the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Oct 2025
Evaluation of evapotranspiration partitioning models in the Amazon forest
Abstract. Although models that simulate actual ET have been widely used globally, their performance in tropical forests is unsatisfactory. The distribution of ET components is one of the key reasons. In this study, we evaluated the ability of three ET models (Forest-CEW, PML-V2, and PT-JPL) in a complex forest by analyzing their components. The data comes from seven ground-based eddy covariance flux towers in Brazil, which are part of the "Large Scale Biosphere- Atmosphere Experiment in Amazonia" (LBA) project. Our study found that the R2 of Forest-CEW was 0.64, that of PT-JPL was 0.43, and that of PML-V2 was only 0.29. The average results of the model show that T/ET=63.2 %±16 %, Ei/ET=32.3 %±16 %, and Es/ET=6 %±5 %. The model simulates better results in Savanna (RMSE=10.4 mm/month) than in the rainforest (RMSE=17.6 mm/month). Rn is the main driving variable of the model ET and T, with a sensitivity of 20 %, temperature is the main driver of Ei, accounting for 17 %, and LAI is the main driver of Es, but it produces a negative effect (-22.5 %). Our analysis emphasizes the differences in the ability of existing models to simulate ET dynamics in complex forests. Improving the formulation of ET components, particularly the canopy interception part, holds significant potential for substantially enhancing the accuracy and reliability of these ET models.
- Preprint
(4109 KB) - Metadata XML
-
Supplement
(1458 KB) - BibTeX
- EndNote
Status: open (until 29 Dec 2025)
-
RC1: 'Comment on egusphere-2025-4325', Anonymous Referee #1, 02 Dec 2025
reply
-
AC1: 'Reply on RC1', Longxiao Luo, 09 Dec 2025
reply
Line 48: Replace “high” by “higher”.
RP1: done
Line 85: The reported LAI value of 6 m²/m² seems inconsistent with Figure 1, where maximum values are around 3. Verify and clarify.
RP2: The sources of these two LAIs are completely different. The LAI mentioned in the text refers to Restrepo-Coupe et al, but they do not explain how it was measured or the size of the area it represents. The LAI dynamics in Figure 1 come from MODIS data, which has a resolution of 500 meters (The data is averaged over several grids near the site).
The LAI values of Restrepo-Coupe et al might represent a specific location, whereas satellite data tend to reflect an average over a range. The reference to R's LAI in the text is meant to indicate that the LAI at the PDG site may exhibit significant spatial heterogeneity. We have revised the manuscript and highlighted the differences between the two LAIs (Line 86).
Line 92: Clarify how monthly averages in Figure 1 were computed. Are they based on a single year or multiple years? How were daily variations in temperature and VPD handled?
RP3: The original flux data is hourly, and we did not apply any special processing when averaging from hourly to daily scales. The data in this study represents multi-year averages. The time span varies between different sites, but generally ranges from 2000 to 2006.
We have added an explanation in the manuscript (Line 79).
Line 108–111: The reduction in Rn during the dry season may be related to the site’s (negative) latitude, which increases the solar-incidence angle. Consider revising the explanation accordingly.
RP4: Thank you very much for your suggestion. This explanation provides important guidance for understanding the unique climatic characteristics of the area, and we have already added it to the manuscript (Line 111).
Line 130: Specify which models have been “well applied” and clarify what is meant by “well applied”.
RP5: Thank you very much for your careful review. We apologize for the wording error here. What we meant to convey was not that the model “well applied”, but that it has already been evaluated. In other studies like Melo et al, some ET models, such as the PT-JPL, have already been evaluated. We have rephrased this sentence in the manuscript (Line 133).
Line 151: Define qs and qp in Equation (2).
RP5: We are sorry for missing this issues. qs and qp respectively represent the saturated humidity and humidity. Already corrected.
Line 152: Units of gv and gs appear inconsistent; they should not be directly added if units differ. Please verify.
RP7: Thank you very much for pointing out the issue. This was a typo; the two units are the same, both describing conduction rate, and it has now been corrected.
Line 167: The text states that equations for Ei and Wwet are presented, but only the Ei equation appears. Include the missing equation or adjust the text.
RP8: Thank you very much for your reminder. We mistakenly described these two variables. Although and are the same in a physical sense, both describing the amount of water on the leaf surface, there is a difference in their units. We have added the relationship between the two in the manuscript (Line 174).
Line 168: Clarify the time step of model simulations (hourly, daily, monthly).
RP9: Thank you for your suggestion. This study is simulated on a daily scale, and we explain the simulation time step in the Methods section (Line 147).
Line 208: The Gash model is referenced but not presented. Provide details or remove the reference.
RP10: Thank you very much for your review. During the previous revisions, we moved the gash model to the supplementary, and we highlighted the detailed citation information in the manuscript (Line 212).
Line 214: Define G in Equation (15).
RP11: Thank you very much for your careful review. G is the soil heat flux and already corrected in the manuscript.
Line 242: The fact that K34 data are used for calibration should be mentioned earlier in the Methods.
RP12: Thank you for your suggestion. We have already addressed this situation at the beginning of the methods section (Line 146).
Line 245: The models appear to overestimate ET rather than underestimate it. Re-evaluate this statement.
RP13: Thank you for your thorough review. This is a typo; all models overestimated ET.
Line 251: In Figure 3, are the plotted points hourly ET values? Clarify in the caption and text.
RP14: The black marks in Figure 3 represent the daily ET results. Since the model simulates on a daily scale, we have provided an explanation in text below the figure (Line 255).
Line 256: The meaning of “easier” is unclear; provide a more precise explanation.
RP15: Thank you very much for your suggestion. We acknowledge that the wording here was unclear. What we actually meant to convey is that the simulation results for the wet season are better than those for the dry season. This has already been revised in the manuscript (Line 259).
Line 259: Replace “provids” with “provides”.
RP16: done
Line 264: Should be 500 m, not 500 km.
RP17: Thank you for your review, we have made corrections.
Line 305: In Figure 6, the trend lines do not seem to offer meaningful information; the slopes lack physical interpretation.
RP18: Due the driest period of the year at these sites occurs in mid-year (around July to August), we examine the consistency between the simulation results and the observations by dividing the year into two stages: from the wet season to the dry season, and from the dry season to the wet season. These two stages represent the increasing and decreasing severity of drought, providing a new point of reference.
In the figure, we added the slope of the trend line to quantify the consistency of ET and T changes at different stages. The more similar the slopes, the less seasonal impact the canopy interception has. We have provided additional explanations for this part in the new manuscript (Line 315-319).
Line 323–329: This discussion seems uninformative. Averaging across sites with fundamentally different ecohydrological dynamics (e.g., rainforests vs. savanna) may be misleading.
RP19: Thank you very much for your suggestions on this section. The purpose of showing the average trends of different sites was to provide a general view of seasonal variations. However, after carefully considering the appropriateness of this part of the analysis, we have decided to remove it. One reason is that averaging across locations with different hydrological dynamics could indeed mislead readers, and another reason is that this section is not closely related to the main content of the study.
Line 364–366: Is this statement supported by model results or observational data? Clarify the basis.
RP20: We sincerely appreciate you raising this important issue. You are correct; the original statement was overly assertive. We intended this sentence to reflect a mechanistic inference rather than a direct research conclusion. We have revised the original text according to your suggestion, emphasizing that this statement is based on inference rather than strong supporting evidence. We believe that the revised wording is more logically consistent and also more scientifically rigorous (Line 362-367).
‘In addition, during the dry season, the sparsity of the vegetation canopy affects the penetration of solar radiation to the ground (Gendron et al., 1998). Combined with the observed dry season microclimate conditions, such as higher temperatures, stronger wind speeds, and lower air humidity, these factors together promote increased soil surface evaporation. Therefore, we speculate that during this period, soil evaporation may account for a larger proportion of total evapotranspiration, although this specific distribution still needs further quantification.’
Line 409: Specify which of the three models this statement refers to.
RP21: Thank you for pointing out the issue. We have emphasized the models of these devices in the manuscript (Line 413).
Line 448–450: Without measurements of the ET components, it is difficult to state that the results are “incorrect”. A more appropriate term might be “inconsistent”.
RP22: Thank you very much for pointing out the inappropriate wording here. We have made corrections to this part, and it is now expressed more appropriately.
Line 463: Clarify which model this refers to. The limitation may apply to all models, as none account for topographical effects.
RP23: Thank you for your additional explanation here. These ET models do not take into account the existing topographical effects, so we have already emphasized this in the manuscript (Line 465).
Citation: https://doi.org/10.5194/egusphere-2025-4325-AC1
-
AC1: 'Reply on RC1', Longxiao Luo, 09 Dec 2025
reply
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 138 | 78 | 19 | 235 | 42 | 12 | 14 |
- HTML: 138
- PDF: 78
- XML: 19
- Total: 235
- Supplement: 42
- BibTeX: 12
- EndNote: 14
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript presents a model intercomparison study involving three approaches for estimating total evapotranspiration (ET) and partitioning it into transpiration (T), canopy-interception evaporation (Ei), and soil evaporation (Es) across seven sites located in and around the Amazon region. The models evaluated include a Penman–Monteith–based formulation (PML-V2), a Priestley–Taylor–based model (PT-JPL), and a new model developed by the authors (Forest-CEW). The study sites span diverse ecosystems—four tropical rainforests, one tropical wet-and-dry forest, one seasonally flooded forest–savanna ecotone, and one savanna—each equipped with eddy-covariance flux towers providing observed ET and key micrometeorological variables.
The topic is interesting and highly relevant, especially given the potentially large contribution of canopy-interception evaporation in dense tropical forests. Because Ei responds to meteorological drivers and canopy characteristics in ways very different from transpiration or soil evaporation, improving its representation remains an important challenge for land-surface and ecohydrological models.
Before evaluating model performance, the authors optimized key parameters for all three models using flux-tower data. Model skill was then assessed against observed ET across the seven sites. While all models captured general seasonal patterns, they performed poorly in reproducing ET magnitudes—particularly during the wet season, when Ei should be most prominent. This systematic bias suggests that interception processes are not adequately represented in any of the models. This is a major limitation of the study, aggravated by the absence of observational constraints for T, Ei, and Es, which restricts the ability to validate the partitioning schemes or identify which process assumptions are responsible for the errors.
Nonetheless, the study offers meaningful scientific insights. By contrasting structurally different models, the authors highlight persistent model biases, expose weaknesses in the treatment of interception losses, and point to model components that require improvement to better capture the ecohydrological dynamics of tropical forest ecosystems. With clearer exposition, careful treatment of uncertainties, and a more explicit discussion of limitations, this work could contribute valuably to the ongoing effort to improve ET partitioning in complex forest environments.
Specific comments (in order of appearance)
The manuscript requires a thorough language and clarity revision. Below are specific comments, organized by line number, to assist the authors in improving precision and readability.
Line 48: Replace “high” by “higher”.
Line 85: The reported LAI value of 6 m²/m² seems inconsistent with Figure 1, where maximum values are around 3. Verify and clarify.
Line 92: Clarify how monthly averages in Figure 1 were computed. Are they based on a single year or multiple years? How were daily variations in temperature and VPD handled?
Line 108–111: The reduction in Rn during the dry season may be related to the site’s (negative) latitude, which increases the solar-incidence angle. Consider revising the explanation accordingly.
Line 130: Specify which models have been “well applied” and clarify what is meant by “well applied”.
Line 151: Define qs and qp in Equation (2).
Line 152: Units of gv and gs appear inconsistent; they should not be directly added if units differ. Please verify.
Line 167: The text states that equations for Ei and Wwet are presented, but only the Ei equation appears. Include the missing equation or adjust the text.
Line 168: Clarify the time step of model simulations (hourly, daily, monthly).
Line 208: The Gash model is referenced but not presented. Provide details or remove the reference.
Line 214: Define G in Equation (15).
Line 242: The fact that K34 data are used for calibration should be mentioned earlier in the Methods.
Line 245: The models appear to overestimate ET rather than underestimate it. Re-evaluate this statement.
Line 251: In Figure 3, are the plotted points hourly ET values? Clarify in the caption and text.
Line 256: The meaning of “easier” is unclear; provide a more precise explanation.
Line 259: Replace “provids” with “provides”.
Line 264: Should be 500 m, not 500 km.
Line 305: In Figure 6, the trend lines do not seem to offer meaningful information; the slopes lack physical interpretation.
Line 323–329: This discussion seems uninformative. Averaging across sites with fundamentally different ecohydrological dynamics (e.g., rainforests vs. savanna) may be misleading.
Line 364–366: Is this statement supported by model results or observational data? Clarify the basis.
Line 409: Specify which of the three models this statement refers to.
Line 448–450: Without measurements of the ET components, it is difficult to state that the results are “incorrect”. A more appropriate term might be “inconsistent”.
Line 463: Clarify which model this refers to. The limitation may apply to all models, as none account for topographical effects.