On the importance of phenology in the evaporative process of the Miombo Woodland: Could it be why satellite-based evaporation estimates differ?

Zimba, Henry; Coenders-Gerrits, Miriam; Banda, Kawawa; Hulsman, Petra; van de Giesen, Nick; Nyambe, Imasiku; Savenije, Hubert H. G.

doi:https://doi.org/10.5194/hess-2023-39

Preprints

https://doi.org/10.5194/hess-2023-39

Preprints

09 Feb 2023

| 09 Feb 2023

Status: a revised version of this preprint is currently under review for the journal HESS.

On the importance of phenology in the evaporative process of the Miombo Woodland: Could it be why satellite-based evaporation estimates differ?

Henry Zimba, Miriam Coenders-Gerrits, Kawawa Banda, Petra Hulsman, Nick van de Giesen, Imasiku Nyambe, and Hubert H. G. Savenije

Abstract. Complex African ecosystems such as the Miombo Woodland, with unique plant phenology, have evaporation dynamics that have not been investigated due to very few, if at all existant, flux tower observations. Furthermore, significant differences have been observed in satellite-based evaporation estimates in the Miombo Woodland especially in the dry season. Therefore, deciding which sattelite evaporation product to use in this ecosystem is difficult, as these products vary in many respects. In this study, the actual evaporation estimates for six satellite-based evaporation estimates are compared across Miombo Woodland phenophases in the Luangwa Basin, in southern Africa. In the absence of basin scale field observations, the actual evaporation estimated using the general water balance is used as reference, to which the six satellite-based evaporation estimates have been compared. Our results show significant variation in actual evaporation estimates in the water limited, high temperature and lower forest canopy cover and leaf chlorophyll conditions in the dormant phenophase. Lowest variation is observed in water abundant, high temperature, high leaf chlorophyll content and high forest canopy cover in the maturity/peak phenophase(s). Compared to the basin scale water balance actual evaporation, all six satellite-based evaporation estimates appear to underestimate evaporation. The results of underestimation at basin scale agrees with local field observations in a dense Miombo Woodland in the Luangwa Basin, which indicates that satellite-based evaporation estimates generally underestimate dry season (dormant phenophase) and early rain season (green-up phenophase) actual evaporation. The discrepancies in dry season satellite-based evaporation estimates may be cuased by the Miombo Woodland species’ phenological adaptation attributes such as: leaf fall, leaf flush, access to deep soil moisture and the within vegetation water storage, coupled with heterogenous plant species response to phenological stimuli. Therefore, it appears that satellite-based evaporation estimates using model structure, processes and inputs that are capable of capturing Miombo species dry season phenological interaction with climate are likely to have actual evaporation estimates closer to field conditions.

Received: 01 Feb 2023 – Discussion started: 09 Feb 2023

Download & links

Henry Zimba, Miriam Coenders-Gerrits, Kawawa Banda, Petra Hulsman, Nick van de Giesen, Imasiku Nyambe, and Hubert H. G. Savenije

Status: final response (author comments only)

RC1:
'Comment on hess-2023-39', Graham Jewitt, 17 Mar 2023
This is a study of importance. The water use of Miombo Woodland is poorly studied and this paper could make a significant contribution to addressing this gap in knowledge. However, in its current form, I do not believe that the paper is publishable. In particular, there are some major assumptions which introduce so much uncertainty in the analysis that the results produced cannot be assumed to be representative - leading to a lack of confidence in the interpretation and conclusions. For the the basin scale ET analysis:
the assumptions made in producing annual estimates of rainfall and runoff need to be much better justified and the way in which the values used better explained. However, even if this is done, I believe that this would highlight a flawed approach, so I suggest that in teh cas eof the basin scale analysis that the author's need to "go back to the drawing board".

the process where an assumed uniformity of Miombo Woodland is assumed and how this influences basin level ET estimtes is quite unclear to me. In addition, the assumptions of extend of Miombo vs Mopane Woodland seem incorrect based on recent studies in the catchment.

This leads to an analysis where ET is derived from (I believe) flawed estimates of rainfall, runoff and land cover leaving little confidence in the annual values produced. To then use this as a basis for comparison with ET estimates from RS products seems a step to far to me.
The link to phenophase through the year and it's spatial variation is a stronger part of the paper, although the assumptions around extent of the Miombo Woodland and it's realtive homegeneity between different parts of the catchment need more careful analysis.
I provide more detailed comments in the annoated attachment.
Citation: https://doi.org/10.5194/hess-2023-39-RC1
- AC2: 'Reply on RC1', Henry Musonda Zimba, 30 Apr 2023
  
  Our responses to reviewer 1 are in the attached document along side the reviewers comments.
  
  Citation: https://doi.org/10.5194/hess-2023-39-AC2
- AC3: 'Reply on RC1', Henry Musonda Zimba, 30 Apr 2023
  
  We made effort to respond to the reviewers comments in the manuscript.
  Our responses are alongside the reviewers comments in the attached document
  
  Citation: https://doi.org/10.5194/hess-2023-39-AC3
RC2:
'Comment on hess-2023-39', Anonymous Referee #2, 24 Mar 2023
This manuscript investigates different products of gridded evapotranspiration (ET) over an extensive but little gauged ecosystem type, the Miombo, in Africa. Little hydrological data is available in the region, but evapotranspiration estimates are required for management decisions. The manuscript compares different ET products most from remote sensing and one from a water balance with each other and adds own qualitative observations on phenology to shed light on why they differ from each other. This is a valid way forward in my opinion although the manuscript remains rather descriptive when it comes to phenology. Local plausibility checks of existing global products are a fist way forward to improve information for management in dats scarce regions and important work. The manuscript is valuable in this respect. There are substantial critical points however, which I have pointed out below. The main point relates to the lack of information on methods, which is so substantial that currently I cannot confirm the validity of the conclusions. This may potentially be cured in a revision. In addition, I think the manuscript could provide much more (i) background on the type ET products (ii) known sources of uncertainty in (i) that could be linked to the unique phenological stages observed in the Miombo, and (iii) provide some discussion about how common this problem is either globally or in neighbouring ecosystem types in Africa. Finally, the manuscript really needs revision for structure and I strongly encourage the experienced co-authors to provide guidance to the first author.
Main comments
(1) I found it difficult to follow the methods, and specific information is required to allow assessment of the results and discussion.
I do not understand the extrapolation in of the runoff data to be used for the water balance model based ET-estimate. This is highly relevant, since the latter is used later as a contrast to compare different ET products. The methodology on how the data was derived needs to be fully stated. Now it is arbitrary and renders this method useless.

Standard deviation and coefficient of variations are used to evaluate difference within and between products. I did not follow how this was done. With ET varying both in space and in time and between datasets, I was unable to decipher in most instances, across which dimension the variance was calculated? Therefore, I am not able to follow the conclusions in several instances. Also, standard deviation and coefficient of variation are redundant, and I propose to use e.g. only the coefficient of variation.

(2) The manuscripts main point is to compare different ET products in an underrepresented ecosystem type and to highlight where they are uncertain. The paper remains very descriptive in parts. Despite the very valuable information delivered at one location, it does not go much into depth. It would be really helpful, if the reasons for the uncertainty and similarity of the those products was discussed in more depth. Some information already appears in the discussion, e.g. mentioning the lack of ground water access in models like GLEAM, but even more background wold be desirable. For example, the methods section could give a short background on how the products are derived focussing on how inherent assumptions on phenology play in, or which LAI products are used.
(3) The manuscript is difficult to navigate. The text is partly fragmented (e.g. similar information is spread out over several sections). The order of presentation was somewhat confusing. There are several repetitions (I have pointed out some below in the detailed comments), and many sentences are complicated and long. It would be good to revise carefully for structure both at the manuscript as well as at the sentence level.
General editorial comments
The Miombo is an extensive savannah ecosystem, but not all readers may be aware of it. It would be good to set the stage right from the start, including in the title, the abstract and introduction. Personally, I have not known this specific ecosystem before (and I am thankful I learnt it here). I propose adding a short introduction about savannah biomes in general, and how they are differentiated in Africa specifically. This would greatly help the reader to understand that this is not a site specific study, but represents an expansive ecosystem class on the African continent. Also, are there other equally underrepresented ecosystems that face similar problems with potential misinterpretation of satellite products due to insufficient representation of the local phenology?
Detailed comments
I had a hard time counting the line numbers. I may be off one line sometimes in the comments. I hope the appropriate places can found found nevertheless.
Abstract
Lines 18-23: Would be good to give a short explanation about what the Miombo is, mentioning its extent and the complex phenology. Would be good to show that this is not a case study in the first sentences. Also, why is ET required.
Lines 26, 29 „variation“: I am unclear what variation is meant and therefore the message of the sentence is also unclear to me.
Line 32-35: The wording of the sentence suggest that ET measurements were done. But those were not presented here. Is this a reference to the phenological information?
Line 36 „caused“
Line 37 „leaf fall, leaf flush“ - Many ecosystems have leaf fall and leaf flush, which aspects indicate the adaptation? Can you add a word to specify?
Line 38: What is meant with „within vegetation water storage“?
Line 38: „heterogenous plant species response“ I undestood only after reading the ms what was meant here. Can you rephrase?
Lines 39-43: I find this message unclear. Do those products exist? Should they be developed, existing ones adapted?
Introduction
As mentioned in the main comment above, it would be good to give a bit of background on the ecosystem, its size, and how it differs from the surrounding ecosystem types.
Line 54: associated with variations
Line 56: which „conductance" is meant ?
Line 56-63: Long and complicated section. Rephrase?
Line 67-68: This is an attribution of a very old hydrological concept to a recent reference. Can you rephrase, e.g. „here we are adopting the same definition as“ and/or find a suitable text book reference.
Line 78: influences
Line 92: also
Line 92-93: „Leaf flushing .. „ I m not sure in which context this is meant?
Line 89-110: If I understand correctly, this section implies that remote sensing products are validated in regions, where ground truth is available. This may cause a bias as it puts a focus on ecosystems other than those located in the global South? If yes, und assuming there are some more examples, it would be good to formulate this paragraph more generally. Otherwise it sounds like a case study.
Material and Methods
Line 111: replace „formulated“ by „conducted“ ?
Lines 138-141: Sounds much more like introduction. Move up?
Lines 145-147: It is not entirely clear what it was „sufficient“. Can you specify.
Section 2.2
I was unclear whether the phenophases were defined as part of this study? The word „field observations“ suggest that. If yes, I propose moving them to the results section. Also explain how they were determined / observed. Alternatively, move content of now section 2.11 up here and merge. There is a lot of overlap in content between those sections.
Figure 1: I found it hard to read this figure, which includes a time axis, definitions and the phases. Would be good to move the definitions out of the figure and into the main text. The different shades of green in the top line (phenophase) and third from top (Period of the hydrological year) do not match and it is therefore unclear what they mean. I propose to leave out the third line and turn it into a simple time axis.
Line 154: Not sure what is meant with „dominant activity“.
Line 170: I do not understand what is meant with the header, also the first sentences do not explain what the delineation does and for what purpose it is required. Please add some more information
Line 181: I do not understand what is meant with „put the basin boundary Miombo Woodland generalisation into context“
Lines 183-185 and 188-89 and Line 197-198: are repetitions. Also, not sure why this comparison is mentioned, but maybe it is the motivation why the delineation is required. Please give more information and move up to the beginning of the section.
Lines 192-194: Better move to introduction.
Line 213: I do not understand „This was because these products are normally applied as is, in their original resolutions.“ Can you be more specific?
Line 220: „acceptable“: Acceptable or not, depends a on the situation. Better state that this is the choice you made, as there were no other option. Also the water balance technically also requires distributed data, please be more specific.
Line 222: What is meant with: „ver-year storage change is neglected“
Sections 2.6 and 2.7: I understand that the precipitation and runoff details given in sections 2.6 and 2.7 are mainly related to the water balance in section 2.5? If yes, it would avoid confusion, if those section were subsections of 2.6.
Line 232: What is meant with: „geographically biased“?
Lines 231-233: The message of this section is unclear. Some are bad, some have some good correlations, but than all are used anyway? Also, this section is a repetition of Lines 236-239. Would be good to merge.
Line 241: Do you mean „to this end“?
Line 245: „that they had desirable spatial and temporal resolutions“ sounds a bit arbitrary. Can you be more specific on what was desirable?
Line 248-249: This extrapolation seems really important. The method is completely unclear. How were the data extrapolated by linear regression? What as the dependent and independent variable in this extrapolation? Furthermore, shortly state why it was necessary.
Line 258: „Denver camera“: Unclear description of methods. Please make sure manufacturers are clearly stated. It sounds like Denver camera is an established name, but what you probably mean is you installed a camera (Product name, Denver A/S) ?
Line 262: Similar comment as above for the fisheye lens - was it added to the same camera?
Eq 2: Mu and sigma are typically used for theoretical distributions, not descriptive stats. I propose using bar over x and s .
Eq. (3) Maybe just use B here
Section 2.10
The header „Data sources“. Can you be me more specific here? Data sources were already previously disussed, e.g. in sections 2.4, 2.6, 2.7, 2.8 ..
The section describes mainly the FLEXTopo model, which is good information to have to interpret the results. It would be good to give background information about the other products here too.
Line 287-288: „FLEXTopo“ used without previous introduction
Lines 293-301: Description of the model is difficult to follow, as it starts with the groundwater cells before stating this is one of several storages modelled, some distributed some apparently not.
Line 292-293: „within a grid cell to which all HRUs are connected.“ To me this is unclear. Are there different HRUs within a grid cell? A bit mit information would help.
Lines 295-296 „Furthermore, each storage component has been structured as a reservoir with matching water balance equations.“ and Lines 301-302: „However, the limitation with the FLEXTopo was the 27.7 km spatial resolution“ - unclear. I do not understand what is meant.
Section 2.11 would be good to have this information much further up. Also as mentions above, there is a great deal of overlap with section 2.2. Would be good to move up and merge.
Lines 396-397: Those references are also the sources for the information in Fig 2, correct? If yes, add them there (too).
Lines 403-414: Specifically here there seems to be a great deal of overlap with Section 2.2 and does not need be mentioned twice.
Results and Discussion
Fig 3: I was really unclear what is plotted here? Does „aggregated“ mean „climatology“, only one year is shown but the year is 2019. Alternatively, is this a spatial aggreagation? Also which precipitation and temperature are shown (Table 1 gives 4 options)? Where is the specific humidity taken from? I think relative humidity would be more intuitive and give a better indication of water deficit. Finally, satellite based soil water content refers to which depth in this case? Does it add information?
Line 451: „climate“ would be better to say „atmospheric“
Line 464-465: unclear „The peak climate and phenological variables values were generally observed in the rain season, green-up and maturity/peak phenophases.“ and „lowest values“ can you specifiy.
Lines 471-483: This is about methodology and should be moved to the Methods section, Lines 476-478 are a repetition of Lines 176-179.
Line 489: Coefficients of variations are very difficult to see in Fig 5, would be good to have a table.
Line 490: Actually, the correlations are rather poor, which points to the opposite? Would be good to discuss those poor correlations as well and potentially move table to main text?
Line 491: I am not sure what is meant with: „variations within each satellite evaporation product“? Spatial or temporal variation? Intra or inter annual?
Lines 495-496: This is a change of topic and interruptive. Phenophases have been discussed in 3.1, merge there.
Line 496: „relatively higher correlation“ - This is difficult to see from the table. Also it is an important message and would be good to have some aggregated information in the main text. For example, you could add a figure showing the correlation coefficients in a box plot (one box per season) and add this to the main text.
Line 436: „(Fig. 5)“ Do you mean Fig. 6?
Lines 538-540: This discussion of Fig 3 should be moved up or a better connection be made, why we need to remember this here.
Line 543: „significant differences“ - I believe you mean “substantial”. Would be good to replace to avoid confusion with the statistical nomenclature requiring significance tests.
Line 549-550: „effectively account for rainfall interception at the commencement of the rain season.“ - It find it difficult to understand the logic of this statement without further information. Can you explore? Do you mean that the interception part of actual ET depends on leaf area index and therefore misrepresentation of LAI may reflect on estimates of actual ET?
Figure 5: Please add y-axis labels. Please state clearly what is shown on left and right panel? What is meant with „aggregated“ and „mean“ ?
Line 580: Please make sure abbreviations / symbols for standard deviation are the same throughout the manuscript, including the equations. Eq. 2 used different ones.
Line 580-581: This information is required earlier, maybe already in the Methods section. Also be please be more specific, e.g. what is the sample when you say between phenological phases? Finally, why is is necessary to use two different measure of variations? Why not just use coefficient of variation throughout?
Lines 582-585: I am left unclear what the percentages in the text refer to? Is this coefficient of variation? Of what? For example, what is meant with „showed the lowest differences (7.83 %)“?
Lines 596-598: Difficult to understand, can you rephrase?
Figure 6: The Figure description does not match the figure content. Please re-organize and make sure to mention the content of panel (c) separately from (b). Also, I found it confusing that (b) shows standard deviation and (c) shows coefficient of variation (CV). I am also not sure, which sample CV is calculated for? Is this inter annual variation?
Line 636-638: Can you explain how you see this from your data?
Line 639: „significant variations in precipitation“ - Which variation is meant? Variation within the same year, between years?
Section 3.2.4 Great to see mentioned which variation is meant (spatial). I found it hard to understand what “variation” meant in the previous section. Now I am assuming it probably meant temporal variation. Correct? Please specify.
Line 647: „forested upland Miombo“ - Can you add basic delineation here, e.g. where is the forested upland region roughly?
Line 679-680: What is meant with „user accuracy“?
Line 678-679: The mentioned drivers of transpiration are textbook knowledge. Please cite a textbook here.
Line 680-681: I think this would be a place to give more background information. For example, can you show those differences and/or have a reference showing in a similar region that (a) that land cover classification can differ substantially between products, and (b) have a substantial influence on the ET estimate. Also, what are probably causes for those different land cover classifications?
Section 3.3 I would have liked to read at least the first part of this section (up to line 718) at the beginning of the results, as it gives a background on the small scale spatial heterogeneity of the canopy over different phenophases. Also, any quantitative information would improve this section, e.g. information changes of greenness or heterogeneity thereof in the fotos in Fig 9 left. The information would help explain the spatial variation and differences between products in sections 3.2. The second part, discussing consequences for ET estimation (e.g. starting at Line 751) could remain here or become a last part of section 3.2.
Line 696-701: This is a repetition to section 3.1 and would be another reason to move this section up.
Lines 749-751: This statement has to be accompanied by a reference or phrased like an assumption.
Fig 9: Please differentiate between left and right panel. Also, what is the sample for the standard deviation in LAI and NDVI? Is this across pixels in the Miombo Forest?
Section 3.4 I would have liked to see the first part (up to Line 893) of this section before section 3.2. It would be much easier to appreciate the discussion on the lack of groundwater access in GLEAM.
Line 828: I do not understand what is meant with „below average aggregated annual mean bias“.
Fig 10: Please ass variables to the y-axis labels. I think you could remove precipitation from panel (b) and thus only show bias there. Please make sure to separately state what is shown in panel (b) and (c).
Conclusions
Line 923-924: I still do not know how the coefficient of variation shows this.
Line 961-964: This part is discussion and not conclusion. Wold be goo to move up.
Citation: https://doi.org/10.5194/hess-2023-39-RC2
- AC1: 'Reply on RC2', Henry Musonda Zimba, 29 Apr 2023
  
  We have made effort to respond to each of the remarks by the reviewer. Our responses are in the attached pdf document
  
  Citation: https://doi.org/10.5194/hess-2023-39-AC1

Henry Zimba, Miriam Coenders-Gerrits, Kawawa Banda, Petra Hulsman, Nick van de Giesen, Imasiku Nyambe, and Hubert H. G. Savenije

Viewed

Total article views: 646 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
478	147	21	646	8	7

HTML: 478
PDF: 147
XML: 21
Total: 646
BibTeX: 8
EndNote: 7

Views and downloads (calculated since 09 Feb 2023)

Month	HTML	PDF	XML	Total
Feb 2023	170	38	3	211
Mar 2023	69	17	5	91
Apr 2023	37	16	7	60
May 2023	30	9	1	40
Jun 2023	17	5	0	22
Jul 2023	44	12	1	57
Aug 2023	36	10	1	47
Sep 2023	23	9	1	33
Oct 2023	13	10	1	24
Nov 2023	11	3	0	14
Dec 2023	3	5	0	8
Jan 2024	20	7	1	28
Feb 2024	5	6	0	11

Cumulative views and downloads (calculated since 09 Feb 2023)

Month	HTML	PDF	XML	Total
Feb 2023	170	38	3	211
Mar 2023	69	17	5	91
Apr 2023	37	16	7	60
May 2023	30	9	1	40
Jun 2023	17	5	0	22
Jul 2023	44	12	1	57
Aug 2023	36	10	1	47
Sep 2023	23	9	1	33
Oct 2023	13	10	1	24
Nov 2023	11	3	0	14
Dec 2023	3	5	0	8
Jan 2024	20	7	1	28
Feb 2024	5	6	0	11

Viewed (geographical distribution)

Total article views: 615 (including HTML, PDF, and XML) Thereof 615 with geography defined and 0 with unknown origin.

Country	#	Views	%

Cited

Discussed

Latest update: 19 Feb 2024

Short summary

In the dry season some Miombo plant species may have no leaves while several others are full of leaves and have access to water in deep soils and within the plants. Therefore, the plants continue to lose water to the atmosphere as they interact with the available energy and water. All the six satellite-based evaporation estimates compared in this study underestimates this plant water loss because they appear to have difficulties capturing the unique dry season Miombo plants attributes.


Total:	0
HTML:	0
PDF:	0
XML:	0

On the importance of phenology in the evaporative process of the Miombo Woodland: Could it be why satellite-based evaporation estimates differ?

Viewed

Viewed (geographical distribution)

Cited

1 citations as recorded by crossref.

Discussed