Hydrological modeling using the SWAT Model in urban and peri-urban environments: The case of Kifissos experimental sub-basin (Athens, Greece)
- 1Laboratory of Engineering Geology and Hydrogeology, School of Mining and Metallurgical Engineering, National Technical University of Athens, Heroon Polytechniou Str. 9, 15780 Zografou, Athens, Greece
- 2Laboratory of Hydrology and Water Resources Development, School of Civil Engineering, National Technical University of Athens, Heroon Polytechneiou Str. 9, 157 80 Zographou, Greece
- 1Laboratory of Engineering Geology and Hydrogeology, School of Mining and Metallurgical Engineering, National Technical University of Athens, Heroon Polytechniou Str. 9, 15780 Zografou, Athens, Greece
- 2Laboratory of Hydrology and Water Resources Development, School of Civil Engineering, National Technical University of Athens, Heroon Polytechneiou Str. 9, 157 80 Zographou, Greece
Abstract. SWAT (Soil and Water Assessment Tool) is a continuous time, semi-distributed river basin model that has been widely used to evaluate the effects of alternative management decisions on water resources. This study, demonstrates the application of SWAT model for streamflow simulation in an experimental basin with daily and hourly rainfall observations to investigate the influence of rainfall resolution on model performance. The model was calibrated for 2018 and validated for 2019 using the SUFI-2 algorithm in the SWAT-CUP program. Daily surface runoff was estimated using the Curve Number method and hourly surface runoff was estimated using the Green and Ampt Mein Larson method. A sensitivity analysis conducted in this study showed that the parameters related to groundwater flow were more sensitive for daily time intervals and channel routing parameters were more influential for hourly time intervals. Model performance statistics and graphical techniques indicated that the daily model performed better than the sub-daily model. The Curve Number method produced higher discharge peaks than the Green and Ampt Mein Larson method and estimated better the observed values. Overall, the general agreement between observations and simulations in both models suggests that the SWAT model appears to be a reliable tool to predict discharge over long periods of time.
Evgenia Koltsida et al.
Status: final response (author comments only)
-
RC1: 'Comment on hess-2021-482', Anonymous Referee #1, 18 Oct 2021
The paper is interesting and pleasant to read. I have only minor doubts that could be addressed before publication.
first of all, in the introduction (lines 33-45) the Authors provide an overview of experimental watersheds dealing with hydrological observations. Although the topic is interesting, I suggest to shorten this paragraph because the cited watersheds ar not considered in the present manuscript and the topic cuold be misleading for the reader.
Second, in introduction (lines 72-73) I believe that aims and novelty of the manuscript should be more enphasized.
Third, in paragraph 2.2, I found a little confusing the instruments description and the data that later are used in the manuscript. The Authors mention (lines 101-106) water level and water velocity sensor that are installed in the experimemental watershed, but specify only units in mm, what about the velocity? is the sensor present and used?
Forth (lines 296-297), the Authors mention an interesting effect of precipitation time step that could affect the result, but in my opinion the results could be affected also by the classic difficulty in obtaining reliable estimations of GAML parameters based on the soil type and heterogeneity. Eventually this issue could be discussed here.
Fifth (line 337) the Authors mention observational errors; could they specify if this could be attributed to the estimation of channel and hillslope flow velocities?
other minor corrections are:
line 81: route?
line 97: specify the acronym.
line 190: changing or constant?
line 198: do the numbers refer to discharge?
-
AC1: 'Reply on RC1', Evgenia Koltsida, 28 Dec 2021
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2021-482/hess-2021-482-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Evgenia Koltsida, 28 Dec 2021
-
RC2: 'Comment on hess-2021-482', Anonymous Referee #2, 29 Oct 2021
Line 26: “Experimental catchments ”What this means? It is not clear what you are trying to say. Rephrase the sentences. Line 28: They are? Who are they? there is two time the a sentece startint with they but it is not clear who or what thery are.
Line 47-49: Which are these models? Why did you choose SWAT?
Line 72: Are you used SWAT+?
Line 72: The aim and innovation of the work need to be better discussed both in introduction and discussion. There are plenty of work which evaluate model performances with different data resolution (soil, morphology and climate) Here some examples which can be useful in discuss the main innovation: How you work give new insight inthe research? I cannot see any novelty or secondary elaboration from the canonical SWAT application such as a susceptibility map or future prediction. I suggest to the authors to focus more on this points
https://doi.org/10.5194/hess-24-3603-2020
https://doi.org/10.1016/j.jenvman.2020.110625
https://doi.org/10.5194/hessd-7-4411-2010-
AC2: 'Reply on RC2', Evgenia Koltsida, 28 Dec 2021
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2021-482/hess-2021-482-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Evgenia Koltsida, 28 Dec 2021
-
RC3: 'Comment on hess-2021-482', Anonymous Referee #3, 04 Nov 2021
General comments:
The authors present a modeling study of an urban catchment with the SWAT. The study aims to compare the performance of daily simulations (using the curve number method) and hourly simulations (using the GAML method). The paper is within the scope of the journal, adress relevant scienfic questions and the presentation is of quality. The Methods and assumptions are well described and understandable. However, Results and Discussion could be developped: at the moment they read a little like a list of comments (especially section 3.3) without enough physical discussion. The authors find that different modeling timesteps and different runoff generation methods impact model parameters and performance, which is somehow expected and established. The novelty of the contribution therefore needs to be detailed and explained.
The reasons why the subdaily model did not perform as well as the daily model deserve more physical discussion about hydrological processes and implications. In addition, is it really fair to compare hourly and daily performance metrics directly? Shouldn't the results of the GAML method be aggregated to the daily timestep to really be compared with the CN method?Specific comments:
Abstract:
The abstract does a good job at summarising the work in concise and clear language. The novetly of study could be explained in the Abstract and numerical values of Results could also be given.
L12: Is 'demonstrate' is the right word? The paper rather "examines" the influence of precipitation timestep on perfomance metrics
L21: "long periods of time": in the paper the modeling period is 3 years which is not very long
1. Introduction:
L74: water level data of water flow data?
2. Materials and Methods:
L95-96: "in different times and under different weather conditions" : is the monitoring continuous?
L101-106: The authors could consider adding a sentence about velocity measurement? Can the real life accuracy of the probes be somehow estimated (especially since observational errors are mentioned later in the discussion)?
L133: and subdaily?
L198: mean and standard deviation of daily discharge? In m3/s ?
3. Results and DiscussionSection 3.1
This section describes the results of the sensitivity analysis. It shows that the daily simulation seems to be more sensitive to runoff generation parameters whereas the sub-daily simulation is more sensitive to channel routing parameters. The section could be better structured: for example it starts with a discussion on CH_N2, then discuss GWQMN and GW_REVAP, then again CH_N2 and in the same paragraph discuss CN2, which makes it hard for the reader to follow the reasoning.
L241-242: repetition from lines 234-239.
L252: is this difference physically meaningful?
L253: CH_N2 or CN2?Section 3.2
This section presents the performance metrics of the daily and subdaily simulations. The authors conclude that the CN method is better than the GAML method. But, as stated above, is it fair to compare daily simulations with hourly simulations? Shouldn't the hourly simulation be aggregated to a daily timestep to have a 'fair' comparison?L272: an explanation for the underestimation?
L274: it is expected that a daily timestep performs better than a subdaily timestep. It could be interesting to compare both methods at the same timestep, as stated above.
L284: performance metrics are satisfactory, but performance metrics also depend on what we want to use the model for: for example, though the model here replicates the timeseries quite well, it could not be trusted for flood analysis (poor performance on hourly peaks).
L289: what is 'ET runoff generation' ?
L296: This is interesting but a little unclear: what is meant by 'too large' ? Would the results be better with, say, 10 min rainfall? Why?Section 3.3
This section focuses on six "heavy" rainfall events in which the authors describe, in text, peak flow values and average flow values during the events. The hourly model underestimates peak flows. This section comes as a surprise for the reader as it is not mentioned in the Methods. Moreover, it is unclear how the events were selected: are they the 6 larger events of the timeries? It would be interesting to have an estimation of their return period to define "heavy"? From line 312 to line 335, the text simply describes hydrographs, without comments or analysis. Maybe the authors could consider a Table instead with rainfall characterstics (totals, duration, return period, etc.) and standard describers of hydrographs (peaks flows, difference between peaks, etc.)? It is concluded that the underestimation of peak flows is due to uncertainty in observed data or input data (rainfall), but without many arguments or any estimation of these uncertainties. How can one be sure that the errors are due to the data and not to the model? There probably exists a parametrization which can replicate high flows much better, with poorer performance on low flows?L336-340: It is correct that errors in a model can be explained by 1/ uncertainty in the observed data 2/ uncertainty in input data or 3/ the model structure/parametrization. But what about this particular study? This could be futher discussed.
4. Conclusions:
L366: 3 years is not really "long time".Comments on Figures:
Figure 1 and Figure 2 could be merged
Figure 3: It could be worth to add rainfall?Minor comments and typos:
L12: this study demonstrates (remove the comma)
L29: they are 'used to monitor', not 'able'
L30: they monitor groundwater
L81: "route" is unclear-
AC3: 'Reply on RC3', Evgenia Koltsida, 28 Dec 2021
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2021-482/hess-2021-482-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Evgenia Koltsida, 28 Dec 2021
Evgenia Koltsida et al.
Data sets
Weather data National Observatory of Athens, NOA https://www.meteo.gr
Discharge data Open Hydrosystem Information Network https://openhi.net/
Soil data Food and Agriculture Organization, FAO http://www.fao.org/
DEM data Shuttle Radar Topography Mission, SRTM https://earthexplorer.usgs.gov/
Land use data Corine Land Cover, CLC 2018 https://land.copernicus.eu/
Model code and software
SWAT code USDA Agricultural Research Service and Texas A&M AgriLife Research http://swat.tamu.edu/
Evgenia Koltsida et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
636 | 298 | 23 | 957 | 9 | 9 |
- HTML: 636
- PDF: 298
- XML: 23
- Total: 957
- BibTeX: 9
- EndNote: 9
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1