the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Multi-scale EO-based agricultural drought monitoring system for operative irrigation networks management
Abstract. Drought prediction is crucial especially where the rainfall regime is irregular and agriculture is mainly based on irrigated crops, such as in Mediterranean countries. In this work, the main objective is to develop an EO-based agricultural drought monitoring system (ADMOS) for operative irrigation networks management at different spatial and temporal scales. Different levels of drought are identified based on an integrated indicator combining anomalies of rainfall, soil moisture, land surface temperature and vegetation indices, allowing to consider the different droughts types and their timing looking on the end-user’s perspective. Multiple remote sensing data, which differ on sensing techniques, spatial and temporal resolutions and electromagnetic frequencies, are used for each anomaly computation. The analyses have been performed over two Irrigation Consortia in Italy (the Chiese and Capitanata ones), which differ for climate, irrigation volumes and techniques, and crop types. The obtained results show a negative correlation between cumulated ADMOS and the irrigation volumes in the Capitanata area, while in the Chiese Consortium a zero correlation is obtained with an almost constant amount of irrigation volumes provided to the crops every year independently from the drought condition. In both areas, crop yields seem to be almost uncorrelated to the drought index, as production is highly sustained by irrigation. Moreover, discrepancies on the anomalies sign is observed, especially when soil moisture is considered. The results also clearly show that asynchronies may exist especially between soil moisture anomalies and vegetation or land surface temperature anomalies.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(13361 KB)
Interactive discussion
Status: closed
-
CC1: 'Comment on nhess-2022-260', Qi Liu, 22 Dec 2022
Authors' aim is to establish a daily system, but the time scales of your input variables are monthly,16 days, 8-days and the like. I think it has great uncertainties that you resample monthly and 16-days scales into daily scale. Here, my suggest is that only use precipitation as the input, there are many daily precipitation data, and you can avoid the monthly scale. Also, the spatial resolustions of data authors used are very different, I think you can clarify how to resolve this problem, because your small study area.
Line 110, this paragraph, I think author should modify it into a flow chart, because you aim is to establish a drought monitoring system. It will be more clear for potential readers. Actually, soil moisture is an effient indicator for vegetation drought, but in this manuscript "2) soil moisture shortage is evaluated by the soil moisture anomaly (SMA), 3) vegetation drying is identified with a land surface temperature anomaly (LSTA)", I think it is unreasonable. And, my suggestion is use soil moisture and root-zone soil moisture as the vegetation water stress, LST anomalies as the heat stress and NDVI anomalies as the vegetation conditions.
Citation: https://doi.org/10.5194/nhess-2022-260-CC1 -
AC1: 'Reply on CC1', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC1-supplement.pdf
-
AC1: 'Reply on CC1', Chiara Corbari, 30 Mar 2023
-
RC1: 'Comment on nhess-2022-260', Anonymous Referee #1, 02 Jan 2023
With interest I have read this manuscript addressing a new drought index. The topic issued by the manuscript “Multi-scale EO-based agricultural drought monitoring system for operative irrigation networks management” is relevant and the structure of the manuscript is well organized. I personally think the manuscript has potential to be published, but there some aspects that need to be clarified by the authors, before I can recommend this work for publication. Based on this is that I recommend major revisions.
General comments:
Based on the title of this manuscript, please further discuss the added information of the ADMOS index for both (Chiese and Capitanata) operative irrigation networks. In particular, how this index can improve the water management? Is it applicable only for irrigated crop regions? On this later aspect it is also important to be clear in regard to the objective of ADMOS, it is meant only for monitoring or also for predicting droughts?
Another aspect which needs to be clarified and justified is related with the performing part of the evaluation of the ADMOS index against crop yield, is this evaluation valid if crops receive water through irrigation?
From the manuscript, it is not clear on what temporal scale the ADMOS index works, is the monthly, weekly? In addition and related to this, discuss what temporal scales which are needed by operative these different irrigation networks.
A particular concern is related with the high vs. low resolution analysis. The results are based on a small sample of data which show a considerable dispersion in the relationship between ADMOS and rainfall and rainfall+irrigation. Why the authors seek for a linear relationship? Should the relationship be linear? How robust or significative are these results? Please discuss the potential drawbacks of all these considerations in the analysis.
Specific comments:
Improve in general the figure caption descriptions.
On several parts of the manuscript the word trend is used, but it remains unclear the particular meaning of it. Like for example “seasonal trend” and the examples listed below. Please clarify this aspect across all the manuscript.
Why is that the authors start the abstract section with “Drought prediction” if they will focus on monitoring? Of course both topics are of major important for a drought early warning system, but in this case I would suggest using the word monitoring.
I suggest modify “electromagnetic frequencies” for spectral bands
Are they Drought monitoring systems for irrigation regions i other regions of the planet? If they are, I consider that a paragraph related to irrigation networks background and how they use drought indices is needed.
On what temporal window is the ADMOS working? Weekly, monthly? Please clarify this
Line 76 Is the Global Integrated Drought Monitoring and Prediction System (GIDMaPS, http://drought.eng.uci.edu) still operational?
Line 79: CDI index should be an indicator and not an index following the WMO definition as it uses different indices separately and not combined in only one index as the SMADI (Soil Moisture Agricultural Index) for example. It is also important to highlight that the CDI uses different time dates for each variable, which is different to the USDM approach.
Line 149: An average irrigation volume of about 1200 mm is provided during the crop season, over a mean precipitation value of 250 mm. How are the irrigations estimated?
Line 283: “SMOS and SMAP anomalies do not show a seasonal trend as clear as that of the ESA-CCI datasets.” But ESA-CCI considers a longer time period. What is meant with seasonal trend?
Line 307: “less peaked SM trend” As the authors don´t mention a trend analysis, please clarify what is meant with trend?
Why the authors use SPI-1 and not SPI-3 or SPI-6?
Please specify what is the SMOS Root zone product.
Line 483: Please modify (Figure not shown) for (not shown)
Figure 12. In this figure is not clear while the ADMOS varies from 0 to -500, please explain clarify with more detail this values as the index form its definition varies from 1 to -4. Add also these details in the figure caption. Is it the accumulation value? Also, use the same amount of decimals for the R2, and I recommend changing the units of m3 to millions of m3 or equivalent.
Line 532: What are the crop impacts of this “too much water is probably provided to the crops “? Please discus how would the ADMOS help to this.
Line 555: Please clarify this sentence: “In particular, seven to eight soil moisture products anomalies have been compared and generally low Pearson correlation values are found with a better correlation in the Chiese area, probably due to higher average yearly rainfalls which correspond with a more stable, less peaked SM trend, easier to reproduce from products working at different resolutions and with different algorithms.” What is meant with 7 to 8 SM products? Is not clear what is meant with less peaked SM trend? What is meant with low Pearson correlation values are found with a better correlation in the Chiese area?
Line 567: Change “plat” for plant
Figure 11. Please clarify on what temporal scale the index was accumulated? Monthly, Yearly?
Citation: https://doi.org/10.5194/nhess-2022-260-RC1 -
AC2: 'Reply on RC1', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Chiara Corbari, 30 Mar 2023
-
RC2: 'Comment on nhess-2022-260', Anonymous Referee #2, 26 Jan 2023
Review of “Multi-scale EO-based agricultural drought monitoring system for operative irrigation networks management”
This study proposes a methodology to assess drought conditions in two irrigation polygons of Italy based on different data sources obtained from satellite data. My opinion is that this manuscript should not be published because it is affected by several formal and methodological problems. The methodology is not well explained and justified and in general all the manuscript it is very difficult to follow. The use of different data sources of different origin makes difficult to know the connection between meteorological and agricultural droughts. The results are also presented in a very confuse way, with different plots in which it is not possible to obtain a clear message about the relationship between metrics and the evolution of the existing anomalies. Below I am providing specific comments that support my general assessment and my suggestion to reject this manuscript.
9-14: Very confuse summary of the results. What is a cumulative drought monitoring system? A drought index can be cumulative, but I wonder what are the authors referring in relation to a drought monitoring system. It is not clear what kind of correlation the authors are referring.
17: There are much better references to refer to drought characteristics and impacts. It seems that the authors have simply cited some papers related to drought… See e.g. IPCC AR6 Chapters 8 and 11 for a summary of drought complexity and implications.
23: In irrigated lands water shortage can be relevant but in rainfed agricultural areas precipitation (but also temperature and atmospheric demand) play a very important role.
25: Definitively this is not the best to refer to land atmosphere feedbacks and droughts. Again the citations are very poorly selected, which gives a very bad impression as it seems that references are only located randomly in the text to justify the use of references. About this topic, I would recommend to read Miralles DG, Gentine P, Seneviratne SI, Teuling AJ. 2019. Land–atmospheric feedbacks during droughts and heatwaves: state of the science and current challenges. Annals of the New York Academy of Sciences. Blackwell Publishing Inc., 1436(1): 19–35. ttps://doi.org/10.1111/nyas.13912.
.25. It should be irrigated agriculture.
.28. I would say better: “during the dry season in water limited regions”. I would not refer to specific regions.
.29. One-sentence paragraph? I also find this very disconnected of the context. I suggest to remove this sentence as it does not provide any relevant message.
.35. Again poorly and non-suitable citations. Vicente-Serrano 2006 analyses spatial pattern of meteorological drought but there is nothing on this study on the dynamic of different types of drought. The authors should revisit all the citations of the manuscript. The poor and unsuitable citation approach is a solid formal argument to suggest the rejection of the manuscript.
.40. Cite the WMO guidelines for SPI in which it is recommended as a reference drought index.
.45. The SPEI is perfectly comparable in time and space (as the SPI) Also the Standard Palmer Drought Index is perfectly comparable spatially, so the argument of the authors is not correct. Why is the use of potential evapotranspiration a limitation? I would say that given atmospheric evaporative demand has a relevant influence on drought severity it should be an advantage.
48-56: If remote sensing soil moisture is affected by so large uncertainties, what is the justification of its used? The low correlations found among soil moisture datasets presented below even justifies more my assessment.
.57. land surface temperature has been widely used. See e.g. TCI developed by Felix Kogan and the drought monitoring systems (and studies) that use it.
.71. The optimal solution is really to relate drought objective metrics with impacts and then select the most suited approach. For this purpose, empirical analysis that relates drought indices and impacts is needed.
.83. I wonder if the authors are proposing a drought monitoring system or a drought index. I believe that they are developing a drought index.
.91. A new drought index should be evaluated with impact data (e.g. crop damages and yields). The volumes of irrigation may be related to several other factors including water availability in reservoir storages, groundwater, etc.
.105. Crop yield is also constrained by VPD anomalies and increases in the atmospheric evaporative demand, particularly under low soil moisture conditions.
.106. Increase in crop temperature can be also caused by decreased leaf stomatal conductance as consequence of increased VPD.
.115. Are the different variables following a normal distribution in order to apply this equation?
.117-120. It is confuse if the authors are using the monthly or daily scales.
.124. Figure 1 is confuse. It is not clear how the different indices are merged in order to generate the ADMOS. What is the criterion followed to select the thresholds?
.130. Are equations 2 and 3 necessary? I do not think necessary to include the equation of the Pearson’s r statistic.
.142. was affected? As the sentence refers to 2012 I think better use the past. Same 143.
.159. How robust is the calculation of SPI and the other drought indices based only on 20 years of data? e.g.in 168 in is indicated that 13 years of data are used. This will provide very uncertain indices. WMO recommends at least 30 years.
Section 22.3. It is very confuse how all these soil moisture indices of different resolution and time span are used together. There is not explanation and justification of why these different soil moisture products are used and what is the advantage of using different datasets if they show low agreement.
.214. Why thermal bands are resampled to 100 meters?
Figure 3. It is impossible to identify the drought periods according to the SPI based on this plot. I would suggest to be replaced all the plots by time series.
Figure 4. Same that for precipitation. I do not think it is possible to compare these different datasets based on these plots. The statistics that compare the datasets suggest strong uncertainties and difficulties for comparison. I do not think that the authors are providing realiable combination of the different datasets and, in addition, validation is not provided.
Same comments are valid for surface temperature and vegetation indices. My impression is that authors have used all the information they have found by different sources, but they have not considered any coherent approach to analyse drought severity, to validate the different products and to stablish uncertainties associated to the datasets. In addition, the information is not showed in a coherent way and it is very difficult to determine the evolution of the anomalies in the different metrics and also to establish comparisons.
Figures 9 and 10: Based on the uncertainties in the datasets and methods indicated above, the uncertainty in the results described based on these figures are very strong. It is not possible to infer on which dataset (e.g. soil moisture, surface temperature and vegetation index) this plot is generated.
.445. I cannot identify how the different products are combined in order to generate the ADMOS and it is very confuse the use of different data products at the same time and in an independent way.
.545: I agree that different indices are compared, but this is not done in this study. There is not validation of different metrics and selection of most suitable according to empirical information.
.549: But the remote sensing information is not used in a coherent way considering a careful validation. Several datasets are put together considering different time periods and I cannot find a coherent message by so confuse merging.
Citation: https://doi.org/10.5194/nhess-2022-260-RC2 -
AC3: 'Reply on RC2', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Chiara Corbari, 30 Mar 2023
-
RC3: 'Comment on nhess-2022-260', Anonymous Referee #3, 03 Feb 2023
I believe the topic is relevant, and I think the paper might merit the chance to be published eventually, but it definitely needs major revisions.
- Some general comments
I believe the purpose of the paper is lost in divagations due to the reporting style and contents. The ADMOS indicator results should be more explained, in particular regarding its capacity to reproduce impacts on yield or its covariation with irrigation. Since the indicator values and thresholds are arbitrary, it is essential to see if such thresholds are capable of marking when impacts on crop yields are to be expected- which seems not the case, or when irrigation inputs are necessary, and in what order of magnitude. It is not clear to me in the end it is possible to use ADMOS to recommend better water management for agriculture.
Since several indices have been calculated to compose the ADMOS, I was expecting a separate comparison between them and the irrigation and yields. Maybe them separately have better prediction capacity than the ADMOS itself, but it was not showed. The conclusions say that you prove that “droughts cannot be described by one single indicator but there is the need first to select the correct physical index for detecting a drought type and secondly to use different drought indices to identify specific conditions”, but I don´t see how you compare the ADMOS predictive capacity (for irrigation or crop yields) and the predictive capacity of all your anomalies series for P, SM, temperature or VI.
Also, it is very difficult to understand the spatial and temporal aggregation scales for the indices and the ADMOS indicator itself. It seems it is calculated daily, but then the yields are annual for the entire consortium, not sure though. What about the irrigation values? They don´t even appear in the data list. How often are them recorded? How are them aggregated over time?
On the other hand, the paper devotes too much space (6 full pages, and 4 chapers, 3.1 to 3.4) to debate the differences between each product (RMSE, r) in each of the variables, when it is not the essential result and could be solved with a summary table and a paragraph of explanation.
I have doubts if- from a statistical point of view- it is recommendable to accumulate an indicator (ADMOS) whose values are categories and not quantifications. For example, accumulating two time steps with ADMOS -1 is -2, but it is not necessarily equivalent to another time step with -2, and still they are added up. That might explain that in summer there are peaks in the accumulation. It should be better justified why it is computed like this.
In any case, does a value of cumulative ADMOS at certain point means that at that moment a certain irrigation volume should be applied? In the conclusions, the papers says “This ADMOS might help irrigation districts managers and farmers to activate the preventive protection actions to try to avoid water volume and crop yield losses.”, but after reading the study I don´t really see how, it would merit an explanation.
Also, the potential lags / delays are not taken into consideration for comparing the evolution of the different variables. Only SPI1 is calculated and confronted with the situation at the same time in other variables, presumably at the daily level, but not sure you can capture the propagation of the anomalies like this. For example, time-steps with very high negative anomalies in SM or Vegetation Index can be concealed by the fact that that day in particular it rained a lot, the ADMOS would show “surplus of water”, but the system has not recovered yet. There is not a way to know if ADMOS then marks real issues in terms of irrigation needs or yield losses at the daily level. With more granular data on these two impacts, tests could be made.
More generally, I do not really see how the ADMOS is helping identify the main agricultural drought problems in the pilots used. More examples, maybe using a particular event, would help strengthening that point.
Last, on a different note, the used references do not seem always the most relevant to justify the points the authors make, sometimes it is just general drought literature, not even focused on reviewing similar efforts.
- Some specific comments
450- “All the curves have a common trend: from the begin of the year till March the curves have gentle slopes, then from March to October they are very steep due to drought conditions, and at the end of the year they return flat.” I think this reveals that the drought indicator is more marking stress than drought, as it points to systematic intensification in summer
Figure 9 and 10. “Synchronicity among the different variables’ anomalies in… “- I don´t think the color code and the graphs are easy to read and interpret.
515- “Following the principles of CAP to improve irrigation management”- Spell CAP.
540- “This methodology improves the traditional analysis, which are generally analysed by considering only soil moisture anomalies”. Many drought analyses rely on other variables.
The writing is confusing in many parts and there are several typos or incongruences, it needs a language revision.
Citation: https://doi.org/10.5194/nhess-2022-260-RC3 -
AC4: 'Reply on RC3', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC4-supplement.pdf
Interactive discussion
Status: closed
-
CC1: 'Comment on nhess-2022-260', Qi Liu, 22 Dec 2022
Authors' aim is to establish a daily system, but the time scales of your input variables are monthly,16 days, 8-days and the like. I think it has great uncertainties that you resample monthly and 16-days scales into daily scale. Here, my suggest is that only use precipitation as the input, there are many daily precipitation data, and you can avoid the monthly scale. Also, the spatial resolustions of data authors used are very different, I think you can clarify how to resolve this problem, because your small study area.
Line 110, this paragraph, I think author should modify it into a flow chart, because you aim is to establish a drought monitoring system. It will be more clear for potential readers. Actually, soil moisture is an effient indicator for vegetation drought, but in this manuscript "2) soil moisture shortage is evaluated by the soil moisture anomaly (SMA), 3) vegetation drying is identified with a land surface temperature anomaly (LSTA)", I think it is unreasonable. And, my suggestion is use soil moisture and root-zone soil moisture as the vegetation water stress, LST anomalies as the heat stress and NDVI anomalies as the vegetation conditions.
Citation: https://doi.org/10.5194/nhess-2022-260-CC1 -
AC1: 'Reply on CC1', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC1-supplement.pdf
-
AC1: 'Reply on CC1', Chiara Corbari, 30 Mar 2023
-
RC1: 'Comment on nhess-2022-260', Anonymous Referee #1, 02 Jan 2023
With interest I have read this manuscript addressing a new drought index. The topic issued by the manuscript “Multi-scale EO-based agricultural drought monitoring system for operative irrigation networks management” is relevant and the structure of the manuscript is well organized. I personally think the manuscript has potential to be published, but there some aspects that need to be clarified by the authors, before I can recommend this work for publication. Based on this is that I recommend major revisions.
General comments:
Based on the title of this manuscript, please further discuss the added information of the ADMOS index for both (Chiese and Capitanata) operative irrigation networks. In particular, how this index can improve the water management? Is it applicable only for irrigated crop regions? On this later aspect it is also important to be clear in regard to the objective of ADMOS, it is meant only for monitoring or also for predicting droughts?
Another aspect which needs to be clarified and justified is related with the performing part of the evaluation of the ADMOS index against crop yield, is this evaluation valid if crops receive water through irrigation?
From the manuscript, it is not clear on what temporal scale the ADMOS index works, is the monthly, weekly? In addition and related to this, discuss what temporal scales which are needed by operative these different irrigation networks.
A particular concern is related with the high vs. low resolution analysis. The results are based on a small sample of data which show a considerable dispersion in the relationship between ADMOS and rainfall and rainfall+irrigation. Why the authors seek for a linear relationship? Should the relationship be linear? How robust or significative are these results? Please discuss the potential drawbacks of all these considerations in the analysis.
Specific comments:
Improve in general the figure caption descriptions.
On several parts of the manuscript the word trend is used, but it remains unclear the particular meaning of it. Like for example “seasonal trend” and the examples listed below. Please clarify this aspect across all the manuscript.
Why is that the authors start the abstract section with “Drought prediction” if they will focus on monitoring? Of course both topics are of major important for a drought early warning system, but in this case I would suggest using the word monitoring.
I suggest modify “electromagnetic frequencies” for spectral bands
Are they Drought monitoring systems for irrigation regions i other regions of the planet? If they are, I consider that a paragraph related to irrigation networks background and how they use drought indices is needed.
On what temporal window is the ADMOS working? Weekly, monthly? Please clarify this
Line 76 Is the Global Integrated Drought Monitoring and Prediction System (GIDMaPS, http://drought.eng.uci.edu) still operational?
Line 79: CDI index should be an indicator and not an index following the WMO definition as it uses different indices separately and not combined in only one index as the SMADI (Soil Moisture Agricultural Index) for example. It is also important to highlight that the CDI uses different time dates for each variable, which is different to the USDM approach.
Line 149: An average irrigation volume of about 1200 mm is provided during the crop season, over a mean precipitation value of 250 mm. How are the irrigations estimated?
Line 283: “SMOS and SMAP anomalies do not show a seasonal trend as clear as that of the ESA-CCI datasets.” But ESA-CCI considers a longer time period. What is meant with seasonal trend?
Line 307: “less peaked SM trend” As the authors don´t mention a trend analysis, please clarify what is meant with trend?
Why the authors use SPI-1 and not SPI-3 or SPI-6?
Please specify what is the SMOS Root zone product.
Line 483: Please modify (Figure not shown) for (not shown)
Figure 12. In this figure is not clear while the ADMOS varies from 0 to -500, please explain clarify with more detail this values as the index form its definition varies from 1 to -4. Add also these details in the figure caption. Is it the accumulation value? Also, use the same amount of decimals for the R2, and I recommend changing the units of m3 to millions of m3 or equivalent.
Line 532: What are the crop impacts of this “too much water is probably provided to the crops “? Please discus how would the ADMOS help to this.
Line 555: Please clarify this sentence: “In particular, seven to eight soil moisture products anomalies have been compared and generally low Pearson correlation values are found with a better correlation in the Chiese area, probably due to higher average yearly rainfalls which correspond with a more stable, less peaked SM trend, easier to reproduce from products working at different resolutions and with different algorithms.” What is meant with 7 to 8 SM products? Is not clear what is meant with less peaked SM trend? What is meant with low Pearson correlation values are found with a better correlation in the Chiese area?
Line 567: Change “plat” for plant
Figure 11. Please clarify on what temporal scale the index was accumulated? Monthly, Yearly?
Citation: https://doi.org/10.5194/nhess-2022-260-RC1 -
AC2: 'Reply on RC1', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Chiara Corbari, 30 Mar 2023
-
RC2: 'Comment on nhess-2022-260', Anonymous Referee #2, 26 Jan 2023
Review of “Multi-scale EO-based agricultural drought monitoring system for operative irrigation networks management”
This study proposes a methodology to assess drought conditions in two irrigation polygons of Italy based on different data sources obtained from satellite data. My opinion is that this manuscript should not be published because it is affected by several formal and methodological problems. The methodology is not well explained and justified and in general all the manuscript it is very difficult to follow. The use of different data sources of different origin makes difficult to know the connection between meteorological and agricultural droughts. The results are also presented in a very confuse way, with different plots in which it is not possible to obtain a clear message about the relationship between metrics and the evolution of the existing anomalies. Below I am providing specific comments that support my general assessment and my suggestion to reject this manuscript.
9-14: Very confuse summary of the results. What is a cumulative drought monitoring system? A drought index can be cumulative, but I wonder what are the authors referring in relation to a drought monitoring system. It is not clear what kind of correlation the authors are referring.
17: There are much better references to refer to drought characteristics and impacts. It seems that the authors have simply cited some papers related to drought… See e.g. IPCC AR6 Chapters 8 and 11 for a summary of drought complexity and implications.
23: In irrigated lands water shortage can be relevant but in rainfed agricultural areas precipitation (but also temperature and atmospheric demand) play a very important role.
25: Definitively this is not the best to refer to land atmosphere feedbacks and droughts. Again the citations are very poorly selected, which gives a very bad impression as it seems that references are only located randomly in the text to justify the use of references. About this topic, I would recommend to read Miralles DG, Gentine P, Seneviratne SI, Teuling AJ. 2019. Land–atmospheric feedbacks during droughts and heatwaves: state of the science and current challenges. Annals of the New York Academy of Sciences. Blackwell Publishing Inc., 1436(1): 19–35. ttps://doi.org/10.1111/nyas.13912.
.25. It should be irrigated agriculture.
.28. I would say better: “during the dry season in water limited regions”. I would not refer to specific regions.
.29. One-sentence paragraph? I also find this very disconnected of the context. I suggest to remove this sentence as it does not provide any relevant message.
.35. Again poorly and non-suitable citations. Vicente-Serrano 2006 analyses spatial pattern of meteorological drought but there is nothing on this study on the dynamic of different types of drought. The authors should revisit all the citations of the manuscript. The poor and unsuitable citation approach is a solid formal argument to suggest the rejection of the manuscript.
.40. Cite the WMO guidelines for SPI in which it is recommended as a reference drought index.
.45. The SPEI is perfectly comparable in time and space (as the SPI) Also the Standard Palmer Drought Index is perfectly comparable spatially, so the argument of the authors is not correct. Why is the use of potential evapotranspiration a limitation? I would say that given atmospheric evaporative demand has a relevant influence on drought severity it should be an advantage.
48-56: If remote sensing soil moisture is affected by so large uncertainties, what is the justification of its used? The low correlations found among soil moisture datasets presented below even justifies more my assessment.
.57. land surface temperature has been widely used. See e.g. TCI developed by Felix Kogan and the drought monitoring systems (and studies) that use it.
.71. The optimal solution is really to relate drought objective metrics with impacts and then select the most suited approach. For this purpose, empirical analysis that relates drought indices and impacts is needed.
.83. I wonder if the authors are proposing a drought monitoring system or a drought index. I believe that they are developing a drought index.
.91. A new drought index should be evaluated with impact data (e.g. crop damages and yields). The volumes of irrigation may be related to several other factors including water availability in reservoir storages, groundwater, etc.
.105. Crop yield is also constrained by VPD anomalies and increases in the atmospheric evaporative demand, particularly under low soil moisture conditions.
.106. Increase in crop temperature can be also caused by decreased leaf stomatal conductance as consequence of increased VPD.
.115. Are the different variables following a normal distribution in order to apply this equation?
.117-120. It is confuse if the authors are using the monthly or daily scales.
.124. Figure 1 is confuse. It is not clear how the different indices are merged in order to generate the ADMOS. What is the criterion followed to select the thresholds?
.130. Are equations 2 and 3 necessary? I do not think necessary to include the equation of the Pearson’s r statistic.
.142. was affected? As the sentence refers to 2012 I think better use the past. Same 143.
.159. How robust is the calculation of SPI and the other drought indices based only on 20 years of data? e.g.in 168 in is indicated that 13 years of data are used. This will provide very uncertain indices. WMO recommends at least 30 years.
Section 22.3. It is very confuse how all these soil moisture indices of different resolution and time span are used together. There is not explanation and justification of why these different soil moisture products are used and what is the advantage of using different datasets if they show low agreement.
.214. Why thermal bands are resampled to 100 meters?
Figure 3. It is impossible to identify the drought periods according to the SPI based on this plot. I would suggest to be replaced all the plots by time series.
Figure 4. Same that for precipitation. I do not think it is possible to compare these different datasets based on these plots. The statistics that compare the datasets suggest strong uncertainties and difficulties for comparison. I do not think that the authors are providing realiable combination of the different datasets and, in addition, validation is not provided.
Same comments are valid for surface temperature and vegetation indices. My impression is that authors have used all the information they have found by different sources, but they have not considered any coherent approach to analyse drought severity, to validate the different products and to stablish uncertainties associated to the datasets. In addition, the information is not showed in a coherent way and it is very difficult to determine the evolution of the anomalies in the different metrics and also to establish comparisons.
Figures 9 and 10: Based on the uncertainties in the datasets and methods indicated above, the uncertainty in the results described based on these figures are very strong. It is not possible to infer on which dataset (e.g. soil moisture, surface temperature and vegetation index) this plot is generated.
.445. I cannot identify how the different products are combined in order to generate the ADMOS and it is very confuse the use of different data products at the same time and in an independent way.
.545: I agree that different indices are compared, but this is not done in this study. There is not validation of different metrics and selection of most suitable according to empirical information.
.549: But the remote sensing information is not used in a coherent way considering a careful validation. Several datasets are put together considering different time periods and I cannot find a coherent message by so confuse merging.
Citation: https://doi.org/10.5194/nhess-2022-260-RC2 -
AC3: 'Reply on RC2', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Chiara Corbari, 30 Mar 2023
-
RC3: 'Comment on nhess-2022-260', Anonymous Referee #3, 03 Feb 2023
I believe the topic is relevant, and I think the paper might merit the chance to be published eventually, but it definitely needs major revisions.
- Some general comments
I believe the purpose of the paper is lost in divagations due to the reporting style and contents. The ADMOS indicator results should be more explained, in particular regarding its capacity to reproduce impacts on yield or its covariation with irrigation. Since the indicator values and thresholds are arbitrary, it is essential to see if such thresholds are capable of marking when impacts on crop yields are to be expected- which seems not the case, or when irrigation inputs are necessary, and in what order of magnitude. It is not clear to me in the end it is possible to use ADMOS to recommend better water management for agriculture.
Since several indices have been calculated to compose the ADMOS, I was expecting a separate comparison between them and the irrigation and yields. Maybe them separately have better prediction capacity than the ADMOS itself, but it was not showed. The conclusions say that you prove that “droughts cannot be described by one single indicator but there is the need first to select the correct physical index for detecting a drought type and secondly to use different drought indices to identify specific conditions”, but I don´t see how you compare the ADMOS predictive capacity (for irrigation or crop yields) and the predictive capacity of all your anomalies series for P, SM, temperature or VI.
Also, it is very difficult to understand the spatial and temporal aggregation scales for the indices and the ADMOS indicator itself. It seems it is calculated daily, but then the yields are annual for the entire consortium, not sure though. What about the irrigation values? They don´t even appear in the data list. How often are them recorded? How are them aggregated over time?
On the other hand, the paper devotes too much space (6 full pages, and 4 chapers, 3.1 to 3.4) to debate the differences between each product (RMSE, r) in each of the variables, when it is not the essential result and could be solved with a summary table and a paragraph of explanation.
I have doubts if- from a statistical point of view- it is recommendable to accumulate an indicator (ADMOS) whose values are categories and not quantifications. For example, accumulating two time steps with ADMOS -1 is -2, but it is not necessarily equivalent to another time step with -2, and still they are added up. That might explain that in summer there are peaks in the accumulation. It should be better justified why it is computed like this.
In any case, does a value of cumulative ADMOS at certain point means that at that moment a certain irrigation volume should be applied? In the conclusions, the papers says “This ADMOS might help irrigation districts managers and farmers to activate the preventive protection actions to try to avoid water volume and crop yield losses.”, but after reading the study I don´t really see how, it would merit an explanation.
Also, the potential lags / delays are not taken into consideration for comparing the evolution of the different variables. Only SPI1 is calculated and confronted with the situation at the same time in other variables, presumably at the daily level, but not sure you can capture the propagation of the anomalies like this. For example, time-steps with very high negative anomalies in SM or Vegetation Index can be concealed by the fact that that day in particular it rained a lot, the ADMOS would show “surplus of water”, but the system has not recovered yet. There is not a way to know if ADMOS then marks real issues in terms of irrigation needs or yield losses at the daily level. With more granular data on these two impacts, tests could be made.
More generally, I do not really see how the ADMOS is helping identify the main agricultural drought problems in the pilots used. More examples, maybe using a particular event, would help strengthening that point.
Last, on a different note, the used references do not seem always the most relevant to justify the points the authors make, sometimes it is just general drought literature, not even focused on reviewing similar efforts.
- Some specific comments
450- “All the curves have a common trend: from the begin of the year till March the curves have gentle slopes, then from March to October they are very steep due to drought conditions, and at the end of the year they return flat.” I think this reveals that the drought indicator is more marking stress than drought, as it points to systematic intensification in summer
Figure 9 and 10. “Synchronicity among the different variables’ anomalies in… “- I don´t think the color code and the graphs are easy to read and interpret.
515- “Following the principles of CAP to improve irrigation management”- Spell CAP.
540- “This methodology improves the traditional analysis, which are generally analysed by considering only soil moisture anomalies”. Many drought analyses rely on other variables.
The writing is confusing in many parts and there are several typos or incongruences, it needs a language revision.
Citation: https://doi.org/10.5194/nhess-2022-260-RC3 -
AC4: 'Reply on RC3', Chiara Corbari, 30 Mar 2023
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2022-260/nhess-2022-260-AC4-supplement.pdf
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
689 | 278 | 64 | 1,031 | 39 | 39 |
- HTML: 689
- PDF: 278
- XML: 64
- Total: 1,031
- BibTeX: 39
- EndNote: 39
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1