the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Root distributions predict shrub-steppe responses to precipitation intensity
Andrew Kulmatiski
Martin C. Holdrege
Cristina Chirvasa
Karen H. Beard
Abstract. Precipitation events are becoming more intense around the world, changing the way water moves through soils and plants. Plants that have, or create, roots that absorb more water under these conditions are likely to become more abundant (e.g., shrub encroachment). Yet, it remains difficult to predict species responses to climate change because we typically do not know where active roots are located or how much water they absorb. Here, we used water tracer injections in a field experiment to describe forb, grass, and shrub root distributions under low and high precipitation intensity treatments. To estimate how much water different active rooting distributions can absorb over time, we used a soil water flow model, and we compared our estimates of water uptake to aboveground plant growth. In low precipitation intensity plots, deep shrub roots were estimated to absorb the most water (93 mm yr−1) and shrubs had the greatest aboveground cover (27 %). Grass root distributions were estimated to absorb an intermediate amount of water (80 mm yr−1) and grasses had intermediate aboveground cover (18 %). Forb root distributions were estimated to absorb the least water (79 mm yr−1) and had the least aboveground cover (12 % cover). In high precipitation intensity plots, shrub and forb roots moved in ways that increased their water uptake relative to grasses, predicting the increased aboveground growth of shrubs and forbs in these plots. In short, water uptake caused by different rooting distributions predicted plant aboveground cover. Our results suggest that detailed descriptions of active plant root distributions can predict plant growth responses to climate change in arid and semi-arid ecosystems.
Andrew Kulmatiski et al.
Status: open (until 12 Apr 2023)
-
RC1: 'Comment on bg-2023-13', Anonymous Referee #1, 16 Mar 2023
reply
The manuscript “Root distributions predict shrub-steppe responses to precipitation intensity” describes results from a multi-year precipitation manipulation experiment and labeled-isotope tracer injection study, combined with soil water flow modeling, to explain how root distributions and water uptake patterns relate to plant abundances in sagebrush steppe. This experiment has been the subject of previous published work as well (e.g. Holdrege et al 2021, cited in the text), and there are places where a bit more could be done to articulate what novel addition is being made by this new contribution, but I also appreciate the way this leverages previous efforts to make the most of a multi-year experiment. I found it to be an interesting and well-conducted study which will make a useful contribution to the literature on global change, ecohydrology, and plant ecophysiology in shrubby ecosystems. I list major and minor concerns and suggestions for revision below.
My greatest overarching concern/question pertains to the plant functional group distinctions made in this study. It makes sense to separate shrubs, forbs, and grasses from each other; however, I was surprised to see annual and perennial grasses lumped this way. Sagebrush systems are home to a variety of grass growth forms, with the most notable contrast being between the native, deep-rooted perennial bunchgrasses, and the shallow-rooted exotic annual grasses (e.g. Bromus species) that are a major ecological threat throughout the biome. It sounds like both annual and perennial grasses were present at the study site, and species-level samples were collected after tracer injections, so would it be possible to separate these for the data analysis? I expect that would be more meaningful than a lumped “grass” category, and could be a very informative in a system where ecologists and land managers are at least as concerned about the future of annual grass invasion under global change as they are about woody encroachment. Some of the seasonal shifts in tracer uptake for grasses between the May and July periods could also perhaps be explained by phenological turnover between different grass species, rather than the same plants changing their rooting behavior – e.g. in Fig. 3a-b, could the high uptake from the shallowest depths be attributed to annual grasses active in May, while the shift toward the second-shallowest depth in July reflects annual grasses senescing while deeper-rooted perennials remain active?
I have a few wording quibbles that apply throughout the manuscript:
- In several spots, root plasticity is described as roots “moving.” I’m not sure of this word choice, as it paints a particular picture of a root actively relocating through the soil, which obscures the variety of potential mechanisms – growth or proliferation of existing roots, or shedding of unused fine roots, or modification of root physiology at different depths, or even seasonal turnover in which species are most active (since species are lumped but may have differences within functional groups in both rooting depths and phenology, see above). Any one of these changes could result in a signal of the distribution of water uptake moving, but come with differing consequences in terms of other factors such as the carbon economy of the plants. I suggest including more discussion and clarity about the different mechanisms that could underlie a net signal of root movement, and perhaps using a more neutral descriptor such as “shift.”
- I also wondered about the use of “competitive advantage” throughout, which seems to imply the potential for competitive exclusion even if that is not what is meant. I suggest more carefully defining this term early on, or perhaps choosing a different term that does not sound so much like one plant winning and the other losing (when what we see here is shifts in relative dominance, but stable coexistence regardless).
Minor comments by line number:
L58-59 – I believe that not all of these studies (e.g. Case et al 2020) used tracer techniques – check which ones are actual examples of this method, vs. examples of water uptake depth estimates along a continuum rather than a shallow vs. deep approach?
L74 – A previous paper is referenced that seems to have had very similar findings from the same experiment. Add a bit more explanation here of what this new paper adds – if it’s just additional years of data, what’s the question being answered by extending the time series? If it’s also that a new method or approach was applied here, explain that novel addition more clearly too.
L109-110 – Clarify whether these adjustments to event sizes meant all events were a fixed size, or if setting a minimum means larger events could also be delivered. A supplemental figure showing event size distributions for the different treatments might be helpful.
L119-122 – On what basis were the high vs. low cutoffs set? At a glance, the 4mm treatment seems much more similar to the 1-3 mm treatments than the 8 or 18 mm treatments, yet it’s lumped with those much higher values. Perhaps there’s extra information that can be shared from the first few years of the experiment to justify this split?
L126 – Clarify whether the increased soil water availability noted here was overall, or at certain depths but not others.
L167-169 – How were the two different seasons of tracer uptake data incorporated into the modeling?
L221-222 – “early season low intensity treatments,” etc., sound like a description of the treatment (low intensity rainfall manipulated early in the growing season) rather than the timing of sampling + event intensity of the treatment. Suggest re-wording to clarify this.
L298-300 – Is this circular, to say that root distributions weighted by existing plant cover correctly predicted the rank order abundance of aboveground plant cover?
Fig 3 – Make line thicknesses consistent among the four panels of this figure (the lower panels have thicker lines, which is distracting). I would also suggest putting panels for different treatments side by side (rather than seasons side by side, treatments in rows) to be consistent with the arrangement of later figures and easier to compare between treatments.
Fig 4 – The legend here seems to be incorrect – should the left panel say “Forb low,” etc.? This is also a place where I wonder about the definition of competitive advantage as simply meaning greater water uptake in a layer, where the other groups may also be using nearly as much water. Does the right panel imply there’s no niche for forbs in the high intensity treatment?
Fig 7 – The forb graphic inset is positioned such that it looks like a label for the vertical dotted line – I suggest moving it elsewhere within the figure panel to avoid confusion.
Citation: https://doi.org/10.5194/bg-2023-13-RC1 -
CC1: 'Reply on RC1', Andrew Kulmatiski, 22 Mar 2023
reply
The manuscript “Root distributions predict shrub-steppe responses to precipitation intensity” describes results from a multi-year precipitation manipulation experiment and labeled-isotope tracer injection study, combined with soil water flow modeling, to explain how root distributions and water uptake patterns relate to plant abundances in sagebrush steppe. This experiment has been the subject of previous published work as well (e.g. Holdrege et al 2021, cited in the text), and there are places where a bit more could be done to articulate what novel addition is being made by this new contribution, but I also appreciate the way this leverages previous efforts to make the most of a multi-year experiment. I found it to be an interesting and well-conducted study which will make a useful contribution to the literature on global change, ecohydrology, and plant ecophysiology in shrubby ecosystems. I list major and minor concerns and suggestions for revision below.
My greatest overarching concern/question pertains to the plant functional group distinctions made in this study. It makes sense to separate shrubs, forbs, and grasses from each other; however, I was surprised to see annual and perennial grasses lumped this way. Sagebrush systems are home to a variety of grass growth forms, with the most notable contrast being between the native, deep-rooted perennial bunchgrasses, and the shallow-rooted exotic annual grasses (e.g. Bromus species) that are a major ecological threat throughout the biome. It sounds like both annual and perennial grasses were present at the study site, and species-level samples were collected after tracer injections, so would it be possible to separate these for the data analysis? I expect that would be more meaningful than a lumped “grass” category, and could be a very informative in a system where ecologists and land managers are at least as concerned about the future of annual grass invasion under global change as they are about woody encroachment. Some of the seasonal shifts in tracer uptake for grasses between the May and July periods could also perhaps be explained by phenological turnover between different grass species, rather than the same plants changing their rooting behavior – e.g. in Fig. 3a-b, could the high uptake from the shallowest depths be attributed to annual grasses active in May, while the shift toward the second-shallowest depth in July reflects annual grasses senescing while deeper-rooted perennials remain active?
Response: Thanks for this comment. This is an important point. Annual grasses are not a large problem at this site, but we have separated annual and perennial grass data where possible. This revealed that perennial grasses decreased in high precipitation intensity treatments and that annual grasses showed no response to treatments. Splitting data also suggested that annual grass growth reflected a one-year lag to wet years in 2016 and 2017.
Unfortunately, it was not possible to separate the grasses by annual perennial for the tracer experiment for two reasons. First, it was often impossible to identify the species of the emerging grass blades in the early season sampling. Second, a species has to be abundant enough across all injection depth plots to produce a tracer uptake profile for that species. However, we can say that annual grass growth was trivial during the tracer injections. This can be seen in the new vegetation survey data that separates annual and perennial growth. This small annual grass growth in 2020 reflected the fact that 2020 was a dry year. We have added some information about annual precipitation to the Methods.
I have a few wording quibbles that apply throughout the manuscript:
- In several spots, root plasticity is described as roots “moving.” I’m not sure of this word choice, as it paints a particular picture of a root actively relocating through the soil, which obscures the variety of potential mechanisms – growth or proliferation of existing roots, or shedding of unused fine roots, or modification of root physiology at different depths, or even seasonal turnover in which species are most active (since species are lumped but may have differences within functional groups in both rooting depths and phenology, see above). Any one of these changes could result in a signal of the distribution of water uptakemoving, but come with differing consequences in terms of other factors such as the carbon economy of the plants. I suggest including more discussion and clarity about the different mechanisms that could underlie a net signal of root movement, and perhaps using a more neutral descriptor such as “shift.”
Response: Thanks, this is a good point. We felt that ‘moved’ conveyed the idea we wanted, but ‘changed’ is more accurate. We have changed ‘moved’ to ‘changed’ or similar throughout.
- I also wondered about the use of “competitive advantage” throughout, which seems to imply the potential for competitive exclusion even if that is not what is meant. I suggest more carefully defining this term early on, or perhaps choosing a different term that does not sound so much like one plant winning and the other losing (when what we see here is shifts in relative dominance, but stable coexistence regardless).
Response: We have replaced ‘competitive advantage’ with ‘competitive advantage for water uptake’ throughout the manuscript.
This is an important and ‘bigger’ question. We do not model or experimentally test for long-term stable coexistence. There are mechanisms of coexistence in our modeling approach that may or may not be large enough to allow coexistence over the long term. We are talking more about the short term ability of a root distribution to absorb more or less water than another. We feel that it is accurate to say that a root distribution that can absorb more water has a competitive advantage for water uptake.
Minor comments by line number:
L58-59 – I believe that not all of these studies (e.g. Case et al 2020) used tracer techniques – check which ones are actual examples of this method, vs. examples of water uptake depth estimates along a continuum rather than a shallow vs. deep approach?
Response: this is a good point, we have removed two of the citations.
L74 – A previous paper is referenced that seems to have had very similar findings from the same experiment. Add a bit more explanation here of what this new paper adds – if it’s just additional years of data, what’s the question being answered by extending the time series? If it’s also that a new method or approach was applied here, explain that novel addition more clearly too.
Response: We have added the following to clarify the contributions of this paper: ‘Here, we add to findings from that paper by reporting vegetation responses for an additional two treatment years and one post-treatment year, and more importantly, by using a tracer experiment to measure forb, grass, and shrub root distributions in experimentally manipulated plots.’
L109-110 – Clarify whether these adjustments to event sizes meant all events were a fixed size, or if setting a minimum means larger events could also be delivered. A supplemental figure showing event size distributions for the different treatments might be helpful.
Response: To clarify, we have added the following: ‘Minimum event sizes were deposited over roughly five to fifteen minutes at an intensity that is in the upper 5% of observed mm min-1 rainfall intensities. A day that received 10 mm of natural precipitation would deposit 5, 3, 2, 1 and 0-1 events in the 2, 3 , 4, 8 and 18 mm treatments.’
L119-122 – On what basis were the high vs. low cutoffs set? At a glance, the 4mm treatment seems much more similar to the 1-3 mm treatments than the 8 or 18 mm treatments, yet it’s lumped with those much higher values. Perhaps there’s extra information that can be shared from the first few years of the experiment to justify this split?
Response: The 4 mm treatments were replicated as a ‘high’ intensity treatment associated with 5C warming.
L167-169 – How were the two different seasons of tracer uptake data incorporated into the modeling?
Response: To clarify, we have added the following: ‘Early-season community-level water uptake (i.e., Hydrus simulation values) were multiplied by early-season tracer uptake proportions by depth and late season community-level water uptake was multiplied by late-season tracer uptake.’
L221-222 – “early season low intensity treatments,” etc., sound like a description of the treatment (low intensity rainfall manipulated early in the growing season) rather than the timing of sampling + event intensity of the treatment. Suggest re-wording to clarify this.
Response: We have added ‘tracer uptake’ after ‘early season’ to clarify.
L298-300 – Is this circular, to say that root distributions weighted by existing plant cover correctly predicted the rank order abundance of aboveground plant cover?
Response: We have included weighted values for two reasons. First, by included water uptake unweighted by leaf area and water uptake weighted by leaf area, it makes it very clear to the reader that our unweighted values do not include leaf area (this has confused readers in the past). Second, the unweighted values provide insight into the physiological effects of root distributions on their own (i.e., competitive ability of a root distributions) and the weighted values provide a more realistic estimate of water flow on the landscape, so we believe there is value in both measures. Further, the weighted values provide some insight into whether the observed abundance is consistent with our estimates of actual total water uptake or not.
Fig 3 – Make line thicknesses consistent among the four panels of this figure (the lower panels have thicker lines, which is distracting). I would also suggest putting panels for different treatments side by side (rather than seasons side by side, treatments in rows) to be consistent with the arrangement of later figures and easier to compare between treatments.
Response: We have changed the figure order as suggested. We have left the thicker lines to highlight the bigger precipitation event treatments.
Fig 4 – The legend here seems to be incorrect – should the left panel say “Forb low,” etc.? This is also a place where I wonder about the definition of competitive advantage as simply meaning greater water uptake in a layer, where the other groups may also be using nearly as much water.
Response: Thanks for catching the type-o. We have changed the left legend to say ‘Low’. We have changed the wording for competitive advantage for water to clarify.
Does the right panel imply there’s no niche for forbs in the high intensity treatment?
Response: Yes, there is almost no depth at which forbs have a competitive advantage under high intensity treatment. However, forb root distributions absorb more water across their profile in high than low intensity treatments.
Fig 7 – The forb graphic inset is positioned such that it looks like a label for the vertical dotted line – I suggest moving it elsewhere within the figure panel to avoid confusion.
Response: We have moved the graphics so they are all on the left side of the figure.
Citation: https://doi.org/10.5194/bg-2023-13-CC1
-
RC2: 'Comment on bg-2023-13', Anonymous Referee #2, 17 Mar 2023
reply
General comments
This manuscript with reference ID bg-2023-13 presents an interesting analysis of root distributions in a shrub-steppe environment. In general, the article is well written, and the authors present their results in a clear and concise way. However, in some places the text could still be improved by adding a few more corrections to the English grammar and style. Most of all, I wondered if the title “Root distributions predict shrub-steppe responses to precipitation intensity” reflects the natural response of vegetation to the climatic drivers, which in this ecosystem appears to be dominated by the annual distribution and magnitude of precipitation and thus could be change to “Root distribution response to precipitation intensity predicts the distribution of vegetation in a shrub-steppe ecosystem”. Nonetheless, the article should be of great interest to the academic readership, and subject to further amendments and modifications (see more specific recommendations in the section below), could be considered for publication.
Specific comments
This manuscript by Kulmatiski and colleagues presents interesting results of a large-scale field experiment conducted in a shrub-steppe ecosystem and highlights the importance of vegetation belowground response to seasonal changes in water availability. The article is generally well written and presents the results concisely, however, I wonder if the presentation of the manuscript could still be improved by including a clear set of hypotheses that could be used to structure the discussion section and address the study findings in light of the available literature on that topic. Currently, the introduction section lists a series of potential vegetation responses to climatic changes in seasonal water availability. However, there appears to be no clear structuring of these findings in relation to the study results obtained by conducting a long-term manipulation experiment. It may be expected that plant belowground investment and root distributions are a major factor determining plant growth and coexistence under current and anticipated climate conditions, but further stressing the importance of respective aspects (i.e., growth, allocation, competition, etc.) investigated in this study might help to elucidate the effect of multiple and interrelated factors in jointly determining the response of the plant community to multiple years of an artificial precipitation treatment. Having said this, it is currently not obvious if this treatment did actually result in significant differences between the factors displayed in Figures 2-7. In line with this, the results section indicates that the authors did not perform statistical analysis due to a lack of spatial replication (which makes it impossible to assess whether the reported responses are actually significant) but I do wonder if having a clear set of hypothesis (i.e. XY will increase to Z) would allow for applying a directional test (i.e., a hypothesis test where a direction is specified), which therefore should be more powerful than a non-directional test. Nonetheless, another thing that struck me is that the article claims that the presented approach could be used to improve predictions of arid and semi-arid ecosystem responses to climate change but fails to deliver on how this could be achieved. Hence, I would strongly recommend revising the manuscript based on the recommendations provided above and further highlight how the results obtained from this study could be applied in vegetation models predicting ecosystem responses to climate change.
Technical corrections
L10: consider rephrasing to: “plant rooting strategies that sustain water uptake are likely to become more abundant”.
L12- 15: consider rephrasing to: “we applied a water tracer experiment to describe forb, grass, and shrub root distributions under low and high precipitation intensity treatments and used a soil water flow model to estimates how much water respective plant root tissue can absorb over time”.
L218-220: Please indicate respective package used in this section, which could be achieved by using: citation("packagename") in the R environment.
Citation: https://doi.org/10.5194/bg-2023-13-RC2 -
CC2: 'Reply on RC2', Andrew Kulmatiski, 22 Mar 2023
reply
General comments
This manuscript with reference ID bg-2023-13 presents an interesting analysis of root distributions in a shrub-steppe environment. In general, the article is well written, and the authors present their results in a clear and concise way. However, in some places the text could still be improved by adding a few more corrections to the English grammar and style. Most of all, I wondered if the title “Root distributions predict shrub-steppe responses to precipitation intensity” reflects the natural response of vegetation to the climatic drivers, which in this ecosystem appears to be dominated by the annual distribution and magnitude of precipitation and thus could be change to “Root distribution response to precipitation intensity predicts the distribution of vegetation in a shrub-steppe ecosystem”. Nonetheless, the article should be of great interest to the academic readership, and subject to further amendments and modifications (see more specific recommendations in the section below), could be considered for publication.
Response: Thanks. We prefer to keep the title as ‘Root distributions predict shrub-steppe responses to precipitation intensity’ because it is shorter and does not emphasize the root response to treatment as much. We think it is important that root distributions predicted both plant growth under both current and manipulated precipitation conditions.
Specific comments
This manuscript by Kulmatiski and colleagues presents interesting results of a large-scale field experiment conducted in a shrub-steppe ecosystem and highlights the importance of vegetation belowground response to seasonal changes in water availability. The article is generally well written and presents the results concisely, however, I wonder if the presentation of the manuscript could still be improved by including a clear set of hypotheses that could be used to structure the discussion section and address the study findings in light of the available literature on that topic. Currently, the introduction section lists a series of potential vegetation responses to climatic changes in seasonal water availability. However, there appears to be no clear structuring of these findings in relation to the study results obtained by conducting a long-term manipulation experiment. It may be expected that plant belowground investment and root distributions are a major factor determining plant growth and coexistence under current and anticipated climate conditions, but further stressing the importance of respective aspects (i.e., growth, allocation, competition, etc.) investigated in this study might help to elucidate the effect of multiple and interrelated factors in jointly determining the response of the plant community to multiple years of an artificial precipitation treatment.
Response: The hypothesis for this research is essentially described by the equations in Hydrus 1D – the soil water flow model. Rather than hypothesis testing, this research essentially parameterizes this model with accurate and precise root distribution data. It isn’t really appropriate to describe the hypothesis in this well established model, though it is appropriate to describe testable predictions that this model (hypothesis) could produce. We do this in the second paragraph of the Introduction as follows: ‘In arid and semi-arid areas, small (~1-5 mm) events are common and often evaporate before reaching the rooting zone (Lauenroth and Bradford, 2009). With larger precipitation events (~5-20 mm), less water is lost to evaporation and water percolates deeper into the soil (Sala et al., 2015). In arid systems, decreased evaporation is likely to increase growth of shallow-rooted plants (e.g., grasses; Post and Knapp, 2021). In semi-arid systems, deeper percolation is likely to increase growth of deeper-rooted species (e.g., shrubs; Gherardi and Sala, 2015; Xu, Medvigy and Rodriguez-Iturbe, 2015; Holdrege, Beard and Kulmatiski, 2021). In mesic systems, however, increasing event size may increase runoff and percolation below the rooting zone. This may have little effect on plant productivity or lead to tree encroachment (Jones et al., 2016; Slette et al., 2022; Berry and Kulmatiski, 2017; Knapp et al., 2008). ‘
Having said this, it is currently not obvious if this treatment did actually result in significant differences between the factors displayed in Figures 2-7. In line with this, the results section indicates that the authors did not perform statistical analysis due to a lack of spatial replication (which makes it impossible to assess whether the reported responses are actually significant) but I do wonder if having a clear set of hypothesis (i.e. XY will increase to Z) would allow for applying a directional test (i.e., a hypothesis test where a direction is specified), which therefore should be more powerful than a non-directional test.
Response: We did perform statistical tests for everything except soil moisture which was only measured in one treated and one control plot.
Statistical test results for tracer uptake (Fig. 3) are shown in Table 1. Statistical results for Figs. 4 and 5 are assumed to be the same because these results essentially multiply the values from Fig. 3 by the results of a deterministic water flow model.
The statistical test results for Fig. 6 are reported in the Figure legend.
Statistical test results for Fig. 7. Are reported in the Results (lines 271-295).
Nonetheless, another thing that struck me is that the article claims that the presented approach could be used to improve predictions of arid and semi-arid ecosystem responses to climate change but fails to deliver on how this could be achieved. Hence, I would strongly recommend revising the manuscript based on the recommendations provided above and further highlight how the results obtained from this study could be applied in vegetation models predicting ecosystem responses to climate change.
Response: Thanks, for this suggestion. We have elaborated on how active root distribution data can be used to forecast vegetation responses to climate change in the Conclusions as follows: ‘It should not be surprising that the ability to absorb soil water can explain plant growth in arid and semi-arid systems, but until now it has been difficult to describe the location and activity of plant roots responsible for this water uptake. As active rooting distribution data become more available, it should be possible to use this data in soil water flow models to predict how much water different plants can absorb under future climate conditions (Holdrege et al. 2022). Inasmuch, it should be possible to forecast forb, grass and shrub growth over time with obvious implications for forecasting forage production, shrub encroachment, the consequences of shrub control, primary productivity and water cycling. ‘
Technical corrections
L10: consider rephrasing to: “plant rooting strategies that sustain water uptake are likely to become more abundant”.
Response: We have made this change.
L12- 15: consider rephrasing to: “we applied a water tracer experiment to describe forb, grass, and shrub root distributions under low and high precipitation intensity treatments and used a soil water flow model to estimates how much water respective plant root tissue can absorb over time”.
Response: We have made this change.
L218-220: Please indicate respective package used in this section, which could be achieved by using: citation("packagename") in the R environment.
Response: We have indicated that the ‘mgcv’ package was used at first appearance of use of GAMM’s.
Citation: https://doi.org/10.5194/bg-2023-13-CC2 -
RC3: 'Reply on CC2', Anonymous Referee #2, 23 Mar 2023
reply
Thank you for responding to my comments, which unfortunately seem to have not been considered thoroughly:
(1) my comment about the title was hinting at the circularity (also highlighted by R1) that having root distributions weighted by existing plant cover may induce some kind of autocorrelation between belowground root distribution and aboveground plant cover?
(2) there is no response to my recommendation of applying a directional test (i.e., a hypothesis test where a direction is specified) in order to achieve higher statistical power than may be achieved with a non-directional test, thus allowing to report a one-tailed p-value.
(3) I still do believe that listing a set of concrete hypothesis (no matter if these were based on an ecological assumption or a model used for testing those) at the end of the introduction section would allow for a more concise presentation of the discussion section, which currently lacks any clear structure.
(4) my recommendation using citation('packagename') has been disregarded, for instance a query for citation("mgcv") would result in the following output: 2011 for generalized additive model method; 2016 for beyond exponential family; 2004 for strictly additive GCV based model method and basics of gamm; 2017 for overview; 2003 for thin plate regression splines, which should point the user to respective references (which have not have been considered in the reference list):
Wood, S.N. (2011) Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B) 73(1):3-36
Wood S.N., N. Pya and B. Saefken (2016) Smoothing parameter and model selection for general smooth models (with discussion). Journal of the American Statistical Association 111:1548-1575.
Wood, S.N. (2004) Stable and efficient multiple smoothing parameter estimation for generalized additive models. Journal of the American Statistical Association. 99:673-686.
Wood, S.N. (2017) Generalized Additive Models: An Introduction with R (2nd edition). Chapman and Hall/CRC.
Wood, S.N. (2003) Thin-plate regression splines. Journal of the Royal Statistical Society (B) 65(1):95-114.
Citation: https://doi.org/10.5194/bg-2023-13-RC3 -
CC3: 'Reply on RC3', Andrew Kulmatiski, 24 Mar 2023
reply
Thank you for responding to my comments, which unfortunately seem to have not been considered thoroughly:
- my comment about the title was hinting at the circularity (also highlighted by R1) that having root distributions weighted by existing plant cover may induce some kind of autocorrelation between belowground root distribution and aboveground plant cover?
Response: Thanks for clarification on your comments. Our conclusions are derived from Figs 3 and 4 which are not weighted by plant cover. Because our weighted calculations appear to cause confusion, we have moved our estimates of weighted water uptake to an appendix. We have also moved the associated measurements of stomatal conductance to the same appendix. This removes a t-test value so that now almost all analyses are GAMMs which also simplifies the paper and helps address the comments below.
We provide estimates of water uptake weighted by plant cover (now in an appendix) as an estimate of current water uptake on the landscape, but this data is not used to support our argument that root distributions predict water uptake and plant growth.
We have also added the following to the end of the Introduction: ‘To be clear, we estimate the amount of water that forb, grass and shrub root distributions can be expected to absorb assuming that each rooting distribution has the same transpiration demand. This approach isolates the effects of root distributions from other effects such as leaf area, stomatal conductance, or aerodynamic resistance.’
More specifically, our tracer uptake values (fig. 3) are not weighted by existing plant cover. These are the proportion of tracer uptake by depth for each plant type. Our niche water uptake values (Fig. 4) are also not weighted by existing plant cover. As explained in the methods and legend for this figure, these values describe the amount of water different root distributions can be expected to absorb if each root distribution is assumed to have the same plant canopy. Our conclusions about how root distributions affect plant growth are derived from these unweighted results.
Figure 5 (now appendix) was the only figure that shows data weighted by existing plant cover. We do not rely on this data for the conclusion that root distributions explain plant growth and responses to treatments. Rather, we present this data because it represents our best estimate of the amount of water that forbs, grasses and shrubs absorb across the landscape given their existing abundances.
We believe the reviewer’s concern regarding ‘autocorrelation’ is that weighting water uptake by current plant abundance is likely to produce the result that the most abundant plants absorb the most water. This is likely to be true, but again, we do not rely on weighted water uptake values to interpret the role of root distributions. Second, these weighted values are important because they represent our best estimate of how water is moving on the landscape. These weighted values also allow us to examine whether total water uptake on the landscape is consistent with the observed cover. For example, if shrubs represent 50% of plant cover and weighted shrub water uptake represents 75% of water uptake, then shrubs may have a low water use efficiency, high herbivory, or have not yet realized their full potential cover.
there is no response to my recommendation of applying a directional test (i.e., a hypothesis test where a direction is specified) in order to achieve higher statistical power than may be achieved with a non-directional test, thus allowing to report a one-tailed p-value.
Response: We used GAMMs to provide support for grouping or splitting tracer uptake or plant growth curves (the two fundamental tests in this paper). For these, we use AIC values to select the model with more support. P-values are not relevant to these tests.
We used t-tests for plant growth differences in the post-treatment year and to compare stomatal conductance in high and low precipitation intensity treatment plots. These tests were not central to our study or conclusions and we have moved the stomatal conductance data and statistical test to an appendix. We did not make hypotheses about post-treatment plant growth responses or stomatal conductance and think doing so would be distracting from the focus of this study.
- I still do believe that listing a set of concrete hypothesis (no matter if these were based on an ecological assumption or a model used for testing those) at the end of the introduction section would allow for a more concise presentation of the discussion section, which currently lacks any clear structure.
Response: To address this comment, we now state our predictions at the end of the Introduction and we have restructured our discussion to summarize how results match those predictions in the Discussion. In the Introduction, we have added the following: ‘We predicted that shrubs dominate in this semi-arid ecosystem because their deeper roots provide access to a larger soil water pool than shallower grass or forb roots. We also predicted that larger precipitation events would ‘push’ water deeper into the soil providing an advantage to plants with deeper and more flexible rooting patterns (i.e., shrubs and forbs). ‘
(4) my recommendation using citation('packagename') has been disregarded, for instance a query for citation("mgcv") would result in the following output: 2011 for generalized additive model method; 2016 for beyond exponential family; 2004 for strictly additive GCV based model method and basics of gamm; 2017 for overview; 2003 for thin plate regression splines, which should point the user to respective references (which have not have been considered in the reference list):
Wood, S.N. (2011) Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B) 73(1):3-36
Wood S.N., N. Pya and B. Saefken (2016) Smoothing parameter and model selection for general smooth models (with discussion). Journal of the American Statistical Association 111:1548-1575.
Wood, S.N. (2004) Stable and efficient multiple smoothing parameter estimation for generalized additive models. Journal of the American Statistical Association. 99:673-686.
Wood, S.N. (2017) Generalized Additive Models: An Introduction with R (2nd edition). Chapman and Hall/CRC.
Wood, S.N. (2003) Thin-plate regression splines. Journal of the Royal Statistical Society (B) 65(1):95-114.
Response: We reference Wood (2004) in the Statistical Analyses section. We have added another reference to Wood (2004) to the first description of GAMMs to make it clearer that this reference relates to the use of GAMMs.
Citation: https://doi.org/10.5194/bg-2023-13-CC3
-
CC3: 'Reply on RC3', Andrew Kulmatiski, 24 Mar 2023
reply
-
RC3: 'Reply on CC2', Anonymous Referee #2, 23 Mar 2023
reply
-
CC2: 'Reply on RC2', Andrew Kulmatiski, 22 Mar 2023
reply
Andrew Kulmatiski et al.
Andrew Kulmatiski et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 191 | 51 | 13 | 255 | 4 | 6 |
- HTML: 191
- PDF: 51
- XML: 13
- Total: 255
- BibTeX: 4
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1