the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Coupled water-carbon modelling in data-limited sites: a new approach to explore future agroforestry scenarios
Abstract. Agroforestry is considered an important strategy for mitigating against, and adapting to, climate change. Questions yet remain regarding the potential impacts of different tree species on water/carbon cycling at different locations, scales and under different climatic conditions. There is an urgent need for numerical models capable of quantifying agroforestry impacts on a host ecosystem services including carbon sequestration and soil water/river flow regulation. A key challenge in modelling agroforestry systems is that they depend heavily on soil moisture as the main driver of many biogeochemical processes. Soil moisture itself is highly variable with soil properties (and therefore with location) but also with depth. Given that target sites for agroforestry are often ungauged, location-specific agroforestry modelling must inevitably rely only on data available from satellites and/or nearby weather stations which do not typically cover the subsurface, i.e., there is an incommensurability between data-availability and system complexity. To overcome this, we propose RSEEP, a new ecohydrological model that only requires rainfall, potential evapotranspiration, and surface soil moisture for its calibration. We demonstrate RSEEP’s capability in water cycling for a site in Scotland where soil moisture observations are available for different depths and vegetation types. We then couple RSEEP to the well-known RothC soil carbon model to (i) test RothC’s sensitivity to water cycling method, and to (ii) simulate water-carbon dynamics of three different silvo-pastoral agroforestry systems (all at 400 stems/ha density) in Scotland; these systems are: with evergreen conifer (Scots Pine), deciduous conifer (Hybrid Larch), and deciduous broadleaf (Sycamore) trees. We find that not including more accurate soil moisture accounting methods in RothC can significantly overestimate soil carbon stocks. Under the current future climate pathway (RCP6.0), 40 years after planting trees, above+below ground carbon storage can be 2–5 times (100–250 t/ha) higher under silvo-pasture than under pasture depending on species, with Larch having the highest potential and Sycamore the lowest. Larch also exhibits the highest potential for preserving soil moisture under drier conditions, but Pine shows the highest potential for river flow regulation under both wet and dry conditions at our site. The choice of species is therefore important and should be made site-specifically and based on the ecosystem service and management priorities/objectives. Examining our scenarios under drought- and flood-relevant conditions and scales is a logical next step.
- Preprint
(9257 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2258', Anonymous Referee #1, 30 Sep 2024
The manuscript deals with a relevant topic -- but in its current form it doesn't quite live up to the expectations raised. Please reconstruct after further attempts to balance the hydrological and soil C components -- the interest is is a substantial range of agroforestry systems, rather than the specific data used for model calibration in the current manuscript. Positioning the new work within the existing agroforestry modelling literature can help.
1. From the title "Coupled water-carbon modelling in data-limited sites: a new approach to explore future agroforestry scenarios" I expected an analysis of the water-carbon coupling at coarser space/time scales, rather than a patch-level daily-time-step model. A Budyko analysis of rainfall, ET and discharge at annual time steps, linked to a water use efficiency concept and relative allocation to woody biomass, annual leaf turnover and soil processing of annual necromass inputs might be more aligned to data availability for current situation plus climate change scenarios. Possibly linked to 'flow persistence' analysis of the dynamics of streamflow data (e.g. https://hess.copernicus.org/articles/21/2321/2017/). Current discussion (5.3.3) on annual water yield vs food risk is rather limited.
2. A more 'macro' modelling effort might operate at an annual time step, but consider the expected number of days that drought or water excess limits soil carbon turnover rates (in the model the specific temporal sequence does not matter for the net effect), but for a same annual precipitation rate the frequency of dry periods can vary with rain intensities -- some climate change models predict changes in intensity/frequency rather than annual total rainfall (line 630). A clearer analysis may help here judging uncertainty embedded in the approach chosen.
3. It would be good if the manuscript could connect to the concept of multiple levels of 'control' over soil Carbon -- e.g. SOMpotential (texture, minerology), SOMattainable (Radiation, Climate) and SOMactual (vegetation, C-input, soil management, weather) as in Batjes, N.H., Ceschia, E., Heuvelink, G.B.M. Demenois, J., Le Maire, G., Cardinael, R., Navarro, C. A., and van Egmond, F. (2023). International review of current MRV initiatives for soil carbon stock change assessment and associated methodologies (Version 1). Deliverable 4.1, EU-ORCaSa project. ISRIC - World Soil Information, INRAE and CIRAD. https://doi.org/10.17027/isric-32Q1-2F50
4. For a P of around 1400 mmm y-1 and Epot of 350 mm (Fig. 1), the occurrence of drought conditions is not immediately obvious -- a clearer annual cycle representation may help here.
For potential evapotranspiration global databases exist -- could be used to compare with your own estimates: Zomer, R.J., Xu, J., Spano, D. and Trabucco, A., 2024. CMIP6-based global estimates of future aridity index and potential evapotranspiration for 2021-2060. Open Research Europe, 4(157), p.157.
Line 229 Wouldn't a classical 'field capacity' concept help here?
5. The righthand Y axis for Fig, 1b is presented as ratio of R and Pet, but its units and values show it is not.
6. The allocation of water uptake over soil layers appears rather rigid and crude, while existing AF models do a better job (with more feedback) -- how relevant is it here? There appears to be a discrepancy between the within-day details of canopy interception and its subsequent evaporation and the hydrological output that is only considered as annual water yield. Of course, realistic streamflow predictions need more spatial hillslope representations.
7. RothC process rates are temperature dependent -- soil temp is influenced by plot microclimate and litter layer -- I don't see that discussed.
8. The role of surface litter layers -- especially in systems without no tillage (essentially different from the data sets on which RotC was developed) -- in soil C dynamics does not get the attention it deserves in much of the soil C literature, but this paper doesn't mention that as an issue worth of consideration.
(see e.g. https://link.springer.com/article/10.1007/s11104-021-05279-z, https://www.annualreviews.org/content/journals/10.1146/annurev-environ-112621-083121 )9. There is a few references to root turnover as source of soil C, but further quantification would help.
10. It is questionable to what degree the 400 trees/ha plots can be called 'agroforestry' ("The understory of the Pine and Grass plots are covered with pasture. In the Larch plots, much of the understory is covered by a dense litter layer, but the plots are still used by sheep/cattle for shelter. The understory in the Sycamore plots are characterised by patches of bare ground and litter that vary in extent seasonally") --
Line 202 Eq 7 specifies a surface cover fraction that only responds to LAI, not surface litter
11. Realistic models of tree stands deal with a gradual diminishing tree population while DBH increases, with joint consequences for LAI and basal area. Are you taking this into account?
12. The relationship between the soil water content of the top 6 cm and the whole soil profile depends on soil and site properties -- the calibration phase embeds the site characteristics of the data rich sites inti that process, but we have no idea how representative the sites are for the intended application domain of the resulting model. A deeper analysis of the underlying processes might help here.
13. It would help if there is an alphabetic list of all symbols used
Line 206 Figure 2 Many hydrological models refer to precipitation (P) rather than rainfall (R) -- following that convention makes the model a bit more generalizable (you don't have snow?)
A next version of the manuscript will need a more careful spell check -- avoiding: understorty (line 127)
Citation: https://doi.org/10.5194/egusphere-2024-2258-RC1 -
AC1: 'Reply on RC1', Salim Goudarzi, 29 Nov 2024
GENERAL COMMENTS:
The manuscript deals with a relevant topic -- but in its current form it doesn't quite live up to the expectations raised. Please reconstruct after further attempts to balance the hydrological and soil C components -- the interest is a substantial range of agroforestry systems, rather than the specific data used for model calibration in the current manuscript. Positioning the new work within the existing agroforestry modelling literature can help. Response: Thank you for the careful reading and useful comments. Better positioning the new work within the existing agroforestry modelling literature is a good point which will help provide clearer context. We will expand our introduction to reflect this point. There may have been a misunderstanding regarding the scope of this article. Our focus on the specific data used for model calibration is merely for demonstration purposes, i.e., that, RESEEP, a newly developed model, can produce reasonable predictions using only surface- and above-surface information. Our proposed approach remains applicable to a substantial range of agroforestry systems where little to no subsurface observations are typically available (and for which our approach is designed). Below is a point-by-point response to the specific comments.
SPECIFIC COMMENTS:
Comment #1: From the title "Coupled water-carbon modelling in data-limited sites: a new approach to explore future agroforestry scenarios" I expected an analysis of the water-carbon coupling at coarser space/time scales, rather than a patch-level daily-time-step model. A Budyko analysis of rainfall, ET and discharge at annual time steps, linked to a water use efficiency concept and relative allocation to woody biomass, annual leaf turnover and soil processing of annual necromass inputs might be more aligned to data availability for current situation plus climate change scenarios. Possibly linked to 'flow persistence' analysis of the dynamics of streamflow data (e.g. https://hess.copernicus.org/articles/21/2321/2017/). Current discussion (5.3.3) on annual water yield vs food risk is rather limited. Response: We note that the spatial and temporal scales at which our model is applied in this study do not reflect the range of scales at which our model can be applied. Patch-level model application here is due to the calibration data which was available at this scale and the aim was to demonstrate our model’s performance. Our 1D approach can be applied at any spatial scale. Regarding the temporal scale, while many studies use annual timesteps, we use finer/daily timesteps, because (1) our simple model structure allows running much finer steps at low CPU cost; (2) Satellite climate datasets today are increasingly available at daily steps. Thus, at least in our case, the loss of accuracy that results from coarser timestepping does not justify the computational savings that it offers. Finally, a flow persistence analysis, as described by van Noordwijk et al. (2017), requires river flow estimates, which is a quantity that our model, being 1D, cannot generate in any real sense (because, as you rightly point out in comment#6, realistic streamflow predictions need more spatial hillslope representations, which is why in this article we refer to “soil water yield” as a proxy for flow, rather than flow itself). In the next version, we will make this point clearer in section 5.3.3, that: although we briefly discuss implications for stream flow, this is more to suggest what may be possible based on our limited results and model structure, and that more elaborate modelling with detailed hillslope representation should be used to explore these possibilities in future studies.
Comment #2: A more 'macro' modelling effort might operate at an annual time step, but consider the expected number of days that drought or water excess limits soil carbon turnover rates (in the model the specific temporal sequence does not matter for the net effect), but for a same annual precipitation rate the frequency of dry periods can vary with rain intensities -- some climate change models predict changes in intensity/frequency rather than annual total rainfall (line 630). A clearer analysis may help here judging uncertainty embedded in the approach chosen. Response: This is a valid point that an intensity-duration-frequency (IDF) analysis is generally more informative than studying annual water balance. However, note that our main objective here has been to compare the different scenarios relative to one another at a given time (e.g., after 40yrs) meaning that the frequency of dry periods will be the same for that time period. Comparisons between two time periods (e.g., +20 yrs and +40yrs) is not relevant to the conclusions we draw from our analysis. Having said that, this is an important point which readers should be made aware of, and we will highlight it in the next version of the manuscript. Specifically, we will clarify in the revision that a lack of IDF analysis is likely impacting any comparison made between two different time periods, because changes in annual PPT/PET values may not fully reflect the intensity and/or duration of dry/wet periods.
Comment #3: It would be good if the manuscript could connect to the concept of multiple levels of 'control' over soil Carbon -- e.g. SOMpotential (texture, minerology), SOMattainable (Radiation, Climate) and SOMactual (vegetation, C-input, soil management, weather) as in Batjes, N.H., Ceschia, E., Heuvelink, G.B.M. Demenois, J., Le Maire, G., Cardinael, R., Navarro, C. A., and van Egmond, F. (2023). International review of current MRV initiatives for soil carbon stock change assessment and associated methodologies (Version 1). Deliverable 4.1, EU-ORCaSa project. ISRIC - World Soil Information, INRAE and CIRAD. https://doi.org/10.17027/isric-32Q1-2F50. Response: Yes, it could be very useful to some readers, and we might consider this for future publications, but this is outside the scope of the current article, which is to introduce a different approach to modelling agroforestry. We will, however, refer to your suggested article in our ‘future work’ section for the interested readers.
Comment #4: For a P of around 1400 mmm y-1 and Epot of 350 mm (Fig. 1), the occurrence of drought conditions is not immediately obvious -- a clearer annual cycle representation may help here. For potential evapotranspiration global databases exist -- could be used to compare with your own estimates: Zomer, R.J., Xu, J., Spano, D. and Trabucco, A., 2024. CMIP6-based global estimates of future aridity index and potential evapotranspiration for 2021-2060. Open Research Europe, 4(157), p.157. Line 229 Wouldn't a classical 'field capacity' concept help here? Response: Regarding clearer annual cycle representation please refer to our response to comment#2. We estimate potential evapotranspiration values using climate variables included in CHESS-SCAPE, a high-resolution dataset for the UK, based on the UKCP18 dataset (UKCP18 is the most widely used climate dataset in the UK). Robinson et al. (reference below), whom we cite, has already provided a detailed comparison with global datasets. In the next draft of the manuscript, we will explicitly refer the readers to this article regarding comparison with global datasets.
REF: Robinson, E. L., Huntingford, C., Semeena, V. S., & Bullock, J. M. (2023). CHESS-SCAPE: High resolution future projections of multiple climate scenarios for the United Kingdom derived from downscaled UKCP18 regional climate model output. Earth System Science Data Discussions, 2023, 151. 2.2.
Comment#5. The righthand Y axis for Fig, 1b is presented as ratio of R and Pet, but its units and values show it is not. Response: Thanks for spotting this (though in Fig, 1d rather 1b). Apologies for causing this confusion but R/Pet is not a ratio. The slash between R and Pet was meant as an ‘or’ rather than a division. We will change the notation to avoid confusion.
Comment #6: The allocation of water uptake over soil layers appears rather rigid and crude, while existing AF models do a better job (with more feedback) -- how relevant is it here? There appears to be a discrepancy between the within-day details of canopy interception and its subsequent evaporation and the hydrological output that is only considered as annual water yield. Of course, realistic streamflow predictions need more spatial hillslope representations. Response: Root water uptake (RWU) is intentionally represented crudely here compared to other AF models. This is to reduce the number of uncertain (calibration) parameters in the model, in line with our objective of developing an AF model that can be applied in data limited sites. The crude representation is still able to conceptually distinguish between short-rooted (grass) and deeper-rooted species (trees) in terms of their ability to adjust their water source based on water availability (see the explanation given in line 266). Finally, hydrological outputs are not just considered as annual water yield, we also consider seasonal water yield values (see section 5.3.3 and Tables A2 and A3). But we fully agree that realistic streamflow predictions need more explicit spatial hillslope representation, which is beyond the capabilities of our current model. As mentioned in our response to the first comment, this is precisely the reason with refer to our hydrological output as ‘water yield’ and not ‘flow’ to avoid giving the wrong impression of what our model can or cannot predict. In this context, we feel that seasonal values are granular enough given our ‘lumped’ model structure.
Comment #7. RothC process rates are temperature dependent -- soil temp is influenced by plot microclimate and litter layer -- I don't see that discussed. Response: This is a valid point that we will acknowledge in the revision. However, representation of microclimate would have required further parameterising the topsoil storage, meaning that, at the very least, one additional calibration parameter would have been added to our list of uncertain parameters. A more elaborate representation would make sense in well-instrumented sites, but that is rarely the case in agroforestry. We therefore see this as a source of uncertainty pertaining to model structure which we have already acknowledged in section 6 (Sources of Uncertainty): “(ix) We have tried to include the main plant-soil-atmospheric interactions in our model, but we have not tested different model structures/complexities to find the best one.” It is not feasible to list all the possible improvements to our model structure, but we will include the lack of sufficient microclimate representation as an example in the revision.
Comment #8. The role of surface litter layers -- especially in systems without no tillage (essentially different from the data sets on which RotC was developed) -- in soil C dynamics does not get the attention it deserves in much of the soil C literature, but this paper doesn't mention that as an issue worth of consideration.(see e.g. https://link.springer.com/article/10.1007/s11104-021-05279-z, https://www.annualreviews.org/content/journals/10.1146/annurev-environ-112621-083121 ). Response: We will acknowledge that we have excluded this process from our modelling and why in section 3.1.1 (Description of RSEEP) (see the previous response). We will also suggest this as an area of future work.
Comment #9: There is a few references to root turnover as source of soil C, but further quantification would help. Response: The limited reference to root C turnover in the submission is commensurate with the level of details with which this process is represented in RothC (which is quite basic). But in the revision, we will highlight this limited representation and suggest that a more detailed representation would be beneficial where possible (i.e., data is available to support it).
Comment #10. It is questionable to what degree the 400 trees/ha plots can be called 'agroforestry' ("The understory of the Pine and Grass plots are covered with pasture. In the Larch plots, much of the understory is covered by a dense litter layer, but the plots are still used by sheep/cattle for shelter. The understory in the Sycamore plots are characterised by patches of bare ground and litter that vary in extent seasonally") -- Line 202 Eq 7 specifies a surface cover fraction that only responds to LAI, not surface litter. Response: This is also a good point which deserves more attention in than it receives in the submission (see paragraph line 569). The UK government considers 100-400 tree/ha range as agroforestry and in fact there are grants available to support farmers to plant at these densities. Further, the observation that little pasture is available at our site 34 years after planting, does not warrant suggesting the exclusion of 400 trees/ha as agroforestry. Observations from more sites would be needed. Additionally, at our site, 400 trees/ha have been planted in 5x5m grid, but they could have been planted in a different pattern, e.g., a densely packed alley, which would have substantially increased light reaching the ground while having similar carbon storage potential. In the next version of the manuscript in section 5.3.1, we will add another paragraph discussing these points with relevant references.
Comment #11. Realistic models of tree stands deal with a gradual diminishing tree population while DBH increases, with joint consequences for LAI and basal area. Are you taking this into account? Response: This is generally more relevant for much higher stand densities. For example, the initial density of UK forests typically range between 2500-3450 trees/ha. In any case, particularly at our study site, field observations show very high survival rate at the 400 density.
Comment #12. The relationship between the soil water content of the top 6 cm and the whole soil profile depends on soil and site properties -- the calibration phase embeds the site characteristics of the data rich sites inti that process, but we have no idea how representative the sites are for the intended application domain of the resulting model. A deeper analysis of the underlying processes might help here. Response: This is a valid good point, and one that we have acknowledged in some detail in section 5.1 (Strengths and weaknesses of RSEEP). There, we show that in sites where RSEEP’s assumptions (of decreasing porosity and hydraulic conductivity with depth) are valid, RSEEP can make very reasonable predictions, otherwise the soil moisture profiles at different depths are not going to be very accurate, although the bulk soil moisture profile (which drives RothC) is likely to be less affected even at those sites.
Comment #13. It would help if there is an alphabetic list of all symbols used. Line 206 Figure 2 Many hydrological models refer to precipitation (P) rather than rainfall (R) -- following that convention makes the model a bit more generalizable (you don't have snow?) A next version of the manuscript will need a more careful spell check -- avoiding: understorty (line 127). Response: Thanks for these suggestions. We did not have snow, in fact this is one of the reasons we chose our study year. However, we can refer to P for precipitation rather than R for rainfall, whilst acknowledging the absence of snow in the revision. In the next version we will also add a list of symbols and perform a more careful spell check.
Citation: https://doi.org/10.5194/egusphere-2024-2258-AC1
-
AC1: 'Reply on RC1', Salim Goudarzi, 29 Nov 2024
-
RC2: 'Comment on egusphere-2024-2258', Anonymous Referee #2, 21 Nov 2024
Review of "Coupled water-carbon modelling in data-limited sites: a new approach to explore future agroforestry scenarios"
Summary
The manuscript presents a soil water model, RSEEP, which when coupled to a carbon model (RothC), should be able to simulate water-carbon dynamics in agroforestry systems in data-limited settings. The authors calibrate and validate RSEEP with two sites in Scotland. They apply the coupled setting to assess the impacts of the silvopastoral agroforestry with three different tree species. Results indicate that soil moisture representation significantly affects carbon stock.
Strengths and Weaknesses
- The introduction establishes the significance of agroforestry as a climate change mitigation and adaptation strategy, supported by up-to-date references.
- The challenges of modeling agroforestry systems, particularly in ungauged or data-scarce regions, are well-articulated in the introduction.
- The justification for developing RSEEP as a parsimonious alternative to complex models is compelling since agroforestry systems are already today in areas with scares monitoring systems and may further expand to other areas with no history of measurements. However, it is a bit surprising that a model designed to be of general character, with few needed inputs and transferable to other regions, is only applied and validated in one region. And this is one of my main criticism: in order to really prove the value of such a model as something new, different (better) than the established, it is necessary to prove its general application (this is at least my opinion).
- The introduction lacks clarity in transitioning between general challenges in agroforestry and the specific aims of the study. Better summarizing how RSEEP uniquely addresses these challenges may improve this.
Methods
- The methodological framework is simple but thorough, with clear explanations of the RSEEP model, including its parameters, calibration, and coupling with RothC. However, please better explain why you only consider the 400 stems/ha site.
- Figure 1 and section 2: please better explain how the two sites are managed. For example, I noticed in fig. 1 that trees are not trimmed in the lower part of the stem. Is there a reason for that? Animals could better use shade and shelter if that would be done. But explain generally how sites are managed (fertilisers, interventions, etc.)
- Fig. 1: explain legend of lower panel in figure caption.
- The assumption of soil saturation during the wettest period is not universally valid and could introduce errors in porosity estimates. Sensitivity to this assumption should have been tested because the beauty of your model (you say) is the potential to be applied at any data scarce region.
- The rationale for specific parameter ranges during calibration is not fully explained, which could raise questions about the model's transferability to other regions. Explain how the ranges make sense (physically).
- I did not review section 3.2 in much detail since or is not my area of expertise.
- Open questions: is soil acidity under the pines an issue? I was surprised to not see any reference to this.
- Name what four climate models you use with RCP 6.0.
- Why taking only RCP 6.0? there is no reason for not offering a range of scenario uncertainty in your model application.
- Line 183: “performance comparison with other soil moisture retrieval methods is beyond our scope.” I am not sure why you think this way. The only reason for having yet another soil water and carbon model is showing that yours is better or different than others. In my opinion comparing with others would be necessary.
- Equation 2 in 3.1.1. explain here or in the discussion how reducing the difference in EVT of trees compared to grass only based in LAI differences adds errors to your results
- Equations 6 and 7: explain A.
- I understand that the CO2-fertilization effect is not considered in this model framework. Am I right? If yes, please thoroughly discuss the implications for your results regarding RCP 6.0.
Results
- The study demonstrates significant improvements in soil moisture prediction with RSEEP, particularly in shallow layers, that’s a strength.
- An important result is the sensitivity of RothC's carbon stock predictions to accurate soil moisture modelling, that’s a good thing.
- The comparative analysis of tree species shows actionable insights, also valuable.
- The degradation of RSEEP's performance in deeper soil layers is not sufficiently analyzed, which undermines its applicability to systems with deep rooting.
- The presentation of uncertainties could be described in more detail in terms of their implications.
Discussion
- The discussion effectively links findings to practical applications, such as species-specific agroforestry strategies for water and carbon management.
- The limitations of RSEEP are openly acknowledged, which adds transparency.
- Some claims about RSEEP's generalizability are speculative and should be better supported by results in other sites and/or comparisons with other models. The lack of validation in other climatic or soil conditions limits confidence in the model's broader applicability.
Typos and minor points
- Line 348: "evapotranpiration" → "evapotranspiration."
- Line 130: mean annual temperatures are in a temperate zone misplaced for the description of any site. Consider shortly explaining the seasonality (max/min winter and summer temperatures)
- Line 157: “Penamn” -> Penman
Conclusion
This study presents an interesting model with potentials to broad application. The manuscript could benefit from clearer articulation of certain methodological choices and a more critical evaluation of uncertainties with other models and on different sites.
Citation: https://doi.org/10.5194/egusphere-2024-2258-RC2 -
AC2: 'Reply on RC2', Salim Goudarzi, 29 Nov 2024
1.Strengths and Weaknesses
Comment #1.1.The justification for developing RSEEP as a parsimonious alternative to complex models is compelling since agroforestry systems are already today in areas with scares monitoring systems and may further expand to other areas with no history of measurements. However, it is a bit surprising that a model designed to be of general character, with few needed inputs and transferable to other regions, is only applied and validated in one region. And this is one of my main criticism: in order to really prove the value of such a model as something new, different (better) than the established, it is necessary to prove its general application (this is at least my opinion). Response: In this study, our primary objective was to introduce RSEEP as a concept and demonstrate its potential by validating it at a single, data-rich site. This allowed us to show that RSEEP has the potential to produce reasonable predictions with minimal inputs. We then also applied it at a second site with limited data availability, although in a similar region. We agree that a broader validation across diverse regions would be needed to demonstrate its general applicability. However, this requires application to a much larger dataset, which was beyond the scope of this initial study. We explicitly highlighted this point in the manuscript as a limitation of the current study and an area for future research: (line 526) “application of our model to different soil types/thicknesses and more sites would provide more confidence that these conclusions generally hold”. Also in the conclusions section: “Further application of our model to different sites would test the generality of this finding.”
Comment #1.2.The introduction lacks clarity in transitioning between general challenges in agroforestry and the specific aims of the study. Better summarizing how RSEEP uniquely addresses these challenges may improve this. Response: Thank you for pointing this out. In retrospect we acknowledge this, and to address the point, we will revise the introduction to more explicitly outline how RSEEP addresses these challenges. Specifically, we will summarize the unique aspects of RSEEP, such as its minimal input requirements and adaptability to data-limited contexts, and link them more directly to the broader gaps in agroforestry modelling. This should provide a stronger rationale for the study and improve the flow of the introduction.
2.Methods
Comment #2.1.The methodological framework is simple but thorough, with clear explanations of the RSEEP model, including its parameters, calibration, and coupling with RothC. However, please better explain why you only consider the 400 stems/ha site. Response: We focused on the 400 stems/ha site (rather than 100/200 stems/ha sites in Glensaugh) because it was the only site instrumented to record soil moisture data, which was essential for calibrating and validating RSEEP. As our primary aim was to demonstrate RSEEP’s capabilities, we felt that this single site provided sufficient data for this purpose. We will clarify this rationale in the revised manuscript.
Comment #2.2.Figure 1 and section 2: please better explain how the two sites are managed. For example, I noticed in fig. 1 that trees are not trimmed in the lower part of the stem. Is there a reason for that? Animals could better use shade and shelter if that would be done. But explain generally how sites are managed (fertilisers, interventions, etc.). Response: Thank you for your observation. Regarding the trees in Figure 1, they are not trimmed in the lower part of the stem because Cruickshank Botanic Garden (CBG) is not an agroforestry site (i.e., animal welfare is not a concern there). Note that the dataset from CBG is used here only to test the strengths and weaknesses of RSEEP in predicting profile soil moisture, rather than testing the coupled RSEEP-RothC model for soil carbon dynamics. For this reason we believe that the details on how this particular site has been managed (fertilizers, interventions, etc. ) are less relevant here. However, we agree that more details on site management at Glensaugh (i.e., the agroforestry site) would have been helpful, but unfortunately such details are unavailable. We have acknowledged this in line 139 and explained that the lack of data on site management further highlights the importance of calculating impacts relative to the base-case: <line139> “But since the details of the management practice are unavailable, we calculate the impact of planting trees on pasture relative to the pasture base-case, to control for unknown effects (more details in section 3.2.4).”
Comment #2.3.Fig. 1: explain legend of lower panel in figure caption. Response: EM01, EM04, EM06 and EM15 are different parameterisations used in the CHESS-SCAPE climate dataset which is briefly introduced in line 155. In the revised manuscript, we will clarify these in the figure caption.
Comment #2.4.The assumption of soil saturation during the wettest period is not universally valid and could introduce errors in porosity estimates. Sensitivity to this assumption should have been tested because the beauty of your model (you say) is the potential to be applied at any data scarce region. Response: We agree that this assumption is not universally valid particularly in regions with drier/warmer climates. However, we expect this to be valid in our site where high annual precipitation and relatively low potential evapotranspiration are common occurrences. In the revision, we will raise this point in section 5.1, Strengths and Weaknesses of RSEEP, to make the readers aware that the assumption of soil saturation during the wettest period is not universally valid and more accurate maximum (surface) porosity estimates should be used where possible.
Comment #2.5.The rationale for specific parameter ranges during calibration is not fully explained, which could raise questions about the model's transferability to other regions. Explain how the ranges make sense (physically). Response: It is generally difficult to make physical sense of the parameter values of any conceptual model, including RSEEP, in relation to, for example, field observations of that parameter. This is because of the inherent incommensurability between the two sets of values, e.g., hydraulic conductivity as represented in RSEEP is unlikely to match the hydraulic conductivity as would be measured in the field. Model calibration is expected to compensate, at least to some degree, for this incommensurability, which is why RSEEP’s parameter ranges and distributions will change depending on the calibration data (i.e., site) and should not be transferred to other sites or regions. In the revised manuscript and in section 3.1.2 Calibration procedure at Cruickshank Botanic Garden, we will highlight these points regarding parameter ranges, their physical meaning and transferability to other regions.
Comment #2.6.Open questions: is soil acidity under the pines an issue? I was surprised to not see any reference to this. Response: Apologies but are unsure what the reviewer is specifically referring to here; is it with regards to lower nutrient availability? Or effect on decomposition rate? etc.. In any case, acidity has not been reported as an issue under any of the species on our site. In the revision, and in section 2.2 (The Glensaugh agroforestry experiment), we will highlight, with references, that although soil acidity might be an issue under pine in some sites, it has not been reported in our site.
Comment #2.7.Name what four climate models you use with RCP 6.0. Response: In the revision, and in line 426, we will clarify which four climate models we mean (which are EM01, EM04, EM06 and EM15 that were introduced earlier on in the paper).
Comment #2.8.Why taking only RCP 6.0? there is no reason for not offering a range of scenario uncertainty in your model application. Response: We selected RCP 6.0 as a ‘middle of the range’ scenario to demonstrate the model’s capabilities, but we agree that including multiple scenarios would better capture uncertainty. We will note this as a limitation in the revised manuscript and suggest future applications to consider a range of emission scenarios.
Comment #2.9.Line 183: “performance comparison with other soil moisture retrieval methods is beyond our scope.” I am not sure why you think this way. The only reason for having yet another soil water and carbon model is showing that yours is better or different than others. In my opinion comparing with others would be necessary. Response: There may have been a misunderstanding regarding our rationale for RSEEP. In paragraph L164, we have explained that existing soil moisture retrieval models either: (a) have too many parameters and thus require more calibration data than is likely to be available in agroforestry sites, or (b) ignore processes that are vital to agroforestry systems (e.g., evapotranspiration). In other words, our claim is not that RSEEP is better than the existing models in estimating profile soil moisture. Instead, our claim is that RSEEP is more suited to agroforestry systems where (a) or (b) would prohibit the application of existing models. In the next draft of the manuscript we will more clearly highlight the rationale in section 3.1, Soil moisture retrieval for ecohydrological modelling.
Comment #2.10.Equation 2 in 3.1.1. explain here or in the discussion how reducing the difference in EVT of trees compared to grass only based in LAI differences adds errors to your results. Response: While this approach aligns with our goal of creating a parsimonious model for data-limited sites, we acknowledge its limitations. We will add a brief discussion in section 3.1.1 of the revised manuscript to explicitly address these potential errors.
Comment #2.11.Equations 6 and 7: explain A. Response: ‘A’ denotes ‘Actual’ and ‘P’ denotes ‘Potential’ through the paper. We will make this clear in the next version.
Comment #2.12.I understand that the CO2-fertilization effect is not considered in this model framework. Am I right? If yes, please thoroughly discuss the implications for your results regarding RCP 6.0. Response: CO2 fertilisation effects are represented in our model in terms of increased evapotranspiration rates over time (which is represented in the form of increased PET with time in RCP6.0). But CO2 fertilisation effects are indeed not represented in terms of increased biocarbon storage rate over time. Though this effect is unlikely to be very significant in our site where plant growth is more nutrient-limited than it is light- or water-limited. This is nonetheless a source of uncertainty in our results. In the revision, we will highlight this point in section 6 (Sources of uncertainty) and clarify that it could have resulted in underestimating biocarbon storage in our model.
3.Results
Comment #3.1.The degradation of RSEEP's performance in deeper soil layers is not sufficiently analyzed, which undermines its applicability to systems with deep rooting. Response: In section 5.1., RSEEP’s limitation in estimating soil moisture at deeper layers is acknowledged. Importantly, rather than being generally true, we have shown that this limitation is only the case in sites where the assumptions of hydraulic conductivity and/or porosity decay with depth are not valid. In the same section (line 520) we have shown that this limitation is unlikely to have impacted our results because the bulk soil moisture behaviour is what drives RothC and the soil carbon dynamics which is predicted well by RSEEP. We believe that this level of analysis is sufficient to broadly show the strengths and weakness of RSEEP in the context in which it is applied. However, in the next version of the manuscript and in section 5.1, we will highlight this point to clarify that, in applications where soil moisture dynamics at different layers is of essence, care should be taken in applying RSEEP if the assumptions of hydraulic conductivity and/or porosity decay with depth are not valid.
Comment #3.2.The presentation of uncertainties could be described in more detail in terms of their implications. Response: Thank you for the suggestion. The current discussion in section 6 was designed to provide a concise yet sufficient overview of uncertainties and their implications. Each identified uncertainty has multiple potential effects, most likely interacting with the other uncertainties (and their multiple potential effects) in complex ways. We feel that further expanding this section would make the paper too long and risk shifting the emphasis away from our primary focus on introducing and validating a different approach to modelling agroforestry systems.
4.Discussion
Comment #4.1.Some claims about RSEEP's generalizability are speculative and should be better supported by results in other sites and/or comparisons with other models. The lack of validation in other climatic or soil conditions limits confidence in the model's broader applicability. Response: This is a valid point, and we fully agree. In fact, we stress this paper is a proof of concept and instead acknowledge (line 526) that “application of our model to different soil types/thicknesses and more sites would provide more confidence that these conclusions generally hold”. Also in the conclusions section: “Further application of our model to different sites would test the generality of this finding.”
5.Typos and minor points
Comment #5.1.Line 348: "evapotranpiration" → "evapotranspiration."
Comment #5.2.Line 130: mean annual temperatures are in a temperate zone misplaced for the description of any site. Consider shortly explaining the seasonality (max/min winter and summer temperatures)
Comment #5.3.Line 157: “Penamn” -> Penman
Response: Thank you for pointing these out, we will implement them and double-check the manuscript for other potential typos.
6.Conclusion
Comment #6.1.This study presents an interesting model with potentials to broad application. The manuscript could benefit from clearer articulation of certain methodological choices and a more critical evaluation of uncertainties with other models and on different sites. Response: Thank you for your thorough and constructive feedback. We will clarify the rationale behind key methodological choices and acknowledge the need for further evaluation of uncertainties, model comparisons, and testing on diverse sites as areas for future work.
Citation: https://doi.org/10.5194/egusphere-2024-2258-AC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
215 | 59 | 99 | 373 | 14 | 17 |
- HTML: 215
- PDF: 59
- XML: 99
- Total: 373
- BibTeX: 14
- EndNote: 17
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1