the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Sep 2023
Drought and radiation explain swings in Amazon rainforest greenness during the 2015–2016 drought
Abstract. The 2015/16 Amazon drought was characterized by below-average regional precipitation for an entire year, which distinguishes it from the dry-season only droughts in 2005 and 2010. Studies of vegetation indices (VI) derived from optical remote sensing over the Amazonian forests indicated three stages in canopy response during the 2015/16 drought, with below-average greenness during the onset and end of the drought, and above-average greenness during the intervening months. So far, a satisfactory explanation for this broad temporal pattern, and spatial variation within the Amazonian forests of this broad response, has not been found. Better understanding of rainforest behaviors during this unusually long drought should help predict their response to future droughts. We hypothesized that below-average greenness could be explained by water deficit and heat stress occurring beyond the tolerance thresholds of rainforest. To test our hypothesis, we used monthly observations of terrestrial water storage (TWS), land surface temperature (LST) and vapor pressure deficit (VPD) for January 2003–December 2016. First, for each 1° grid cell, we determined the ‘normal’ range of monthly TWS, LST and VPD during non-drought years (i.e. 2003–2016, excluding 2005, 2010, 2015 and 2016), and identified the extreme values of ‘normal’ range, i.e. minimum TWS, maximum LST and maximum VPD. We considered the normal hydrological and thermal ranges to have been exceeded when (1) two or three of these variables were simultaneously beyond their extreme values, or (2) only one variable was beyond the extreme value, but the other two were significantly (p<0.05) different from the average for non-drought years. Using these criteria, regions experiencing hydrological and thermal conditions beyond the ‘normal’ range during different stages of the 2015/16 event were delineated. The results showed a gradual southward shift of these regions: from the north-eastern Amazon in August–October 2015, to the north-central part in November 2015–February 2016 and finally to the southern Amazon in July 2016. The majority of forests within the delimited regions experienced below-average greenness. Conversely, outside of these regions, greenness responded positively to radiation anomalies, as is expected under normal conditions. The opposing influences of drought and radiation anomalies together explained more than 70 % of the observed spatiotemporal patterns in greenness. These results suggest that our exceeding ‘normal’ ranges based approach, combining water storage, temperature and atmospheric moisture demand drivers, can reasonably identify the most likely drought-affected regions at monthly to seasonal time scales. Using observation-based hydrological and thermal condition thresholds can help with interpreting the response of Amazon rainforest to future drought events.
- Preprint
(2197 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on bg-2023-155', Anonymous Referee #1, 09 Oct 2023
The manuscript of Liu et al. is an interesting study about identifying the different environmental drivers of the drought-affected regions of 2015-2016 within the Amazon forest. They show that the regions where water storage, temperature and atmospheric moisture demand exceeded their ‘normal’ ranges agreed with more than 70% of the observed patterns in vegetation greenness. This manuscript has potential and could be a useful contribution to the drought research in the tropics. I do have some questions that I would like to see addressed.
General comments
I am not convinced by how the authors defined the occurrence of extreme values of LST, VPD and TWS in 2015-2016. Extreme values of ‘normal’ years are defined as the lowest or highest mean monthly value, not as the actual lowest or highest value that a ‘normal’ month reached for a certain variable. If there is a lot of variation in monthly TWS, LST or VPD between different years, the mean value doesn’t catch that. Comparing a raw value to a mean without taking into account the normal variation will easily give ‘extreme’ results. The authors combine this metric of extremeness with a significant difference from the monthly mean using a Wilcoxon rank test. I am curious how different the results would be when only one of the requirements is used (i.e. only using the requirement of at least two variables that are more extreme than the normal extreme, or of one extreme variable and the other two being significantly different)? Do they both give similar results or is one more in agreement with the vegetation anomalies than the other? This analysis could be added in the appendix.
It is not clearly explained in the methodology or the results how the regions with VI and PAR anomalies are compared with the extreme VPD, TWS and LST regions. This is the main analysis that leads to the conclusions of the manuscript, so it is important that this part is clear.
Specific comments
I would change the word ‘swings’ in the title. It is not a very common scientific word and it does not get repeated once within the manuscript, making it a strange wording choice.
The abstract feels too technical and methodological. I would make the explanation of the approach shorter, so that the conclusion of the manuscript is more clear. As a reader, my attention would get lost in the technicalities of the abstract and it would not convince me to read the paper (which would be a pity, because the research is valuable).
The authors use a combination of NDVI and EVI to quantify the greenness anomalies of the forest, but there is no mention of the problem with NDVI saturation in high biomass regions.
Line 191-193: Why did you use the non-parametric Wilcoxon rank test instead of calculating standardized anomalies and using them to say when values were significantly different?
Line 195-198: How were these requirements decided? Based on literature (if so, add references)? Based on which requirements gave the highest agreement with the vegetation indices (if so, add the other methods that were tried)?
Line 208-212: Step 3 in methods is not very clear. It might be better to add an explanatory figure such as figure 2? Too much mixed vocabulary used: ‘positive greenness anomaly’, ‘below-average greenness’, ‘exceeding the normal range’ à this makes it difficult to follow. Maybe better to use one word for ‘regions exceeding the normal ranges’, to make the explanation less wordy.
Line 211-212 is not clear: ‘compared with that of only considering VI-ano and PAR-ano’. What does this mean?
Figure 4 does not completely convince me that there are, for example, four months with an extremely low TWS in 2015-2016. How extreme is this compared to the normal variation in non-drought years? Figure 2b-d shows that there are indeed large variations in the monthly values. Maybe add the standard deviation in lines around the monthly dots? This will probably make the figure too crowded to be pretty, but this might be something for the appendix.
Table 2: it is not clear how the percentages are calculated. Is it 72% of all pixels with VI and PAR anomalies in same direction that also exceeded the ND range, or 72% of all pixels that exceeded the ND range that had VI and PAR anomalies in same direction, or …? This could be better explained in the table caption. I don’t understand what the current table caption means.
Technical corrections
Line 52: Add ‘the’ to the Amazon forest response.
Line 56: Add ‘the’ to the NDVI or remove ‘the’ from the EVI.
Line 101: Add ‘the’ to the 2005 drought.
Line 197: Remove ‘being’ from sentence.
Line 214-219: I think this part refers to Figure 3 instead of Figure 4?
Citation: https://doi.org/10.5194/bg-2023-155-RC1 -
AC1: 'Reply on RC1', Yi Y. Liu, 09 Dec 2023
We would like to thank Referee #1 for the constructive comments and suggestions, which will be very helpful in improving the quality of this manuscript. Below are our responses and plans to revise. The comments are in black, and our responses are in italics.
The manuscript of Liu et al. is an interesting study about identifying the different environmental drivers of the drought-affected regions of 2015-2016 within the Amazon forest. They show that the regions where water storage, temperature and atmospheric moisture demand exceeded their ‘normal’ ranges agreed with more than 70% of the observed patterns in vegetation greenness. This manuscript has potential and could be a useful contribution to the drought research in the tropics. I do have some questions that I would like to see addressed.
General comments
R1C1: I am not convinced by how the authors defined the occurrence of extreme values of LST, VPD and TWS in 2015-2016. Extreme values of ‘normal’ years are defined as the lowest or highest mean monthly value, not as the actual lowest or highest value that a ‘normal’ month reached for a certain variable. If there is a lot of variation in monthly TWS, LST or VPD between different years, the mean value doesn’t catch that. Comparing a raw value to a mean without taking into account the normal variation will easily give ‘extreme’ results. The authors combine this metric of extremeness with a significant difference from the monthly mean using a Wilcoxon rank test. I am curious how different the results would be when only one of the requirements is used (i.e. only using the requirement of at least two variables that are more extreme than the normal extreme, or of one extreme variable and the other two being significantly different)? Do they both give similar results or is one more in agreement with the vegetation anomalies than the other? This analysis could be added in the appendix.
Response: Following the referee’s suggestion, we will conduct a sensitivity analysis during the revision to further explore the influences of different definitions of extreme values on our study outcomes. The corresponding figures and tables, illustrating the impact of these variations, will be added in the appendix. In the Discussion section, we will introduce a new paragraph dedicated to the insights gained from this sensitivity analysis.
R1C2: It is not clearly explained in the methodology or the results how the regions with VI and PAR anomalies are compared with the extreme VPD, TWS and LST regions. This is the main analysis that leads to the conclusions of the manuscript, so it is important that this part is clear.
Response: We conducted a comparative analysis between the outcomes derived from two distinct scenarios; they are outlined below.
Scenario 1: It is assumed that VI anomalies are exclusively driven by PAR anomalies, leading to changes in the same direction. Accordingly, we created a map depicting the anticipated direction of VI anomalies (either positive or negative) for each grid cell across the Amazonian forests.
Scenario 2: We first utilized extreme values of TWS, VPD and LST to categorize regions into two groups: (a) those within normal ranges and (b) those exceeding normal ranges. For regions within normal ranges, we hypothesized that VI anomalies would align with PAR anomalies, exhibiting changes in the same direction. In regions exceeding normal ranges, negative VI anomalies are expected, irrespective of the direction of PAR anomalies. Accordingly, we generated another map illustrating the anticipated direction of VI anomalies (either positive or negative) for each grid cell.
By comparing these anticipated VI anomalies with MODIS-observed VI anomalies for all grid cells, we calculate the percentage of observed VI anomalies aligning with the anticipated direction in Scenario 1 and 2, respectively. This comparative analysis allows us to quantify the enhancements (%) achieved in explaining observed VI anomalies through the incorporation of our exceeding normal ranges-based approach.
In the revised manuscript, an explanatory figure depicting this experimental design will be added to the Methods section for enhanced clarity. The figure will contain numbered boxes and these numbers will be referred to in the relevant text to help guide the reader through the analysis method. The new figure will provide a visual ‘road-map’ with the associated text providing full details thus allowing others to replicate our method (a central tenet of our collective scientific endeavour).
Specific comments
R1C3: I would change the word ‘swings’ in the title. It is not a very common scientific word and it does not get repeated once within the manuscript, making it a strange wording choice.
Response: Good point; we will replace the word ‘swings’ with ‘fluctuations’.
R1C4: The abstract feels too technical and methodological. I would make the explanation of the approach shorter, so that the conclusion of the manuscript is more clear. As a reader, my attention would get lost in the technicalities of the abstract and it would not convince me to read the paper (which would be a pity, because the research is valuable).
Response: Following the referee’s suggestions, we plan to use one sentence ‘we proposed an approach to categorize regions into two groups: (1) those within normal hydrological and thermal ranges and (2) those exceeding normal ranges’ to replace the technical parts from line 21 to 26 in the current abstract. We will then use those saved words to ensure that the conclusion / lessons learnt from this contribution is clearer and more accessible to those that read the abstract.
R1C5: The authors use a combination of NDVI and EVI to quantify the greenness anomalies of the forest, but there is no mention of the problem with NDVI saturation in high biomass regions.
Response: In the revised manuscript, we will acknowledge the problem of NDVI signal saturation in high biomass regions and discuss its potential impacts on our study outcomes.
R1C6: Line 191-193: Why did you use the non-parametric Wilcoxon rank test instead of calculating standardized anomalies and using them to say when values were significantly different?
Response: As many hydrologic variables are not normally distributed, using the non-parametric Wilcoxon rank test offers the advantage of not assuming that data are normally distributed.
R1C7: Line 195-198: How were these requirements decided? Based on literature (if so, add references)? Based on which requirements gave the highest agreement with the vegetation indices (if so, add the other methods that were tried)?
Response: The criteria for these requirements were established partially through a review of the existing literature and partially through empirical testing using alternative methods, such as the utilization of the TWS anomaly alone. In the revised manuscript, we will improve the clarity of our methodology by incorporating corresponding references that informed our decisions. Furthermore, we will provide a more explicit rationale for the selection of the combination of all TWS, VPD and LST as key factors in our analysis.
R1C8: Line 208-212: Step 3 in methods is not very clear. It might be better to add an explanatory figure such as figure 2? Too much mixed vocabulary used: ‘positive greenness anomaly’, ‘below-average greenness’, ‘exceeding the normal range’ à this makes it difficult to follow. Maybe better to use one word for ‘regions exceeding the normal ranges’, to make the explanation less wordy.
Response: We will add an explanatory figure in the revised manuscript to demonstrate the comparison analysis in Step 3. Meanwhile, we will reduce the mixed vocabulary and try to only use one or two words and terms in the revised manuscript.
R1C9: Line 211-212 is not clear: ‘compared with that of only considering VI-ano and PAR-ano’. What does this mean?
Response: In this study, we conducted a comparative analysis between the outcomes derived from two distinct scenarios. Please see our response to R1C2 above.
R1C10: Figure 4 does not completely convince me that there are, for example, four months with an extremely low TWS in 2015-2016. How extreme is this compared to the normal variation in non-drought years? Figure 2b-d shows that there are indeed large variations in the monthly values. Maybe add the standard deviation in lines around the monthly dots? This will probably make the figure too crowded to be pretty, but this might be something for the appendix.
Response: During the revision, we will incorporate standard deviation lines around the monthly dots, particularly to highlight the extreme nature of TWS, LST and VPD in 2015-2016 compared to the normal variation in non-drought years.
R1C11: Table 2: it is not clear how the percentages are calculated. Is it 72% of all pixels with VI and PAR anomalies in same direction that also exceeded the ND range, or 72% of all pixels that exceeded the ND range that had VI and PAR anomalies in same direction, or …? This could be better explained in the table caption. I don’t understand what the current table caption means.
Response: We will refine the methodology in the Methods and Results sections during the revision and improve the caption for Table 2.
In reference to the percentages presented in Table 2, a concise explanation is provided here. In this study, we conducted a comparative analysis between the outcomes derived from two distinct scenarios.
Scenario 1: It is assumed that VI anomalies are exclusively driven by PAR anomalies, leading to changes in the same direction. Accordingly, we created a map depicting the anticipated direction of VI anomalies (either positive or negative) for each grid cell across the Amazonian forests.
Scenario 2: We first utilized extreme values of TWS, VPD and LST to categorize regions into two groups: (1) those within normal ranges and (2) those exceeding normal ranges. For regions within normal ranges, we hypothesized that VI anomalies would align with PAR anomalies, exhibiting changes in the same direction. In regions exceeding normal ranges, negative VI anomalies are expected, irrespective of the direction of PAR anomalies. Accordingly, we generated another map illustrating the anticipated direction of VI anomalies (either positive or negative) for each grid cell.
By comparing these anticipated VI anomalies with MODIS-observed VI anomalies for all grid cells, we calculated the percentage of observed VI anomalies aligning with the anticipated direction in Scenario 1 and 2, respectively. This comparative analysis allows us to quantify the enhancements (%) achieved in explaining observed VI anomalies through the incorporation of our exceeding normal ranges-based approach.
Technical corrections
R1C12: Line 52: Add ‘the’ to the Amazon forest response.
Response: Will do.
R1C13: Line 56: Add ‘the’ to the NDVI or remove ‘the’ from the EVI.
Response: Will do.
R1C14: Line 101: Add ‘the’ to the 2005 drought.
Response: Will do.
R1C15: Line 197: Remove ‘being’ from sentence.
Response: Will do.
R1C16: Line 214-219: I think this part refers to Figure 3 instead of Figure 4?
Response: Yes, thanks for your careful review (and apologies for our error); we will change to Figure 3.
Citation: https://doi.org/10.5194/bg-2023-155-AC1
-
AC1: 'Reply on RC1', Yi Y. Liu, 09 Dec 2023
-
RC2: 'Comment on bg-2023-155', Ian Baker, 25 Oct 2023
Article: Drought and radiation explain swings in Amazon rainforest greenness during the 2015-2016 drought
Authors: Liu et al.
Summary: The authors use a combination of TWS, VPD and LST (abbreviations explained in the manuscript, so I’m not going to rehash them here) to explore correlation between drought and greenness, as expressed by MODIS NDVI and EVI. They set a criteria of 2 of 3 variables beyond extreme values, or 1 variable beyond extreme and 2 significantly different from average as being outside of normal range. The idea (as I’m interpreting it) is that Amazonian forests have evolved to maintain ecophysiological function within certain bounds of water, temperature, and humidity stress, and the authors have defined a metric to determine when those bounds have been exceeded. They find that these metrics “explained more than 70% of the observed spatiotemporal patterns in greenness”. When applied to the El Nino of 2015/2016 they find that drought as expressed in VIs in general moves from north to south during the event, and from August 2016 through July 2016 the regional VI progresses from below-normal, near-normal, above-normal, and below-normal stages.
This paper is interesting and well-written, although a bit dense at times. It took me several readings to get my head around the method, but once I did I found it an interesting and thought-provoking paper. Initially I wondered if the authors were neglecting the precipitation variability in the region (total annual precipitation, length of dry season), but I realized that this variability is accounted for in the construction of the ‘normal’ cycles for each gridcell shown in Figure 2. The authors might spell this out explicitly-wouldn’t take more than a couple of words or a sentence added. I do have a lot of questions about the methods and results, but I do not have any objections that would lead me to recommend rejection. As I don’t believe that a major overhaul is required to address my comments, my formal recommendation is that this paper be accepted for publication with minor revisions.
Ian Baker
Colorado State University
October 2023
Ian.baker@colostate.edu
Review: Let’s get started…
There has been an ongoing discussion (or debate, if you want) around the notion of greenness increasing with (mild) drought (Saleska, Huete) or not (Samanta, Morton) for the last 10 years or so. Saleska claims the argument over with the results of Wu, Albert, Restrepo-Coupe and others who find that there is (in parts of the Amazon) a drop and reflush of leaves at the end of the dry season, but the new leaves have reduced photosynthetic capacity until they have ‘matured’ for somewhere around 60-90 days. The authors mention this element of regional ecophysiology, but do so in a rather oblique manner. This paper would have much more impact if the issue was met head-on. The community knows about the debate and recent results, the subject matter of this paper is related to this topic, so why not make a direct comment on it? Here are some specific thoughts about this:
- What do your results suggest about the resolution of this debate?
- Can you relate greenness to GPP in the context of your findings of VI correlation with TWS, LST and VPD, or not?
- If not, do your results suggest something about leaf demography (drop, flush) and how it relates to drought?
- Are there field studies of leaf drop/flush to support an attempt to explain regional behavior?
- Can you relate your conclusions to other studies, such as those that look at SIF in the region (Doughty, Koren)?
I see TRMM precipitation mentioned in section 2.1 and Table 1, but don’t recall seeing mention of precipitation elsewhere, as TWS is the variable used. There are interesting questions around the use of TWS from GRACE and TRMM precip. Neither product is perfect. Precipitation may evaporate off leaves (if light), and if heavy may run off before infiltration. TWS will have a significant contribution from soil well below maximum rooting depth, and that soil water might be irrelevant to the analysis. Furthermore, there may be lags between precipitation and plant function, but these lags may be accounted for by using TWS. If only one precipitation/soil moisture metric is used, then the other need not be listed in Table 1. Additionally, the authors should explain the reasoning behind the choice of one ‘wetness’ product over the other.
The calculation of seasonal cycle (various ND-Ave values) was calculated using the notion of a calendar year (Jan-Dec), but the analysis of the drought event was performed over the time of a ‘water year’ from August-July. In fact, this second methodology makes more sense, as the change in the calendar year comes in the middle of the wet season. My recollection is that many Amazonian researchers perform calculations over the scale of a water year. Why wasn’t the notion of water year used consistently? Does it change the results when compared to the calendar year calculation?
On a related note, the years 2005, 2010 (and for that matter 2015/2016) were not drought everywhere. Were these denoted drought years just because other publications have said so, or was there an actual calculation of the fraction of the target gridcells during the (calendar) year that met drought criteria, and these years had the largest area under drought? Or was it that the metrics used to define drought was most severe?
Specific Comments
- I think the references in lines 214-217 are for figure 3.
- Line 227: Figure 4a shows TWS above non-drought values for the first 2 months, not 3.
- Line 228: “During Stage III, only a small area with TWS<TWSND-Min occurred in the north-east.” This is confusing. Figure 4a shows TWS well below non-drought TWS, with a consistent amplitude, over the period from December 2015 through July 2016. The maps in Figure 5 seem to contradict this. Is it a difference between a few very dry gridcells and a lot of ‘sort of’ dry gridcells? Some explanation of this apparent discrepancy might be helpful.
- Lines 294-299. Are your results consistent with the findings of this study? Do you find that you don’t see a depletion of TWS in the region of the field studies of Fontes? If you do see a TWS depletion, and Fontes says it doesn’t matter, what does that suggest about your method?
References
Albert, L.P., Wu, J., Prohaska, N., de Camargo, P.B., Huxman, T.E., Tribuzy, E.S., Ivanov, V.Y., Oliveira, R.S., Garcia, S., Smith, M.N. and Junior, R.C.O., 2018. Age-dependent leaf physiology and consequences for crown-scale carbon uptake during the dry season in an Amazon evergreen forest. New Phytologist, 219(3), pp.870-884.
Doughty, R., Köhler, P., Frankenberg, C., Magney, T.S., Xiao, X., Qin, Y., Wu, X. and Moore III, B., 2019. TROPOMI reveals dry-season increase of solar-induced chlorophyll fluorescence in the Amazon forest. Proceedings of the National Academy of Sciences, 116(44), pp.22393-22398.
Doughty, R., Xiao, X., Qin, Y., Wu, X., Zhang, Y. and Moore III, B., 2021. Small anomalies in dry-season greenness and chlorophyll fluorescence for Amazon moist tropical forests during El Niño and La Niña. Remote Sensing of Environment, 253, p.112196.
Huete, A.R., Didan, K., Shimabukuro, Y.E., Ratana, P., Saleska, S.R., Hutyra, L.R., Yang, W., Nemani, R.R. and Myneni, R., 2006. Amazon rainforests green‐up with sunlight in dry season. Geophysical research letters, 33(6).
Koren, G., Van Schaik, E., Araújo, A.C., Boersma, K.F., Gärtner, A., Killaars, L., Kooreman, M.L., Kruijt, B., Van Der Laan-Luijkx, I.T., Von Randow, C. and Smith, N.E., 2018. Widespread reduction in sun-induced fluorescence from the Amazon during the 2015/2016 El Niño. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1760), p.20170408.
Morton, D.C., Nagol, J., Carabajal, C.C., Rosette, J., Palace, M., Cook, B.D., Vermote, E.F., Harding, D.J. and North, P.R., 2014. Amazon forests maintain consistent canopy structure and greenness during the dry season. Nature, 506(7487), pp.221-224.
Restrepo‐Coupe, N., Levine, N.M., Christoffersen, B.O., Albert, L.P., Wu, J., Costa, M.H., Galbraith, D., Imbuzeiro, H., Martins, G., da Araujo, A.C. and Malhi, Y.S., 2017. Do dynamic global vegetation models capture the seasonality of carbon fluxes in the Amazon basin? A data‐model intercomparison. Global change biology, 23(1), pp.191-208.
Restrepo-Coupe, N., da Rocha, H.R., Hutyra, L.R., da Araujo, A.C., Borma, L.S., Christoffersen, B., Cabral, O.M., de Camargo, P.B., Cardoso, F.L., da Costa, A.C.L. and Fitzjarrald, D.R., 2013. What drives the seasonality of photosynthesis across the Amazon basin? A cross-site analysis of eddy flux tower measurements from the Brasil flux network. Agricultural and Forest Meteorology, 182, pp.128-144.
Citation: https://doi.org/10.5194/bg-2023-155-RC2 -
AC2: 'Reply on RC2', Yi Y. Liu, 09 Dec 2023
We extend our sincere gratitude to Dr Ian Baker for his valuable comments and insightful suggestions, leading us to delve into thought and improve the quality of this manuscript. Below are our responses and plans to improve our revised manuscript. The comments are in black, and our responses are in italics.
Summary: The authors use a combination of TWS, VPD and LST (abbreviations explained in the manuscript, so I’m not going to rehash them here) to explore correlation between drought and greenness, as expressed by MODIS NDVI and EVI. They set a criteria of 2 of 3 variables beyond extreme values, or 1 variable beyond extreme and 2 significantly different from average as being outside of normal range. The idea (as I’m interpreting it) is that Amazonian forests have evolved to maintain ecophysiological function within certain bounds of water, temperature, and humidity stress, and the authors have defined a metric to determine when those bounds have been exceeded. They find that these metrics “explained more than 70% of the observed spatiotemporal patterns in greenness”. When applied to the El Nino of 2015/2016 they find that drought as expressed in VIs in general moves from north to south during the event, and from August 2016 through July 2016 the regional VI progresses from below-normal, near-normal, above-normal, and below-normal stages.
R2C1: This paper is interesting and well-written, although a bit dense at times. It took me several readings to get my head around the method, but once I did I found it an interesting and thought-provoking paper. Initially I wondered if the authors were neglecting the precipitation variability in the region (total annual precipitation, length of dry season), but I realized that this variability is accounted for in the construction of the ‘normal’ cycles for each gridcell shown in Figure 2. The authors might spell this out explicitly-wouldn’t take more than a couple of words or a sentence added. I do have a lot of questions about the methods and results, but I do not have any objections that would lead me to recommend rejection. As I don’t believe that a major overhaul is required to address my comments, my formal recommendation is that this paper be accepted for publication with minor revisions.
Response: We will add the suggested sentence about precipitation variability (total annual precipitation, length of dry season) in the Methods section.
Review: Let’s get started…
R2C2: There has been an ongoing discussion (or debate, if you want) around the notion of greenness increasing with (mild) drought (Saleska, Huete) or not (Samanta, Morton) for the last 10 years or so. Saleska claims the argument over with the results of Wu, Albert, Restrepo-Coupe and others who find that there is (in parts of the Amazon) a drop and reflush of leaves at the end of the dry season, but the new leaves have reduced photosynthetic capacity until they have ‘matured’ for somewhere around 60-90 days. The authors mention this element of regional ecophysiology, but do so in a rather oblique manner. This paper would have much more impact if the issue was met head-on. The community knows about the debate and recent results, the subject matter of this paper is related to this topic, so why not make a direct comment on it? Here are some specific thoughts about this:
• What do your results suggest about the resolution of this debate?
• Can you relate greenness to GPP in the context of your findings of VI correlation with TWS, LST and VPD, or not?
• If not, do your results suggest something about leaf demography (drop, flush) and how it relates to drought?
• Are there field studies of leaf drop/flush to support an attempt to explain regional behavior?
• Can you relate your conclusions to other studies, such as those that look at SIF in the region (Doughty, Koren)?Response: We thank Dr Baker for posing this suite of insightful questions, prompting us to delve deeper into our method and results. We will carefully review the suggested (and related) references and, where most relevant, we will incorporate these questions, and the answers our analysis provides, into the revised manuscript.
R2C3: I see TRMM precipitation mentioned in section 2.1 and Table 1, but don’t recall seeing mention of precipitation elsewhere, as TWS is the variable used. There are interesting questions around the use of TWS from GRACE and TRMM precip. Neither product is perfect. Precipitation may evaporate off leaves (if light), and if heavy may run off before infiltration. TWS will have a significant contribution from soil well below maximum rooting depth, and that soil water might be irrelevant to the analysis. Furthermore, there may be lags between precipitation and plant function, but these lags may be accounted for by using TWS. If only one precipitation/soil moisture metric is used, then the other need not be listed in Table 1. Additionally, the authors should explain the reasoning behind the choice of one ‘wetness’ product over the other.
Response: We propose to incorporate one additional soil moisture product into the revised manuscript. While TWS represents the water dynamics of the entire column, the soil moisture product can characterise the soil water dynamics in the top few meters. The combination of these two independent ‘wetness’ products is anticipated to enhance the robustness of this study.
R2C4: The calculation of seasonal cycle (various ND-Ave values) was calculated using the notion of a calendar year (Jan-Dec), but the analysis of the drought event was performed over the time of a ‘water year’ from August-July. In fact, this second methodology makes more sense, as the change in the calendar year comes in the middle of the wet season. My recollection is that many Amazonian researchers perform calculations over the scale of a water year. Why wasn’t the notion of water year used consistently? Does it change the results when compared to the calendar year calculation?
Response: Following the referee’s suggestion and to make our analysis and results as consistent as possible (both internally consistent and externally consistent with our peers), we will use the water year (from August to July) instead of the calendar year in the revised manuscript.
R2C5: On a related note, the years 2005, 2010 (and for that matter 2015/2016) were not drought everywhere. Were these denoted drought years just because other publications have said so, or was there an actual calculation of the fraction of the target gridcells during the (calendar) year that met drought criteria, and these years had the largest area under drought? Or was it that the metrics used to define drought was most severe?
Response: We adopted the term ‘2015-2016 drought’ primarily aligning with other publications. We conducted some calculations using the TRMM precipitation data ourselves. We found the regional precipitation from August 2015 through July 2016 was consistently below non-drought years. Below 100 mm monthly precipitation was mainly observed over the north Amazon in the second half of 2015 and over the south Amazon in July 2016. For the months from March to June 2016, nearly no region was identified exceeding normal ranges. Our study supports the fact that not everywhere experienced drought during severe drought years. We will emphasize this in the revised manuscript.
Specific Comments
R2C6: I think the references in lines 214-217 are for figure 3.
Response: Yes, thanks for your careful review (and apologies for our mistake) this will be changed to Figure 3.
R2C7: Line 227: Figure 4a shows TWS above non-drought values for the first 2 months, not 3.
Response: Yes, correct. We will change ‘first three months’ to ‘first two months’.
R2C8: Line 228: “During Stage III, only a small area with TWS<TWSND-Min occurred in the north-east.” This is confusing. Figure 4a shows TWS well below non-drought TWS, with a consistent amplitude, over the period from December 2015 through July 2016. The maps in Figure 5 seem to contradict this. Is it a difference between a few very dry gridcells and a lot of ‘sort of’ dry gridcells? Some explanation of this apparent discrepancy might be helpful.
Response: We acknowledge that this part may be somewhat confusing. To enhance clarity, we plan to introduce an explanatory figure in the Methods section, illustrating the distinction between TWSND-Ave and TWSND-Min. In brief, for each grid cell, TWSND-Min is generally lower than TWSND-Ave. Specifically, during Stage III (March – June 2016), TWS was observed to be below TWSND-Ave across most grid cells, but only a few of these grid cells registered values below TWSND-Min. The explanatory figure will be included in the revised manuscript to provide a clearer understanding of the differences between the two metrics. Similar figures will also be developed for LST and VPD.
R2C9: Lines 294-299. Are your results consistent with the findings of this study? Do you find that you don’t see a depletion of TWS in the region of the field studies of Fontes? If you do see a TWS depletion, and Fontes says it doesn’t matter, what does that suggest about your method?
Response: Our method uses the extreme values of TWS, LST and VPD to identify the regions exceeding normal ranges. For the study area of Fontes, our data showed that the region experienced high LST, high VPD and low TWS, which is consistent with their field measurements. Fontes says the low soil moisture does not matter, which means: (1) the high LST and VPD played a more important role in limiting the function of plants; and (2) the low soil moisture and TWS were still within the tolerance thresholds of rainforest.
During the revision, we plan to conduct further temporal analysis to examine the impacts of TWS levels on VI anomalies. We will select two grid cells. One is the grid cell where the Fontes’ study is located (i.e., Fontes’ grid cell) while the other one has a much lower TWS value than the Fontes’ grid cell. If VI anomalies in the second grid cell were much more severe than Fontes’ grid cell, its TWS value was very likely beyond the tolerance thresholds of rainforest. The results of this analysis will be added in the Discussion section and in the Appendix.
Citation: https://doi.org/10.5194/bg-2023-155-AC2
Status: closed
-
RC1: 'Comment on bg-2023-155', Anonymous Referee #1, 09 Oct 2023
The manuscript of Liu et al. is an interesting study about identifying the different environmental drivers of the drought-affected regions of 2015-2016 within the Amazon forest. They show that the regions where water storage, temperature and atmospheric moisture demand exceeded their ‘normal’ ranges agreed with more than 70% of the observed patterns in vegetation greenness. This manuscript has potential and could be a useful contribution to the drought research in the tropics. I do have some questions that I would like to see addressed.
General comments
I am not convinced by how the authors defined the occurrence of extreme values of LST, VPD and TWS in 2015-2016. Extreme values of ‘normal’ years are defined as the lowest or highest mean monthly value, not as the actual lowest or highest value that a ‘normal’ month reached for a certain variable. If there is a lot of variation in monthly TWS, LST or VPD between different years, the mean value doesn’t catch that. Comparing a raw value to a mean without taking into account the normal variation will easily give ‘extreme’ results. The authors combine this metric of extremeness with a significant difference from the monthly mean using a Wilcoxon rank test. I am curious how different the results would be when only one of the requirements is used (i.e. only using the requirement of at least two variables that are more extreme than the normal extreme, or of one extreme variable and the other two being significantly different)? Do they both give similar results or is one more in agreement with the vegetation anomalies than the other? This analysis could be added in the appendix.
It is not clearly explained in the methodology or the results how the regions with VI and PAR anomalies are compared with the extreme VPD, TWS and LST regions. This is the main analysis that leads to the conclusions of the manuscript, so it is important that this part is clear.
Specific comments
I would change the word ‘swings’ in the title. It is not a very common scientific word and it does not get repeated once within the manuscript, making it a strange wording choice.
The abstract feels too technical and methodological. I would make the explanation of the approach shorter, so that the conclusion of the manuscript is more clear. As a reader, my attention would get lost in the technicalities of the abstract and it would not convince me to read the paper (which would be a pity, because the research is valuable).
The authors use a combination of NDVI and EVI to quantify the greenness anomalies of the forest, but there is no mention of the problem with NDVI saturation in high biomass regions.
Line 191-193: Why did you use the non-parametric Wilcoxon rank test instead of calculating standardized anomalies and using them to say when values were significantly different?
Line 195-198: How were these requirements decided? Based on literature (if so, add references)? Based on which requirements gave the highest agreement with the vegetation indices (if so, add the other methods that were tried)?
Line 208-212: Step 3 in methods is not very clear. It might be better to add an explanatory figure such as figure 2? Too much mixed vocabulary used: ‘positive greenness anomaly’, ‘below-average greenness’, ‘exceeding the normal range’ à this makes it difficult to follow. Maybe better to use one word for ‘regions exceeding the normal ranges’, to make the explanation less wordy.
Line 211-212 is not clear: ‘compared with that of only considering VI-ano and PAR-ano’. What does this mean?
Figure 4 does not completely convince me that there are, for example, four months with an extremely low TWS in 2015-2016. How extreme is this compared to the normal variation in non-drought years? Figure 2b-d shows that there are indeed large variations in the monthly values. Maybe add the standard deviation in lines around the monthly dots? This will probably make the figure too crowded to be pretty, but this might be something for the appendix.
Table 2: it is not clear how the percentages are calculated. Is it 72% of all pixels with VI and PAR anomalies in same direction that also exceeded the ND range, or 72% of all pixels that exceeded the ND range that had VI and PAR anomalies in same direction, or …? This could be better explained in the table caption. I don’t understand what the current table caption means.
Technical corrections
Line 52: Add ‘the’ to the Amazon forest response.
Line 56: Add ‘the’ to the NDVI or remove ‘the’ from the EVI.
Line 101: Add ‘the’ to the 2005 drought.
Line 197: Remove ‘being’ from sentence.
Line 214-219: I think this part refers to Figure 3 instead of Figure 4?
Citation: https://doi.org/10.5194/bg-2023-155-RC1 -
AC1: 'Reply on RC1', Yi Y. Liu, 09 Dec 2023
We would like to thank Referee #1 for the constructive comments and suggestions, which will be very helpful in improving the quality of this manuscript. Below are our responses and plans to revise. The comments are in black, and our responses are in italics.
The manuscript of Liu et al. is an interesting study about identifying the different environmental drivers of the drought-affected regions of 2015-2016 within the Amazon forest. They show that the regions where water storage, temperature and atmospheric moisture demand exceeded their ‘normal’ ranges agreed with more than 70% of the observed patterns in vegetation greenness. This manuscript has potential and could be a useful contribution to the drought research in the tropics. I do have some questions that I would like to see addressed.
General comments
R1C1: I am not convinced by how the authors defined the occurrence of extreme values of LST, VPD and TWS in 2015-2016. Extreme values of ‘normal’ years are defined as the lowest or highest mean monthly value, not as the actual lowest or highest value that a ‘normal’ month reached for a certain variable. If there is a lot of variation in monthly TWS, LST or VPD between different years, the mean value doesn’t catch that. Comparing a raw value to a mean without taking into account the normal variation will easily give ‘extreme’ results. The authors combine this metric of extremeness with a significant difference from the monthly mean using a Wilcoxon rank test. I am curious how different the results would be when only one of the requirements is used (i.e. only using the requirement of at least two variables that are more extreme than the normal extreme, or of one extreme variable and the other two being significantly different)? Do they both give similar results or is one more in agreement with the vegetation anomalies than the other? This analysis could be added in the appendix.
Response: Following the referee’s suggestion, we will conduct a sensitivity analysis during the revision to further explore the influences of different definitions of extreme values on our study outcomes. The corresponding figures and tables, illustrating the impact of these variations, will be added in the appendix. In the Discussion section, we will introduce a new paragraph dedicated to the insights gained from this sensitivity analysis.
R1C2: It is not clearly explained in the methodology or the results how the regions with VI and PAR anomalies are compared with the extreme VPD, TWS and LST regions. This is the main analysis that leads to the conclusions of the manuscript, so it is important that this part is clear.
Response: We conducted a comparative analysis between the outcomes derived from two distinct scenarios; they are outlined below.
Scenario 1: It is assumed that VI anomalies are exclusively driven by PAR anomalies, leading to changes in the same direction. Accordingly, we created a map depicting the anticipated direction of VI anomalies (either positive or negative) for each grid cell across the Amazonian forests.
Scenario 2: We first utilized extreme values of TWS, VPD and LST to categorize regions into two groups: (a) those within normal ranges and (b) those exceeding normal ranges. For regions within normal ranges, we hypothesized that VI anomalies would align with PAR anomalies, exhibiting changes in the same direction. In regions exceeding normal ranges, negative VI anomalies are expected, irrespective of the direction of PAR anomalies. Accordingly, we generated another map illustrating the anticipated direction of VI anomalies (either positive or negative) for each grid cell.
By comparing these anticipated VI anomalies with MODIS-observed VI anomalies for all grid cells, we calculate the percentage of observed VI anomalies aligning with the anticipated direction in Scenario 1 and 2, respectively. This comparative analysis allows us to quantify the enhancements (%) achieved in explaining observed VI anomalies through the incorporation of our exceeding normal ranges-based approach.
In the revised manuscript, an explanatory figure depicting this experimental design will be added to the Methods section for enhanced clarity. The figure will contain numbered boxes and these numbers will be referred to in the relevant text to help guide the reader through the analysis method. The new figure will provide a visual ‘road-map’ with the associated text providing full details thus allowing others to replicate our method (a central tenet of our collective scientific endeavour).
Specific comments
R1C3: I would change the word ‘swings’ in the title. It is not a very common scientific word and it does not get repeated once within the manuscript, making it a strange wording choice.
Response: Good point; we will replace the word ‘swings’ with ‘fluctuations’.
R1C4: The abstract feels too technical and methodological. I would make the explanation of the approach shorter, so that the conclusion of the manuscript is more clear. As a reader, my attention would get lost in the technicalities of the abstract and it would not convince me to read the paper (which would be a pity, because the research is valuable).
Response: Following the referee’s suggestions, we plan to use one sentence ‘we proposed an approach to categorize regions into two groups: (1) those within normal hydrological and thermal ranges and (2) those exceeding normal ranges’ to replace the technical parts from line 21 to 26 in the current abstract. We will then use those saved words to ensure that the conclusion / lessons learnt from this contribution is clearer and more accessible to those that read the abstract.
R1C5: The authors use a combination of NDVI and EVI to quantify the greenness anomalies of the forest, but there is no mention of the problem with NDVI saturation in high biomass regions.
Response: In the revised manuscript, we will acknowledge the problem of NDVI signal saturation in high biomass regions and discuss its potential impacts on our study outcomes.
R1C6: Line 191-193: Why did you use the non-parametric Wilcoxon rank test instead of calculating standardized anomalies and using them to say when values were significantly different?
Response: As many hydrologic variables are not normally distributed, using the non-parametric Wilcoxon rank test offers the advantage of not assuming that data are normally distributed.
R1C7: Line 195-198: How were these requirements decided? Based on literature (if so, add references)? Based on which requirements gave the highest agreement with the vegetation indices (if so, add the other methods that were tried)?
Response: The criteria for these requirements were established partially through a review of the existing literature and partially through empirical testing using alternative methods, such as the utilization of the TWS anomaly alone. In the revised manuscript, we will improve the clarity of our methodology by incorporating corresponding references that informed our decisions. Furthermore, we will provide a more explicit rationale for the selection of the combination of all TWS, VPD and LST as key factors in our analysis.
R1C8: Line 208-212: Step 3 in methods is not very clear. It might be better to add an explanatory figure such as figure 2? Too much mixed vocabulary used: ‘positive greenness anomaly’, ‘below-average greenness’, ‘exceeding the normal range’ à this makes it difficult to follow. Maybe better to use one word for ‘regions exceeding the normal ranges’, to make the explanation less wordy.
Response: We will add an explanatory figure in the revised manuscript to demonstrate the comparison analysis in Step 3. Meanwhile, we will reduce the mixed vocabulary and try to only use one or two words and terms in the revised manuscript.
R1C9: Line 211-212 is not clear: ‘compared with that of only considering VI-ano and PAR-ano’. What does this mean?
Response: In this study, we conducted a comparative analysis between the outcomes derived from two distinct scenarios. Please see our response to R1C2 above.
R1C10: Figure 4 does not completely convince me that there are, for example, four months with an extremely low TWS in 2015-2016. How extreme is this compared to the normal variation in non-drought years? Figure 2b-d shows that there are indeed large variations in the monthly values. Maybe add the standard deviation in lines around the monthly dots? This will probably make the figure too crowded to be pretty, but this might be something for the appendix.
Response: During the revision, we will incorporate standard deviation lines around the monthly dots, particularly to highlight the extreme nature of TWS, LST and VPD in 2015-2016 compared to the normal variation in non-drought years.
R1C11: Table 2: it is not clear how the percentages are calculated. Is it 72% of all pixels with VI and PAR anomalies in same direction that also exceeded the ND range, or 72% of all pixels that exceeded the ND range that had VI and PAR anomalies in same direction, or …? This could be better explained in the table caption. I don’t understand what the current table caption means.
Response: We will refine the methodology in the Methods and Results sections during the revision and improve the caption for Table 2.
In reference to the percentages presented in Table 2, a concise explanation is provided here. In this study, we conducted a comparative analysis between the outcomes derived from two distinct scenarios.
Scenario 1: It is assumed that VI anomalies are exclusively driven by PAR anomalies, leading to changes in the same direction. Accordingly, we created a map depicting the anticipated direction of VI anomalies (either positive or negative) for each grid cell across the Amazonian forests.
Scenario 2: We first utilized extreme values of TWS, VPD and LST to categorize regions into two groups: (1) those within normal ranges and (2) those exceeding normal ranges. For regions within normal ranges, we hypothesized that VI anomalies would align with PAR anomalies, exhibiting changes in the same direction. In regions exceeding normal ranges, negative VI anomalies are expected, irrespective of the direction of PAR anomalies. Accordingly, we generated another map illustrating the anticipated direction of VI anomalies (either positive or negative) for each grid cell.
By comparing these anticipated VI anomalies with MODIS-observed VI anomalies for all grid cells, we calculated the percentage of observed VI anomalies aligning with the anticipated direction in Scenario 1 and 2, respectively. This comparative analysis allows us to quantify the enhancements (%) achieved in explaining observed VI anomalies through the incorporation of our exceeding normal ranges-based approach.
Technical corrections
R1C12: Line 52: Add ‘the’ to the Amazon forest response.
Response: Will do.
R1C13: Line 56: Add ‘the’ to the NDVI or remove ‘the’ from the EVI.
Response: Will do.
R1C14: Line 101: Add ‘the’ to the 2005 drought.
Response: Will do.
R1C15: Line 197: Remove ‘being’ from sentence.
Response: Will do.
R1C16: Line 214-219: I think this part refers to Figure 3 instead of Figure 4?
Response: Yes, thanks for your careful review (and apologies for our error); we will change to Figure 3.
Citation: https://doi.org/10.5194/bg-2023-155-AC1
-
AC1: 'Reply on RC1', Yi Y. Liu, 09 Dec 2023
-
RC2: 'Comment on bg-2023-155', Ian Baker, 25 Oct 2023
Article: Drought and radiation explain swings in Amazon rainforest greenness during the 2015-2016 drought
Authors: Liu et al.
Summary: The authors use a combination of TWS, VPD and LST (abbreviations explained in the manuscript, so I’m not going to rehash them here) to explore correlation between drought and greenness, as expressed by MODIS NDVI and EVI. They set a criteria of 2 of 3 variables beyond extreme values, or 1 variable beyond extreme and 2 significantly different from average as being outside of normal range. The idea (as I’m interpreting it) is that Amazonian forests have evolved to maintain ecophysiological function within certain bounds of water, temperature, and humidity stress, and the authors have defined a metric to determine when those bounds have been exceeded. They find that these metrics “explained more than 70% of the observed spatiotemporal patterns in greenness”. When applied to the El Nino of 2015/2016 they find that drought as expressed in VIs in general moves from north to south during the event, and from August 2016 through July 2016 the regional VI progresses from below-normal, near-normal, above-normal, and below-normal stages.
This paper is interesting and well-written, although a bit dense at times. It took me several readings to get my head around the method, but once I did I found it an interesting and thought-provoking paper. Initially I wondered if the authors were neglecting the precipitation variability in the region (total annual precipitation, length of dry season), but I realized that this variability is accounted for in the construction of the ‘normal’ cycles for each gridcell shown in Figure 2. The authors might spell this out explicitly-wouldn’t take more than a couple of words or a sentence added. I do have a lot of questions about the methods and results, but I do not have any objections that would lead me to recommend rejection. As I don’t believe that a major overhaul is required to address my comments, my formal recommendation is that this paper be accepted for publication with minor revisions.
Ian Baker
Colorado State University
October 2023
Ian.baker@colostate.edu
Review: Let’s get started…
There has been an ongoing discussion (or debate, if you want) around the notion of greenness increasing with (mild) drought (Saleska, Huete) or not (Samanta, Morton) for the last 10 years or so. Saleska claims the argument over with the results of Wu, Albert, Restrepo-Coupe and others who find that there is (in parts of the Amazon) a drop and reflush of leaves at the end of the dry season, but the new leaves have reduced photosynthetic capacity until they have ‘matured’ for somewhere around 60-90 days. The authors mention this element of regional ecophysiology, but do so in a rather oblique manner. This paper would have much more impact if the issue was met head-on. The community knows about the debate and recent results, the subject matter of this paper is related to this topic, so why not make a direct comment on it? Here are some specific thoughts about this:
- What do your results suggest about the resolution of this debate?
- Can you relate greenness to GPP in the context of your findings of VI correlation with TWS, LST and VPD, or not?
- If not, do your results suggest something about leaf demography (drop, flush) and how it relates to drought?
- Are there field studies of leaf drop/flush to support an attempt to explain regional behavior?
- Can you relate your conclusions to other studies, such as those that look at SIF in the region (Doughty, Koren)?
I see TRMM precipitation mentioned in section 2.1 and Table 1, but don’t recall seeing mention of precipitation elsewhere, as TWS is the variable used. There are interesting questions around the use of TWS from GRACE and TRMM precip. Neither product is perfect. Precipitation may evaporate off leaves (if light), and if heavy may run off before infiltration. TWS will have a significant contribution from soil well below maximum rooting depth, and that soil water might be irrelevant to the analysis. Furthermore, there may be lags between precipitation and plant function, but these lags may be accounted for by using TWS. If only one precipitation/soil moisture metric is used, then the other need not be listed in Table 1. Additionally, the authors should explain the reasoning behind the choice of one ‘wetness’ product over the other.
The calculation of seasonal cycle (various ND-Ave values) was calculated using the notion of a calendar year (Jan-Dec), but the analysis of the drought event was performed over the time of a ‘water year’ from August-July. In fact, this second methodology makes more sense, as the change in the calendar year comes in the middle of the wet season. My recollection is that many Amazonian researchers perform calculations over the scale of a water year. Why wasn’t the notion of water year used consistently? Does it change the results when compared to the calendar year calculation?
On a related note, the years 2005, 2010 (and for that matter 2015/2016) were not drought everywhere. Were these denoted drought years just because other publications have said so, or was there an actual calculation of the fraction of the target gridcells during the (calendar) year that met drought criteria, and these years had the largest area under drought? Or was it that the metrics used to define drought was most severe?
Specific Comments
- I think the references in lines 214-217 are for figure 3.
- Line 227: Figure 4a shows TWS above non-drought values for the first 2 months, not 3.
- Line 228: “During Stage III, only a small area with TWS<TWSND-Min occurred in the north-east.” This is confusing. Figure 4a shows TWS well below non-drought TWS, with a consistent amplitude, over the period from December 2015 through July 2016. The maps in Figure 5 seem to contradict this. Is it a difference between a few very dry gridcells and a lot of ‘sort of’ dry gridcells? Some explanation of this apparent discrepancy might be helpful.
- Lines 294-299. Are your results consistent with the findings of this study? Do you find that you don’t see a depletion of TWS in the region of the field studies of Fontes? If you do see a TWS depletion, and Fontes says it doesn’t matter, what does that suggest about your method?
References
Albert, L.P., Wu, J., Prohaska, N., de Camargo, P.B., Huxman, T.E., Tribuzy, E.S., Ivanov, V.Y., Oliveira, R.S., Garcia, S., Smith, M.N. and Junior, R.C.O., 2018. Age-dependent leaf physiology and consequences for crown-scale carbon uptake during the dry season in an Amazon evergreen forest. New Phytologist, 219(3), pp.870-884.
Doughty, R., Köhler, P., Frankenberg, C., Magney, T.S., Xiao, X., Qin, Y., Wu, X. and Moore III, B., 2019. TROPOMI reveals dry-season increase of solar-induced chlorophyll fluorescence in the Amazon forest. Proceedings of the National Academy of Sciences, 116(44), pp.22393-22398.
Doughty, R., Xiao, X., Qin, Y., Wu, X., Zhang, Y. and Moore III, B., 2021. Small anomalies in dry-season greenness and chlorophyll fluorescence for Amazon moist tropical forests during El Niño and La Niña. Remote Sensing of Environment, 253, p.112196.
Huete, A.R., Didan, K., Shimabukuro, Y.E., Ratana, P., Saleska, S.R., Hutyra, L.R., Yang, W., Nemani, R.R. and Myneni, R., 2006. Amazon rainforests green‐up with sunlight in dry season. Geophysical research letters, 33(6).
Koren, G., Van Schaik, E., Araújo, A.C., Boersma, K.F., Gärtner, A., Killaars, L., Kooreman, M.L., Kruijt, B., Van Der Laan-Luijkx, I.T., Von Randow, C. and Smith, N.E., 2018. Widespread reduction in sun-induced fluorescence from the Amazon during the 2015/2016 El Niño. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1760), p.20170408.
Morton, D.C., Nagol, J., Carabajal, C.C., Rosette, J., Palace, M., Cook, B.D., Vermote, E.F., Harding, D.J. and North, P.R., 2014. Amazon forests maintain consistent canopy structure and greenness during the dry season. Nature, 506(7487), pp.221-224.
Restrepo‐Coupe, N., Levine, N.M., Christoffersen, B.O., Albert, L.P., Wu, J., Costa, M.H., Galbraith, D., Imbuzeiro, H., Martins, G., da Araujo, A.C. and Malhi, Y.S., 2017. Do dynamic global vegetation models capture the seasonality of carbon fluxes in the Amazon basin? A data‐model intercomparison. Global change biology, 23(1), pp.191-208.
Restrepo-Coupe, N., da Rocha, H.R., Hutyra, L.R., da Araujo, A.C., Borma, L.S., Christoffersen, B., Cabral, O.M., de Camargo, P.B., Cardoso, F.L., da Costa, A.C.L. and Fitzjarrald, D.R., 2013. What drives the seasonality of photosynthesis across the Amazon basin? A cross-site analysis of eddy flux tower measurements from the Brasil flux network. Agricultural and Forest Meteorology, 182, pp.128-144.
Citation: https://doi.org/10.5194/bg-2023-155-RC2 -
AC2: 'Reply on RC2', Yi Y. Liu, 09 Dec 2023
We extend our sincere gratitude to Dr Ian Baker for his valuable comments and insightful suggestions, leading us to delve into thought and improve the quality of this manuscript. Below are our responses and plans to improve our revised manuscript. The comments are in black, and our responses are in italics.
Summary: The authors use a combination of TWS, VPD and LST (abbreviations explained in the manuscript, so I’m not going to rehash them here) to explore correlation between drought and greenness, as expressed by MODIS NDVI and EVI. They set a criteria of 2 of 3 variables beyond extreme values, or 1 variable beyond extreme and 2 significantly different from average as being outside of normal range. The idea (as I’m interpreting it) is that Amazonian forests have evolved to maintain ecophysiological function within certain bounds of water, temperature, and humidity stress, and the authors have defined a metric to determine when those bounds have been exceeded. They find that these metrics “explained more than 70% of the observed spatiotemporal patterns in greenness”. When applied to the El Nino of 2015/2016 they find that drought as expressed in VIs in general moves from north to south during the event, and from August 2016 through July 2016 the regional VI progresses from below-normal, near-normal, above-normal, and below-normal stages.
R2C1: This paper is interesting and well-written, although a bit dense at times. It took me several readings to get my head around the method, but once I did I found it an interesting and thought-provoking paper. Initially I wondered if the authors were neglecting the precipitation variability in the region (total annual precipitation, length of dry season), but I realized that this variability is accounted for in the construction of the ‘normal’ cycles for each gridcell shown in Figure 2. The authors might spell this out explicitly-wouldn’t take more than a couple of words or a sentence added. I do have a lot of questions about the methods and results, but I do not have any objections that would lead me to recommend rejection. As I don’t believe that a major overhaul is required to address my comments, my formal recommendation is that this paper be accepted for publication with minor revisions.
Response: We will add the suggested sentence about precipitation variability (total annual precipitation, length of dry season) in the Methods section.
Review: Let’s get started…
R2C2: There has been an ongoing discussion (or debate, if you want) around the notion of greenness increasing with (mild) drought (Saleska, Huete) or not (Samanta, Morton) for the last 10 years or so. Saleska claims the argument over with the results of Wu, Albert, Restrepo-Coupe and others who find that there is (in parts of the Amazon) a drop and reflush of leaves at the end of the dry season, but the new leaves have reduced photosynthetic capacity until they have ‘matured’ for somewhere around 60-90 days. The authors mention this element of regional ecophysiology, but do so in a rather oblique manner. This paper would have much more impact if the issue was met head-on. The community knows about the debate and recent results, the subject matter of this paper is related to this topic, so why not make a direct comment on it? Here are some specific thoughts about this:
• What do your results suggest about the resolution of this debate?
• Can you relate greenness to GPP in the context of your findings of VI correlation with TWS, LST and VPD, or not?
• If not, do your results suggest something about leaf demography (drop, flush) and how it relates to drought?
• Are there field studies of leaf drop/flush to support an attempt to explain regional behavior?
• Can you relate your conclusions to other studies, such as those that look at SIF in the region (Doughty, Koren)?Response: We thank Dr Baker for posing this suite of insightful questions, prompting us to delve deeper into our method and results. We will carefully review the suggested (and related) references and, where most relevant, we will incorporate these questions, and the answers our analysis provides, into the revised manuscript.
R2C3: I see TRMM precipitation mentioned in section 2.1 and Table 1, but don’t recall seeing mention of precipitation elsewhere, as TWS is the variable used. There are interesting questions around the use of TWS from GRACE and TRMM precip. Neither product is perfect. Precipitation may evaporate off leaves (if light), and if heavy may run off before infiltration. TWS will have a significant contribution from soil well below maximum rooting depth, and that soil water might be irrelevant to the analysis. Furthermore, there may be lags between precipitation and plant function, but these lags may be accounted for by using TWS. If only one precipitation/soil moisture metric is used, then the other need not be listed in Table 1. Additionally, the authors should explain the reasoning behind the choice of one ‘wetness’ product over the other.
Response: We propose to incorporate one additional soil moisture product into the revised manuscript. While TWS represents the water dynamics of the entire column, the soil moisture product can characterise the soil water dynamics in the top few meters. The combination of these two independent ‘wetness’ products is anticipated to enhance the robustness of this study.
R2C4: The calculation of seasonal cycle (various ND-Ave values) was calculated using the notion of a calendar year (Jan-Dec), but the analysis of the drought event was performed over the time of a ‘water year’ from August-July. In fact, this second methodology makes more sense, as the change in the calendar year comes in the middle of the wet season. My recollection is that many Amazonian researchers perform calculations over the scale of a water year. Why wasn’t the notion of water year used consistently? Does it change the results when compared to the calendar year calculation?
Response: Following the referee’s suggestion and to make our analysis and results as consistent as possible (both internally consistent and externally consistent with our peers), we will use the water year (from August to July) instead of the calendar year in the revised manuscript.
R2C5: On a related note, the years 2005, 2010 (and for that matter 2015/2016) were not drought everywhere. Were these denoted drought years just because other publications have said so, or was there an actual calculation of the fraction of the target gridcells during the (calendar) year that met drought criteria, and these years had the largest area under drought? Or was it that the metrics used to define drought was most severe?
Response: We adopted the term ‘2015-2016 drought’ primarily aligning with other publications. We conducted some calculations using the TRMM precipitation data ourselves. We found the regional precipitation from August 2015 through July 2016 was consistently below non-drought years. Below 100 mm monthly precipitation was mainly observed over the north Amazon in the second half of 2015 and over the south Amazon in July 2016. For the months from March to June 2016, nearly no region was identified exceeding normal ranges. Our study supports the fact that not everywhere experienced drought during severe drought years. We will emphasize this in the revised manuscript.
Specific Comments
R2C6: I think the references in lines 214-217 are for figure 3.
Response: Yes, thanks for your careful review (and apologies for our mistake) this will be changed to Figure 3.
R2C7: Line 227: Figure 4a shows TWS above non-drought values for the first 2 months, not 3.
Response: Yes, correct. We will change ‘first three months’ to ‘first two months’.
R2C8: Line 228: “During Stage III, only a small area with TWS<TWSND-Min occurred in the north-east.” This is confusing. Figure 4a shows TWS well below non-drought TWS, with a consistent amplitude, over the period from December 2015 through July 2016. The maps in Figure 5 seem to contradict this. Is it a difference between a few very dry gridcells and a lot of ‘sort of’ dry gridcells? Some explanation of this apparent discrepancy might be helpful.
Response: We acknowledge that this part may be somewhat confusing. To enhance clarity, we plan to introduce an explanatory figure in the Methods section, illustrating the distinction between TWSND-Ave and TWSND-Min. In brief, for each grid cell, TWSND-Min is generally lower than TWSND-Ave. Specifically, during Stage III (March – June 2016), TWS was observed to be below TWSND-Ave across most grid cells, but only a few of these grid cells registered values below TWSND-Min. The explanatory figure will be included in the revised manuscript to provide a clearer understanding of the differences between the two metrics. Similar figures will also be developed for LST and VPD.
R2C9: Lines 294-299. Are your results consistent with the findings of this study? Do you find that you don’t see a depletion of TWS in the region of the field studies of Fontes? If you do see a TWS depletion, and Fontes says it doesn’t matter, what does that suggest about your method?
Response: Our method uses the extreme values of TWS, LST and VPD to identify the regions exceeding normal ranges. For the study area of Fontes, our data showed that the region experienced high LST, high VPD and low TWS, which is consistent with their field measurements. Fontes says the low soil moisture does not matter, which means: (1) the high LST and VPD played a more important role in limiting the function of plants; and (2) the low soil moisture and TWS were still within the tolerance thresholds of rainforest.
During the revision, we plan to conduct further temporal analysis to examine the impacts of TWS levels on VI anomalies. We will select two grid cells. One is the grid cell where the Fontes’ study is located (i.e., Fontes’ grid cell) while the other one has a much lower TWS value than the Fontes’ grid cell. If VI anomalies in the second grid cell were much more severe than Fontes’ grid cell, its TWS value was very likely beyond the tolerance thresholds of rainforest. The results of this analysis will be added in the Discussion section and in the Appendix.
Citation: https://doi.org/10.5194/bg-2023-155-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 321 | 119 | 32 | 472 | 23 | 20 |
- HTML: 321
- PDF: 119
- XML: 32
- Total: 472
- BibTeX: 23
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1