the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Apr 2022
Water partitioning in a Neotropical Savanna forest (Cerrado s.s.): interception responses at different time-scales using adapted versions of the Rutter and the Gash models
Abstract. Cerrado is the broadest Savanna ecosystem of South America and has an important role in our global climate. How rainfall finds it way through the vegetation layers of the undisturbed Cerrado forest is of utmost importance to understand the evaporation process and the water availability in this unique ecosytem. Nonetheless, only few studies consider the partitioning of rainfall in the Cerrado. And if they do, these studies are limited by only considering interception by the canopy, while the forest floor can intercept a significant amount as well. Additionally, the studies often apply canopy interception models that were calibrated on short term monitoring. Hence evaluating how interception models perform at different time-scales and how the interception process responds to seasonal changes is poorly understood for the Cerrado forest. In this study we aimed to evaluate the canopy and forest floor interception estimates at different time-scales and its seasonal response for an undisturbed Cerrado s.s. forest in Brazil. Two commonly used interception models (Rutter and Gash) were adapted to include forest floor interception using observations of both canopy and forest floor interception during a 32 months study period. Our results show that the models are suitable to estimate throughfall and infiltration at daily basis, but not the evaporative processes. We confirmed that both models had limitations to simulate very high interception rates on an event scale. Nonetheless, both models are able to reproduce the total interception well at monthly scale (R2 = 0.7–0.97, NSE = 0.63–0.85), and they can represent seasonal trends in the interception process in Cerrado s.s. forests. Nevertheless, the Rutter model seems to perform better when seasonal parameters are used than the Gash model, but both models are equally valuable to inter-annual analysis when non-seasonal parameters are used.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Hydrology and Earth System Sciences.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(1893 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on hess-2022-59', Anonymous Referee #1, 11 May 2022
This manuscript describes measurement and modeling of interception loss from the canopy and forest floor. As is typical in published applications of Rutter-Gash models, the models performed poorly at the scale of individual events and better when accumulated over longer timescales as presented in Table 5. The manuscript also presents some potentially important modifications to the standard Rutter-Gash approach by modifying the model of canopy storage, modifying the evaporation rate from trunks, and adding explicit measurement of evaporation from wet forest floor. The manuscript does not evaluate the behavior of the models in relation to the standard ones, so the importance of the modifications is difficult to understand. In addition, the manuscript omits some details about parameter estimation and presents some model modifications vaguely and in terms of programming techniques, so the methods are not possible to follow comprehensively. The discussion touches on some important questions pertaining to the applicability of the modeling assumptions, such as time scale of forest-floor evaporation and assumptions embedded in the Gash model, and mismatches between data needs and data availability, such as radiation budgets relevant for the forest floor. However, the manuscript mostly disregards these important points and the poor by-storm performance, and makes detailed interpretation of the effects of seasonal variation of parameters. Given that the models mostly cannot reproduce per-storm interception loss, and the time-integrated performance is driven by essentially the regression between rainfall and throughfall expressed in the calibrated canopy storage capacity, I do not agree with the conclusions that the modeling is useful for understanding canopy processes. These are not unique problems for this study and field experience is valuable for developing the next generation of models, but the modeling presented in this manuscript is not well described or developed.
This manuscript needs a thorough edit for English: there are many errors in word choice, grammar, and spelling.
L34 these are rainforest numbers that are not likely relevant to the field site here
L74 The logic of the theoretical work should be justified here. The objective based on the study site, L72, is not as important as the theoretical implications of behavior of the model, and there are modifications to the basic models that are not explained here at all. Similarly, the model introduction L55-64 should focus on the shortcomings of existing models that need to be overcome, not on basics.
L94 15,000 stems per ha? This is 1.5 trees per square meter; is that correct?
L98 What use is PAR for this study?
L114 why “also”?
L119 according to DBH, but by what criteria and why?
L120 this reference does not provide details of this device. A few critical details about how a LID works are needed for this part to be understandable
L123 But there was only one August during the study, correct?
L137 upper envelopes and linear regressions are not the same. Stem storage capacity has generally not been estimated by this technique in the past for several important reasons. This needs justification and explanation.
L150 These two ad-hoc modifications to the Rutter model are not sufficiently described or justified. There is no way to understand exactly what "priority" means in terms of the water and energy balances.
L152 What about the Tf records indicate dynamic storage?
L153 this term needs explicit definition. I do not know what Cc is supposed to represent
L156 What is the purpose of Ccmax exactly and how was it implemented? Rutter's model explicitly allows storage C to exceed storage capacity S, with no explicit upper limit. Adding a cap here appears to reduce the dynamic storage, not enable it. Why only "before the end of the storm"?
L159 what is Cmax. the same as Ccmax?
L160 I don't understand this parameter identification technique
L176 I do not understand how data were used to estimate Cfmax
L179 Most of the parameters referred to in this sentence are not defined
L185 I can't find where these parameters were estimated
L199 something is wrong. This describes an experiment 8 months long but fig 3 says 3 years
L204 The original Gash model was applied at the monthly time scale. There is no time scale assumed by Gash
L214 This conflicts slightly with L145. I assume this means stomatal resistance was zero. There are no details about assumptions and about parameters relating to aerodynamic resistance
L220 Fig 3: L218 told us to expect net radiation here
L224 more similar than what?
L242 Why does the Gash model predict essentially the same canopy evaporation every storm except for very small storms?
L264 I think it is unlikely that measurements of forest-floor evaporation were better than the throughfall measurements.
L265 It is not obvious to me that canopy evaporation is modeled better.
L280 As described earlier, it did exactly the opposite: it prevented accumulation
L284 How? Shouldn't mean daily Rutter be the same as daily Gash?
L344 This problem is severe for the modeling assumptions and needs detailed theoretical discussion
L348 I do not understand this sentence
L353 I agree. I think this makes the Gash model irrelevant for this process. Lack of data to estimate energy budgets and vapor exchange at the scale of the forest floor are also major problems for applying a Rutter model. Given that Rutter does not seem to be helpful for estimating canopy evaporation on a storm scale, it seems even less likely to be helpful for forest-floor evaporation.
L397 I don’t think the evaporation model was useful for small storms, either. There was not systematic negative bias in estimates, so that means the small storms were overestimated.
L425 I would agree if the models were useful at short timescales, but they are not. If the processes are not correctly modeled, how can we extrapolate using the model?
L510 I do not agree. The models get major masses correct, but due to calibration (e.g, of canopy storage). The models are not likely to be of any predictive value should, for example, vegetation structure change.
The journal names are missing for many References.
Citation: https://doi.org/10.5194/hess-2022-59-RC1 -
AC1: 'Reply on RC1', Livia Rosalem, 15 Nov 2022
Author’s Response to Referee #1
We would like to thank the anonymous Referee #1 for the detailed review and for the time spent reviewing our text. We replied to the referee’s comments, which are highly useful to improve the quality of our manuscript. Note that the original referee’s comments are identified as R1C and written in bold, and the authors’ responses are labeled as AR. In addition, all comments are numbered (e.g., R1C-01).
This manuscript describes measurements and modeling of interception loss from the canopy and forest floor. As is typical in published applications of Rutter-Gash models, the models performed poorly at the scale of individual events and better when accumulated over longer timescales as presented in Table 5. The manuscript also presents some potentially important modifications to the standard Rutter-Gash approach by modifying the model of canopy storage, modifying the evaporation rate from trunks, and adding explicit measurements of evaporation from the wet forest floor.
The manuscript does not evaluate the behavior of the models in relation to the standard ones, so the importance of the modifications is difficult to understand. In addition, the manuscript omits some details about parameter estimation and presents some model modifications vaguely and in terms of programming techniques, so the methods are not possible to follow comprehensively.
R1C-01: We presented modified versions of the Rutter and Gash models by adding the forest floor interception process. The canopy and trunk processes were maintained similar to the original versions; the only difference was the addition of a dynamic storage coefficient (Ccmax) to the canopy storage. This coefficient does not change the storage and evaporative processes modelling. However, it allows more water being hold on the canopy during the rainfall, which reduces the drainage picks and enables to increase the evaporation.
The results showed that Rutter’s model canopy evaporation continued underestimated despite adding the Ccmax coefficient. For the Gash model, the results should be the same for canopy and trunk evaporation since the dynamic storage was not included. Thus, the main differences observed should be only on the total interception, which for the original versions should be always lower since they do not account for the forest floor interception.
The parameter estimation was following the same approach as for the original models (explanation started on line 134). Intending not to be unduly long, the well-known approaches to determine the models’ parameters were shortly explained and referenced. We developed a code in Python which we can add to the Appendix.
R1C-02: The discussion touches on some important questions pertaining to the applicability of the modeling assumptions, such as time scale of forest-floor evaporation and assumptions embedded in the Gash model, and mismatches between data needs and data availability, such as radiation budgets relevant for the forest floor. However, the manuscript mostly disregards these important points and the poor by-storm performance and makes a detailed interpretation of the effects of seasonal variation of parameters.
AR: As it was mentioned, we addressed the critical issues related to the modelling assumptions and the need of some important information as net radiation at the forest floor level to improve modelling. The poor daily event performance for the interception processes was highlighted in the discussion of each interception process result (line 271 to canopy; line 298 to trunk; line 339 to forest floor) besides being present in the abstract and conclusions too.
Despite the poorly by-storm performance, as the forest in the study has a pronounced canopy leaf off during the dry season the models should be able to at least indicate the lower canopy interception observed even if it was in a biweekly or monthly time scale. It is important information to evaluate the models’ performance, and it was not verified in woody savannah areas, like Cerrado s.s.
We consider our results corroborate that the interception models present limitations to simulate very low and extremely elevated levels of interception on an individual event scale (Linhoss and Siegert, 2020). By applying seasonal coefficients, we could see that the models can represent the seasonal variation occasioned by the phenology variation (represented by the different storage capacities coefficients). And the lowest performance during summer (lower NSE values than the other seasons) could be another indication of the limitations of the models to represent the interception during intense rainfall events.
R1C-03: Given that the models mostly cannot reproduce per-storm interception loss, and the time-integrated performance is driven by essentially the regression between rainfall and throughfall expressed in the calibrated canopy storage capacity, I do not agree with the conclusions that the modeling is useful for understanding canopy processes. These are not unique problems for this study and field experience is valuable for developing the next generation of models, but the modeling presented in this manuscript is not well described or developed.
AR: Our conclusion was that "
“[…] the adapted models applied here are valuable to modelling the interception processes in Neotropical savannas like Cerrado s.s. forests. The models had problems to simulate the evaporation processes at daily time-scale, but not to others processes. Both models are suitable to estimate the total interception at monthly basis and could be used to inter annual analysis, but for seasonal differences in Cerrado s.s. forests the Rutter model seems to be more appropriate.”
In fact, despite not to be able to simulate well at a daily event-scale, the models had a good performance at a monthly scale and can be useful to estimate the total interception annually.
Regarding model development, our intent was to maintain the concepts and assumptions of the original models. As there is not a Gash model version that includes the forest floor interception, and it is the first time we have forest floor interception direct measurements in a forest in South America, we thought the first step was to work with the two more used interception models.
R1C-04: This manuscript needs a thorough edit for English: there are many errors in word choice, grammar, and spelling.
AR: We will do a thorough English language check in the revised version.
R1C-05: L34 these are rainforest numbers that are not likely relevant to the field site here
AR: Thank you for your comment. Our intention was to highlight the high feedback values in tropical regions, as these are close to our study site. But we agree that it will be better to add more specific information, regarding the same forest type. So we suggest changing the sentence in lines 34-35 to “In tropical savanna, the feedback to the atmosphere can be high (Oliveira et al., 2015; Anache, 2018), reaching up to 82% of annual rainfall through evapotranspiration process (Cabral et al., 2015).”
R1C-06: L74 The logic of the theoretical work should be justified here. The objective based on the study site, L72, is not as important as the theoretical implications of behavior of the model, and there are modifications to the basic models that are not explained here at all. Similarly, the model introduction L55-64 should focus on the shortcomings of existing models that need to be overcome, not on basics.
AR: We agree that there are some models’ issues that should be overcome. But our main objective here was not to show that we can model interception for a Cerrado forest. Our aim was to extend the existing Rutter and Gash models by adding the forest floor interception and verifying the models’ performance to simulate seasonal variations as it is pronounced for the Cerrado s.s. forest.
R1C-07: L94 15,000 stems per ha? This is 1.5 trees per square meter; is that correct?
AR: Yes, it is correct. The density and richness of species in the study area were evaluated by Reys et al. (2013).
R1C-08: L98 What use is PAR for this study?
AR: This was additional information that can be removed.
R1C-09: L114 why “also”?
AR: It was because at the beginning of the item “2.2. Experimental setting” was mentioned some monitored variables recorded each 10 min including precipitation. We can rewrite this item to explain better that all automatic measurements were recorded each 10 min.
A suggestion of alteration to this part of the text is present in the answer to the comment R1C-11.
R1C-10: L119 according to DBH, but by what criteria and why?
AR: Indeed, this information is missing in the text. A suggestion of alteration to this part of the text is present in the answer to the comment R1C-11
R1C-11: L120 this reference does not provide details of this device. A few critical details about how a LID works are needed for this part to be understandable
AR: Hereby follows the revised item 2.2.
“This study comprises the monitoring period between June 01st of 2017 and February 06th of 2020. Precipitation, temperature, relative humidity, wind velocity, net solar radiation, and others environmental variables were collected in the site through a meteorological tower of 11 m of height. In case of missing data of the main monitoring weather station, we used the nearest available meteorological data from site 1 (Anache et al., 2019), which was 1.7 km away from our site.
Figure 1. Location of the study area (from ©Google Maps, 2022).
To investigate the interception process in the Cerrado s.s., canopy, trunk and forest floor interception were measured. Canopy and trunk interception were indirectly determined by the difference between the rainfall (𝑃𝑔) and throughfall (𝑇𝑓) and stemflow (𝑇𝑠), respectively. By including forest floor interception, the total forest interception is measured by the difference between the 𝑃𝑔 and the infiltration (𝐹), as in Eq. (1).
𝐸𝑖,𝑐+ d𝑆𝑐d𝑡+𝐸𝑖,𝑡+ d𝑆𝑡d𝑡 +𝐸𝑖,𝑓+ d𝑆𝑓d𝑡=𝑃𝑔−𝐹 (1)
where 𝑆𝑐, 𝑆𝑡 and 𝑆𝑓 are the storage capacities (mm) of the canopy, the trunks, and the forest floor, respectively, and 𝐸𝑖,𝑐 , 𝐸𝑖,𝑡 , and 𝐸𝑖,𝑓 are the evaporation from these components (mm) in a certain period of time (𝑡).
𝑇𝑓 was measured through four pluviographs of 0.254 mm resolution. Additionally, three gutters linked to the pluviographs (tipping bucket resolution of 0.048 mm or 0.029 L) were used and five additional gutters were directly connected to reservoirs for manual measurements after accumulated events (> 15 mm). For 𝑇𝑠 monitoring, we installed three automatic collectors. Plastic hoses were wrapped around six trees (two per each collector) at breast height that channeled the 𝑇𝑠 through pluviographs (tipping bucket volume of 5 mL) to reservoirs Selected trees to be monitored were divided into three groups according to the DBH (DBH ≤ 10 cm, 11 cm≤ DBH≤ 30 cm e DBH ≥ 31 cm). In addition, 𝑇𝑠 was measured through more 12 manual collectors installed by Oliveira et al. (2015). More details about the equipment are given in Table 1.
Forest floor interception and infiltration were measured through LIDs – Litter Interception Devices. This device allows the continuous direct measurement of the weight (1 g of resolution) of a forest litter sample in situ and at ground level. Moreover, it also allows infiltration monitoring through a pluviograph (5 ml of tipping bucket resolution) installed inside the device, below the tray wherein the litter sample takes place (Rosalem et al., 2019). We installed three of these weighing LID devices, but due to problems with a load cell only the interception records of two LIDs were used. The LIDs were installed close to the meteorological tower and were filled with quasi-undisturbed litter samples of the thickness of about 6 cm. The forest litter samples were changed each August considering its high decomposition rate in dense Cerrado areas (half-life for the decomposing material around 1.8 year) (Cianciaruso et al., 2006). All automatic measurements were recorded each 10 minutes.”
R1C-12: L123 But there was only one August during the study, correct?
AR: No. The monitoring was conducted between June 01st of 2017 and February 06th of 2020, which included three August months. But, as the first samples were taken in June, they were no changes in the first August.
R1C-13: L137 upper envelopes and linear regressions are not the same. Stem storage capacity has generally not been estimated by this technique in the past for several important reasons. This needs justification and explanation.
AR: The sentence in line L137 was not so clear. The applied approach was the same as explained in the previous sentence (L136). To our knowledge, interception modelling studies do apply this approach (Távora and Koide, 2020; Deng, 2020; Gerrits et al., 2010). We followed the procedures recommended by Rutter et al. (1971) and Gash and Morton (1978), and used the mean method (Klaassen et al., 1998) to get the Sc, St, and Sf parameters.
The description of the methodology to obtain the parameters, presented between L134-L141, will be re-written as:
“The storage capacities parameters were determined following the procedures by Rutter et al. (1971) and Gash and Morton (1978), applying the mean method (Klaassen et al., 1998; Gerrits et al., 2010; Sadeghi et al., 2015). They are determined by the negative intercept of the upper envelop scatter diagram between observed values of an input process (e.g., Pg) and its subsequent process (e.g., Tf) (Robins, 1969, apud (Rutter et al., 1971). The partitioning factors of the rain, free throughfall (p) and trunk input (pt), were determined by the slope of the lower envelop line for the same scatter diagram. Thus, the 𝑆𝑐, p, 𝑆𝑡, and pt parameters were determined using Pg and Tf records, and Pg and Ts records, respectively (Távora and Koide, 2020; Deng, 2020). By the same procedure, the storage capacity of the forest floor (𝑆𝑓) and the partitioning factor of the preferential flow (pf), were determined through Tf and F values (Gerrits et al., 2010). These analyses were carried out based on independent rain storms (each one preceded by at least 24 hours without rainfall) out the total of 236 rain days during the calibration period.”
R1C-14: L150 These two ad-hoc modifications to the Rutter model are not sufficiently described or justified. There is no way to understand exactly what "priority" means in terms of water and energy balances.
AR: When Rutter’s model is run, the first water balance happens in the canopy wherein the volume drained will join to free throughfall, forming the throughfall. Similarly, for the trunk component, we have to choose which process is going to be the first to remove water from the reservoir (i.e., through evaporation, drainage, or stemflow). So, the term “priority” was used in the text to explain that we choose to first calculate evaporation (based on the storage) whereafter we calculate the drainage, stemflow, or infiltration (to the forest floor reservoir). This ordering of calculation steps was also done by Gerrits et al. (2010). By choosing to prioritize the evaporative process, we were trying to minimize the low evaporative rate during the rainfall (L281). Besides, it indirectly could account for other processes, such as splash evaporation (Murakami, 2006; Bassette and Bussière, 2008) and vertical updrafts during rainfall (van Dijk and Bruijnzeel, 2001).
R1C-15: L152 What about the Tf records indicate dynamic storage?
AR: Yes. This is what we wanted to say in L152 by saying “These observations”. They mean the Tf records.
R1C-16: L153 this term needs explicit definition. I do not know what Cc is supposed to represent
AR: We respond to this comment in the following comment response (R1C-17).
R1C-17: L156 What is the purpose of Ccmax exactly and how was it implemented? Rutter's model explicitly allows storage C to exceed storage capacity S, with no explicit upper limit. Adding a cap here appears to reduce the dynamic storage, not enable it. Why only "before the end of the storm"?
AR: Cc has the same meaning as C in Rutter’s original model, which represents the amount of water on the canopy. We added the c minor just to specify that this amount was in the canopy. In the text (L153) we said “[..] some dynamic storage on the canopy (Cc) is present.”, declaring the Cc meaning in the text. However, to avoid confusion, we will rewrite L153-L157."
The purpose of Ccmax was to allow a greater water amount to remain in the canopy among time steps of model’s running. If the Ccmax is not added as a maximum threshold, the difference Cc – Sc will be drainage and only Sc remains in the next step. And this maximum dynamic storage (Ccmax) parameter also allows to haven’t water amount physically impossible on the canopy during the rain event.
“These observations show that apparently some dynamic storage on the canopy is present. The results of Aston (1979) and Lloyd et al. (1988) agree with this, and the authors recommended, as Valente et al. (1997), that a limiting parameter should be used to prevent a build-up of water on the canopy. Therefore, besides the dynamic storage parameters, here specified as Cc, Ct, or Cf, to canopy, trunk, and forest floor, respectively, not only C as in (Rutter et al., 1971), we added a maximum dynamic storage parameter (𝐶𝑐𝑚𝑎𝑥) as a threshold to the retained water amount on the canopy before the end of the rainfall storm. Klaassen et al. (1998) observed by using microwave transmission to measure the water storage on the canopy, that during rain events the dynamic storage is affected by rain intensity in such a way that could lead sometimes to higher dynamic storage along the rain than at end of the rain event. For the original model, the maximum water amount on the reservoir is the storage coefficient. So, the Ccmax was necessary to allow greater water amount than Sc before the end of the rainfall, and to prevent a build-up on the canopy.”
R1C-18: L159 what is Cmax. the same as Ccmax?
AR: Yes. It was a typing error. It should be Ccmax.
R1C-19: L160 I don't understand this parameter identification technique
AR: I think you are asking about the representation meaning of Cc< Sc. This represents the amount of water on the canopy, which is below the storage coefficient amount. For this condition, we said that besides the evaporation, the drainage could still happen by the leaves shaking off (Gerrits et al., 2010), and it was modelled using the same exponential equation as proposed by (Rutter et al., 1971) to the drainage process.
R1C-20: L176 I do not understand how data were used to estimate Cfmax
AR: Ccmax and Cfmax represent the maximum dynamic storage on the canopy and the forest floor, respectively. Since we have continuous monitoring of the water content in the LIDs’ samples, the Cfmax is found by the peaks of water retention during the rain events.
The forest litter amount varies along the year due to decomposition and the litter input, which causes some shifts in the amount of litter that can be noted in the time series of water content. Thus, we selected the higher peaks observed among these shifts to the two time series used (of LID 1 and LID 2), and used the average value as Cfmax.
R1C-21: L179 Most of the parameters referred to in this sentence are not defined
AR: Is and f are the exponential coefficients used to simulate the drainage through the forest litter sample that will contribute to the infiltration when the dynamic storage of the forest floor (Cf) is higher than the storage coefficient (Sf) and lower than the maximum dynamic storage of the forest floor (Cfmax). As mentioned in L179, Is and f are just exponential coefficients, like Ds and b used in the Rutter model to simulate the drainage process on the canopy.
R1C-22: L185 I can't find where these parameters were estimated
AR: How the parameter Sf is obtained was presented in L138. However, in the response to comment 13 (R1C-13), we proposed a revision of our text between L134 and L141 that it will make clearer the approach and add how the partitioning coefficients (p, pt, and pf) were obtained.
R1C-23: L199 something is wrong. This describes an experiment 8 months long but fig 3 says 3 years
AR: Yes, there is a typing error in L199. It was supposed to be “From June of 2017 […]”, such as the monitoring period mentioned in L100.
R1C-24: L204 The original Gash model was applied at the monthly time scale. There is no time scale assumed by Gash
AR: Indeed in the first paper with the Gash model (Gash, 1979) the model was applied at the monthly time scale and there is no time scale assumed by Gash. But it can be run on a daily time-step (Valente et al., 1997), by assuming the occurrence of a single rainfall event per rainfall day. And it is hard to select a time-step lower than a day due to the “drying phase” that should last from the end of the rainfall until the canopy and trunks are completely dry.
We suggest changing the sentence in L204, since it gives the idea that the Gash model is applied only at a daily time scale. We suggest the following change in the text (L204):
“Because the Gash model was applied at a daily time-scale, the Rutter model results were evaluated a minimum on a daily time-scale, thereto the models’ results could be compared.”
R1C-25: L214 This conflicts slightly with L145. I assume this means stomatal resistance was zero. There are no details about assumptions and about parameters relating to aerodynamic resistance
AR: Yes, you are right. We calculated Ep through the Penman equation, and not through the Penman-Monteith equation. The two are equal in case we take a stomatal resistance of zero. To be consistent we will say that we calculate Ep with Penman.
R1C-26: L220 Fig 3: L218 told us to expect net radiation here
AR: One sentence should be added here to explain what is going to be presented in Figure 3. We suggest the following sentence:
“The observed relative air humidity and net radiation corresponded on average to 66.8 (± 20 %) and 154.6 (± 239 W m-2), respectively. The daily potential evaporation and income radiation along the monitored period are presented in Figure 3”.
R1C-27: L224 more similar than what?
AR: Thank you for correcting us on this. The word “more” has to be removed from this sentence.
R1C-28: L242 Why does the Gash model predict essentially the same canopy evaporation every storm except for very small storms?
AR: As the Gash model was run at a daily timescale the canopy storage refill is not simulated. And, as most storms had precipitation greater than the 𝑃′𝑔 (the amount of water necessary to saturate the canopy) the canopy evaporation reached this threshold (𝑃′𝑔) because the model also assumes that the whole daily potential evaporation value is used to dry the canopy and trunk.
R1C-29: L264 I think it is unlikely that measurements of forest-floor evaporation were better than the throughfall measurements.
AR: In our view, in this case, it could be possible due to the greater collection area of the LID (0.16 m² each LID) than the pluviographs collecting throughfall (0.0214 m² each pluviograph), and the high sensitivity of the load cells used (0.001 kg). However, it is possible that the main cause for the overestimation is being fixed throughfall gauges.
R1C-30: L265 It is not obvious to me that canopy evaporation is modeled better.
AR: In L265, we explained why we couldn’t use the Ec indirect estimations considering the forest floor evaporation measurements (𝐸𝑐 =𝑃𝑔−𝐹−𝑇𝑠−𝐸𝑓) to compare the daily modelled Ec. Despite the difference in Ec indirect measurements commented on in the previous paragraph (L260-264), the Ec indirectly observed by using the automatic throughfall records had to be used to analyze the daily time scale performance.
R1C-31: L280 As described earlier, it did exactly the opposite: it prevented accumulation
AR: We couldn’t understand exactly what you meant here, but as it is related to Ccmax, maybe you will have a different opinion after our more detailed explanation in R1C-17 and R1C-20.
R1C-32: L284 How? Shouldn't mean daily Rutter be the same as daily Gash?
AR: By the Gash model assumption that “the meteorological conditions prevailing during any wetting-up of the canopy are sufficiently similar to those prevailing for the rest of the storms”, the rate factor is used in the calculus. When the potential evaporation is used, like in Rutter model, there is a decrease in the potential evaporation during the storm. In our study, we observed that it happened mainly due to the decrease in income radiation during the storm. It is worth mentioning that 43% of the rain event occurred during the night. This may cause a greater difference between the daily predictions by these models since the interception ratio (interception/precipitation) in the first half of the rain event has shown to be greater than the ratio in the second half (Iida et al., 2017).
Another crucial point is that the Rutter model allows to simulate the refill of canopy storage capacity if more than one rain event occurs during the day, which is not possible for the Gash model.
As many rain events occurred during the night, the models will have different responses at a daily time scale.
R1C-33: L344 This problem is severe for the modeling assumptions and needs detailed theoretical discussion
AR: The answer to this comment is presented in the following AR (comment R1C-34).
R1C-34: L348 I do not understand this sentence
AR: Actually, prioritizing the canopy and trunk evaporation over the forest floor evaporation was not an assumption. It was not our intention to reformulate the original model, instead, our intention here was to add the forest floor component to the original model. So, what we did was to add this component and change it as minimum as possible the original structure. We believe that seeing the possible issues for including the forest floor in the model, will allow us and others to improve the measurements needed (e.g., energy budget at soil level in the forest) and subsequently the modelling.
Thus, we said in L344-L347 what is known about the different weather conditions between the above and below the canopy. And, despite the potential evaporation being lower below the canopy, the approach modelling applied caused delays onset of the forest floor evaporation process.
As you said you didn’t understand the sentence in L348, here it is a suggestion to change the paragraph (L343-L348).
“Due to the average rate factors (𝐸̅/𝑅̅), the Gash model could simulate higher values than the Rutter model during the rain. In our adapted Rutter model, the canopy and trunk evaporation are prioritized over the forest floor evaporation. As the 𝐸𝑝 below the canopy is lower than above, and due to different weather conditions near the understory (Coenders-Gerrits et al., 2020). In addition, the forest floor and the canopy differ in physical structure, resource availability and biotic conditions (Yanoviak and Kaspari, 2000), which implies different energy and water fluxes. Hence, aiming to not change the original structure of the Rutter model, the remained Ep was used to modelling the forest floor evaporation. So, the canopy and trunk evaporation process were dominant over forest floor evaporation, which caused delays onset on the forest floor evaporation for this adapted version.”
R1C-35: L353 I agree. I think this makes the Gash model irrelevant for this process. Lack of data to estimate energy budgets and vapor exchange at the scale of the forest floor are also major problems for applying a Rutter model. Given that Rutter does not seem to be helpful for estimating canopy evaporation on a storm scale, it seems even less likely to be helpful for forest-floor evaporation.
AR: For the Gash model, by including the forest floor interception component maybe it will be better to not consider the daily time scale. Also, a modification added by (Valente et al., 1997) in their reformulated version for sparse forests to canopy evaporation should be considered “Although evaporation from the saturated trunks is not used explicitly in this model it is subjacent to the calculation of interception loss from the trunks. Thus, a small modification was made to the model of Gash et al. (1995) so that Ec is now replaced with (1 - e) Ec.”. So, besides taking water residence time into consideration, further improvements to Gash model for non-sparse forests should also consider the fraction of concern to forest floor evaporation.
We agree that both models present poor storm scale predictions, especially in Tropical areas like Cerrado forests with more intense rainfall events. However, the monthly performance to total interception estimates for these models was good (NSE: 0.63 – 0.85 and R²: 0.78 – 0.97), and good accumulated interception estimates for the Rutter model to the almost 3 years of monitoring presented in Figure 5 in the text (underestimation of 11.5%, L417). In Fig 5 is also possible to see that the total volume estimated for the forest floor interception was particularly good for both models, varying between -2% and 9% when seasonal coefficients were applied.
R1C-36: L397 I don’t think the evaporation model was useful for small storms, either. There was not systematic negative bias in estimates, so that means the small storms were overestimated.
AR: The adapted Rutter model had poor estimations on daily basis, even for small storm events. However, the adapted Gash model presented daily canopy evaporation not so bad for small storm events, meaning rainfall lower than 7 mm.day-1 (NSE = 0.65, R²=0.90, MBE=+0.11).
R1C-37: L425 I would agree if the models were useful at short timescales, but they are not. If the processes are not correctly modeled, how can we extrapolate using the model?
AR: In this sentence we were referring to broad sense, meaning that interception modelling could help to elucidate the role of different factors.
R1C-38: L510 I do not agree. The models get major masses correct, but due to calibration (e.g, of canopy storage). The models are not likely to be of any predictive value should, for example, vegetation structure change.
AR: As we said in L511, the adapted models could be used on monthly basis, or inter-annual analysis since the accumulated volumes were not so far from the observed ones to the adapted Rutter model. The throughfall and infiltration were well modeled for both models, even on a daily time scale (NSE: 0.81 – 0.94), which elucidate that the problem remains on the evaporation modelling.
Regarding canopy storage, the coefficient was not calibrated but the maximum dynamic storage (Ccmax, added to the adapted Rutter model) was calibrated. This parameter had a major effect on the throughfall and, consequently, on the infiltration estimates, but didn’t compensate for the problem with low potential evaporation during the rainfall to the canopy evaporation for the Rutter model.
R1C-39: The journal names are missing for many References.
AR: Thanks for the observation. We are going to check it for the final paper version.
References
Anache, J. A. A.: Alterações no ciclo hidrológico e na perda de solo devido aos diferentes usos do solo e variações climáticas em área de Cerrado, Doutorado em Hidráulica e Saneamento, Universidade de São Paulo, São Carlos, https://doi.org/10.11606/T.18.2018.tde-17042018-110107, 2018.
Bassette, C. and Bussière, F.: Partitioning of splash and storage during raindrop impacts on banana leaves, Agricultural and Forest Meteorology, 148, 991–1004, https://doi.org/10.1016/j.agrformet.2008.01.016, 2008.
Cabral, O. M. R., da Rocha, H. R., Gash, J. H., Freitas, H. C., and Ligo, M. A. V.: Water and energy fluxes from a woodland savanna (cerrado) in southeast Brazil, Journal of Hydrology: Regional Studies, 4, 22–40, https://doi.org/10.1016/j.ejrh.2015.04.010, 2015.
Deng, J.: Fitting the revised Gash analytical model of rainfall interception to Mongolian Scots pines in Mu Us Sandy Land, China, Trees, Forests and People, 1, 100007, https://doi.org/10.1016/j.tfp.2020.100007, 2020.
van Dijk, A. I. J. M. and Bruijnzeel, L. A.: Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 1. Model description, Journal of Hydrology, 247, 230–238, https://doi.org/10.1016/S0022-1694(01)00392-4, 2001.
Gash, J. H. C.: An analytical model of rainfall interception by forests, Quarterly Journal of the Royal Meteorological Society, 105, 43–55, https://doi.org/10.1002/qj.49710544304, 1979.
Gash, J. H. C. and Morton, A. J.: An application of the Rutter model to the estimation of the interception loss from Thetford Forest, Journal of Hydrology, 38, 49–58, https://doi.org/10.1016/0022-1694(78)90131-2, 1978.
Gerrits, A. M. J., Pfister, L., and Savenije, H. H. G.: Spatial and temporal variability of canopy and forest floor interception in a beech forest, Hydrological Processes, 24, 3011–3025, https://doi.org/10.1002/hyp.7712, 2010.
Iida, S., Levia, D. F., Shimizu, A., Shimizu, T., Tamai, K., Nobuhiro, T., Kabeya, N., Noguchi, S., Sawano, S., and Araki, M.: Intrastorm scale rainfall interception dynamics in a mature coniferous forest stand, Journal of Hydrology, 548, 770–783, https://doi.org/10.1016/j.jhydrol.2017.03.009, 2017.
Klaassen, W., Bosveld, F., and de Water, E.: Water storage and evaporation as constituents of rainfall interception, Journal of Hydrology, 212–213, 36–50, https://doi.org/10.1016/S0022-1694(98)00200-5, 1998.
Murakami, S.: A proposal for a new forest canopy interception mechanism: Splash droplet evaporation, Journal of Hydrology, 319, 72–82, https://doi.org/10.1016/j.jhydrol.2005.07.002, 2006.
Oliveira, P. T. S., Wendland, E., Nearing, M. A., Scott, R. L., Rosolem, R., and da Rocha, H. R.: The water balance components of undisturbed tropical woodlands in the Brazilian cerrado, Hydrology and Earth System Sciences, 19, 2899–2910, https://doi.org/10.5194/hess-19-2899-2015, 2015.
Reys, P., Camargo, M. G. G. de, Grombone-Guaratini, M. T., Teixeira, A. de P., Assis, M. A., and Morellato, L. P. C.: Estrutura e composição florística de um Cerrado sensu stricto e sua importância para propostas de restauração ecológica, Hoehnea, 40, 449–464, https://doi.org/10.1590/S2236-89062013000300005, 2013.
Rutter, A. J., Kershaw, K. A., Robins, P. C., and Morton, A. J.: A predictive model of rainfall interception in forests, 1. Derivation of the model from observations in a plantation of Corsican pine, Agricultural Meteorology, 9, 367–384, https://doi.org/10.1016/0002-1571(71)90034-3, 1971.
Sadeghi, S. M. M., Attarod, P., Van Stan, J. T., Pypker, T. G., and Dunkerley, D.: Efficiency of the reformulated Gash’s interception model in semiarid afforestations, Agricultural and Forest Meteorology, 201, 76–85, https://doi.org/10.1016/j.agrformet.2014.10.006, 2015.
Távora, B. E. and Koide, S.: Event-Based Rainfall Interception Modeling in a Cerrado Riparian Forest—Central Brazil: An Alternative Approach to the IS Method for Parameterization of the Gash Model, Water, 12, 2128, https://doi.org/10.3390/w12082128, 2020.
Valente, F., David, J. S., and Gash, J. H. C.: Modelling interception loss for two sparse eucalypt and pine forests in central Portugal using reformulated Rutter and Gash analytical models, Journal of Hydrology, 190, 141–162, https://doi.org/10.1016/S0022-1694(96)03066-1, 1997.
Citation: https://doi.org/10.5194/hess-2022-59-AC1
-
AC1: 'Reply on RC1', Livia Rosalem, 15 Nov 2022
-
RC2: 'Comment on hess-2022-59', Anonymous Referee #2, 28 Jun 2022
Comments on "Water partitioning in a Neotropical Savanna forest (Cerrado s.s.): interception responses at different time-scales using adapted versions of the Rutter and the Gash models" submitted to HESS-D by Rosalem et al.
[General comment]
This paper measured the interception processes including forest floor interception loss for a period of 32 months, and applied Rutter and Gash models. Although the methodology used in this study should be evaluated carefully, I recognized that the efforts to conduct measurements would be very intensive, and would agree the importance to understand interception phenomena in Cerrado region. However, the logic of this manuscript is not constructed well, so readers cannot understand what is the most important findings in this research. The main description seems to be the model performances for daily, biweekly and monthly basis. I understood the differences in performance among different time scales, but I do not think that the differences are important to clarify the interception process. The important thing must be what factors resulted in the differences and what processes affect the model performance. I would like to recommend that authors, at first, write a paper to show interception process in this site in detail based only on measurement data without applying any models. The interception loss in this site would be affected by rainfall characteristics (e.g., rainfall duration, rainfall amount, rainfall intensity, etc)? Or, other micrometeorological factors (e.g., wind speed, vapor pressure deficit, net radiation, etc) influence the interception process? Based on these knowledges, I believe that authors could develop suitable models to simulate the interception process. I hope my comments will help to do substantial revisions.
[Specific comments]
Could you add the LAI data? (line 92-93)
I cannot catch up how to calculate the spatially representative amount of throughfall. Is that the average of four Davis, five manual gutters and three automatic gutters? (line 100-119)
Because the canopy in this site is discontinuous (line 93), I could expect that the spatial heterogeneity of throughfall would be very high. Please show the differences in throughfall amount measured by four Davis, five manual gutters and three automatic gutters. The differences are related to the canopy openness? Also, a total of 12 measurements of throughfall are safely enough to obtain spatial representative value?
As Rosalem et al. (2018), published in Ecohydrology, pointed out, Davis gauge underestimates the inflow of water flux with increasing intensity (please see FIGURE 3 in Rosalem et al., 2018). I would like to confirm that authors applied the same correction to throughfall measurements in this paper. If not, application must be required.
Three gutters are connected to three Davis gauge? If so, I am wondering that the one tip amount of 0.048 mm is too small to detect the correct amount. As Rosalem et al. (2018) showed, the underestimation by Davis gauge is relatively high. How many pulses generated by the gauge were recorded in 10-min intervals? The time between tips, equivalent to 600 second divided by accumulated pulse count, should be more than 1.0-1.5 second. Please note that, if authors used other rainfall gauge, similar issue exists and should be investigated.
How did authors calculate stemflow amount in the stand scale?
Three automatic collectors of stemflow is connected to Davis gauge? The same issue mentioned above, underestimation of inflow with increasing intensity, must be checked.
Forest floor interception was measured by two LIDs (line 121-122). Please show the evidence indicating two LIDs could safely measure the spatial representative value of forest floor interception. This is very critical, because high spatial heterogeneity of throughfall could be expected from the disconnected canopy in this site.
Figure 2: The forest floor evaporation was calculated considering potential evaporation (Ep), but how did you calculate Ep above forest floor? Did you measure net radiation above the forest floor?
Please add the description to explain how to calculate the aerodynamic conductance above canopy and forest floor (line 214-215).
How did you obtain the interception ratio of 33%? (line 253) Throughfall was described as 70-72% (line 228), so 33% interception is too large. There is a description of 40% interception in conclusion section (line 481). Maybe the target rainfall events are different among parts, but it is difficult to understand.
I cannot understand why Ec is calculated as the difference Pg and the sum of F, Ts and Ef. I recommend that basic equation showing rainwater balance should be added in the M&M section.
Discussion about the stemflow is not directly related to this paper (line 300-313). If authors show the data of canopy structure, bark, and so on, it is useful. Unfortunately, the current MS did not include any data, so I recommend to remove this part.
I felt that the much sentences in the current MS are related to model performance (line 237-245, 268-273, 293-299, 320-328, 335-342, 368-380, 435-454, Table 6, 7, 8, 9). Similar descriptions are found among parts, so I recommend reconstruction of the logic. In my opinion, it would be better that "discussion" should be separated from the "result" section. Then, descriptions of model performance should move to discussion section, and more concise discussion is recommended. Rather than differences in model performance, the reason for the difference and factors affecting it are more important to understand the interception process in this site.
Looking at appendix C, there are high correlations between observation and model output for throughfall and stemflow. However, the correlation for interception loss is very low. Could you explain this point?
[Technical corrections]
Line 98: "average PAR (photosynthetic active radiation) of 1041.8 ± 427.4 μmol.m-2.s-1", I cannot understand the duration of average.
Line 199-200: "From June of 2019 to May of 2019, we used to calibrate our Rutter and Gash models, while the second part, from June of 2019 up to of 07th February of 2020, .." Please check the consistency of months. The current description is strange.
Figure 3, y-axis title of the upper panel: "Potencial" should be "Potential".
Figure 3, lower panel: Is this net radiation? Solar radiation was measured at 2 m height (Table 1). More than 600 W m-2 value is too high for solar radiation above the forest floor. Please check carefully.
Citation: https://doi.org/10.5194/hess-2022-59-RC2 -
AC2: 'Reply on RC2', Livia Rosalem, 15 Nov 2022
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-59/hess-2022-59-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Livia Rosalem, 15 Nov 2022
Status: closed
-
RC1: 'Comment on hess-2022-59', Anonymous Referee #1, 11 May 2022
This manuscript describes measurement and modeling of interception loss from the canopy and forest floor. As is typical in published applications of Rutter-Gash models, the models performed poorly at the scale of individual events and better when accumulated over longer timescales as presented in Table 5. The manuscript also presents some potentially important modifications to the standard Rutter-Gash approach by modifying the model of canopy storage, modifying the evaporation rate from trunks, and adding explicit measurement of evaporation from wet forest floor. The manuscript does not evaluate the behavior of the models in relation to the standard ones, so the importance of the modifications is difficult to understand. In addition, the manuscript omits some details about parameter estimation and presents some model modifications vaguely and in terms of programming techniques, so the methods are not possible to follow comprehensively. The discussion touches on some important questions pertaining to the applicability of the modeling assumptions, such as time scale of forest-floor evaporation and assumptions embedded in the Gash model, and mismatches between data needs and data availability, such as radiation budgets relevant for the forest floor. However, the manuscript mostly disregards these important points and the poor by-storm performance, and makes detailed interpretation of the effects of seasonal variation of parameters. Given that the models mostly cannot reproduce per-storm interception loss, and the time-integrated performance is driven by essentially the regression between rainfall and throughfall expressed in the calibrated canopy storage capacity, I do not agree with the conclusions that the modeling is useful for understanding canopy processes. These are not unique problems for this study and field experience is valuable for developing the next generation of models, but the modeling presented in this manuscript is not well described or developed.
This manuscript needs a thorough edit for English: there are many errors in word choice, grammar, and spelling.
L34 these are rainforest numbers that are not likely relevant to the field site here
L74 The logic of the theoretical work should be justified here. The objective based on the study site, L72, is not as important as the theoretical implications of behavior of the model, and there are modifications to the basic models that are not explained here at all. Similarly, the model introduction L55-64 should focus on the shortcomings of existing models that need to be overcome, not on basics.
L94 15,000 stems per ha? This is 1.5 trees per square meter; is that correct?
L98 What use is PAR for this study?
L114 why “also”?
L119 according to DBH, but by what criteria and why?
L120 this reference does not provide details of this device. A few critical details about how a LID works are needed for this part to be understandable
L123 But there was only one August during the study, correct?
L137 upper envelopes and linear regressions are not the same. Stem storage capacity has generally not been estimated by this technique in the past for several important reasons. This needs justification and explanation.
L150 These two ad-hoc modifications to the Rutter model are not sufficiently described or justified. There is no way to understand exactly what "priority" means in terms of the water and energy balances.
L152 What about the Tf records indicate dynamic storage?
L153 this term needs explicit definition. I do not know what Cc is supposed to represent
L156 What is the purpose of Ccmax exactly and how was it implemented? Rutter's model explicitly allows storage C to exceed storage capacity S, with no explicit upper limit. Adding a cap here appears to reduce the dynamic storage, not enable it. Why only "before the end of the storm"?
L159 what is Cmax. the same as Ccmax?
L160 I don't understand this parameter identification technique
L176 I do not understand how data were used to estimate Cfmax
L179 Most of the parameters referred to in this sentence are not defined
L185 I can't find where these parameters were estimated
L199 something is wrong. This describes an experiment 8 months long but fig 3 says 3 years
L204 The original Gash model was applied at the monthly time scale. There is no time scale assumed by Gash
L214 This conflicts slightly with L145. I assume this means stomatal resistance was zero. There are no details about assumptions and about parameters relating to aerodynamic resistance
L220 Fig 3: L218 told us to expect net radiation here
L224 more similar than what?
L242 Why does the Gash model predict essentially the same canopy evaporation every storm except for very small storms?
L264 I think it is unlikely that measurements of forest-floor evaporation were better than the throughfall measurements.
L265 It is not obvious to me that canopy evaporation is modeled better.
L280 As described earlier, it did exactly the opposite: it prevented accumulation
L284 How? Shouldn't mean daily Rutter be the same as daily Gash?
L344 This problem is severe for the modeling assumptions and needs detailed theoretical discussion
L348 I do not understand this sentence
L353 I agree. I think this makes the Gash model irrelevant for this process. Lack of data to estimate energy budgets and vapor exchange at the scale of the forest floor are also major problems for applying a Rutter model. Given that Rutter does not seem to be helpful for estimating canopy evaporation on a storm scale, it seems even less likely to be helpful for forest-floor evaporation.
L397 I don’t think the evaporation model was useful for small storms, either. There was not systematic negative bias in estimates, so that means the small storms were overestimated.
L425 I would agree if the models were useful at short timescales, but they are not. If the processes are not correctly modeled, how can we extrapolate using the model?
L510 I do not agree. The models get major masses correct, but due to calibration (e.g, of canopy storage). The models are not likely to be of any predictive value should, for example, vegetation structure change.
The journal names are missing for many References.
Citation: https://doi.org/10.5194/hess-2022-59-RC1 -
AC1: 'Reply on RC1', Livia Rosalem, 15 Nov 2022
Author’s Response to Referee #1
We would like to thank the anonymous Referee #1 for the detailed review and for the time spent reviewing our text. We replied to the referee’s comments, which are highly useful to improve the quality of our manuscript. Note that the original referee’s comments are identified as R1C and written in bold, and the authors’ responses are labeled as AR. In addition, all comments are numbered (e.g., R1C-01).
This manuscript describes measurements and modeling of interception loss from the canopy and forest floor. As is typical in published applications of Rutter-Gash models, the models performed poorly at the scale of individual events and better when accumulated over longer timescales as presented in Table 5. The manuscript also presents some potentially important modifications to the standard Rutter-Gash approach by modifying the model of canopy storage, modifying the evaporation rate from trunks, and adding explicit measurements of evaporation from the wet forest floor.
The manuscript does not evaluate the behavior of the models in relation to the standard ones, so the importance of the modifications is difficult to understand. In addition, the manuscript omits some details about parameter estimation and presents some model modifications vaguely and in terms of programming techniques, so the methods are not possible to follow comprehensively.
R1C-01: We presented modified versions of the Rutter and Gash models by adding the forest floor interception process. The canopy and trunk processes were maintained similar to the original versions; the only difference was the addition of a dynamic storage coefficient (Ccmax) to the canopy storage. This coefficient does not change the storage and evaporative processes modelling. However, it allows more water being hold on the canopy during the rainfall, which reduces the drainage picks and enables to increase the evaporation.
The results showed that Rutter’s model canopy evaporation continued underestimated despite adding the Ccmax coefficient. For the Gash model, the results should be the same for canopy and trunk evaporation since the dynamic storage was not included. Thus, the main differences observed should be only on the total interception, which for the original versions should be always lower since they do not account for the forest floor interception.
The parameter estimation was following the same approach as for the original models (explanation started on line 134). Intending not to be unduly long, the well-known approaches to determine the models’ parameters were shortly explained and referenced. We developed a code in Python which we can add to the Appendix.
R1C-02: The discussion touches on some important questions pertaining to the applicability of the modeling assumptions, such as time scale of forest-floor evaporation and assumptions embedded in the Gash model, and mismatches between data needs and data availability, such as radiation budgets relevant for the forest floor. However, the manuscript mostly disregards these important points and the poor by-storm performance and makes a detailed interpretation of the effects of seasonal variation of parameters.
AR: As it was mentioned, we addressed the critical issues related to the modelling assumptions and the need of some important information as net radiation at the forest floor level to improve modelling. The poor daily event performance for the interception processes was highlighted in the discussion of each interception process result (line 271 to canopy; line 298 to trunk; line 339 to forest floor) besides being present in the abstract and conclusions too.
Despite the poorly by-storm performance, as the forest in the study has a pronounced canopy leaf off during the dry season the models should be able to at least indicate the lower canopy interception observed even if it was in a biweekly or monthly time scale. It is important information to evaluate the models’ performance, and it was not verified in woody savannah areas, like Cerrado s.s.
We consider our results corroborate that the interception models present limitations to simulate very low and extremely elevated levels of interception on an individual event scale (Linhoss and Siegert, 2020). By applying seasonal coefficients, we could see that the models can represent the seasonal variation occasioned by the phenology variation (represented by the different storage capacities coefficients). And the lowest performance during summer (lower NSE values than the other seasons) could be another indication of the limitations of the models to represent the interception during intense rainfall events.
R1C-03: Given that the models mostly cannot reproduce per-storm interception loss, and the time-integrated performance is driven by essentially the regression between rainfall and throughfall expressed in the calibrated canopy storage capacity, I do not agree with the conclusions that the modeling is useful for understanding canopy processes. These are not unique problems for this study and field experience is valuable for developing the next generation of models, but the modeling presented in this manuscript is not well described or developed.
AR: Our conclusion was that "
“[…] the adapted models applied here are valuable to modelling the interception processes in Neotropical savannas like Cerrado s.s. forests. The models had problems to simulate the evaporation processes at daily time-scale, but not to others processes. Both models are suitable to estimate the total interception at monthly basis and could be used to inter annual analysis, but for seasonal differences in Cerrado s.s. forests the Rutter model seems to be more appropriate.”
In fact, despite not to be able to simulate well at a daily event-scale, the models had a good performance at a monthly scale and can be useful to estimate the total interception annually.
Regarding model development, our intent was to maintain the concepts and assumptions of the original models. As there is not a Gash model version that includes the forest floor interception, and it is the first time we have forest floor interception direct measurements in a forest in South America, we thought the first step was to work with the two more used interception models.
R1C-04: This manuscript needs a thorough edit for English: there are many errors in word choice, grammar, and spelling.
AR: We will do a thorough English language check in the revised version.
R1C-05: L34 these are rainforest numbers that are not likely relevant to the field site here
AR: Thank you for your comment. Our intention was to highlight the high feedback values in tropical regions, as these are close to our study site. But we agree that it will be better to add more specific information, regarding the same forest type. So we suggest changing the sentence in lines 34-35 to “In tropical savanna, the feedback to the atmosphere can be high (Oliveira et al., 2015; Anache, 2018), reaching up to 82% of annual rainfall through evapotranspiration process (Cabral et al., 2015).”
R1C-06: L74 The logic of the theoretical work should be justified here. The objective based on the study site, L72, is not as important as the theoretical implications of behavior of the model, and there are modifications to the basic models that are not explained here at all. Similarly, the model introduction L55-64 should focus on the shortcomings of existing models that need to be overcome, not on basics.
AR: We agree that there are some models’ issues that should be overcome. But our main objective here was not to show that we can model interception for a Cerrado forest. Our aim was to extend the existing Rutter and Gash models by adding the forest floor interception and verifying the models’ performance to simulate seasonal variations as it is pronounced for the Cerrado s.s. forest.
R1C-07: L94 15,000 stems per ha? This is 1.5 trees per square meter; is that correct?
AR: Yes, it is correct. The density and richness of species in the study area were evaluated by Reys et al. (2013).
R1C-08: L98 What use is PAR for this study?
AR: This was additional information that can be removed.
R1C-09: L114 why “also”?
AR: It was because at the beginning of the item “2.2. Experimental setting” was mentioned some monitored variables recorded each 10 min including precipitation. We can rewrite this item to explain better that all automatic measurements were recorded each 10 min.
A suggestion of alteration to this part of the text is present in the answer to the comment R1C-11.
R1C-10: L119 according to DBH, but by what criteria and why?
AR: Indeed, this information is missing in the text. A suggestion of alteration to this part of the text is present in the answer to the comment R1C-11
R1C-11: L120 this reference does not provide details of this device. A few critical details about how a LID works are needed for this part to be understandable
AR: Hereby follows the revised item 2.2.
“This study comprises the monitoring period between June 01st of 2017 and February 06th of 2020. Precipitation, temperature, relative humidity, wind velocity, net solar radiation, and others environmental variables were collected in the site through a meteorological tower of 11 m of height. In case of missing data of the main monitoring weather station, we used the nearest available meteorological data from site 1 (Anache et al., 2019), which was 1.7 km away from our site.
Figure 1. Location of the study area (from ©Google Maps, 2022).
To investigate the interception process in the Cerrado s.s., canopy, trunk and forest floor interception were measured. Canopy and trunk interception were indirectly determined by the difference between the rainfall (𝑃𝑔) and throughfall (𝑇𝑓) and stemflow (𝑇𝑠), respectively. By including forest floor interception, the total forest interception is measured by the difference between the 𝑃𝑔 and the infiltration (𝐹), as in Eq. (1).
𝐸𝑖,𝑐+ d𝑆𝑐d𝑡+𝐸𝑖,𝑡+ d𝑆𝑡d𝑡 +𝐸𝑖,𝑓+ d𝑆𝑓d𝑡=𝑃𝑔−𝐹 (1)
where 𝑆𝑐, 𝑆𝑡 and 𝑆𝑓 are the storage capacities (mm) of the canopy, the trunks, and the forest floor, respectively, and 𝐸𝑖,𝑐 , 𝐸𝑖,𝑡 , and 𝐸𝑖,𝑓 are the evaporation from these components (mm) in a certain period of time (𝑡).
𝑇𝑓 was measured through four pluviographs of 0.254 mm resolution. Additionally, three gutters linked to the pluviographs (tipping bucket resolution of 0.048 mm or 0.029 L) were used and five additional gutters were directly connected to reservoirs for manual measurements after accumulated events (> 15 mm). For 𝑇𝑠 monitoring, we installed three automatic collectors. Plastic hoses were wrapped around six trees (two per each collector) at breast height that channeled the 𝑇𝑠 through pluviographs (tipping bucket volume of 5 mL) to reservoirs Selected trees to be monitored were divided into three groups according to the DBH (DBH ≤ 10 cm, 11 cm≤ DBH≤ 30 cm e DBH ≥ 31 cm). In addition, 𝑇𝑠 was measured through more 12 manual collectors installed by Oliveira et al. (2015). More details about the equipment are given in Table 1.
Forest floor interception and infiltration were measured through LIDs – Litter Interception Devices. This device allows the continuous direct measurement of the weight (1 g of resolution) of a forest litter sample in situ and at ground level. Moreover, it also allows infiltration monitoring through a pluviograph (5 ml of tipping bucket resolution) installed inside the device, below the tray wherein the litter sample takes place (Rosalem et al., 2019). We installed three of these weighing LID devices, but due to problems with a load cell only the interception records of two LIDs were used. The LIDs were installed close to the meteorological tower and were filled with quasi-undisturbed litter samples of the thickness of about 6 cm. The forest litter samples were changed each August considering its high decomposition rate in dense Cerrado areas (half-life for the decomposing material around 1.8 year) (Cianciaruso et al., 2006). All automatic measurements were recorded each 10 minutes.”
R1C-12: L123 But there was only one August during the study, correct?
AR: No. The monitoring was conducted between June 01st of 2017 and February 06th of 2020, which included three August months. But, as the first samples were taken in June, they were no changes in the first August.
R1C-13: L137 upper envelopes and linear regressions are not the same. Stem storage capacity has generally not been estimated by this technique in the past for several important reasons. This needs justification and explanation.
AR: The sentence in line L137 was not so clear. The applied approach was the same as explained in the previous sentence (L136). To our knowledge, interception modelling studies do apply this approach (Távora and Koide, 2020; Deng, 2020; Gerrits et al., 2010). We followed the procedures recommended by Rutter et al. (1971) and Gash and Morton (1978), and used the mean method (Klaassen et al., 1998) to get the Sc, St, and Sf parameters.
The description of the methodology to obtain the parameters, presented between L134-L141, will be re-written as:
“The storage capacities parameters were determined following the procedures by Rutter et al. (1971) and Gash and Morton (1978), applying the mean method (Klaassen et al., 1998; Gerrits et al., 2010; Sadeghi et al., 2015). They are determined by the negative intercept of the upper envelop scatter diagram between observed values of an input process (e.g., Pg) and its subsequent process (e.g., Tf) (Robins, 1969, apud (Rutter et al., 1971). The partitioning factors of the rain, free throughfall (p) and trunk input (pt), were determined by the slope of the lower envelop line for the same scatter diagram. Thus, the 𝑆𝑐, p, 𝑆𝑡, and pt parameters were determined using Pg and Tf records, and Pg and Ts records, respectively (Távora and Koide, 2020; Deng, 2020). By the same procedure, the storage capacity of the forest floor (𝑆𝑓) and the partitioning factor of the preferential flow (pf), were determined through Tf and F values (Gerrits et al., 2010). These analyses were carried out based on independent rain storms (each one preceded by at least 24 hours without rainfall) out the total of 236 rain days during the calibration period.”
R1C-14: L150 These two ad-hoc modifications to the Rutter model are not sufficiently described or justified. There is no way to understand exactly what "priority" means in terms of water and energy balances.
AR: When Rutter’s model is run, the first water balance happens in the canopy wherein the volume drained will join to free throughfall, forming the throughfall. Similarly, for the trunk component, we have to choose which process is going to be the first to remove water from the reservoir (i.e., through evaporation, drainage, or stemflow). So, the term “priority” was used in the text to explain that we choose to first calculate evaporation (based on the storage) whereafter we calculate the drainage, stemflow, or infiltration (to the forest floor reservoir). This ordering of calculation steps was also done by Gerrits et al. (2010). By choosing to prioritize the evaporative process, we were trying to minimize the low evaporative rate during the rainfall (L281). Besides, it indirectly could account for other processes, such as splash evaporation (Murakami, 2006; Bassette and Bussière, 2008) and vertical updrafts during rainfall (van Dijk and Bruijnzeel, 2001).
R1C-15: L152 What about the Tf records indicate dynamic storage?
AR: Yes. This is what we wanted to say in L152 by saying “These observations”. They mean the Tf records.
R1C-16: L153 this term needs explicit definition. I do not know what Cc is supposed to represent
AR: We respond to this comment in the following comment response (R1C-17).
R1C-17: L156 What is the purpose of Ccmax exactly and how was it implemented? Rutter's model explicitly allows storage C to exceed storage capacity S, with no explicit upper limit. Adding a cap here appears to reduce the dynamic storage, not enable it. Why only "before the end of the storm"?
AR: Cc has the same meaning as C in Rutter’s original model, which represents the amount of water on the canopy. We added the c minor just to specify that this amount was in the canopy. In the text (L153) we said “[..] some dynamic storage on the canopy (Cc) is present.”, declaring the Cc meaning in the text. However, to avoid confusion, we will rewrite L153-L157."
The purpose of Ccmax was to allow a greater water amount to remain in the canopy among time steps of model’s running. If the Ccmax is not added as a maximum threshold, the difference Cc – Sc will be drainage and only Sc remains in the next step. And this maximum dynamic storage (Ccmax) parameter also allows to haven’t water amount physically impossible on the canopy during the rain event.
“These observations show that apparently some dynamic storage on the canopy is present. The results of Aston (1979) and Lloyd et al. (1988) agree with this, and the authors recommended, as Valente et al. (1997), that a limiting parameter should be used to prevent a build-up of water on the canopy. Therefore, besides the dynamic storage parameters, here specified as Cc, Ct, or Cf, to canopy, trunk, and forest floor, respectively, not only C as in (Rutter et al., 1971), we added a maximum dynamic storage parameter (𝐶𝑐𝑚𝑎𝑥) as a threshold to the retained water amount on the canopy before the end of the rainfall storm. Klaassen et al. (1998) observed by using microwave transmission to measure the water storage on the canopy, that during rain events the dynamic storage is affected by rain intensity in such a way that could lead sometimes to higher dynamic storage along the rain than at end of the rain event. For the original model, the maximum water amount on the reservoir is the storage coefficient. So, the Ccmax was necessary to allow greater water amount than Sc before the end of the rainfall, and to prevent a build-up on the canopy.”
R1C-18: L159 what is Cmax. the same as Ccmax?
AR: Yes. It was a typing error. It should be Ccmax.
R1C-19: L160 I don't understand this parameter identification technique
AR: I think you are asking about the representation meaning of Cc< Sc. This represents the amount of water on the canopy, which is below the storage coefficient amount. For this condition, we said that besides the evaporation, the drainage could still happen by the leaves shaking off (Gerrits et al., 2010), and it was modelled using the same exponential equation as proposed by (Rutter et al., 1971) to the drainage process.
R1C-20: L176 I do not understand how data were used to estimate Cfmax
AR: Ccmax and Cfmax represent the maximum dynamic storage on the canopy and the forest floor, respectively. Since we have continuous monitoring of the water content in the LIDs’ samples, the Cfmax is found by the peaks of water retention during the rain events.
The forest litter amount varies along the year due to decomposition and the litter input, which causes some shifts in the amount of litter that can be noted in the time series of water content. Thus, we selected the higher peaks observed among these shifts to the two time series used (of LID 1 and LID 2), and used the average value as Cfmax.
R1C-21: L179 Most of the parameters referred to in this sentence are not defined
AR: Is and f are the exponential coefficients used to simulate the drainage through the forest litter sample that will contribute to the infiltration when the dynamic storage of the forest floor (Cf) is higher than the storage coefficient (Sf) and lower than the maximum dynamic storage of the forest floor (Cfmax). As mentioned in L179, Is and f are just exponential coefficients, like Ds and b used in the Rutter model to simulate the drainage process on the canopy.
R1C-22: L185 I can't find where these parameters were estimated
AR: How the parameter Sf is obtained was presented in L138. However, in the response to comment 13 (R1C-13), we proposed a revision of our text between L134 and L141 that it will make clearer the approach and add how the partitioning coefficients (p, pt, and pf) were obtained.
R1C-23: L199 something is wrong. This describes an experiment 8 months long but fig 3 says 3 years
AR: Yes, there is a typing error in L199. It was supposed to be “From June of 2017 […]”, such as the monitoring period mentioned in L100.
R1C-24: L204 The original Gash model was applied at the monthly time scale. There is no time scale assumed by Gash
AR: Indeed in the first paper with the Gash model (Gash, 1979) the model was applied at the monthly time scale and there is no time scale assumed by Gash. But it can be run on a daily time-step (Valente et al., 1997), by assuming the occurrence of a single rainfall event per rainfall day. And it is hard to select a time-step lower than a day due to the “drying phase” that should last from the end of the rainfall until the canopy and trunks are completely dry.
We suggest changing the sentence in L204, since it gives the idea that the Gash model is applied only at a daily time scale. We suggest the following change in the text (L204):
“Because the Gash model was applied at a daily time-scale, the Rutter model results were evaluated a minimum on a daily time-scale, thereto the models’ results could be compared.”
R1C-25: L214 This conflicts slightly with L145. I assume this means stomatal resistance was zero. There are no details about assumptions and about parameters relating to aerodynamic resistance
AR: Yes, you are right. We calculated Ep through the Penman equation, and not through the Penman-Monteith equation. The two are equal in case we take a stomatal resistance of zero. To be consistent we will say that we calculate Ep with Penman.
R1C-26: L220 Fig 3: L218 told us to expect net radiation here
AR: One sentence should be added here to explain what is going to be presented in Figure 3. We suggest the following sentence:
“The observed relative air humidity and net radiation corresponded on average to 66.8 (± 20 %) and 154.6 (± 239 W m-2), respectively. The daily potential evaporation and income radiation along the monitored period are presented in Figure 3”.
R1C-27: L224 more similar than what?
AR: Thank you for correcting us on this. The word “more” has to be removed from this sentence.
R1C-28: L242 Why does the Gash model predict essentially the same canopy evaporation every storm except for very small storms?
AR: As the Gash model was run at a daily timescale the canopy storage refill is not simulated. And, as most storms had precipitation greater than the 𝑃′𝑔 (the amount of water necessary to saturate the canopy) the canopy evaporation reached this threshold (𝑃′𝑔) because the model also assumes that the whole daily potential evaporation value is used to dry the canopy and trunk.
R1C-29: L264 I think it is unlikely that measurements of forest-floor evaporation were better than the throughfall measurements.
AR: In our view, in this case, it could be possible due to the greater collection area of the LID (0.16 m² each LID) than the pluviographs collecting throughfall (0.0214 m² each pluviograph), and the high sensitivity of the load cells used (0.001 kg). However, it is possible that the main cause for the overestimation is being fixed throughfall gauges.
R1C-30: L265 It is not obvious to me that canopy evaporation is modeled better.
AR: In L265, we explained why we couldn’t use the Ec indirect estimations considering the forest floor evaporation measurements (𝐸𝑐 =𝑃𝑔−𝐹−𝑇𝑠−𝐸𝑓) to compare the daily modelled Ec. Despite the difference in Ec indirect measurements commented on in the previous paragraph (L260-264), the Ec indirectly observed by using the automatic throughfall records had to be used to analyze the daily time scale performance.
R1C-31: L280 As described earlier, it did exactly the opposite: it prevented accumulation
AR: We couldn’t understand exactly what you meant here, but as it is related to Ccmax, maybe you will have a different opinion after our more detailed explanation in R1C-17 and R1C-20.
R1C-32: L284 How? Shouldn't mean daily Rutter be the same as daily Gash?
AR: By the Gash model assumption that “the meteorological conditions prevailing during any wetting-up of the canopy are sufficiently similar to those prevailing for the rest of the storms”, the rate factor is used in the calculus. When the potential evaporation is used, like in Rutter model, there is a decrease in the potential evaporation during the storm. In our study, we observed that it happened mainly due to the decrease in income radiation during the storm. It is worth mentioning that 43% of the rain event occurred during the night. This may cause a greater difference between the daily predictions by these models since the interception ratio (interception/precipitation) in the first half of the rain event has shown to be greater than the ratio in the second half (Iida et al., 2017).
Another crucial point is that the Rutter model allows to simulate the refill of canopy storage capacity if more than one rain event occurs during the day, which is not possible for the Gash model.
As many rain events occurred during the night, the models will have different responses at a daily time scale.
R1C-33: L344 This problem is severe for the modeling assumptions and needs detailed theoretical discussion
AR: The answer to this comment is presented in the following AR (comment R1C-34).
R1C-34: L348 I do not understand this sentence
AR: Actually, prioritizing the canopy and trunk evaporation over the forest floor evaporation was not an assumption. It was not our intention to reformulate the original model, instead, our intention here was to add the forest floor component to the original model. So, what we did was to add this component and change it as minimum as possible the original structure. We believe that seeing the possible issues for including the forest floor in the model, will allow us and others to improve the measurements needed (e.g., energy budget at soil level in the forest) and subsequently the modelling.
Thus, we said in L344-L347 what is known about the different weather conditions between the above and below the canopy. And, despite the potential evaporation being lower below the canopy, the approach modelling applied caused delays onset of the forest floor evaporation process.
As you said you didn’t understand the sentence in L348, here it is a suggestion to change the paragraph (L343-L348).
“Due to the average rate factors (𝐸̅/𝑅̅), the Gash model could simulate higher values than the Rutter model during the rain. In our adapted Rutter model, the canopy and trunk evaporation are prioritized over the forest floor evaporation. As the 𝐸𝑝 below the canopy is lower than above, and due to different weather conditions near the understory (Coenders-Gerrits et al., 2020). In addition, the forest floor and the canopy differ in physical structure, resource availability and biotic conditions (Yanoviak and Kaspari, 2000), which implies different energy and water fluxes. Hence, aiming to not change the original structure of the Rutter model, the remained Ep was used to modelling the forest floor evaporation. So, the canopy and trunk evaporation process were dominant over forest floor evaporation, which caused delays onset on the forest floor evaporation for this adapted version.”
R1C-35: L353 I agree. I think this makes the Gash model irrelevant for this process. Lack of data to estimate energy budgets and vapor exchange at the scale of the forest floor are also major problems for applying a Rutter model. Given that Rutter does not seem to be helpful for estimating canopy evaporation on a storm scale, it seems even less likely to be helpful for forest-floor evaporation.
AR: For the Gash model, by including the forest floor interception component maybe it will be better to not consider the daily time scale. Also, a modification added by (Valente et al., 1997) in their reformulated version for sparse forests to canopy evaporation should be considered “Although evaporation from the saturated trunks is not used explicitly in this model it is subjacent to the calculation of interception loss from the trunks. Thus, a small modification was made to the model of Gash et al. (1995) so that Ec is now replaced with (1 - e) Ec.”. So, besides taking water residence time into consideration, further improvements to Gash model for non-sparse forests should also consider the fraction of concern to forest floor evaporation.
We agree that both models present poor storm scale predictions, especially in Tropical areas like Cerrado forests with more intense rainfall events. However, the monthly performance to total interception estimates for these models was good (NSE: 0.63 – 0.85 and R²: 0.78 – 0.97), and good accumulated interception estimates for the Rutter model to the almost 3 years of monitoring presented in Figure 5 in the text (underestimation of 11.5%, L417). In Fig 5 is also possible to see that the total volume estimated for the forest floor interception was particularly good for both models, varying between -2% and 9% when seasonal coefficients were applied.
R1C-36: L397 I don’t think the evaporation model was useful for small storms, either. There was not systematic negative bias in estimates, so that means the small storms were overestimated.
AR: The adapted Rutter model had poor estimations on daily basis, even for small storm events. However, the adapted Gash model presented daily canopy evaporation not so bad for small storm events, meaning rainfall lower than 7 mm.day-1 (NSE = 0.65, R²=0.90, MBE=+0.11).
R1C-37: L425 I would agree if the models were useful at short timescales, but they are not. If the processes are not correctly modeled, how can we extrapolate using the model?
AR: In this sentence we were referring to broad sense, meaning that interception modelling could help to elucidate the role of different factors.
R1C-38: L510 I do not agree. The models get major masses correct, but due to calibration (e.g, of canopy storage). The models are not likely to be of any predictive value should, for example, vegetation structure change.
AR: As we said in L511, the adapted models could be used on monthly basis, or inter-annual analysis since the accumulated volumes were not so far from the observed ones to the adapted Rutter model. The throughfall and infiltration were well modeled for both models, even on a daily time scale (NSE: 0.81 – 0.94), which elucidate that the problem remains on the evaporation modelling.
Regarding canopy storage, the coefficient was not calibrated but the maximum dynamic storage (Ccmax, added to the adapted Rutter model) was calibrated. This parameter had a major effect on the throughfall and, consequently, on the infiltration estimates, but didn’t compensate for the problem with low potential evaporation during the rainfall to the canopy evaporation for the Rutter model.
R1C-39: The journal names are missing for many References.
AR: Thanks for the observation. We are going to check it for the final paper version.
References
Anache, J. A. A.: Alterações no ciclo hidrológico e na perda de solo devido aos diferentes usos do solo e variações climáticas em área de Cerrado, Doutorado em Hidráulica e Saneamento, Universidade de São Paulo, São Carlos, https://doi.org/10.11606/T.18.2018.tde-17042018-110107, 2018.
Bassette, C. and Bussière, F.: Partitioning of splash and storage during raindrop impacts on banana leaves, Agricultural and Forest Meteorology, 148, 991–1004, https://doi.org/10.1016/j.agrformet.2008.01.016, 2008.
Cabral, O. M. R., da Rocha, H. R., Gash, J. H., Freitas, H. C., and Ligo, M. A. V.: Water and energy fluxes from a woodland savanna (cerrado) in southeast Brazil, Journal of Hydrology: Regional Studies, 4, 22–40, https://doi.org/10.1016/j.ejrh.2015.04.010, 2015.
Deng, J.: Fitting the revised Gash analytical model of rainfall interception to Mongolian Scots pines in Mu Us Sandy Land, China, Trees, Forests and People, 1, 100007, https://doi.org/10.1016/j.tfp.2020.100007, 2020.
van Dijk, A. I. J. M. and Bruijnzeel, L. A.: Modelling rainfall interception by vegetation of variable density using an adapted analytical model. Part 1. Model description, Journal of Hydrology, 247, 230–238, https://doi.org/10.1016/S0022-1694(01)00392-4, 2001.
Gash, J. H. C.: An analytical model of rainfall interception by forests, Quarterly Journal of the Royal Meteorological Society, 105, 43–55, https://doi.org/10.1002/qj.49710544304, 1979.
Gash, J. H. C. and Morton, A. J.: An application of the Rutter model to the estimation of the interception loss from Thetford Forest, Journal of Hydrology, 38, 49–58, https://doi.org/10.1016/0022-1694(78)90131-2, 1978.
Gerrits, A. M. J., Pfister, L., and Savenije, H. H. G.: Spatial and temporal variability of canopy and forest floor interception in a beech forest, Hydrological Processes, 24, 3011–3025, https://doi.org/10.1002/hyp.7712, 2010.
Iida, S., Levia, D. F., Shimizu, A., Shimizu, T., Tamai, K., Nobuhiro, T., Kabeya, N., Noguchi, S., Sawano, S., and Araki, M.: Intrastorm scale rainfall interception dynamics in a mature coniferous forest stand, Journal of Hydrology, 548, 770–783, https://doi.org/10.1016/j.jhydrol.2017.03.009, 2017.
Klaassen, W., Bosveld, F., and de Water, E.: Water storage and evaporation as constituents of rainfall interception, Journal of Hydrology, 212–213, 36–50, https://doi.org/10.1016/S0022-1694(98)00200-5, 1998.
Murakami, S.: A proposal for a new forest canopy interception mechanism: Splash droplet evaporation, Journal of Hydrology, 319, 72–82, https://doi.org/10.1016/j.jhydrol.2005.07.002, 2006.
Oliveira, P. T. S., Wendland, E., Nearing, M. A., Scott, R. L., Rosolem, R., and da Rocha, H. R.: The water balance components of undisturbed tropical woodlands in the Brazilian cerrado, Hydrology and Earth System Sciences, 19, 2899–2910, https://doi.org/10.5194/hess-19-2899-2015, 2015.
Reys, P., Camargo, M. G. G. de, Grombone-Guaratini, M. T., Teixeira, A. de P., Assis, M. A., and Morellato, L. P. C.: Estrutura e composição florística de um Cerrado sensu stricto e sua importância para propostas de restauração ecológica, Hoehnea, 40, 449–464, https://doi.org/10.1590/S2236-89062013000300005, 2013.
Rutter, A. J., Kershaw, K. A., Robins, P. C., and Morton, A. J.: A predictive model of rainfall interception in forests, 1. Derivation of the model from observations in a plantation of Corsican pine, Agricultural Meteorology, 9, 367–384, https://doi.org/10.1016/0002-1571(71)90034-3, 1971.
Sadeghi, S. M. M., Attarod, P., Van Stan, J. T., Pypker, T. G., and Dunkerley, D.: Efficiency of the reformulated Gash’s interception model in semiarid afforestations, Agricultural and Forest Meteorology, 201, 76–85, https://doi.org/10.1016/j.agrformet.2014.10.006, 2015.
Távora, B. E. and Koide, S.: Event-Based Rainfall Interception Modeling in a Cerrado Riparian Forest—Central Brazil: An Alternative Approach to the IS Method for Parameterization of the Gash Model, Water, 12, 2128, https://doi.org/10.3390/w12082128, 2020.
Valente, F., David, J. S., and Gash, J. H. C.: Modelling interception loss for two sparse eucalypt and pine forests in central Portugal using reformulated Rutter and Gash analytical models, Journal of Hydrology, 190, 141–162, https://doi.org/10.1016/S0022-1694(96)03066-1, 1997.
Citation: https://doi.org/10.5194/hess-2022-59-AC1
-
AC1: 'Reply on RC1', Livia Rosalem, 15 Nov 2022
-
RC2: 'Comment on hess-2022-59', Anonymous Referee #2, 28 Jun 2022
Comments on "Water partitioning in a Neotropical Savanna forest (Cerrado s.s.): interception responses at different time-scales using adapted versions of the Rutter and the Gash models" submitted to HESS-D by Rosalem et al.
[General comment]
This paper measured the interception processes including forest floor interception loss for a period of 32 months, and applied Rutter and Gash models. Although the methodology used in this study should be evaluated carefully, I recognized that the efforts to conduct measurements would be very intensive, and would agree the importance to understand interception phenomena in Cerrado region. However, the logic of this manuscript is not constructed well, so readers cannot understand what is the most important findings in this research. The main description seems to be the model performances for daily, biweekly and monthly basis. I understood the differences in performance among different time scales, but I do not think that the differences are important to clarify the interception process. The important thing must be what factors resulted in the differences and what processes affect the model performance. I would like to recommend that authors, at first, write a paper to show interception process in this site in detail based only on measurement data without applying any models. The interception loss in this site would be affected by rainfall characteristics (e.g., rainfall duration, rainfall amount, rainfall intensity, etc)? Or, other micrometeorological factors (e.g., wind speed, vapor pressure deficit, net radiation, etc) influence the interception process? Based on these knowledges, I believe that authors could develop suitable models to simulate the interception process. I hope my comments will help to do substantial revisions.
[Specific comments]
Could you add the LAI data? (line 92-93)
I cannot catch up how to calculate the spatially representative amount of throughfall. Is that the average of four Davis, five manual gutters and three automatic gutters? (line 100-119)
Because the canopy in this site is discontinuous (line 93), I could expect that the spatial heterogeneity of throughfall would be very high. Please show the differences in throughfall amount measured by four Davis, five manual gutters and three automatic gutters. The differences are related to the canopy openness? Also, a total of 12 measurements of throughfall are safely enough to obtain spatial representative value?
As Rosalem et al. (2018), published in Ecohydrology, pointed out, Davis gauge underestimates the inflow of water flux with increasing intensity (please see FIGURE 3 in Rosalem et al., 2018). I would like to confirm that authors applied the same correction to throughfall measurements in this paper. If not, application must be required.
Three gutters are connected to three Davis gauge? If so, I am wondering that the one tip amount of 0.048 mm is too small to detect the correct amount. As Rosalem et al. (2018) showed, the underestimation by Davis gauge is relatively high. How many pulses generated by the gauge were recorded in 10-min intervals? The time between tips, equivalent to 600 second divided by accumulated pulse count, should be more than 1.0-1.5 second. Please note that, if authors used other rainfall gauge, similar issue exists and should be investigated.
How did authors calculate stemflow amount in the stand scale?
Three automatic collectors of stemflow is connected to Davis gauge? The same issue mentioned above, underestimation of inflow with increasing intensity, must be checked.
Forest floor interception was measured by two LIDs (line 121-122). Please show the evidence indicating two LIDs could safely measure the spatial representative value of forest floor interception. This is very critical, because high spatial heterogeneity of throughfall could be expected from the disconnected canopy in this site.
Figure 2: The forest floor evaporation was calculated considering potential evaporation (Ep), but how did you calculate Ep above forest floor? Did you measure net radiation above the forest floor?
Please add the description to explain how to calculate the aerodynamic conductance above canopy and forest floor (line 214-215).
How did you obtain the interception ratio of 33%? (line 253) Throughfall was described as 70-72% (line 228), so 33% interception is too large. There is a description of 40% interception in conclusion section (line 481). Maybe the target rainfall events are different among parts, but it is difficult to understand.
I cannot understand why Ec is calculated as the difference Pg and the sum of F, Ts and Ef. I recommend that basic equation showing rainwater balance should be added in the M&M section.
Discussion about the stemflow is not directly related to this paper (line 300-313). If authors show the data of canopy structure, bark, and so on, it is useful. Unfortunately, the current MS did not include any data, so I recommend to remove this part.
I felt that the much sentences in the current MS are related to model performance (line 237-245, 268-273, 293-299, 320-328, 335-342, 368-380, 435-454, Table 6, 7, 8, 9). Similar descriptions are found among parts, so I recommend reconstruction of the logic. In my opinion, it would be better that "discussion" should be separated from the "result" section. Then, descriptions of model performance should move to discussion section, and more concise discussion is recommended. Rather than differences in model performance, the reason for the difference and factors affecting it are more important to understand the interception process in this site.
Looking at appendix C, there are high correlations between observation and model output for throughfall and stemflow. However, the correlation for interception loss is very low. Could you explain this point?
[Technical corrections]
Line 98: "average PAR (photosynthetic active radiation) of 1041.8 ± 427.4 μmol.m-2.s-1", I cannot understand the duration of average.
Line 199-200: "From June of 2019 to May of 2019, we used to calibrate our Rutter and Gash models, while the second part, from June of 2019 up to of 07th February of 2020, .." Please check the consistency of months. The current description is strange.
Figure 3, y-axis title of the upper panel: "Potencial" should be "Potential".
Figure 3, lower panel: Is this net radiation? Solar radiation was measured at 2 m height (Table 1). More than 600 W m-2 value is too high for solar radiation above the forest floor. Please check carefully.
Citation: https://doi.org/10.5194/hess-2022-59-RC2 -
AC2: 'Reply on RC2', Livia Rosalem, 15 Nov 2022
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-59/hess-2022-59-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Livia Rosalem, 15 Nov 2022
Data sets
Dataset of "Water partitioning in a Neotropical Savanna forest (Cerrado s.s.): seasonal and non-seasonal responses at different time-scales using adapted versions of the Rutter and the Gash models Lívia Rosalem https://www.hydroshare.org/resource/9134e6dc7cc94a999ee005966d0399f5/
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,172 | 292 | 47 | 1,511 | 54 | 54 |
- HTML: 1,172
- PDF: 292
- XML: 47
- Total: 1,511
- BibTeX: 54
- EndNote: 54
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1