the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 May 2023
Introducing a new floodplain scheme in ORCHIDEE (version 7885): validation and evaluation over the Pantanal wetlands
Anthony Schrapffer
Jan Polcher
Anna Sörensson
Lluís Fita
Abstract. Adapting and improving the hydrological processes in Land Surface Models is crucial given the increase of the resolution of the Climate Models to correctly represent the hydrological cycle. The present paper introduces a floodplains scheme adapted to the higher resolution river routing of the ORCHIDEE Land Surface Model. The scheme is based on a sub-tile parameterization of the hydrological units, Hydrological Transfer Unit concept (HTUs), based on high resolution hydrologically-coherent Digital Elevation Models which can be used for all types of resolutions and projections. The floodplain scheme was developed and evaluated for different atmospheric forcings and resolutions (0.5° and 25 km) over one of the world’s largest floodplains: the Pantanal, located in Central South America.
The floodplains scheme is validated based on the river discharge at the outflow of the Pantanal which represents the hydrological cycle over the basin, the temporal evolution of the water mass over the region assessed by the anomaly of Total Water Storage in Gravity Recovery And Climate Experiment (GRACE) and the temporal evaluation of the flooded areas compared to the Global Inundation Extent from Multi-Satellites dataset (GIEMS-2). The hydrological cycle is satisfactorily simulated, however, the base flow may be underestimated. The temporal evolution flooded area is coherent with the observations although the size of the is underestimated in comparison to GIEMS-2.
The presence of floodplains increases the soil moisture up to 50 % and decreases average temperature with 3 °C and with 6 °C during the dry season. The higher soil moisture increases the vegetation density and, along with the presence of open water surfaces due to the floodplains, it affects the surface energy budget by increasing the latent flux at the expense of the sensible flux. This is linked to the increase of the evapotranspiration related to the increased water availability. The effect of the floodplains scheme on the land surface conditions highlights that coupled simulations using the floodplains scheme may influence local and regional precipitation and regional circulation.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(7634 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(7634 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Anthony Schrapffer et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-549', Anonymous Referee #1, 12 Jun 2023
This manuscript presents a new scheme for floodplains, adapted to a high spatial resolution river routing in Orchidee. The mechanism is described, and tests are performed, using two atmospheric forcing over the Pantanal wetland, between 1990 and 2013. The scheme is evaluated with river discharge in situ measurements, as well as with GRACE data and satellite-derived surface water extent. The impact of the new scheme is tested, on the soil moisture, on the surface temperature, and on the vegetation density, and on the evapotranspiration. Before being publishable, the paper has to undergo a major revision.
Major comments:
1) How sensitive is the scheme to the dataset (here GLWD) used as a maximum mask for the inundation? A test should be performed to assess its effect, as this dataset is certainly valuable, but not perfect. There is a comment about the use of GLWD at lines 334 and following, but it is not said how the relevance of the dataset is tested (and possibly modified).
2) Figure 2 shows an evaluation of the mean annual cycle for the discharge and the models. It would be interesting to test the inter-annual variations (directly plotting the long time series or better by calculating some de-seasonalized anomalies). Is the model able to capture these changes from a year to the next? Same question for the water masses. Is the model able to capture the inter-annual variations observed by Grace?
3) Between the two forcing datasets, the differences in term of water masses are particularly striking (Figure 3), and as large as the difference between the cases with and without floodplains for the WFDEI case (see for soil moisture or for the slow reservoir for instance). That casts some doubts on the validity of the model / forcing combination. Can you comment?
4) Comparisons of the surface water extent are presented for different satellite-derived surface water. We need a few sentences for each dataset, to know how they have been derived and assess their possible limitation. Otherwise, there is no interest to compare to multiple products. For instance, the sensitivity of the different products to open water / vegetated water should be discussed.
5) Some mechanisms are mentioned that cannot be considered by this river flooding scheme (l. 550). Add a paragraph in the model description to mention them (section 2)?
6) For the soil moisture estimates, would it be possible to add some SMOS or SMAP retrieval? For the vegetation, any tests with NDVI or other proxy for the vegetation, in terms of seasonality and inter-annuality?
7) Applying the scheme to another region and evaluating it would certainly strengthen the paper. It is rather frustrating to have global model and datasets only applied to one specific case. At least another basin that is in the same type of environment (the Orinoco?) and for one common forcing?
Minor comments:
High spatial resolution river routing is mentioned at many occasions, but the reviewer could not find the information about that spatial resolution. That has to be clearly mentioned right away in the paper.
l.61: ‘such as such as’
l.196: the notations are confusing. Clarify.
l.208: ‘thRough’
l.327: ‘the routine graphS’
l.339: it would help to have a map of the area, with the river, its tributaries, and the location of the reference station.
l.440: ‘Depending on the period simulated, the SIMULATED flooded area simulated was…’
Table 2: indicate the meaning of the *. It is done in Table 3, but not here.
l.564-565: Surfaces of point 2) are not seen by GIEMS-2. Are they seen by the mNDWI estimates?
Figure 5: Add some comparisons with the other satellite-derived estimates. Especially the one the authors are themselves deriving.
l.631: ‘relativeS’
l.798: ’assess flooded area…principally in areas covered by floods’????
l.806: All the satellite-products do not only consider the open-water surfaces. In this work, the model is expected to be evaluated for wetlands. Most wetlands are vegetated surface water. If the satellite-products you use are only sensitive to open-water, it seems that the paper is missing its goal. Clarify.
Citation: https://doi.org/10.5194/egusphere-2023-549-RC1 - AC1: 'Reply on RC1', Anthony Schrapffer, 07 Aug 2023
-
RC2: 'Comment on egusphere-2023-549', Dai Yamazaki, 14 Jun 2023
General Comments
This manuscript describes the new floodplain scheme implemented in ORCHDEE model, evaluate the validity of the new scheme, and analyzed its impact on other land surface variables. Even though it’s still a case study simulation over Pantanal, I feel the paper very carefully analyzed how floodplain is important for land surface modelling.
The modeling strategy seems to be a bit complicated, while I feel the complexity is necessary given that the floodplain inundation itself is a complex physical process. I suggest the authors to provide more kind explanations about floodplain parameterization scheme, for example by using schematic figures, to help readers to understand how the proposed floodplain scheme works. However, the manuscript is overall well written, while minor revision is needed before acceptance.
Major concerns
[1] I feel the manuscript is too long. It might be unavoidable as a model description paper, but readability might increase if not-so-important parts are moved to supplements.
[2] So many variables/symbols are used to parameterize proposed floodplain scheme, and I feel difficulty following the explanations and equations. I suggest to create one schematic figure which represents the parameterization concept of floodplain scheme (with explicit description of which symbols corresponds to which variables). Visual explanation must help readers to understand about the new floodplain scheme.
Specific comments:
L193: whether the floodplains are activated or not.
This should be “regardless of whether …”
In addition, please explain what slow and fast reservoir represents. It is explained in results section that they represent aquifer and shallow groundwater, but this should be stated here. Otherwise, readers cannot know why they have limited relationship to floodplain scheme.
L235: The floodplains scheme allows a specific HTU to "overflow" the content of its floodplains reservoir into connected upstream HTUs with floodplains.
This is very interesting scheme. I wonder what is the impact of this overflow scheme on simulated water and energy budget. If space allows, please include some analysis.
L284 2.4.1: Case S_f,I < S_fmax,i
I recommend you to explain the case in plain language in the section title, not by the equation.
L285: height of the floodplain
This term is ambiguous. Do you mean “water surface elevation of the floodplain”?
L331: in order to define a mask of potentially flooded areas based on the following categories:
Could you please explain in which case this floodplain mask is required, and what is the impact of using this floodplain mask?
L355: before using the scheme over another region to evaluate if this parameterization is the more appropriate.
In many part of the world, there is no observation data for calibration. If possible, it’s better to perform some sensitivity tests of parameters (confirm results are not so sensitive to parameters, or specify which parameter has larger impact).
L360: Methodology of Validation and Analysis
Please also provide some description of the simulation domain. Probably, a figure showing the simulation domain (with location of the gauges) is better to be provided.
L419: forced with ERA5 re-analysis data.
I assume this is regional atmospheric simulation, and in that case ERA5 must be “boundary condition” rather than “forcing”.
Figure 2:
Could you please analyze the mechanics of river discharge delay? E.g. where water stays before reaching to the river gauge? Did they stay in floodplain as surface water? Or did they stay in soil by infiltration? Given that the difference between FP and NOFP simulation is large, it’s better to provide detailed analysis on the mechanism which cause the difference.
L507: soil moisture and in the stream reservoir increases slightly
Considering the magnitude of change, compared to other storage variables, I feel the soil moisture was “significantly” increased by floodplain scheme (it’s not slight increase).
L508: This increase is even more important in the fast and slow reservoirs.
Please also reconsider this statement. The relative increase could be large, but absolute change is larger in soil moisture.
Figure 3:
I suggest it’s better to make some discussion on the water volume change and annual river discharge (by converting annual discharge to volume unit). How large the volumetric change in each reservoir is, compared to the annual discharge? This analysis must be essential to understand why discharge seasonality changed significantluy.
L551: divergent flows which very sensitive to the orography and cannot be represented in this model
Please explain why divergent flow cannot be represented. (i.e. because only one downstream is assumed for model’s river network).
L639: vegetation fraction decrease
I think vegetation fraction can decrease also due to water logging along floodplains (too much water). It seems this impact is not considered in the proposed model, so better to be mentioned as limitation.
L816: . The divergent processes are not represented in the Hydrological DEM and, therefore, are not implemented in ORCHIDEE.
Divergent flow is represented in MGB-IPH and CaMa-Flood by analyzing high-resolution topography data (Pontes et al. 2017; Yamazaki et al 2014). Given that representation is possible, I think it’s better to mention about the possibility.
L860: IMaps
What is IMaps? Please explain.
- AC2: 'Reply on RC2', Anthony Schrapffer, 07 Aug 2023
-
RC3: 'Comment on egusphere-2023-549', Anonymous Referee #3, 23 Jun 2023
This paper describes a new floodplain scheme developed within the framework of the land surface
modeling platform ORCHIDEE. The main applications of this new model development are intended to
be used at the regional-to-global scale in so-called “offline mode” (decoupled from a regional climate,
RCM, or global-scale earth system/climate model, GCM) or coupled to an atmospheric model, thus the
level of complexity, process representation and input data are adapted for such applications. As noted
by the authors, RCM and GCM spatial resolutions are constantly increasing, thus there is a need to
adapt the hydrological parameterizations in such models accordingly. Rather than using a classic grid
structure (as many GCMs currently use) dictated by the atmospheric model, the current scheme is
based on the Hydrological Transfer Unit (HTU) concept. The implementation of this scheme benefits
from numerous relatively high spatial resolution topographical and geomorphological off-the-shelf
datasets now available to hydrologists. This paper describes the methodology and mathematical
underpinnings of this new floodplain scheme and how it interacts with other components of
ORCHIDEE (such as evaporation, river flow, runoff, etc.). The scheme is next used to simulate the
floodplains along with the other main components of the surface hydrological cycle over a recent
multi-year period over the Pantanal basin in South America, which contains one of the world’s largest
floodplains thus making it a very pertinent case study. The model is evaluated at two spatial scales, one
representing the approximate scale still used by many GCMs (i.e. 0.5 o) and another representing a scale
comparable to RCMs and what more and more GCMs are (or plan to) move to in upcoming years (~25
km). As boundary conditions, so-called atmospheric forcing must be prescribed in offline mode but
there are many such products and there are considerable differences among them, especially at different
spatial scales as herein. The authors have addressed these uncertainties by using a very standard
analysis product as forcing at the more coarse resolution, along with a forcing which has been
developed specifically for this region at a higher spatial resolution. The model simulations, notably the
floodplain outputs, are evaluated using several standard satellite-based products along with in-situ
discharge measurements. Convincing statistical results are used to summarize the performance of the
model using the new parameterization for the two input forcing compared to the baseline model
(without the new floodplain scheme). A discussion of errors (in terms of the model input, output and
the evaluation data), limitations, and gained insights are presented. I find the organization of the paper
to be quite good, it is well written: the overall presentation is clear, the results are presented in a very
pragmatic manner and future perspectives are discussed. I recommend publication after only some
minor revisions as this paper is an important contribution to the rapidly developing region-to-global
scale hydrological modeling field, notably improved terrestrial water cycle simulations in RCMs and
GCMs.
General Comments:
1. Lines 339-358: In my opinion, the only part of this paper which needs some improvement is
Calibration of the Parameters. There are no graphics (for example, showing the discharge
performance at the calibration station) and only limited statistics (Table 1.). Lines 347-349
mention that The best combination of parameters has been established through a grid search
method which consists in evaluating the different combinations of parameters within theirrespective interval of definition. I find this a bit vague and it seems to gloss over a very
important part of any new parameterization: parameter calibration/estimation/determination. I
feel the authors should just give a slightly more detailed description of how exactly the
parameters were calibrated. There is some limited information, but more details would be
appreciated. Also, plots of discharge before and after calibration would be informative. Also,
1991-1996 was the calibration period: why these 6 years? Is the natural variability adequately
represented over these years? And so on. Again, just a few more details on the methods and
results. Parameter sensitivity analysis is a critical part of any new model development and a bit
more information would be very informative to readers.
2. The quality of the English is good, however there are a certain number of very small errors,
notably the use or lack thereof of “a” or “s” at the end of some words, e.g. Line 48: a South
American tropical floodplains. There are just a few small errors like this on nearly every page,
so they do not detract from the reading or result in a lack of understanding. But I’d recommend
a quick filtering to catch them.
More specific:
Line 121: I suggest changing ruling to governing
Line 130: Referring to the text: HTU only flows into a single HTU and is acyclic as water cannot
return to the original HTU: I assume that backwater effects can be neglected at the spatial resolutions
you are modeling here?
Eq.2 for evaporation from the floodplains: water surfaces have very low roughness lengths compared to
land surfaces: typically Charnock-type parameterizations are used for water bodies. I assume that
floodplains are generally fairly smooth...should this effect (or is it?) somehow incorporated into this
computation? I suspect that using such a roughness length could reduce the evaporation from
floodplains (?).
Line 172: I am surprised that soil water infiltration can be larger outside of floodplains than within
them. Can the authors present some sort of physical arguments or an observational basis for this
assumption?
Line 188: Referring to the text: The time constant of the floodplains (τf) is slower than the stream
reservoir time constant (τstream) and faster than the fast reservoir time constant. Can the authors give
some sort of physical argument or explanation for this (frictional effects of flooded riparian vegetation
and non-riparian vegetation in flooded zones for example? Or some other reason? Or just a reference
justifying this choice?)
Eq.5: It is not quite clear to me why when Sfmax,i > 0 there is no contribution from the upstream stream
reservoir to the local stream flow (it is just from the upstream floodplain...)...I am missing something
here.
Eq.8: It seems that a term is missing on the RHS...the possible addition of overflow from the
downstream reservoir?
Lines 271-273 should probably be placed after Eq.13 since “beta” doesn’t seem to be mentioned until
Eq.13.Lines 312-313: Referring to the text: The different values of standard deviation are bounded by
lowlim_std = 0.05m and uplim_std = 20m . Why these particular values? Is the model very sensitive to
this range?
Line 465: It seems that there are only roughly 1 to 2 GRACE pixels covering your zone, likely not with
a perfect overlap. Is this really sufficient? Can you say a bit more about the errors involved in this
comparison to justify this for readers?
Line 475: Maybe I missed it, but I assume the statistics were made using monthly model outputs and
observations?
Fig.3 showing the multi-year monthly averages as a single annual cycle is indeed an informative way to
convey the quality of the climatological performance of the scheme. But aside from the statistics, it
would be good to see some graphical information on the year-to-year variability per month in the main
paper: some sort of spread (standard deviation or quantiles, etc.) on these plots would be most
informative. Indeed we wish to see the climatological (average annual cycle), but it is of course the
improvement or degradation in terms of model vs observed variability that is also of interest.
Line 522: Referring to the text: AmSud_FP seems to have more runoff. This sounds a bit speculative
and it seems that it would be easy to verify by comparing the modeled runoff with and without
floodplains? The authors could just include some numerical values here within the text for example.
End of Section 4.2: After reading this section, I am left wondering whether it possible to give a number
or show a figure of the contribution of Eflood to the total E? The total E with floodplains will almost
certainly increase (when using the same prescribed forcing) over soils which have been wetted once
floodwaters retreat, so increases in E will be at least related to this, as discussed in the paper. But Eflood
seems to be rather uncertain/difficult to model and observe, I wonder how much Eflood is contributing to
the overall E increase. I say this because I wonder if a more surface-water adapted approach for E
might be in order, especially if this flux is significant compared to the other E components.
Lines 615-620: If I understand correctly, rainfall can lead to greater soil water infiltration that when the
same grid element is flooded. This seems a bit counter-intuitive, to me anyway. Are there observational
studies which can be referenced etc. to justify this?
Lines 670-674: What about the increase in net radiation over the flood waters? The typical albedo for
water surfaces is generally around 0.07, far lower than vegetation or soil. Is this considered?
Line 676-677: typo, a phrase is repeated → depending on vegetation type and on soil types (Clay,
Sand, Silt). depending on vegetation and soil types (clay, sand, silt)
Lines 730-731: Can the authors just provide a phrase describing the specific sub-surface component
and how this could help solve the mentioned issues?Citation: https://doi.org/10.5194/egusphere-2023-549-RC3 - AC3: 'Reply on RC3', Anthony Schrapffer, 07 Aug 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-549', Anonymous Referee #1, 12 Jun 2023
This manuscript presents a new scheme for floodplains, adapted to a high spatial resolution river routing in Orchidee. The mechanism is described, and tests are performed, using two atmospheric forcing over the Pantanal wetland, between 1990 and 2013. The scheme is evaluated with river discharge in situ measurements, as well as with GRACE data and satellite-derived surface water extent. The impact of the new scheme is tested, on the soil moisture, on the surface temperature, and on the vegetation density, and on the evapotranspiration. Before being publishable, the paper has to undergo a major revision.
Major comments:
1) How sensitive is the scheme to the dataset (here GLWD) used as a maximum mask for the inundation? A test should be performed to assess its effect, as this dataset is certainly valuable, but not perfect. There is a comment about the use of GLWD at lines 334 and following, but it is not said how the relevance of the dataset is tested (and possibly modified).
2) Figure 2 shows an evaluation of the mean annual cycle for the discharge and the models. It would be interesting to test the inter-annual variations (directly plotting the long time series or better by calculating some de-seasonalized anomalies). Is the model able to capture these changes from a year to the next? Same question for the water masses. Is the model able to capture the inter-annual variations observed by Grace?
3) Between the two forcing datasets, the differences in term of water masses are particularly striking (Figure 3), and as large as the difference between the cases with and without floodplains for the WFDEI case (see for soil moisture or for the slow reservoir for instance). That casts some doubts on the validity of the model / forcing combination. Can you comment?
4) Comparisons of the surface water extent are presented for different satellite-derived surface water. We need a few sentences for each dataset, to know how they have been derived and assess their possible limitation. Otherwise, there is no interest to compare to multiple products. For instance, the sensitivity of the different products to open water / vegetated water should be discussed.
5) Some mechanisms are mentioned that cannot be considered by this river flooding scheme (l. 550). Add a paragraph in the model description to mention them (section 2)?
6) For the soil moisture estimates, would it be possible to add some SMOS or SMAP retrieval? For the vegetation, any tests with NDVI or other proxy for the vegetation, in terms of seasonality and inter-annuality?
7) Applying the scheme to another region and evaluating it would certainly strengthen the paper. It is rather frustrating to have global model and datasets only applied to one specific case. At least another basin that is in the same type of environment (the Orinoco?) and for one common forcing?
Minor comments:
High spatial resolution river routing is mentioned at many occasions, but the reviewer could not find the information about that spatial resolution. That has to be clearly mentioned right away in the paper.
l.61: ‘such as such as’
l.196: the notations are confusing. Clarify.
l.208: ‘thRough’
l.327: ‘the routine graphS’
l.339: it would help to have a map of the area, with the river, its tributaries, and the location of the reference station.
l.440: ‘Depending on the period simulated, the SIMULATED flooded area simulated was…’
Table 2: indicate the meaning of the *. It is done in Table 3, but not here.
l.564-565: Surfaces of point 2) are not seen by GIEMS-2. Are they seen by the mNDWI estimates?
Figure 5: Add some comparisons with the other satellite-derived estimates. Especially the one the authors are themselves deriving.
l.631: ‘relativeS’
l.798: ’assess flooded area…principally in areas covered by floods’????
l.806: All the satellite-products do not only consider the open-water surfaces. In this work, the model is expected to be evaluated for wetlands. Most wetlands are vegetated surface water. If the satellite-products you use are only sensitive to open-water, it seems that the paper is missing its goal. Clarify.
Citation: https://doi.org/10.5194/egusphere-2023-549-RC1 - AC1: 'Reply on RC1', Anthony Schrapffer, 07 Aug 2023
-
RC2: 'Comment on egusphere-2023-549', Dai Yamazaki, 14 Jun 2023
General Comments
This manuscript describes the new floodplain scheme implemented in ORCHDEE model, evaluate the validity of the new scheme, and analyzed its impact on other land surface variables. Even though it’s still a case study simulation over Pantanal, I feel the paper very carefully analyzed how floodplain is important for land surface modelling.
The modeling strategy seems to be a bit complicated, while I feel the complexity is necessary given that the floodplain inundation itself is a complex physical process. I suggest the authors to provide more kind explanations about floodplain parameterization scheme, for example by using schematic figures, to help readers to understand how the proposed floodplain scheme works. However, the manuscript is overall well written, while minor revision is needed before acceptance.
Major concerns
[1] I feel the manuscript is too long. It might be unavoidable as a model description paper, but readability might increase if not-so-important parts are moved to supplements.
[2] So many variables/symbols are used to parameterize proposed floodplain scheme, and I feel difficulty following the explanations and equations. I suggest to create one schematic figure which represents the parameterization concept of floodplain scheme (with explicit description of which symbols corresponds to which variables). Visual explanation must help readers to understand about the new floodplain scheme.
Specific comments:
L193: whether the floodplains are activated or not.
This should be “regardless of whether …”
In addition, please explain what slow and fast reservoir represents. It is explained in results section that they represent aquifer and shallow groundwater, but this should be stated here. Otherwise, readers cannot know why they have limited relationship to floodplain scheme.
L235: The floodplains scheme allows a specific HTU to "overflow" the content of its floodplains reservoir into connected upstream HTUs with floodplains.
This is very interesting scheme. I wonder what is the impact of this overflow scheme on simulated water and energy budget. If space allows, please include some analysis.
L284 2.4.1: Case S_f,I < S_fmax,i
I recommend you to explain the case in plain language in the section title, not by the equation.
L285: height of the floodplain
This term is ambiguous. Do you mean “water surface elevation of the floodplain”?
L331: in order to define a mask of potentially flooded areas based on the following categories:
Could you please explain in which case this floodplain mask is required, and what is the impact of using this floodplain mask?
L355: before using the scheme over another region to evaluate if this parameterization is the more appropriate.
In many part of the world, there is no observation data for calibration. If possible, it’s better to perform some sensitivity tests of parameters (confirm results are not so sensitive to parameters, or specify which parameter has larger impact).
L360: Methodology of Validation and Analysis
Please also provide some description of the simulation domain. Probably, a figure showing the simulation domain (with location of the gauges) is better to be provided.
L419: forced with ERA5 re-analysis data.
I assume this is regional atmospheric simulation, and in that case ERA5 must be “boundary condition” rather than “forcing”.
Figure 2:
Could you please analyze the mechanics of river discharge delay? E.g. where water stays before reaching to the river gauge? Did they stay in floodplain as surface water? Or did they stay in soil by infiltration? Given that the difference between FP and NOFP simulation is large, it’s better to provide detailed analysis on the mechanism which cause the difference.
L507: soil moisture and in the stream reservoir increases slightly
Considering the magnitude of change, compared to other storage variables, I feel the soil moisture was “significantly” increased by floodplain scheme (it’s not slight increase).
L508: This increase is even more important in the fast and slow reservoirs.
Please also reconsider this statement. The relative increase could be large, but absolute change is larger in soil moisture.
Figure 3:
I suggest it’s better to make some discussion on the water volume change and annual river discharge (by converting annual discharge to volume unit). How large the volumetric change in each reservoir is, compared to the annual discharge? This analysis must be essential to understand why discharge seasonality changed significantluy.
L551: divergent flows which very sensitive to the orography and cannot be represented in this model
Please explain why divergent flow cannot be represented. (i.e. because only one downstream is assumed for model’s river network).
L639: vegetation fraction decrease
I think vegetation fraction can decrease also due to water logging along floodplains (too much water). It seems this impact is not considered in the proposed model, so better to be mentioned as limitation.
L816: . The divergent processes are not represented in the Hydrological DEM and, therefore, are not implemented in ORCHIDEE.
Divergent flow is represented in MGB-IPH and CaMa-Flood by analyzing high-resolution topography data (Pontes et al. 2017; Yamazaki et al 2014). Given that representation is possible, I think it’s better to mention about the possibility.
L860: IMaps
What is IMaps? Please explain.
- AC2: 'Reply on RC2', Anthony Schrapffer, 07 Aug 2023
-
RC3: 'Comment on egusphere-2023-549', Anonymous Referee #3, 23 Jun 2023
This paper describes a new floodplain scheme developed within the framework of the land surface
modeling platform ORCHIDEE. The main applications of this new model development are intended to
be used at the regional-to-global scale in so-called “offline mode” (decoupled from a regional climate,
RCM, or global-scale earth system/climate model, GCM) or coupled to an atmospheric model, thus the
level of complexity, process representation and input data are adapted for such applications. As noted
by the authors, RCM and GCM spatial resolutions are constantly increasing, thus there is a need to
adapt the hydrological parameterizations in such models accordingly. Rather than using a classic grid
structure (as many GCMs currently use) dictated by the atmospheric model, the current scheme is
based on the Hydrological Transfer Unit (HTU) concept. The implementation of this scheme benefits
from numerous relatively high spatial resolution topographical and geomorphological off-the-shelf
datasets now available to hydrologists. This paper describes the methodology and mathematical
underpinnings of this new floodplain scheme and how it interacts with other components of
ORCHIDEE (such as evaporation, river flow, runoff, etc.). The scheme is next used to simulate the
floodplains along with the other main components of the surface hydrological cycle over a recent
multi-year period over the Pantanal basin in South America, which contains one of the world’s largest
floodplains thus making it a very pertinent case study. The model is evaluated at two spatial scales, one
representing the approximate scale still used by many GCMs (i.e. 0.5 o) and another representing a scale
comparable to RCMs and what more and more GCMs are (or plan to) move to in upcoming years (~25
km). As boundary conditions, so-called atmospheric forcing must be prescribed in offline mode but
there are many such products and there are considerable differences among them, especially at different
spatial scales as herein. The authors have addressed these uncertainties by using a very standard
analysis product as forcing at the more coarse resolution, along with a forcing which has been
developed specifically for this region at a higher spatial resolution. The model simulations, notably the
floodplain outputs, are evaluated using several standard satellite-based products along with in-situ
discharge measurements. Convincing statistical results are used to summarize the performance of the
model using the new parameterization for the two input forcing compared to the baseline model
(without the new floodplain scheme). A discussion of errors (in terms of the model input, output and
the evaluation data), limitations, and gained insights are presented. I find the organization of the paper
to be quite good, it is well written: the overall presentation is clear, the results are presented in a very
pragmatic manner and future perspectives are discussed. I recommend publication after only some
minor revisions as this paper is an important contribution to the rapidly developing region-to-global
scale hydrological modeling field, notably improved terrestrial water cycle simulations in RCMs and
GCMs.
General Comments:
1. Lines 339-358: In my opinion, the only part of this paper which needs some improvement is
Calibration of the Parameters. There are no graphics (for example, showing the discharge
performance at the calibration station) and only limited statistics (Table 1.). Lines 347-349
mention that The best combination of parameters has been established through a grid search
method which consists in evaluating the different combinations of parameters within theirrespective interval of definition. I find this a bit vague and it seems to gloss over a very
important part of any new parameterization: parameter calibration/estimation/determination. I
feel the authors should just give a slightly more detailed description of how exactly the
parameters were calibrated. There is some limited information, but more details would be
appreciated. Also, plots of discharge before and after calibration would be informative. Also,
1991-1996 was the calibration period: why these 6 years? Is the natural variability adequately
represented over these years? And so on. Again, just a few more details on the methods and
results. Parameter sensitivity analysis is a critical part of any new model development and a bit
more information would be very informative to readers.
2. The quality of the English is good, however there are a certain number of very small errors,
notably the use or lack thereof of “a” or “s” at the end of some words, e.g. Line 48: a South
American tropical floodplains. There are just a few small errors like this on nearly every page,
so they do not detract from the reading or result in a lack of understanding. But I’d recommend
a quick filtering to catch them.
More specific:
Line 121: I suggest changing ruling to governing
Line 130: Referring to the text: HTU only flows into a single HTU and is acyclic as water cannot
return to the original HTU: I assume that backwater effects can be neglected at the spatial resolutions
you are modeling here?
Eq.2 for evaporation from the floodplains: water surfaces have very low roughness lengths compared to
land surfaces: typically Charnock-type parameterizations are used for water bodies. I assume that
floodplains are generally fairly smooth...should this effect (or is it?) somehow incorporated into this
computation? I suspect that using such a roughness length could reduce the evaporation from
floodplains (?).
Line 172: I am surprised that soil water infiltration can be larger outside of floodplains than within
them. Can the authors present some sort of physical arguments or an observational basis for this
assumption?
Line 188: Referring to the text: The time constant of the floodplains (τf) is slower than the stream
reservoir time constant (τstream) and faster than the fast reservoir time constant. Can the authors give
some sort of physical argument or explanation for this (frictional effects of flooded riparian vegetation
and non-riparian vegetation in flooded zones for example? Or some other reason? Or just a reference
justifying this choice?)
Eq.5: It is not quite clear to me why when Sfmax,i > 0 there is no contribution from the upstream stream
reservoir to the local stream flow (it is just from the upstream floodplain...)...I am missing something
here.
Eq.8: It seems that a term is missing on the RHS...the possible addition of overflow from the
downstream reservoir?
Lines 271-273 should probably be placed after Eq.13 since “beta” doesn’t seem to be mentioned until
Eq.13.Lines 312-313: Referring to the text: The different values of standard deviation are bounded by
lowlim_std = 0.05m and uplim_std = 20m . Why these particular values? Is the model very sensitive to
this range?
Line 465: It seems that there are only roughly 1 to 2 GRACE pixels covering your zone, likely not with
a perfect overlap. Is this really sufficient? Can you say a bit more about the errors involved in this
comparison to justify this for readers?
Line 475: Maybe I missed it, but I assume the statistics were made using monthly model outputs and
observations?
Fig.3 showing the multi-year monthly averages as a single annual cycle is indeed an informative way to
convey the quality of the climatological performance of the scheme. But aside from the statistics, it
would be good to see some graphical information on the year-to-year variability per month in the main
paper: some sort of spread (standard deviation or quantiles, etc.) on these plots would be most
informative. Indeed we wish to see the climatological (average annual cycle), but it is of course the
improvement or degradation in terms of model vs observed variability that is also of interest.
Line 522: Referring to the text: AmSud_FP seems to have more runoff. This sounds a bit speculative
and it seems that it would be easy to verify by comparing the modeled runoff with and without
floodplains? The authors could just include some numerical values here within the text for example.
End of Section 4.2: After reading this section, I am left wondering whether it possible to give a number
or show a figure of the contribution of Eflood to the total E? The total E with floodplains will almost
certainly increase (when using the same prescribed forcing) over soils which have been wetted once
floodwaters retreat, so increases in E will be at least related to this, as discussed in the paper. But Eflood
seems to be rather uncertain/difficult to model and observe, I wonder how much Eflood is contributing to
the overall E increase. I say this because I wonder if a more surface-water adapted approach for E
might be in order, especially if this flux is significant compared to the other E components.
Lines 615-620: If I understand correctly, rainfall can lead to greater soil water infiltration that when the
same grid element is flooded. This seems a bit counter-intuitive, to me anyway. Are there observational
studies which can be referenced etc. to justify this?
Lines 670-674: What about the increase in net radiation over the flood waters? The typical albedo for
water surfaces is generally around 0.07, far lower than vegetation or soil. Is this considered?
Line 676-677: typo, a phrase is repeated → depending on vegetation type and on soil types (Clay,
Sand, Silt). depending on vegetation and soil types (clay, sand, silt)
Lines 730-731: Can the authors just provide a phrase describing the specific sub-surface component
and how this could help solve the mentioned issues?Citation: https://doi.org/10.5194/egusphere-2023-549-RC3 - AC3: 'Reply on RC3', Anthony Schrapffer, 07 Aug 2023
Peer review completion
Journal article(s) based on this preprint
Anthony Schrapffer et al.
Anthony Schrapffer et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 242 | 77 | 19 | 338 | 6 | 7 |
- HTML: 242
- PDF: 77
- XML: 19
- Total: 338
- BibTeX: 6
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(7634 KB) - Metadata XML