Impact of Climate and Land Use/Land Cover Change on the Water Resources of a Tropical Inland Valley Catchment in Uganda, East Africa

Round 1

Reviewer 1 Report

see attached file.

I selected 'major revisions' but what I suggest is at the cusp between major and minor, and does not require heavy new work

Comments for author File: Comments.pdf

Author Response

Response to Reveiwer #1 comments

 

 Major concerns

1. The difficulty with the climate projections used (which is very common), is that some projections are contradictory: in table 3, two models project rainfall trends opposite to those of the other four. Rather than considering the individual climate projections and the associated hydrological response, I suggest the authors mainly stick to an ensemble approach, by considering the mean, the standard deviation and the range (for example) of the model ensemble as a measure of the trend and its uncertainty. I think that the "big picture" would appear more clearly than at present, as the reader is sometimes lost in the face of contradictory trends. Additionally, a measure of the number of runs giving a change of the same sign may be used to further qualify the consensus on future changes (as in IPCC reports). By the way, either individual runs (table 4, fogs. 8, 10-12) or only the ensemble mean (fig 8, 13, table 6) are considered. Please clarify the underlying logic.

Response:

It is important not only to concentrate on the ensemble mean. First, calculating ensemble mean daily precipitation will result in completely wrong exceedance probabilities (see Fig. 5). Therefore, we need all GCMs/RCMs to run the SWAT model. From the results of all hydrological model runs we can calculate an ensemble mean. Nevertheless, low and high flow have to be analyzed for each GCM/RCM separately. For water balance, ensemble mean may give interesting results but do not point out important issues like extremes. Therefore, we analyzed mainly the six GCMs/RCMs and only for a few information we aggregate that for an ensemble mean.

 

2. The second concern is perhaps linked to the first one. I often I draw opposite conclusions from the graphs than those the authors write in the text. I do not always see clearly if the author refer to a particular model result or to the model average. For example, see my comments for lines 300- 311 in the detailed list below.

In the same vein, from fig 13 and table 6, I conclude that hydrological changes are mainly driven by LULC changes, while the authors claim that climate change has a dominant role. This is  embarrassing because it is a key conclusion of the study. Precisely, looking at fig 13, the change in water balance is larger between LUs for one RCP than between RCPs for one LU, (e.g. LU1), and it seems that shifting from RCP4.5 to RCP 8.5 further amplifies the changes (LU1, LU2). Except if I misunderstood something, I can not conclude, “Compared to the LULC change

scenarios, climate change will dominantly influence the hydrological processes of the inland valley.” (lines 611-612 for example). Please better explain how you get to these conclusions or clarify.

Response: We partly agree with the reviewer that LULC play a significant role in hydrological changes in the study area. We have rephrased the sentence and conclusion. From Table 3 and Table 5 it is clear that CC (range -44 to +136 %) has a higher impact on water yield than LUCC. (-25 to +4 %), respectively. For single components like surface runoff in % change is higher for LUCC than for CC but one has to keep in mind that the absolute values are rather low.

 

 

 

3. In line with 2., I suggest the author make a real attribution study, by including in table 6 the water

balance changes for current LU with future climate (fig 7) and the future LU with historical climate

(table 5), and modelling uncertainties from the GCM/RCP ensemble (see # 4 below)

 

Response: The new Table 3 now shows the change in selected water balance components due to the current LU with future climate impact. We thought if we include these changes in Table 6 (Table 5 in the revised manuscript), it would make the table congested and unclear to read and interprete easily. This also applies to Table 5 (Table 4 in revised manuscript), which shows the impact of future LU with historical climate on selected water balance.

 

4. From my experience, I can say that uncertainties and errors brought by models (moreover by cascade of models such as a GCM/RCM/hydro model) are larger than those brought by ground observations (there may be rare exceptions), and hence I think that the author can not write “The  uncertainties in the model performance especially the overestimation of some peaks can be explained by measurement errors in discharge and rainfall data” (lines 500-501) , which is a convenient by very likely wrong reason, unless this bias has been clearly attributed to bias in

observations. Biases in peak flows are most often due to modelling uncertainties (eg manning coefficient or other routing parameter).

More generally, I suggest the author better account for modelling uncertainties, or at least discuss them. For example, in the attribution effort (see point 3 above), I suggest the authors also consider the uncertainty in this ensemble and, making appropriate assumptions, use it to update the estimate of water balance changes. Given the large differences in rainfall for some runs (table 3), this uncertainty may compete (table 4, figure 7) with changes due to the climate/LU factors.

Response: We agree that a cascade of models introduce large uncertainties in the predictions. The SWAT model was calibrated using observed climate, land use, discharge etc. From our own field experience we know that observed rainfall and discharge data are uncertain which cannot be compensated by model calibration. By calibrating models using e.g. observed discharge parameter uncertainty is reduced although maybe for the wrong reason. We added a sentence for explanation.

When applying the model to CC and LUCC studies additional uncertainties have to be considered, especially when using different GCMs/RCMs for projections. This has been discussed. Although it would be interesting to perform a complete uncertainty study. it is beyond the scope of this investigation because we used a satisfactory calibrated model for CC and LUCC impact studies and discussed uncertainty in the driving forces.

 

 

5. I think the discussion section should be revised. Currently it is often more an extended summary of results than a real discussion. Some interesting synthetic points should be move to section.

 

Response: The discussion section has been revised and improved.

 

 

 

 

Detailed comments.

Section 2.1. I suggest you briefly describe the dominant hydrological processes at play in the basin

(Horton vs Dunne flow, origin of river flow (baseflow, interflow). It will provide a better

understanding of the simulated impact of LU on the components of the water balance (e.g., runoff,

table 6).

Response: A brief description of the dominant hydrological processes at play in the catchment have been included in section 2.1. The major hydrological processes at play are evapotranspiration and runoff (surface runoff and lateral flow) and these are influenced by variability in land use/land cover, soil properties and slope gradient.  

 

lines 29, 219, 255, …, 560 (also elsewhere). What do you mean exactly by “mixed” change or

trend ? Contrasted ? Contradictory ?

Response: The word “mixed” change and trend have been revised in the document.

 

lines 60:62 : “Inland valleys ...”. Please check this sentence.

Response: The sentence has been revised. It now reads better as “Inland valleys are highly diverse and complex systems of variable ecosystems from the upland areas through the hydromorphic fringe to the valley bottom, with each valley characterized with a typical hydrology”

 

Lines 63, last word: “poses” instead of “possesses” ?

Response: This has been revised.

 

lines 133 onwards (section 2.3). I have noticed that most model work was published in a previous

paper but the reader may need some basic facts to be reminded here. For example : why T° is

needed for hydrological simulations (used in scenario, I guess to compute PET/AET ?) and how T°

was considered in the calibration/validation work (historical period) ?

Response: Yes, T is required to compute potential evapotranspiration. We added this information.

line 144 : Regarding the topic of the journal, the readers of the paper (Climate Community) may not be used to calibrating/validating models. May be add a few words explaining why hydrologists do so ?

Response: A statement explaining why hydrologist carry out calibration and validation for the models has been added. See section 2.3, paragraph two (2)

 

line 170 : the space resolution of the disaggregated field is unclear to me : is it 0.44 ° (which is far larger than the basin), or do you used a kind of statistical sub-grid distribution ? Please clarify

 

Response: The resolution of the driving RCMs is 0.44°. By using the bias correction as described the data are downscaled to the local station level. We added a short sentence.

 

lines 175-185: At which time-step was bias correction done ? It is written that min and max

temperature were corrected on a monthly basis but what about daily T° (if used, unclear to me)

What about rainfall. It seems from section 2.3 that the model was run at the daily time step for

calibration/validation. Please clarify.

 

Response: Bias correction was conducted at daily time series for the precipitation and for temperature.   While for precipitation the probability density function of daily rainfall was considered, bias correction for temperature was performed using a mean monthly correction factor applied to daily values.

 

Lines 225-231 : this section should be moved to the results section

Response: The section has been moved to the results section 3.4

 

Fig 2. I guess these are daily values. Please update the caption.

 

Response: The word “daily” has been added to the caption

 

Fig 3. Change the color of Ens. Mean (yellow) : not easily distinguishable.

 

Response: The color has been changed to yellowgreen. It is now clearly distinguishable

 

ref 27 : Journal name repeated twice. I noticed several other refs with the same proble.m Please

check.

Response: We checked the list of references

 

Fig. 4.5 Looking at the values on the Y axis, I guess the figure deals with daily precipitation ?

Please clarify in the caption.

Response: the world “daily” has been added to the caption

 

Moreover, this figure looks strange to me :

• I cannot find any reason why the exceedance probability of the ensemble mean could be so

different than that of each model (given, for example, that it is not the case in fig 3b).

 

Response: The reason is that it does not rain every day in the climate models and not all at the same day. Therefore, the daily average (/ensemble mean) is computed with some 0 values. Summing up the mean values results in an appropriate monthly value but the intensity (mm/d) is completely different.

 

• Did you compute the probability on the non-zero (and non missing) values or did you

include dry days ?

 

Response: Dry days were included in the computation of the exceedance probability. This statement has been included in the caption for Figure 5 in the revised manuscript

 

• If I read the graph well, 20 % of rainfall is larger than ~ 5 mm and about 34 % is larger than 0,1 mm in the RCMs, which means that 66 % of rainfall in the models are lower than 0,1 mm/day. Is it correct ?. A frequent flaw of GCM and often RCM is to produce too many small daily rainfall (0,1mm/day is already very small, and corresponds to the rainfall measurement error). Is it the case in your dataset ? Please comment on that.

 

Please enlarge the X-axis beyond 35 % in order to display the whole range of all curves,

specially the green one.

 

• In my understanding, the green line most probably refers to observations. Please check.

Otherwise you have to give explanations to all the points above.

Response: the world “daily” has been added to the caption

 The green line is for the Ensemble mean for the six individual models. An explanation has been included. By averaging over all members of course you will have a weaker maximum and much more lower values, since the RCMs and our bias-correction only aim at accurately reproducing the regional climate. This means, for example, that it is raining in the right seasons, but it does not mean that it will rain in all models at the same day! By this, you will have more rainy days and your maximum will be decreased.

 

Also, rainfall intensity (mm/d) beyond 35% is lower than 0.1 mm/d which is  hydrologically not important. The reason for the style of the green line is the calculation of the daily average with results in much lower rainfall intensities and significant amount of drizzle.

 

Fig 6 : Better label X-axis with month names for easier reading (as the main text refers to month

names) ; for a better understanding, materialize the wet/dry seasons on the graphs with e.g.

shaded areas ? ; “arithmetic mean” is missing in c) and d)

 

Response: The X-axis has been labelled with month names and the wet/dry seasons have been shaded for easy interpretation. Arithmetic mean has been included in Fig c) and d).

 

lines 300-311. I can hardly see the announced trends, except if the authors look at the ensemble mean. For example : ≪ All models show a decrease in mean precipitation during the dry season (JJA) except for CanESM – CanRCM model and CanESM – RCA.. (l 302:303) ≫. As far as the whole JJA period is concerned, it is true only for the blue and grey curves (fig 6a) and the purple and grey curves (fig 6 b). The ensemble mean is very close to 0 for these months indicating no trends with respect to the historical period. Similarly: ≪ In the short rainy season (SON), all models project an increase in precipitation ( l 307-308 ) ≫. It is not true for the blue and grey curves in fig 6.a and 6.b (change is negative in Sep-Oct). This whole section needs to be rephrased more precisely to make the conclusions match what is seen on the graphs. I suggest the author only consider the ensemble mean +- 1 standard deviation, as no systematic signal can be derived from the individual runs (far too noisy).

 

Response: The paragraph under section 3.2.2. has been revised and now reads better

 

lines 312 -320 . Same as above. Moreover, I am surprised that the temperature change (yearly mean) is far lower than 0,2°C even for RCP8.5 in which the global mean T° is expected to increase by several degrees (up to 7 in some extreme CMIP6 simulations) in 2100, which means that a part of that rise should already be there in 2050, even in CMIP5 runs. Yet, no T° increase (on average, over the year ) can be seen in fig 6 c and d). Is this realistic ? Does this mean that the projected T° rise is modest over this region or does it come from the bias correction method and /or from the seasonal bias in the simulated T° ( too cold in Sep-Dec, fig 5 a) ? Please comment on that.

 

Response: The temperature data are taken from the GCMs/RCMs. It seems to be a local phenomenon especially for East Africa and cannot be compared to the global mean.

 

lines 328 onwards (whole section 3.3.1). Is it really surprising that changes in rainfall induce changes in the water balance components ? May be it would be more informative to evaluate the magnitude of the changes. The CanESM-CanRCM run simulates + 27 % rainfall (RCP4.5, table 3), and fig7 suggests that it translates into a ~3-fold increase in runoff,~ +75% in recharge, etc. May be plotting the change in water balance components vs the change in rainfall (same for T°) would be more informative ? i.e. comment on the magnitude of the changes rather than on the trend which is obvious ?

 

Response: Table 3 (in the revised manuscript) indicating the magnitude of change in the selected water balance components has been included with description of the results and discussion in the revised manuscript. Nevertheless, this does only make sense for some of the water balance components as small changes in e.g. surface runoff (small values, e.g. 2 mm/a instead of 1 mm/a is 100 % increase)) will result in very large relative changes which are not relevant.

 

By the way it seems to me that the variables shown on fig 7 are not of the same nature with respect to the annual water balance, which can write for example “rainfall=discharge+ET+soil water storage +GW storage”. It is unclear how (in this model), water yield, lateral flow,  relate to the other components.

 

Response: Fig 7 has been deleted since it shows only trends. This has been replaced with Table 3 which shows the magnitude of change for the selected water balance components such as ET, Surface runoff, Water yield, deep aquifer recharge and precipitation. In this model, water yield is simulated as a summation of lateral flow+ surface runoff+ groundwater flow (baseflow). Thus, lateral flow is a component of water yield in the model.

 

lines 340-341 : Is PET calculated by the SWAT model ? I guess it is mainly dependent on T° which

explains why PET virtually does not change (see above). Then, the change in ET is mainly due to

the fact that it rains more ? But in this region ET is limited by the available energy. Did you use net radiation in the forcing ?

 

Response: The SWAT model calculates PET and for this study Penman monteith method was applied to derive PET. This method requires solar radiation as one of the inputs and therefore, solar radiation was used as a forcing in the model.

 

Fig 7 (caption) spell check : “Simulations based on BIAS-CORRECTED precipitation…”

 

Response: Figure 7 has been deleted

 

Table 3 and Table 4 : the values in the last line of both tables (arithmetic mean) are not equal to the average of the 6 values of the lines above (except two). Please check or explain.

 

Response: Results in Table 3 of the old manuscript have been merged in Table 3 for the revised manuscript. Thus, Table 3 for the old manuscript has been deleted. This has been corrected and the arithmentic mean no corresponds to the average of the 6 individual models for Table 3 and Table 4

 

lines 359-364 and Fig 8. How discharge compares with the variables of fig. 7 (surface runoff,

lateral flow and water yield, ) ; how is it deriving from them ?

 

Response: Discharge is the same as water yield. It is the sum of surface runoff, lateral flow and groundwater flow (baseflow). 

 

line 361 : grammar check “The projected monthly change in discharge follows a similar trend of

the projected monthly precipitation change.

 

Response: Figure 8 has been deleted and thus, the statement “The projected monthly change in discharge follows a similar trend of the projected monthly precipitation change”

 

lines 362-363. “This indicates that precipitation is a key factor in determining discharge in the

investigated inland valley.” Dealing with a watershed, this result was rather expected ! I am not

sure it is worth noticing that way...

 

Response: The whole paragraph has been deleted since Fig 8 was also deleted.

 

Is fig 8. really useful given that the seasonal patterns are more detailed in fig 10, and that

ensemble results (fig 10) are more robust that individual ones (fig 8) when dealing with

GCM/RCMs. ?

 

Response: We agree with the reviewer that Figure 8 is not necessary more so it does not give conclusive results therefore, it has been deleted

 

lines 369-382. Please rephrase this section, considering the following points :

• How are defined low and high flow ?

 

Response: Definition for low and high flow has been included in section 3.3.2

• It can not find the 30-35 % (line 377) proportion of annual flow on fig. 9.

 

Response: The whole sentence with “30-35%” has been deleted since it was abit confusing.

• I would not talk about over- and under-estimations, as it is not a matter of bias or

estimation error..

 

Response: The phrase “over- and under-estimations” has been deleted

 

• It seems from fig 9 that the highest discharge values where drawn from the observations

(black dots in the upper-left side of the plot). Does that mean that the extreme peak flows

are not higher in the projected series ? Accordingly, I do not understand “However, the

projected extreme peaks are scarce (~0-2%) compared to the observed discharge.” (l

376).

 

Response: Fig 9 now Fig 8 in the revised manuscript has been revised hence, the peak flows for the projections are visible. The sentence ““However, the projected extreme peaks are scarce (~0-2%) compared to the observed discharge” has been revised and now reads better.

• line 378 “extreme” is not a synonym for “high flow”.

 

Response: The word “extreme” has been replaced with the word “high” and now the statement reads better

 

Fig 9. I guess these are mean annual values. Please add this information in the caption

 

Response: These are daily values. We have added the word “daily” in the caption.

 

Fig. 10. Please precise how to read the box plots (mean, sd, outliers, …) .

 

Response: These are standard Box-Whisker plots.

 

Fig. 11.

• I guess it represents the return period of the highest daily discharge over the year. Please

precise.

 

Response: We agree with the reviewer that Fig 11 represents the return period of the highest daily discharge over the year. Therefore, this has been effected into the write up for Fig 11.

 

• Is there any reason why the authors use “return level” instead of the more usual “return

period” expression ?

 

Response: The word “ return level  has been changed to “ return period” which is commonly used.

 

• I suggest you simplify the labels on the X-axis, moving “return period” in the plot title, and in the legend as well (let only the run names) ?

Response: The word “return period” has been removed from the plot title and the legend. The graph is now simplified and clearer

 

• Better replace the light colors with darker ones in both plots (hardly distinguishable)

 

Response: The light colors have been replaced with more clear colors.

 

• Why the sharp contrast between RCPs and historical series in fig 11 is not as obvious in fig 9. Do both graphs deal with the same variables (daily, annual Q) ? . Please clarify.

 

Response: Fig 8 (fig.9 old) shows the ensemble mean while Fig. 11 shows the different GCMs/RCMs

 

• It seems that the 50-year return flow is higher for CNRM-CCLM and EC-EARTH-CCLM (fig

11 a) than in the historical period (fig 11 b), whilst those two GCM/RCM runs simulate a

decrease in mean annual precipitation and discharge in both RCPs (table 3 and 4). Is it

correct and how can you explain it ?

 

Response: This is the result of the analysis. This means that less rainfall cause less discharge but this reduced rainfall has higher intensities (mm/d). Also total surface runoff is lower in the two GCMs/RCMs. Therefore, the change in the 50 years return flow is caused by single events, which are much higher.

 

Fig. 12.

• Please update X-axis labels with run names (as in fig 7)

 

Response: The X-axis labels have been replaced with run names

 

• Please precise how to read the box plots (the layout differs from Fig 10)

 

Response: The box plot layout has been modified. Now matches with Fig 10.

lines 434-440 and Table 5. It is still unclear to me what are water yield and lateral flow, and I can not figure out which one/ones is/are related to the discharge at the basin outlet. In the table, the sum of all the water balance terms does not equal rainfall. Please check or clarify.

 

Response: Water yield is a summation of lateral flow, surface runoff, base flow/groundwater thus lateral flow is a component of water yield. Water yield is the same as discharge since both are a summation of lateral flow, surface runoff and groundwater flow/base flow. Therefore, water yield is directly related to discharge at the outlet basin.  

 

lines 439-440. How do you explain that ET increases slightly ( 8 mm) in the Exploitation scenario,

which should be the one with the lowest vegetation cover ? Is it within the modelling uncertainties ?  In this case we can not consider it increases.

 

Response: Even in the current land use there is a lot of agriculture. Agriculture is not bare soil and causes also significant transpiration although the largest part is evaporation in the wetland. Together with the high water use by the eucalyptus trees, the results are realistic.

 

By the way the water balance terms only change a little in this scenario. Does it mean that the

basin is already close to this exploited state ?

 

Response: Yes, the catchment is close to its exploited state whereby its natural state has been converted to agriculture (accounting for about 65% of the total catchment area). A statement on the status of the catchment has been included in this section

 

Table 6. These results seem to be mean values. Please precise what has been averaged and

consider concern #4.

Response: We have added some sentences in order to clarify that issue

 

 

l 508 : systemqtic bias pas synonyme d’incertitude , si on peut le corriger : une simul peute tre

biaise mais certaine. Incertoitude a cause du consensus entre les moedles. Tres commun

Systematic biais aussi du a des biais dans les GCM/RCM. Ex trop de petitres pluies, peu

d’ameliorartion depuis CMIP3 → correction de biais = necessaire .

S’il y a eu pdf matching, commela distrib de pluies reste aussi biaisee ?

 

line 520 : figure 6 suggests that T° change is about 0°C +- 0,1°C, which means nearly no change.

Response:

The uncertainty in the climate projections is due to the GCMs/RCMs and not due to bias correction. With bias correction the simulated history is corrected towards the observed history and the correction function is applied to the future. Even after bias correction, the future is uncertain.

 

line 522. This conclusion is not so obvious in section 3.2.2 and Fig. 6, specifically regarding T°

change. Please check and make it clearer. By the way this sentence could be a conclusion of section 3.2.2

Response: We have reformulated that part and it fits now better the complexity of the climate signal from figure 6.

 

line 528 : “Noteworthy, in the long rainy season, March has the highest increase in projected

precipitation,… “. Contradictory with fig 6 where the highest rainfall increase in the ensemble mean occurs in Oct-Nov. Please check / explain.

Response: We adjusted the sentence and added another one to clarify. It was referring only to the time within the long rains, while Oct/Nov belongs to the short rains.

 

 

lines 531-533 : “Comparisons of climate scenarios, RCP8.5 has...”. Please check and rephrase the  whole sentence. Plus : this trend in T° is not obvious at all from fig 6 c and d, where T° change seems more or less comparable in both RCPs, and where the trend on rainfall is mainly due to 1-2 model runs. Please check / better align the conclusions with the graphs ?.

Response: This has been revised

 

 

lines 540 onwards Section 3.1. Please refer to the associated graph/table (fig 7 ?)

Response: This has been revised

 

 

line 541 : why was “aquifer recharge” removed from the list of water balance components ?

 

Response: Aquifer recharge has been included

 

Section 4.3.1. These arguments were rather expected (more/less rain → better/worse agricultural conditions) and not very useful to me. I suggest the authors discuss the uncertainties arising for the climate x LULC projections, and the resulting uncertainties in policy advices.

Response:  We have tried to discuss the uncertainties arising from climate projections

 

Section 4.3.1 and 4.3.2. I suggest you merge discussion on discharge (4.3.2) with that of the other water balance components, given that they behave the same way.

 

Response: The sections (4.3.1 and 4.3.2) have been merged

 

lines 578-581. I do not understand how uncertainties in rainfall measurements from one station

could affect the flow return period simulated with a rainfall forcing from a particular GCM/RCM

model. Please clarify ? It seems that the difference in the estimated return period probably lies in difference in the probability of large daily rainfall events. I think the discussion should address this point.

 

Response: Observed rainfall was used to bias correct rainfall simulated by the GCMs/RCMs. Therefore, uncertainties in the observations are directly transferred to the bias correction.

 

lines 604-606. I have a problem with this sentence : “It is recommended to use EC-EARTH – CCLM model for low flow analyses and CanESM – CanRCM and MIROC-RCA models for high flows and potential flood risk assessments in order to cover the spatial dimension for the uncertainty among these climate change scenarios.” You assume implicitly that your GCM/RCM ensemble covers the whole range of possible situations, and that you have correctly sampled the lowest low flows and the highest high flows. But are you sure that another combination of GCM/RCM, or simulations from another CMIP6 exercise, would not give a different range of possibilities ? More fundamentally, this issue is linked to the way you handle uncertainties in the work. Considering one particular model is dangerous. Considering ensemble statistics is more robust, and facilitates the explanations, I think.

 

Response: Our analysis is based on six GCMs/RCMs. Of course this cannot represent all GCMs/RCMs which will ever be available but covers already a large range of possible developments. For developing adaptation strategies one should look at the extremes concerning high and low flow from the investigated climate models instead of concentrating on a mean (ensemble) future. This is a worst-case approach even though other climate models may be produce more extreme results. 

 

 

 

 

Author Response File: Author Response.docx

Reviewer 2 Report

General comments:

The authors present a very compelling case in the Introduction on the combined effects of climate change and LULC changes on the hydrology of a catchment area in East Africa. Although not completely novel, these type of studies need to be tailored to specific regions owing to the fact that climate change reflects regionally/locally differently in different regions, and LULC practices are also geographically-dependent. So this study is of importance to scientists, managers, governments, and conservation institutions a like. One major drawback of the manuscript is that it presents data with a high degree of uncertainties, consequently, all results related to this data contains further uncertainties and they vary wildly. Any type of meaningful conclusion would be impossible to infer from this data, unless they discard some of the models by deeming them unfit, and them focusing the discussion/conclusions on the more reliable models. One key question I have, and stated below with other comments, is why the authors didn't test other bias-correction techniques? Including one that they mention could potentially improve their results.

Please send your manuscript to an English editing service for a complete revision of language, grammar, and readability, it also needs to be proofread thoroughly for minor issues (e.g., double space, missing space, comma placement, etc).

Specific comments:

• Title (and first use in the Abstract and Introduction) - the appropriate term is "land use/land cover (LULC)", please correct this.
• Abstract: Please revise the statement regarding increasing and decreasing RCM temperatures, are these two trends for the two RCPs or for historical vs future. It is not clear.
• Values should have a space between the number and units.
• L66-67: Please clarify wether GCMs and RCMs predict precipitation increase or decrease for East Africa, and wether this is for total yearly accumulated precipitation.
• L139: The study area presents a varying terrain, yet only one weather station was used to provide input data into the SWAT mode. Please justify the use of a single point data source for such a large area in irregular terrain. Is this because of lack of data sources or a limitation of the model?
• L166: Please state that 4 GCMs were used to drive 4 RCMs in different combinations to produce the 6-member ensemble.
• L201-202: Please mention what probability distribution was adopted (Weibull?).
• Table 2: Please revise the first two rows, as there seems to be some inconsistencies related to the RCP used in combination with LU1.
• L363-364: Why are the projected monthly discharge change of the two models highlighted noteworthy and not the others? Why are conclusions drawn using these results and not the others? Are these two models considered more reliable? If this type of assessment statement are further elaborated on in the Discussion section, then eliminate them here, they seem out of place and based on cherry picking.
• L379-380: Please revise the statement given on the RCP comparison, it should be 8.5 predicting higher total discharge than 4.5.
• The authors attribute the high uncertainties associated with their results, among other reasons, the bias-correction technique applied and how a different approach could improve results. Why didn't they tested these other bias-correction techniques.

Author Response

Response to Reviewer #2 comments

General comments:

The authors present a very compelling case in the Introduction on the combined effects of climate change and LULC changes on the hydrology of a catchment area in East Africa. Although not completely novel, these type of studies need to be tailored to specific regions owing to the fact that climate change reflects regionally/locally differently in different regions, and LULC practices are also geographically-dependent. So this study is of importance to scientists, managers, governments, and conservation institutions a like. One major drawback of the manuscript is that it presents data with a high degree of uncertainties, consequently, all results related to this data contains further uncertainties and they vary wildly. Any type of meaningful conclusion would be impossible to infer from this data, unless they discard some of the models by deeming them unfit, and them focusing the discussion/conclusions on the more reliable models. One key question I have, and stated below with other comments, is why the authors didn't test other bias-correction techniques? Including one that they mention could potentially improve their results.

Response: The highest degree of uncertainty is due to climate projections. All other data used in this study do have the common uncertainty which do not differ from other studies worldwide. How can one decide which model describes a possible future correctly? We used six GCMs/RCMs which show a large range of possible futures as is it often observed for West and East Africa.  Because we analyzed the impact of CC and LUICC on the hydrology of the investigated catchment we were able to study how the uncertainty on diving forces transform into uncertainty in the hydrology.

There exist a number of bias corrections methods. Even when using different methods, the uncertainties in the driving GCMs/RCMs still exist. In accordance with literature, we found the chosen method appropriate for bias correction as it considers the probability density function and results in appropriate exceedance probabilities of daily rainfall. Applying another bias correction method would only slightly influence the results but not the general findings.

Please send your manuscript to an English editing service for a complete revision of language, grammar, and readability, it also needs to be proofread thoroughly for minor issues (e.g., double space, missing space, comma placement, etc).

Response: the language, grammar and readability has been improved

Specific comments:

• Title (and first use in the Abstract and Introduction) - the appropriate term is "land use/land cover (LULC)", please correct this.

Response: This has been corrected

• Abstract: Please revise the statement regarding increasing and decreasing RCM temperatures, are these two trends for the two RCPs or for historical vs future. It is not clear.

Response:

The increase/decrease is given as future temperature compared to historic temperature.

• Values should have a space between the number and units.

Response: the number and the units have been spaced in the revised version

• L66-67: Please clarify wether GCMs and RCMs predict precipitation increase or decrease for East Africa, and wether this is for total yearly accumulated precipitation.

Response: Clarification have been made about the GCMs and RCMs precipitation predication over East Africa. “Reference to Endris et al., 2019 study indicates that GCM and RCM data project a decrease in seasonal  rainfall over most parts of the East African region during JJAS and MAM seasons, while an increase in rainfall over equatorial and southern part of the region during OND, with higher changes in the equatorial region”

• L139: The study area presents a varying terrain, yet only one weather station was used to provide input data into the SWAT mode. Please justify the use of a single point data source for such a large area in irregular terrain. Is this because of lack of data sources or a limitation of the model?

Response: The catchment had only one weather station and therefore, the study was limited by the data sources although the model can accommodate as so many weather stations. The good news is that the model captured well the trend in the observed streamflow of the catchment indicating a good representation of the climate data.  This has been clarified in the revised manuscript.

• L166: Please state that 4 GCMs were used to drive 4 RCMs in different combinations to produce the 6-member ensemble.

Response: The statement has been added in the revised version of the manuscript. See sub section 2.4, paragraph two.

• L201-202: Please mention what probability distribution was adopted (Weibull?).

Response: We have included the probability distributions, which were adopted, and these include the Weibull, Frechet and Gumbel distributions

• Table 2: Please revise the first two rows, as there seems to be some inconsistencies related to the RCP used in combination with LU1.

Response: The first two rows have been revised and the first row with RCP8.5+LU1 changed to RCP4.5+LU1 correctly aligning with the abbreviation on column three.

• L363-364: Why are the projected monthly discharge change of the two models highlighted noteworthy and not the others? Why are conclusions drawn using these results and not the others? Are these two models considered more reliable? If this type of assessment statement are further elaborated on in the Discussion section, then eliminate them here, they seem out of place and based on cherry picking.

Response: The statement has been eliminated and shifted to the discussion section to strengthen the discussion. See subsection 4.3.2.

• L379-380: Please revise the statement given on the RCP comparison, it should be 8.5 predicting higher total discharge than 4.5.

Response: The statement has been revised and now reads better

• The authors attribute the high uncertainties associated with their results, among other reasons, the bias-correction technique applied and how a different approach could improve results. Why didn't they tested these other bias-correction techniques.

Response: We are attributing the high uncertainties in precipitation due to non-bias correction but after applying the bias correction technique, the precipitation bias is significantly reduced indicating that the technique performs better to reduce the bias in the results. Therefore, we think there is no need to testing other bias-correction techniques since this approach yielded good results. See section 4.2.1 in the revised manuscript

 

 

Author Response File: Author Response.docx

Reviewer 3 Report

Climate and land use / cover (LULC) change have an important impact on water resources. This study assessed climate and LULC change impacts on the hydrological processes of a tropical headwater inland valley catchment of Uganda. The potential implications of climate and land use management change for water balance and total discharge in a tropical inland valley of Namulonge, central Uganda were simulated using the SWAT model. The topic has merit and seems to fit well with the scope of the journal. The article is easy to read and gives a well-presented overview of change in the study area. There are some minor concerns as follows:

1. Line 25, RCP should give its full name in English.
2. Line 50, Km2 should be km².
3. Line 99, Figure 1 may be better to show remote sensing images, because the study is about LULC, so that readers can better understand the types of LULC in the study area.

Author Response

Response to Reviewer#3 comments

 

Comments and Suggestions for Authors

Climate and land use / cover (LULC) change have an important impact on water resources. This study assessed climate and LULC change impacts on the hydrological processes of a tropical headwater inland valley catchment of Uganda. The potential implications of climate and land use management change for water balance and total discharge in a tropical inland valley of Namulonge, central Uganda were simulated using the SWAT model. The topic has merit and seems to fit well with the scope of the journal. The article is easy to read and gives a well-presented overview of change in the study area. There are some minor concerns as follows:

1. Line 25, RCP should give its full name in English.

Response: The abbreviation “RCP” has been written in full as “Representative Concentration Pathways”

1. Line 50, Km2 should be km².

Response: It has been corrected and well written

1. Line 99, Figure 1 may be better to show remote sensing images, because the study is about LULC, so that readers can better understand the types of LULC in the study area.

Response: A land use map has been added. See Figure 2 in the revised version of the manuscript

 

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have substantially improved the manuscript, and considered my main concerns. Some sentences must be checked for english here and then. Apart from these editing problems, I think the manuscript is now ready for publication

 

Author Response

Response to Reveiwer #1 comments

Many thanks for your helpful comments on our manuscript. We were able to incorporate all your suggestions during the review. Please, find below our detailed answer to your comments. Your suggestions and remarks have really improved our paper. We hope that you find your comments sufficiently considered and we would like to thank you for your support.

Comments and Suggestions for Authors: The authors have substantially improved the manuscript, and considered my main concerns. Some sentences must be checked for english here and then. Apart from these editing problems, I think the manuscript is now ready for publication

Response: We have cross-checked the English language and spellings. The manuscript has been improved

Author Response File: Author Response.docx

Reviewer 2 Report

No further specific comments.

Author Response

Response to Reviewer #2 comments

Many thanks for your helpful comments on our manuscript. We were able to incorporate all your suggestions during the review. Please, find below our detailed answer to your comments. Your suggestions and remarks have really improved our paper. We hope that you find your comments sufficiently considered and we would like to thank you for your support.

Comment 1: English language and style are fine/minor spell check required 

Response: We have improved the flow and English language of the manuscript.

Comments/suggestions 2: Research design and conclusion must be improved

Response: The methodology and conclusion have been improved. For instance, sub-section 2.3 and 2.6 for the methodology section have been modified.

Author Response File: Author Response.docx