evaluate the performance of the approach described in this study, the rainfall erosivity estimates should be evaluated against a set of true data, or, better said, rainfall erosivity values (annual average or single storms) estimated in several meteorological stations. Adopting a statistical significant number of meteorological stations around the globe and across different climate zones. As it was done to evaluate GloREDa and other studies using satellite data at global (Bezak et al. 2022; Liu et al. 2020) or regional scale (Kim et al. 2020).
Without a proper evaluation/validation of the results against true data, in my opinion the study cannot be published. The comparison against GloREDa is a good exercise but cannot assess the performance of the estimates provided in this study. Therefore, I would recommend to the author to evaluation/validation his results against a dataset of measured rainfall erosivity data. Once this is done, and the results fully evaluated and described in the paper, the paper will make a fine contribution to current literature.
Citation: https://doi.org/10.5194/egusphere-2022-1315-RC3 - AC2: 'Response to reviewers', Robert Emberson, 28 Apr 2023
-
CC1: 'Comment on egusphere-2022-1315', Panos Panagos, 06 Mar 2023
A very interesting article which focus on a topic getting a lot of attention. Obviously, the erosivity Is better estimated with high temporal rainfall data.
According to our experience, the satellite products are not yet “mature” enough to capture the variability of rainfall erosivity. According to Bezak (2021), 11% of the erosive events contribute to around 50% of the total erosivity. This was done based on the detailed rainfall erosivity records in GloREDa.
Similar observations have been done by Matthews et al. (2022) based on 300,000 erosive events.
Therefore, based on measured erosivity data, very few events are the ones who contribute to major part of total erosivity. Unfortunately, the satellite products are not yet mature enough to capture the high erosive events.
The satellite products are tending to smooth the high erosive values.
Therefore, there will be differences between your results and the ones of GloREDa in high erosive areas (continents). That is the case for Table 2.
I would also to propose a comparison per climatic zone where you will find big discrepancies in the tropical zones.
The global erosivity map (Panagos et al., 2017) has been tested and evaluated against the 3,625 measured R-factor values. Please see the figure 4 in the Panagos et al. (2017) and the excellent performance of Global assessment.
As you state in the manuscript, seems that you have estimated better than GloREDa which is not the case. If you insist your statement, you should prove that you perform better than GloREDa in the measured 3,625 stations. The GloREDa measured stations data will be available soon with a data paper.
Another remark: The MFI is much problematic and this has been shown in a recent review of Chen et al (2023) .
In my opinion, a mixture of satellite products with measured GloREDa would be an ideal and operational solution. That is why the EU Soil Observatory launched also a data collection campaign to get more measured stations data for GloREDa. More info about this call for data in the European Soil Data Centre 2.0 newsletter (February 2023).
Lines 30-35: the cost of soil erosion at global scale, at least for agricultural productivity losses has been estimated to about 8 billion dollars per year. You can find more information in Sartori et al. (2019).
In Europe, “Ballabio et al (2017) – Mapping Monthly erosivity in Europe” have developed monthly erosivity maps and datasets based on REDES. Similar has done also for GloREDa at global scale and an article is under preparation to present the monthly erosivity maps. Therefore, measured R-factor data on GloREDa can derive monthly erosivity maps at global scale.
References:
Bezak, N., Mikoš, M., Borrelli, P., Liakos, L. and Panagos, P., 2021. An in-depth statistical analysis of the rainstorms erosivity in Europe. Catena, 206, p.105577.
Matthews, F., Panagos, P. and Verstraeten, G., 2022. Simulating event-scale rainfall erosivity across European climatic regions. Catena, 213, p.106157.
Chen, W., Huang, Y.C., Lebar, K. and Bezak, N., 2023. A systematic review of the incorrect use of an empirical equation for the estimation of the rainfall erosivity around the globe. Earth-Science Reviews, p.104339.
Citation: https://doi.org/10.5194/egusphere-2022-1315-CC1 -
AC1: 'Reply on CC1', Robert Emberson, 27 Apr 2023
Note: in replying to this comment, I have added my response in Italics.
A very interesting article which focus on a topic getting a lot of attention. Obviously, the erosivity Is better estimated with high temporal rainfall data.
I appreciate the time taken by the reader to comment on the article – this is additional input that is certainly welcome.
According to our experience, the satellite products are not yet “mature” enough to capture the variability of rainfall erosivity. According to Bezak (2021), 11% of the erosive events contribute to around 50% of the total erosivity. This was done based on the detailed rainfall erosivity records in GloREDa.
Similar observations have been done by Matthews et al. (2022) based on 300,000 erosive events.
Therefore, based on measured erosivity data, very few events are the ones who contribute to major part of total erosivity. Unfortunately, the satellite products are not yet mature enough to capture the high erosive events.
The satellite products are tending to smooth the high erosive values.
This is overall an excellent point. I want to provide a full answer, but first I want to stress that I think the value of satellite-based and gauge-based data is very different, and any discussion of the two side-by-side does need to acknowledge their relative merits.
First – I am not sure I completely agree that satellite rainfall is guaranteed to underperform gauges when it comes to estimating very large rainfall events. For areas where gauge density is very high, this is likely to be true, but the local rainfall intensity in extreme rainfall events can vary widely and associated wind can also interrupt effective gauge measurement (e.g. Medlin et al. 2007). Satellite rainfall can underestimate extreme rainfall if the overpass of the Microwave satellite does not coincide with peak rainfall intensity, but studies have also shown that the uncertainty of satellite rainfall products is lower at higher rainfall intensities (e.g. Tian and Peters-Lidard (2010)).
Using Bezak et al. 2021 as inspiration, I have explored the storm histories of 4 locations from around the world; two in areas of concern for soil erosion (Near Wichita, USA, and Lucknow, North India) one in a critical region of degradation where the IMERG estimate exceeds GloREDa (Central Sierra Leone) and also near San Pedro de Atacama, in the Atacama desert of Northern Chile. Bezak et al. (2021) clearly show that the 11% of erosive events are responsible for the bulk of erosivity, but I wanted to test what maximum rainfall values would correspond to the storms most responsible for erosivity. In Lucknow and Wichita, more than 80% of the total rainfall kinetic energy calculated from the satellite data comes from storms with a peak rainfall intensity of <30mm/hr. In Sierra Leone, this value is much higher (<70mm/hr), but the same overall pattern emerges. In the Atacama, only three rainfall storm events were found over 20 years of data.
Although one could argue that the satellite products may just be missing or underestimating the large storms, the fact that the Sierra Leone case shows the same patterns as in Lucknow and Wichita suggests that the satellite is capable of capturing rainfall events far in excess of the 30mm/hr events (or lower) that are relevant in those locations (since it does so in Sierra Leone). I suggest that although the larger rainfall events do contribute the bulk of the erosivity, those rainfall events are in many locations are not reaching a maximum intensity that would lead them to be significantly underestimated; the key types of rainfall event that are underestimated by satellite products are major tropical storms (Marc et al. 2022), occurring in areas of extreme topography, where rainfall may exceed 100-200mm/hr – somewhat less relevant to agricultural soil degradation. Since the underestimation of extreme rainfall by satellites is generally because peak rainfall is missed by the microwave satellite overpass, we would still expect to sample some of the largest events - i.e., I would not expect the IMERG based estimates to be systematically lower for every storm. Curves for all 4 locations are shown below.
Overall – I agree that there are challenges to using satellite precipitation data, and in revising this study I want to avoid giving the sense that one product is ‘better’ than the other. It is clear that GloREDa is highly valuable in large parts of the globe. In revising this study I hope to provide a more nuanced perspective about the value of satellite vs gauge based products, but not to suggest that one or the other is superior.
Therefore, there will be differences between your results and the ones of GloREDa in high erosive areas (continents). That is the case for Table 2.
I would also to propose a comparison per climatic zone where you will find big discrepancies in the tropical zones.
This is certainly interesting, and something I will be pursuing in future study.
The global erosivity map (Panagos et al., 2017) has been tested and evaluated against the 3,625 measured R-factor values. Please see the figure 4 in the Panagos et al. (2017) and the excellent performance of Global assessment.
As you state in the manuscript, seems that you have estimated better than GloREDa which is not the case. If you insist your statement, you should prove that you perform better than GloREDa in the measured 3,625 stations. The GloREDa measured stations data will be available soon with a data paper.
I look forward to seeing the data! Thank you for letting me know. In revision, as mentioned above, I want to rephrase to ensure it is clear I am not stating that the satellite product is ‘better’ that GloREDa – this is more a comparison of the two for illustrative purposes.
Another remark: The MFI is much problematic and this has been shown in a recent review of Chen et al (2023) .
In this case, I entirely concur. In revising the study, I want to ensure that it is clear to the reader that although I am testing MFI here I agree that it should not be widely used. Since prior authors have used TRMM era satellite precipitation products to estimate MFI, it seems important to at least mention it.
In my opinion, a mixture of satellite products with measured GloREDa would be an ideal and operational solution. That is why the EU Soil Observatory launched also a data collection campaign to get more measured stations data for GloREDa. More info about this call for data in the European Soil Data Centre 2.0 newsletter (February 2023).
I think this is an excellent overall point. I address it more fully above, but I should stress that a combined model is likely to be the best performing output. In this case, the intent of the study is to discuss the value and limitations of a solely IMERG-derived model, but a more general assessment would include diverse input datasets.
Lines 30-35: the cost of soil erosion at global scale, at least for agricultural productivity losses has been estimated to about 8 billion dollars per year. You can find more information in Sartori et al. (2019).
Thank you – an excellent point – I will add this in revision.
In Europe, “Ballabio et al (2017) – Mapping Monthly erosivity in Europe” have developed monthly erosivity maps and datasets based on REDES. Similar has done also for GloREDa at global scale and an article is under preparation to present the monthly erosivity maps. Therefore, measured R-factor data on GloREDa can derive monthly erosivity maps at global scale.
Again – I am very much looking forward to reading the upcoming study and comparing results.
References:
Bezak, N., Mikoš, M., Borrelli, P., Liakos, L. and Panagos, P., 2021. An in-depth statistical analysis of the rainstorms erosivity in Europe. Catena, 206, p.105577.
Matthews, F., Panagos, P. and Verstraeten, G., 2022. Simulating event-scale rainfall erosivity across European climatic regions. Catena, 213, p.106157.
Chen, W., Huang, Y.C., Lebar, K. and Bezak, N., 2023. A systematic review of the incorrect use of an empirical equation for the estimation of the rainfall erosivity around the globe. Earth-Science Reviews, p.104339.
Citation: https://doi.org/10.5194/egusphere-2022-1315-AC1
-
AC1: 'Reply on CC1', Robert Emberson, 27 Apr 2023
Robert Alexander Emberson
Robert Alexander Emberson
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 405 | 152 | 29 | 586 | 64 | 13 | 15 |
- HTML: 405
- PDF: 152
- XML: 29
- Total: 586
- Supplement: 64
- BibTeX: 13
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1