Research article 03 Nov 2021
Research article | 03 Nov 2021
Land use and climate change effects on water yield from East African forested water towers
Charles Nduhiu Wamucii et al.
Download
- Final revised paper (published on 03 Nov 2021)
- Preprint (discussion started on 29 Mar 2021)
Interactive discussion
Status: closed
-
CC1: 'Comment on hess-2021-151', Steve Lyon, 21 Apr 2021
This study investigates climate and human impacts on water towers in East Africa. The analysis is conducted in a Budyko framework. The target region is often considered vulnerable to changes in water resources making this investigation warranted and the result likely informative. Overall, the study is well conceived. However, I feel there are some considerable limitations in the structure of presentation. Further, some of the mechanistic interpretations are not fully supported given the potential of confounding impacts and potential uncertainty in data and analysis.
In the introduction (especially around P2,L15-L20), I would expect to see some more consideration of the strengths and weaknesses of various approaches for assessing climate and land-use change on water resources. Do we have some results or previous work that are relevant for the region? What is the motivation for selecting the Budyko approach over other approaches? There is not much review of current science offered up in the introduction. This should be expanded to help the reader understand the motivations for the current study and approach.
In addition, the lack of framing the study in a research question or a hypothesis is a major weakness. The result is that the study is some exploration of data that does not seem to address a problem or help advance the science. Such exploration (“can-we-do-it” type of work) is fine for a technical report but more would be needed for publication in a peer-reviewed journal. I am confident the authors can put this study in a research framework and present a clean and testable hypothesis or a some societally relevant research question.
The study mixes direct observation data interpolated across sites and remotely sensed data at various scales. I’m wondering if there is any potential impact of the various assumptions and approaches in each dataset? Synthesizing across various approach can often compound huge amounts of uncertainties and errors as we build composite analysis (in space and time). How has uncertainty been considered in your analysis and what role would data error have on your results/interpretation? Some consideration and discussion of uncertainty impacts must be presented to help the reader understand how robust the findings are in this study. This should be fairly straight forward given how the water balances were constructed using 100 random points. Perhaps perform a re-selection of random point and assess the difference or use some sort of calibration/validation approach on a sub-division of the 100 points (like a boot strap).
Further, I am not sure about the 100 random points in the methodology. Why was this done? Is it just too difficult to define the spatial extent of the water towers (which would allow using all the spatial data in the area)? Seems there would be some value in conducting this experiment at various elevations to assess the impact of elevation (as temperature proxy) on the results. Please outline why the method of 100 random point was selected and what the impacts would be on the results relative to another method.
There appears to be a large amount of mechanistic speculation on why points depart from the Budyko curve. There has been ample research over recent decades explaining how we can see variations along and from the curve. Further, many different explanations have been offered as to why catchments would deviate from theoretical curves with time. Could you outline some motivation for how you can be certain you are isolating mechanisms with your analysis? We would anticipate much interaction and coupled response that could be masked in the movement of points in Budkyo space (see van der Velde et al., 2014). It i likely that this lack of consideration of complexity relates back to the weakness and lack of thorough literature review seen in the introduction.
Along these same lines, what role would other factors such as CO2 increase and/or human alteration to water usage have in these regions? I could envision shifts in water cycling due to an intensification of plant activity through increased NPP or agricultural intensification. Warmer and CO2 richer climates could behave differently. Further, how much pumping and/or movement through irrigation schemes takes place in some of these systems? I understand they should be pristine or high-elevation forest without impact, but are they really without abstraction or other anthropogenic impacts?
In general, the results as presented are dense and not easy to follow. Read things a few times and not sure I can understand all the nuance of what is being shown here due to how things are being presented. This is not helped by poorly constructed figures with overlapping number, limited axis labels, and multiple colors to track. A major effort to organize the results into a concise section is required. Start by group the various results into sub-sections and cleaning up the figures. Structuring this section could also be aided by a more thoughtful research question or hypothesis setup. Then the results could be organized into how they answer the research question(s).
The discussion section is lacking rigor. At best it repeats the results with more interpretation. I miss a connection to the literature and how the results help inform and advance the science. Also, what are the strengths and limitations of the approach considered and how do these impact interpretation? Could not see what value the discussion added to the paper overall. Rather, it felt like the results were being explained again and the assumptions behind interpretation being ignored. Lastly, while there are no rules, the length of the discussion is rather short relative to the length of the results presented. In my experience, that can be indicative of a study that is exploring data rather than an experiment to test a hypothesis.
Minor edits
P1,L23: “atmospheric demand” is a bit wonky language for the abstract – could you phrase this differently?
P1,L35: Consider changing to “Mountain forests capture, store, purify and release water” to avoid ambiguity. Also, was “they” in reference to “mountain forests” or something else?
P2,L40: Are these all the water towers in the region? If so, state that. If not, justify why these towers.
P3,L4: The CRU data set is fairly course and known to contain rather few observations in Africa. Can you justify the use of these data here? Could another remote sensing product provide more accurate data?
P3,L4: I do not know how CRU gets PET. Could you provide some more information on how these data are prepared? This holds for all the data sets considered.
P3,L16: Break these longer sections up into sub-section to help the reader follow along.
P3,L31: What is “FU”?
P4,L11: 2011-2019?
References
van der Velde, Y., Vercauteren, N., Jaramillo, F., Dekker, S., Destouni, G., Lyon, S.W. (2014), Exploring hydroclimatic change disparity via the Budyko framework, Hydrological Processes, 28, 4110-4118.
-
RC3: 'CC1 again as RC', Steve Lyon, 26 Apr 2021
This study investigates climate and human impacts on water towers in East Africa. The analysis is conducted in a Budyko framework. The target region is often considered vulnerable to changes in water resources making this investigation warranted and the result likely informative. Overall, the study is well conceived. However, I feel there are some considerable limitations in the structure of presentation. Further, some of the mechanistic interpretations are not fully supported given the potential of confounding impacts and potential uncertainty in data and analysis.
In the introduction (especially around P2,L15-L20), I would expect to see some more consideration of the strengths and weaknesses of various approaches for assessing climate and land-use change on water resources. Do we have some results or previous work that are relevant for the region? What is the motivation for selecting the Budyko approach over other approaches? There is not much review of current science offered up in the introduction. This should be expanded to help the reader understand the motivations for the current study and approach.
In addition, the lack of framing the study in a research question or a hypothesis is a major weakness. The result is that the study is some exploration of data that does not seem to address a problem or help advance the science. Such exploration (“can-we-do-it” type of work) is fine for a technical report but more would be needed for publication in a peer-reviewed journal. I am confident the authors can put this study in a research framework and present a clean and testable hypothesis or a some societally relevant research question.
The study mixes direct observation data interpolated across sites and remotely sensed data at various scales. I’m wondering if there is any potential impact of the various assumptions and approaches in each dataset? Synthesizing across various approach can often compound huge amounts of uncertainties and errors as we build composite analysis (in space and time). How has uncertainty been considered in your analysis and what role would data error have on your results/interpretation? Some consideration and discussion of uncertainty impacts must be presented to help the reader understand how robust the findings are in this study. This should be fairly straight forward given how the water balances were constructed using 100 random points. Perhaps perform a re-selection of random point and assess the difference or use some sort of calibration/validation approach on a sub-division of the 100 points (like a boot strap).
Further, I am not sure about the 100 random points in the methodology. Why was this done? Is it just too difficult to define the spatial extent of the water towers (which would allow using all the spatial data in the area)? Seems there would be some value in conducting this experiment at various elevations to assess the impact of elevation (as temperature proxy) on the results. Please outline why the method of 100 random point was selected and what the impacts would be on the results relative to another method.
There appears to be a large amount of mechanistic speculation on why points depart from the Budyko curve. There has been ample research over recent decades explaining how we can see variations along and from the curve. Further, many different explanations have been offered as to why catchments would deviate from theoretical curves with time. Could you outline some motivation for how you can be certain you are isolating mechanisms with your analysis? We would anticipate much interaction and coupled response that could be masked in the movement of points in Budkyo space (see van der Velde et al., 2014). It i likely that this lack of consideration of complexity relates back to the weakness and lack of thorough literature review seen in the introduction.
Along these same lines, what role would other factors such as CO2 increase and/or human alteration to water usage have in these regions? I could envision shifts in water cycling due to an intensification of plant activity through increased NPP or agricultural intensification. Warmer and CO2 richer climates could behave differently. Further, how much pumping and/or movement through irrigation schemes takes place in some of these systems? I understand they should be pristine or high-elevation forest without impact, but are they really without abstraction or other anthropogenic impacts?
In general, the results as presented are dense and not easy to follow. Read things a few times and not sure I can understand all the nuance of what is being shown here due to how things are being presented. This is not helped by poorly constructed figures with overlapping number, limited axis labels, and multiple colors to track. A major effort to organize the results into a concise section is required. Start by group the various results into sub-sections and cleaning up the figures. Structuring this section could also be aided by a more thoughtful research question or hypothesis setup. Then the results could be organized into how they answer the research question(s).
The discussion section is lacking rigor. At best it repeats the results with more interpretation. I miss a connection to the literature and how the results help inform and advance the science. Also, what are the strengths and limitations of the approach considered and how do these impact interpretation? Could not see what value the discussion added to the paper overall. Rather, it felt like the results were being explained again and the assumptions behind interpretation being ignored. Lastly, while there are no rules, the length of the discussion is rather short relative to the length of the results presented. In my experience, that can be indicative of a study that is exploring data rather than an experiment to test a hypothesis.
Minor edits
P1,L23: “atmospheric demand” is a bit wonky language for the abstract – could you phrase this differently?
P1,L35: Consider changing to “Mountain forests capture, store, purify and release water” to avoid ambiguity. Also, was “they” in reference to “mountain forests” or something else?
P2,L40: Are these all the water towers in the region? If so, state that. If not, justify why these towers.
P3,L4: The CRU data set is fairly course and known to contain rather few observations in Africa. Can you justify the use of these data here? Could another remote sensing product provide more accurate data?
P3,L4: I do not know how CRU gets PET. Could you provide some more information on how these data are prepared? This holds for all the data sets considered.
P3,L16: Break these longer sections up into sub-section to help the reader follow along.
P3,L31: What is “FU”?
P4,L11: 2011-2019?
References
van der Velde, Y., Vercauteren, N., Jaramillo, F., Dekker, S., Destouni, G., Lyon, S.W. (2014), Exploring hydroclimatic change disparity via the Budyko framework, Hydrological Processes, 28, 4110-4118.
- AC1: 'Reply on RC3', Charles Wamucii, 27 May 2021
-
RC3: 'CC1 again as RC', Steve Lyon, 26 Apr 2021
-
RC1: 'Comment on hess-2021-151', M. van Noordwijk, 21 Apr 2021
General
1. The manuscript provides an interesting comparative study of the 'water towers' in East Africa and the change in terms of a simple water balance that can be inferred from a combination of various spatial data sources
2. The description of the quantitative framework can be improved, including a more consistent use of acronyms (especially for actual evapotranspiration) and equations
3. The study relies heavily on the use of a link between NDVI and the omega parameter in the Budyko framework, while the text acknowledges many factors (including soil, topography and seasonality) influence the relationship. At least in the discussion this needs some further work to see how much this could have influenced results and conclusions.4. The eight water towers are most described as 'replicates', rather than each having a specific geographic, ecological and social context: this may be the limit of what is currently possible, but at least some of the contrasts noted call for further analysis and attribution (e.g. in relation to human population density within and surrounding the water tower.
5. It would help the paper if sharper questions would be formulated at the end of the introduction that gives structure to the subsequent discussion
6. Beyond the supply of blue water to downstream parts of the watershed, the high actual evapotranspiration in water towers plays a role in regional rainfall recycling -- at least some discussion of this aspect would be relevant.
Minor
p1, Line 17 Mention 'steady state' assumption of Budyko framework at an annual time scale
p1, Line 24 'non-resilient' suggests a binary classification, is there a more gradual description on the degree of resilience
p1, Line 29 but mountains also cause 'rainshadows' that don't get the rainfall they might have had without the presence of a mountain...
p1, Line 31 more quantitative criteria are needed to get the type of delineation that you use here
p1, Line 34 in glaciated mountain chains water flow depends primarily on temperature, without ice cap on recent rainfall -- so the temporal variability will differ and dependence on land cover increase
p1, Line 35 'receive' is a rather passive description -- isn't it 'convert atmospheric moisture into rainfall'
p1, Line 37 Some reference to Africa as geologically old shield, but rift valley plate tectonics are associated with younger and higher mountains
p1, Line 39-40 If you introduce more quantitative P/Epot criteria in line 31, this discussion on E African water towers becomes more meaningful, as it relates to both the P and the Epot side of the ratio.
p1, Line 41 rainfall distribution is meager? what do you mean
p1, Line 42 Early work on rainfall in Sudan (El Tom, 1972) showed that the standard deviation of annual rainfall is nearly independent of mean annual value, showing that dry areas are highly variable in relative terms, with decadal variation super-imposed (Hulme, 1990) and not easily distinguishable from trended global climate change.p2, Line 5 Please unpack the sentence
p2. Line 6 Possibly relevant: ET estimates for SS Africa
p2, Line 8 For corrections on common deforestation discourses, see Aleman et al. 2018
p2, Line 16 The methods of Ma et al. 2010, 2014 combine these two categories by running rainfall statistics and recorded land-use change patterns in reverse order in calibrated process-based models
P2 Line 22 Maybe mention the steady state assumptions at annual time-scale upfront. A simple equation might help here.
p2 Line 26 It would help the subsequent discussion if you formulate some clear questions here that you try to answer in the results section
P2 line 30 if you want to avoid use of 'we', please find a less abstract passive formulation...Fig. 1 As 'montane forests' and 'water towers' only partially overlap, please give the quantitative definitions of both;
Public discussion on the Mau forests in Kenya described these as 'water towers', you don't; again clarifying the quantitative criteria can helpTable 1 In the Dewi et al. water tower delineation no fixed contour was used for the delineation, but one relative to the watershed as a whole. Please clarify your choice here, esp regarding the two (Aberdare and Bale) that were adjusted to the surrounding areas...
p3 Line 11 Here you seem to shift from PET to ET or ETa -- the preceding paragraph only mentions Epot.
p3 Line 14 This section may be clearer if you first present a water balance equation...
p3 Line 17 Deserts tend to have wadi's -- even in zones with low average rainfall, runoff occurs and rainfall intensity exceeds instantaneous infiltration capacity. Your Budyko-based description here needs some empirical adjustment (and scale considerations)
p3 Line 17 Not only under very dry conditions... About 50% of tropics has a P/PET ratio below 0.65; only a quarter has P/PET above 1.0
p3 line 31 Fu in stead of FU
p3 Equation 1 -- please specify the time step (1 year?)
Wouldn't it be better to include a DeltaS storage term, and then make explicit that you assume this is zero at the time scale of your analysis (but this is a considerable source of uncertainty and error...)
Fig 1A please settle on a single acronym for AET = ETa = ET
p3 Line 37 The seasonality effect is linked to the DeltaS term that you're hiding...
p4 line 6 As this is an empirical result, please describe the data set on which it was calibrated (from which it was derived)
p4 Line 8 So what about other influences on omega (soil types, and topography, climate seasonality, ...) that you just mentioned? You assume that these are at the average values in the Li et al. dataset? This will require some further justification, especially as you operate in the relatively rare bimodal rainfall part of the world.
p4 Line 17 Please remove ;
p4 line 12 Please indicate what you treat as 'known' inputs here and what as parameters to be estimated
p4 Line 36 So the EIBud is based on the NDVI relationship? It would help if you give more formal definitions of the terms here
p4 Line 6 Is the DeltaEI here the same as d in Eq 5?p5 Wouldn't it be easier and more informative to present the ETa/PET ratios?
p7 line 1 where omega values 'observed'? maybe 'derived'
Fig 7 How can Q estimates of 1000 mm/year be obtained for places with P hardly above 1000 mm/year?
p14 Discussion: A clearer structure of the discussion is needed.
p14 Line 19 As you used NDVI data, you used land cover rather than land use change as basis...
p14 Line 21 The sensitivity to land cover change reflects the limited degree of actual change (due to existing institutional arrangements) rather than the lack of response if such rules would be relaxed. Please distinguish these two aspects.
p14 line 24. An alternative to describing deviations along the Y axis (vertical) is to attribute them along the X-axis (horizontal): would such an approach be feasible?
p14 Discussion: Can you imagine doing the same analysis on the basis of ET/PET ratios attributed to NDVI, rather than the more complex Budyko route that involves P in the estimation of omega?
p14 Line 27: what do you mean by naturally occurring oscillations in this context? Does the occurrence of fire (partially anthropogenic) play a role: it changes NDVI for one or more years, increasing water yield; it may be more common on e.g. Mt Kenya and in the Imatong mountains
p15 Line 4 Please clarify 'resilience' as bouncing back in relation to 'elasticity' that refers to the degree of initial change, rather than its temporal dimension.p15 line 10 Isn't this a consequence of the way water towers are defined?
Suggested additional references
Aleman, J.C., et al. 2018. Forest extent and deforestation in tropical Africa since 1900. Nature ecology & evolution, 2(1), pp.26-33.
El Tom, M.A., 1972. The reliability of rainfall over the Sudan. Geografiska Annaler: Series A, Physical Geography, 54(1), pp.28-31.
Hulme, M., 1990. The changing rainfall resources of Sudan. Transactions of the Institute of British Geographers, pp.21-34.
Ma, X. et al. 2014. Attribution of climate change, vegetation restoration, and engineering measures to the reduction of suspended sediment in the Kejie catchment, southwest China. Hydrology and Earth System Sciences, 18(5), pp.1979-1994.
Ma, X. et al. 2010. Sensitivity of streamflow from a Himalayan catchment to plausible changes in land cover and climate. Hydrological Processes: An International Journal, 24(11), pp.1379-1390.
Marshall, M. et al. 2013. Improving operational land surface model canopy evapotranspiration in Africa using a direct remote sensing approach. Hydrology and Earth System Sciences, 17(3), pp.1079-1091.- AC2: 'Reply on RC1', Charles Wamucii, 27 May 2021
-
RC2: 'Comment on hess-2021-151', Anonymous Referee #2, 22 Apr 2021
General comments:
This is a very interesting study; however, I feel that it suffers from two main deficiencies that would need to be addresed in a revised version of the manuscript. The first is that there does not appear to be an overarching research question that is being addressed. Given what the authors know about the general climate of eastern Africa, it should be possible to suggest where the various water towers should plot on the baseline Budyko curve, and then test to see if in fact that was the case. The second issue is the absence of any specific consideration of the uncertainties associated with the estimates of the variables (P, PET, NDVI) used in the analysis. My concern here is that the authors spend considerable time discussing temporal changes in water balance components that may or may not fall outside the range of uncertainty associated with these components.
Specific comments:
Page/line
3/3 How was potential evaporation estimated?
3/2-8 What are the uncertainties associated with the estimates of P, PET and NDVI?
3/27-30 Can catchments deviate from the Budyko curve under stationary conditions?
7/Figure 3c How significant are these changes given the uncertainty associated with w?
12/Figure 8 Why is there a general overprediction of Q?
12/3-4 How much of the greater sensitivity of water yield to climate changes rather than land use changes is due to the form of equation 1? Is the differential sensitivity simply a function of the formulation of the Budyko curve, or is it real?
15/2-4 “… which further proves the presence of anthropogenic influence …” – does it really “prove” it?
19/Figure A4 How can you get different shapes for the baseline Budyko curves for the same w value (e.g. Aberdare Ranges vs. Mt Meru)?
I have attached other specific comments and suggested edits on the manuscript.
- AC3: 'Reply on RC2', Charles Wamucii, 27 May 2021
Peer review completion





