Using a Vegetation Model and Stakeholder Input to Assess the Climate Change Vulnerability of Tribally Important Ecosystem Services

Round 1

Reviewer 1 Report

The study is generally very good; the metodhology is interesting and well-presented, the language is correct and the results clear and significant.

The main flaw of the paper is the controversial ecosystem service classification which is introduced. The rationale of this classification is not clear, and it lets an inherent confusion emerge between the ecosystem service itself and the resource/species/element that is providing it. I therefore suggest to modify this classification in a way coherent with the most common classifications available (e.g. Millenium Ecosystem Assessment).

This and other comments can be found in the file attached.

Comments for author File: Comments.pdf

Author Response

Point 1: This sentence here is out of place. You should move it to the previous praragraph, when you were already discussing some examples of ecosystem service (or at least use always the same example thoughout the document). 


Response 1: Thank you, I moved the examples within the sentence in question (line 83-84) to the previous paragraph and removed the Labrador tea example to be more consistent throughout the document..

Point 2: put "dinamyc global vegetation model" with capital letters to show that DGVM is based on these words

Response 2: Revised as suggested.

Point 3: this is methodology, not introduction. Here you should staten the aims of your work not how you did it

Response 3: Thank you, this paragraph introduces the overall approach of our analysis and states the intent of our work. Although we appreciate that introductions should not contain methods, we do not agree that the sentence in question belongs in the methodology.

Point 4: this is results, not introduction. Move it to another section

Response 4: We agree and removed the section in question and focused this paragraph on the overall approach, not the results. See lines 98-114.

Point 5: this is part of the discussion, not of introduction. move it

Response 5: We agree and removed the section in question.

Point 6: It would be great to have more info on stakeholders selection and involvement ... how were these people involved and selected? how was the interview structured? how were the different tribal interests considered and balanced? etc

Response 6: Good questions, we included more information on the stakeholder selection process, see lines 125-129.

Point 7: Is this terminology correct? I can't find it anywhere else. Please check

Response 7: Yes, the terminology is correct and is used later in the manuscript. I clarified this sentence, see line 136.

Point 8: You are mixing up the resources and the ecosystem service, which are not quite the same thing!

Response 8: We appreciate that there was some confusion in the manuscript about our definition of ecosystem services. We clarified the definition (e.g., see lines 71-74) and removed the term “ecosystem services” when not appropriate.

Point 9: It is really not clear what is the rationale behind this classification... is the use? the vegetation type? the location?

"timber production" or "grazing quality" are ecosystem service, but "mammals" are  not...

I really don't like the classification you propose because it originates misunderstandings in the reader, since it basically considers the service and the resource that produces it as the same thing (which is definitely not true). I strongly suggest you to change the rationale of the classification into something less debatable.

An helpful example of distinction between ecosystem service demand and supply could come from this study: Accastello C., Blanc S., Brun F.; 2019. A Framework for the Integration of Nature-Based Solutions into Environmental Risk Management Strategies. Sustainability; 11 (2): 489.

Response 9: We understand there was confusion about our groups and have clarified our categorization. See lines 134-152 and Response 8 for more.

Point 10: Another flaw of this classification: aren't "timber production" and "culturally significant trees" two conflicting service? If you had classified the services based on the Millenium Ecosystem Assessment (which you cited, so you know it) this would have not happened

Response 10: Interestingly, no, timber production and culturally significant trees are not necessarily conflicting services. For instance, one can manage a single stand of trees for both high-quality timber production and retain culturally significant trees within that stand.

Point 11: isn't "grazing quality" possibly overlapping with "aridland plants"? and "mammals" and Birds" really cannot be a "ecosystem service category"!

This categorisation is just not working

Response 10: We clarified that our groups were chosen by our stakeholders and thus are not the pre-defined Millennium Ecosystem Assessment categories. Yes, some of our species and resources overlap, for example, some bird species overlap in habitat with some tree species. However, these groups are relevant in that our stakeholders determined their relevance. We clarified this in the Methods.

Point 11: describes how "meaningful" was defined: which factors were considered to set such timeframe?

Response 10: We clarified this section and referred to ‘appropriate timeframes’ instead of ‘meaningful’.

Point 12: Is fire the main natural hazard affecting this area? or are there any other hazards which are not considered by the model but that are relevant for the area?

Response 12: Yes, fire is the main natural disturbance and we clarified this section, see line 202.

Point 13: I don't like this heading because the huckelberry is not an Ecosystem Service per se, but rather a supplier of ES, so I 'd change the heading to avoid misinterpretation

Response 13: Updated the heading as suggested, see Table 1.

Point 14: The "vegetation types" you present here could also work very well to classify your 78 ecosystem services! consider to change the current categorisation and use this one

Response 14: Although we appreciate your suggestion, our stakeholders helped decide how best to categorize the resources in this study. That said, we have changed the reference of ‘ecosystem services’ to ‘species and resources’ to avoid confusion.

Point 15: so you could categorize your ecosystem services directly in relation to vegetation type instead of using your 6 categories with no clear rationale

Response 15: Our stakeholders helped decide how best to categorize the species and resources in this study, please see lines 134-152.

Point 16: been ... been correct

Response 16: We corrected this.

Point 17: identified

Response 17: Updated the word use.

Point 18: I'd reduce the size of the symbols to better distinguish them and reduce overlapping. And also put white background

Response 18:  We updated the figure as suggested, see Figure 5.

Point 19: I don't like this way of mixing up animal species, vegetation and the services they provide (and you do it all over the paper). They should be clearly differentiated, as services don't have "habitat", but species do. Please check  your terminology all over the paper in order to distinguish the species that provide a service from the service itself

Response 19: We clarified this and changed the use of terminology from ecosystem service to resource, for example, see lines 324-339.

Point 20: change the background, it is not helpful

Response 20: We’re not sure why the backgrounds of our figures are green in the PDF, however, the original figures do not have a background color.

Point 21: here again you mess up between services and their providing species. the sentence "potential loss of suitable habitat" can work for the "birds" category, but not for the "grazing quality". I believe this confusion is mainly originated by the incoherent categorisation of your ecosystem services

Response 21: We’ve clarified that our study examines species and resources. And we argue that there is habitat for grazing quality – specifically the vegetation type, grasslands, as projected by our DGVM simulations.

Point 22: this "historical habitat" line should not be here... maybe if you switch columns and rows you can organise better this table (put 2050 and 2080 at the top, expansion and contraction as sub-heading ot columns and huckleberry and bitterbrush as the two lines)

Response 22: Although we appreciate the comment, we feel strongly that including the area of historical habitat helps the reader understand the context of expansion, stable, contraction, and the overall change in habitat.

Point 23: the "stable" line is not very informative... you could skip it.

If you want to improve the quantity of information delivered by this table, think of including % values

Response 23: We also feel that “stable” line is important to include because and it’s more informative to report area instead of % values because huckleberry habitat covers 40% more area than bitterbrush. Reporting % values tells a different story than area.

Point 24: use another colour for expansion (blue is not visible enough beside dark green)

Response 24: Noted and updated.

Point 25: Don't need to change page between results and discussion

Response 25: Noted and updated.

Point 26: You always stated this approach was "generalizable" but you've never clearly explained why you consider it to. Add few sentences about it here in the discussion

Response 26: We further explained why our approach is generalizable within the discussion, see lines 379-284.

Point 27: You talk about "resource managers", but it is not clear the role of the native americans in this mitigation/adaptation strategies. Shouldn't they be involved as well? weren't these results given back to them?

If you don't explore this topic it looks like you investigated this tribally important service but then you do not consider the tribe in the response to these scenarios

Response 27: We recognize that we need to be more explicit with our management strategies and clarified this section, see lines 463-464.

Point 28: as said above, this ecosystem service classification is not correct and should be changed

Response 28: We updated this section, see Table A1.

Point 29: How did you model the grazing quality without knowing the species that provide this ecosystem service?

Response 29: We did not model grazing quality per se, we related it to our DGVM grasslands vegetation type and modelled future changes to grasslands. There were no specific species that were identified for grazing quality by our stakeholders. We clarified this on lines 148-150.

Reviewer 2 Report

Manuscript Forests-799025 by Case et al. described potential impacts of 4 climate change scenarios on ecosystem service using the MC2 vegetation model.  I found the topic very timely and of broad interest and the manuscript overall well-written.  I do have some concerns, primarily regarding the presentation of the MC2 model, and a few more minor concerns that are listed below.

Major Concerns

• My biggest concern is that there needs to be more transparency about the DGVM used in this study. Not all DGVMs operate the same (as acknowledged briefly at the end of the discussion) and it’s important to have more clarity about the key aspects of MC2 early in the manuscript (at the very least in methods) to be better able to interpret the results and potential strengths and caveats.  Based on reading between the lines, it sounds like MC2 simulates ecosystem types rather than plant functional types that could be used to classify an ecosystem type.  Is this accurate?  Similarly, more information is needed regarding what dictates ecosystem distribution across the landscape.  Are distributions based on ecophysiological processes and some metric of competition or is it based on bioclimatic indices?  Relatedly, which meteorological variables from CMIP5 were used to produce change and at what temporal time step?  Was it just temperature and precipitation or does MC2 use a suite of additional variables such solar radiation and humidity?  Does MC2 have spatial interactions or are dynamics in pixels independent of the surrounding ones?  At least a general overview is important for placing results into the context of projections from other DGVMs.  
• I think the short paragraph at the end of the discussion (lines 435-445) highlights why I consider this lack of detail problematic.  In this paragraph, the authors attempt to address some of MC2’s shortcomings, but the context of what MC2’s key features are and what it DOES simulate have not been described at a high level.  I do think this paragraph is important, but I think these caveats can be placed into better context both with how they may impact the predictions and how that places MC2 in the context of other models.
• This lack of information about the model becomes most problematic in the discussion paragraph starting in line 372 lacks presented evidence to support their assertions about specific impacts of climate change driving patterns.  For example, if lengthening of growing season is a key mechanism for vegetation shifts (lines 373-375), then I expect there to be an analysis demonstrating that the selected drivers had an impact on vegetation phenology that connects these.  At the very least, a citation that connects these using MC2 is important.  Similarly lines 381-383 make assertions about declining summer soil moisture in specific vegetation types, but no supporting evidence linking meteorological drivers, environmental variables, and vegetation response.  Drawing this connection seems particularly problematic given that in line 443, the authors state the MC2 does not simulate plant hydraulics. Description of geographic shifts in ecosystem type distribution alone are insufficient to support claims about specific causing the change.
• I appreciate that the authors have done a commendable job of translating model output into ecosystem services for an important and often overlooked stakeholder group.  Nonetheless because all of the results and discussion rely on a particular model, it is important that more context be provided to understand the caveats.  Reliance on MC2 is not a problem -- the key information about the model and how it impacts the results, and particularly potentially difference among the chosen climate scenarios is important. 

More minor concerns

• I am not sure that just the mean GCM for the analyses in 3.2 is sufficiently explained in the methods and discussion.  This choice could be justified if it was in the middle of all the other GCMs for change in veg types, but this doesn’t appear to be really true, so I feel it might be important to portray some measure of the variability or sensitivity of these changes among GCMs.  I understand computational limitations might be at play here, but it seems like this would then warrant greater discussion of how different GCMs could impact species-specific interpretations in the discussion, such as in the paragraph starting in line 384.
• Why just RCP 8.5 and not the generally standard contrast of RCP 4.5 and 8.5?  The discussion in lines 421-434 hints at some of the caveats with this choice, but I don’t think goes far enough in acknowledging the potential bias in results by choice of RCP or cherry-picking of scenarios.  Choosing scenarios from RCP8.5 is likely to be the most dramatic representation of change, which might be useful for decision making, but this was not an argument made and is generally not considered a robust representation of uncertainty. I recommend being more up front with some choices such as this in the methods.
• Maybe add table or label Fig 2 with the representative condition of the chosen GCMs and use lingo throughout results to refer to “hot-wet” scenario etc. rather than GCM name.  (I work with many GCMs and can’t ever keep the name straight with the relative projection.)  I think this would really help make sense of trends in Fig 5 where seeing a large scatter makes it hard to make sense of trends.

Specific Points

• Figure 3, 4 -- black outlines for reservations are not clearly visible -- please make borders thicker or something to make them easier to find.
• Fig 5 -- In this figure, I’m finding it hard to disentangle scenario trends from the baseline mean percent of the study area.  I think it’s because the “historical” gets lost in the wide range of colors shapes.  Might make more sense to standardize everything to baseline to better show what shows gain/loss and what is consistent -- especially because the “historical” point should be the same in both graphs.  If you want to show differences in baseline cover you could add a third panel with just bars for historical so that the current balance of biome types is clearer.  If leave as is, I recommend changing your symbols and making historical the line (and possibly bigger) since right now MRI-CGCM3 looks like the oddball shape-wise rather than your reference condition.
• Figure 5 -- It looks like the biomes are ordered by aridity. If so, I recommend stating this in the legend just to make the reason a bit clearer.
• Figure 6 -- is there a reason this is oriented in its current form as opposed to 90-degrees?  Also, the y-axis label and caption need a better description.  I spent quite a while thinking this was a histogram, but the y-axis is the net number of scenarios showing change, right?  This is basically some sort of score, but doesnt give an indicator of the variability among scenarios or ecosystem services.  The double-variability (variability among plants within service + variability among GCM scenarios) make how to show this tricky and I think the result is a pretty good approach, but at least this reader needed some reminder about your index.
• Get rid of the green background on plots.  I realize this was an initial submission, but it was very distracting.  Also, please double check for color-blind friendly pallets -- most figures did not render well on my non-color printer. I think the base colors are close, but the darkness is too similar to translate well.  In Figure 6, I was unable to distinguish among birds, arid plants, and grazing quality as well as birds and first foods.  Even on my color monitor, I have trouble distinguishing the color difference between mammals and aridland plants.

Author Response

Point 1: My biggest concern is that there needs to be more transparency about the DGVM used in this study. Not all DGVMs operate the same (as acknowledged briefly at the end of the discussion) and it’s important to have more clarity about the key aspects of MC2 early in the manuscript (at the very least in methods) to be better able to interpret the results and potential strengths and caveats.  Based on reading between the lines, it sounds like MC2 simulates ecosystem types rather than plant functional types that could be used to classify an ecosystem type.  Is this accurate?  Similarly, more information is needed regarding what dictates ecosystem distribution across the landscape.  Are distributions based on ecophysiological processes and some metric of competition or is it based on bioclimatic indices?  Relatedly, which meteorological variables from CMIP5 were used to produce change and at what temporal time step?  Was it just temperature and precipitation or does MC2 use a suite of additional variables such solar radiation and humidity?  Does MC2 have spatial interactions or are dynamics in pixels independent of the surrounding ones?  At least a general overview is important for placing results into the context of projections from other DGVMs.  

Response 1: We agree that the manuscript requires more information about MC2 and have added more text in section 2.3: Vegetation Projections. We also added more details about the climate data in section 2.2: Climate Change Projections.

Point 2: I think the short paragraph at the end of the discussion (lines 435-445) highlights why I consider this lack of detail problematic.  In this paragraph, the authors attempt to address some of MC2’s shortcomings, but the context of what MC2’s key features are and what it DOES simulate have not been described at a high level.  I do think this paragraph is important, but I think these caveats can be placed into better context both with how they may impact the predictions and how that places MC2 in the context of other models.

Response 2: We agree and have added more information in the methods about MC2 and have expanded the discussion of MC2’s shortcomings and limitations in the Discussion, see lines 494-511.

Point 3: This lack of information about the model becomes most problematic in the discussion paragraph starting in line 372 lacks presented evidence to support their assertions about specific impacts of climate change driving patterns.  For example, if lengthening of growing season is a key mechanism for vegetation shifts (lines 373-375), then I expect there to be an analysis demonstrating that the selected drivers had an impact on vegetation phenology that connects these.  At the very least, a citation that connects these using MC2 is important.  Similarly lines 381-383 make assertions about declining summer soil moisture in specific vegetation types, but no supporting evidence linking meteorological drivers, environmental variables, and vegetation response.  Drawing this connection seems particularly problematic given that in line 443, the authors state the MC2 does not simulate plant hydraulics. Description of geographic shifts in ecosystem type distribution alone are insufficient to support claims about specific causing the change.

Response 3: We have added a number of citations that illustrate these driving mechanisms within MC2, see lines 429, 432, and 437.

Point 4: I appreciate that the authors have done a commendable job of translating model output into ecosystem services for an important and often overlooked stakeholder group.  Nonetheless because all of the results and discussion rely on a particular model, it is important that more context be provided to understand the caveats.  Reliance on MC2 is not a problem -- the key information about the model and how it impacts the results, and particularly potentially difference among the chosen climate scenarios is important. 

Response 4: We agree and have expanded our discussion to highlight MC2’s limitations and shortcomings, see lines 494-511.

Point 5: I am not sure that just the mean GCM for the analyses in 3.2 is sufficiently explained in the methods and discussion.  This choice could be justified if it was in the middle of all the other GCMs for change in veg types, but this doesn’t appear to be really true, so I feel it might be important to portray some measure of the variability or sensitivity of these changes among GCMs.  I understand computational limitations might be at play here, but it seems like this would then warrant greater discussion of how different GCMs could impact species-specific interpretations in the discussion, such as in the paragraph starting in line 384.

Response 5: We clarified our use of the CESM1-CAM5 GCM for this portion of the analysis. The decision to use the ‘mean’ GCM was made in consultation with our stakeholders and we also clarify this in the methods. We further highlight how other GCMs could impact species-specific interpretations in the discussion.

Point 6: Why just RCP 8.5 and not the generally standard contrast of RCP 4.5 and 8.5?  The discussion in lines 421-434 hints at some of the caveats with this choice, but I don’t think goes far enough in acknowledging the potential bias in results by choice of RCP or cherry-picking of scenarios.  Choosing scenarios from RCP8.5 is likely to be the most dramatic representation of change, which might be useful for decision making, but this was not an argument made and is generally not considered a robust representation of uncertainty. I recommend being more up front with some choices such as this in the methods.

Response 6: We appreciate this comment and have clarified that our choice of using RCP 8.5 was largely a result of discussions of our stakeholders.

Point 7: Maybe add table or label Fig 2 with the representative condition of the chosen GCMs and use lingo throughout results to refer to “hot-wet” scenario etc. rather than GCM name.  (I work with many GCMs and can’t ever keep the name straight with the relative projection.)  I think this would really help make sense of trends in Fig 5 where seeing a large scatter makes it hard to make sense of trends.

Response 7: We use the GCM names instead of the informal descriptions (e.g., “hot-wet”) throughout the manuscript because previous reviewers have suggested this is more accurate and appropriate given that “hot-wet” is a relative description.

Point 8: Figure 3, 4 -- black outlines for reservations are not clearly visible -- please make borders thicker or something to make them easier to find.

Response 8: The figures embedded within the manuscript are unfortunately lower resolution than the final, separate figures submitted to the journal. The black outlines are much more visible in the final, high-resolution versions.

Point 9: Fig 5 -- In this figure, I’m finding it hard to disentangle scenario trends from the baseline mean percent of the study area.  I think it’s because the “historical” gets lost in the wide range of colors shapes.  Might make more sense to standardize everything to baseline to better show what shows gain/loss and what is consistent -- especially because the “historical” point should be the same in both graphs.  If you want to show differences in baseline cover you could add a third panel with just bars for historical so that the current balance of biome types is clearer.  If leave as is, I recommend changing your symbols and making historical the line (and possibly bigger) since right now MRI-CGCM3 looks like the oddball shape-wise rather than your reference condition.

Response 9: Modified as suggested.

Point 10: Figure 5 -- It looks like the biomes are ordered by aridity. If so, I recommend stating this in the legend just to make the reason a bit clearer.

Response 10: Figure 5 is not ordered by aridity.

Point 11: Figure 6 -- is there a reason this is oriented in its current form as opposed to 90-degrees?  Also, the y-axis label and caption need a better description.  I spent quite a while thinking this was a histogram, but the y-axis is the net number of scenarios showing change, right?  This is basically some sort of score, but doesnt give an indicator of the variability among scenarios or ecosystem services.  The double-variability (variability among plants within service + variability among GCM scenarios) make how to show this tricky and I think the result is a pretty good approach, but at least this reader needed some reminder about your index.

Response 11: Figure 6 is not a histogram. We have added text to further describe the figure.

Point 12: Get rid of the green background on plots.  I realize this was an initial submission, but it was very distracting.  Also, please double check for color-blind friendly pallets -- most figures did not render well on my non-color printer. I think the base colors are close, but the darkness is too similar to translate well.  In Figure 6, I was unable to distinguish among birds, arid plants, and grazing quality as well as birds and first foods.  Even on my color monitor, I have trouble distinguishing the color difference between mammals and aridland plants.

Response 12: We have updated the color schemes on figures (e.g., Figure 6) and there is no background green color on plots, this was an unfortunate legacy of converting the manuscript from a word document to a PDF.

Reviewer 3 Report

This study utilizes a global dynamic vegetation model to assess the vulnerability of tribally relevant ecosystem services to climate change in the Pacific Northwest. The authors demonstrate a transferrable approach to such an analysis by first translating cultural resources to vegetation types modeled by the DGVM MC2. Using GCM’s representing four plausible future scenarios, the study maps projected changes in potential natural vegetation distributions and associated ecosystem services. The results identify potentially vulnerable regions to guide preservation and adaptation efforts. The paper tackles societally relevant questions using scientific tools which creates challenges but is highly valuable and I recommend revisions.

Main comments:

1. The four GCM’s were chosen to represent a range of projected change in annual mean temperature and precipitation. However, the annual mean temperature and precipitation are not highly relevant climate measures for ecosystems. The timing of precipitation and annual cycle in temperature are more influential on which vegetation types can be supported by a region. Therefore, projected changes in timing and variability should also be taken into account when selecting GCM’s to represent a range of possible futures. To support the choice of GCM’s used in this study I would suggest adding text describing how projected change in seasonality and variability can also influence vegetation distributions and adding two figures to the SI: 1) Reproduce figure 2 for each season and 2) reproduce figure 2 for each season showing the projected change in variability of temperature and precipitation.
2. It is not clear how the vegetation distributions simulated with each GCM were averaged across GCMs and differenced between time periods (line 196 & Figures 3/4/5). The projected changes in potential natural vegetation type may not represent the change due to greenhouse gas concentrations, but may be confounded by climate model biases, and/or distorted by averaging.
3. For each time period, was the simulated fractional coverage of each vegetation type averaged over the four simulations, then the dominant vegetation type determined? If not, then how was averaging performed with discrete dominant vegetation types?
4. Were the historical simulations performed with observed historical climate or with the GCM simulated climate for the historical period?
5. Regardless of the answer to a and b, there are issues with comparing average future projections to historical simulations in Figures 3, 4, & 5. The changes in potential natural vegetation type for each GCM should be calculated with respect to the simulated vegetation type using the same GCM during the historical period, and then averaged. Taking the difference and then averaging isolates the change in vegetation distribution due to future climate change. Each GCM has biases in climate, by comparing the future climate simulated by one GCM to the average results from 4 GCMs historical climate the delta represents both the change due to greenhouse gas concentrations and the model bias. Model biases are often on the same order of magnitude as projected climate changes, particularly over small domains. Presently the way projected change in vegetation distribution is presented it is confounded by climate model biases (and also biases due to averaging over the historical period). To illustrate this point, the future climate projections with CANESM2 minus the average historical climate from all 4 models would likely not represent the “hot and wet” future scenario anymore, depending on the model biases in historical climate. The future vegetation distribution should be compared directly to the historical vegetation distribution simulated using the same GCM, and then the average of the vegetation changes can be taken. Future vegetation distributions shown in Figures 3 and 4 should be paired with their respective historical simulation. Figure 5 should show the change in percent coverage for each model relative to the respective model simulation of the historical period.
6. The metric used to indicate a directional change (Figure 6) does not account for the relative size of each vegetation type and can have misleading results. For example, by end-of-century simulations with CESM1-CAM5 a species associated with moist coniferous forest (contracted ~8%), coniferous forest (expanded 1%), and dry coniferous forest (expanded 1%) would receive a score of +1 indicating a potential expansion of suitable habitat, but the percent of the study area covered by these vegetation types actually contracted by ~6%. Converting the expansions/contractions to binary distorts the results. It also distorts comparisons among ecosystem services in figure 6. An alternative would be to retain the % change in habitat, particularly since this study is not aiming to identify where specific changes are projected to occur.
7. Mapping ecosystem services to MC2 vegetation types introduces considerable uncertainties which impact the results presented in this study. For example, huckleberries may be associated with moist coniferous forests but not all moist coniferous forests provide suitable habitat for huckleberries. Expansion/contraction of moist coniferous forests that historically did not support huckleberries do not result in changes in suitable habitat. Can additional information be used to constrain the range of suitable habitat for particular species? Are soils data informative? Are there known temperature thresholds for specific species? If not, more text should be devoted to acknowledging these uncertainties and discussing the potential implications for results.

Minor comments

Please clarify how uncertainty was leveraged in this study. Variability among model projections was used to examine several plausible future scenarios but this does not leverage uncertainty. I would suggest rewording to communicate that model-to-model variability can be a tool to explore and communicate a range of plausible futures (lines 21/32/99).

 “Suitable habitat” may be too precise of a term to use when referring to expansion/contraction of potential natural vegetation type. Suitable habitat implies that a specific area can sustain a species, but the method used in this study relates species occurrence to potential natural vegetation type rather than rigorously determining an area as suitable habitat.

The discussion could benefit from a section on how anthropogenic influence may affect results with respect to changes in potential natural vegetation.

Table 1 only includes examples of the species associated with each MC2 vegetation type. The results cannot be fully reproduced without a table showing which vegetation types each species in Table B1 was associated with.

For bitterbrush, areas lower than 700m are excluded from the definition of suitable habitat. Is this elevation limit is due to a temperature threshold; and would the elevation limit be lower under future climate conditions? Is potential suitable habitat excluded by setting the elevation limit?

Please clarify if MC2 simulations show changes in dominant vegetation type or density, and whether more than one vegetation type can occupy a grid cell (no co-occurrence).

Line 21: Typo: emphasize

Line 30: Typo: inherent uncertainty associated with using model output

Line 83: Ref. 20 does not explicitly address climate impacts on huckleberry and bitterbrush so the following sentence beginning on line 83 should either provide a reference (perhaps ref. 51/52) or be reworded to identify that this is the first study to investigate the climate risk to huckleberry and bitterbrush.

Line 99: Typo: emphasize

Line 112: Section 2.1 Since this is intended to be a reproducible methodology, it may be worth adding more description of the stakeholder engagement process. Were the methods of ref. 34 explicitly followed? Were specific methods or consultation formats more useful? Did any aspects of the consultation process present challenges? How were differing perspectives reconciled? Did stakeholders provide feedback on the process or the outcomes? Acknowledging the complexity associated with stakeholder engagement would make this study a more valuable methodological guide when applied in other regions. 

Line 112: Section 2.1 Stakeholder viewpoints are generally referred to as a monolith throughout this section. Changing the language to acknowledge the diversity of input and perspectives would better describe the depth of breadth of knowledges incorporated into the study.  

Line 138: Missing references after 34: Could not review references from this point on. 

Line 147: Typo: four GCMs

Line 147: Reword for accuracy. The four GCM’s do not span the range of future climate space, only a range in annual mean temperature and precipitation climatology. For example, seasonality, modes of variability, and trends were not taken into account.

Line 153: Section 2.2 – Were stakeholders involved in these decisions? If so please describe, if not please clarify.

Line 167: Edit “the range” to “a range”

Line 200: Typo projected

Line 201: Typo projected

Line 213: Typo: medicinal purposes

Line 243: typo: identified

Line 251: Suggested edit for accuracy: Potential natural vegetation types

Line 264: Figures 3 & 4: Please clarify what “historical” represents. Are these simulations performed under historical observed climate conditions or is it the ensemble mean of simulations under GCM historical climate?

Table 2. An estimate of the uncertainty associated with projected changes in suitable habitat for huckleberry and bitterbrush can be added by repeating the analysis with the 3 other GCM’s. This would add considerable credibility to the magnitude and direction of projected changes.

Figure 7. The blue regions of expansion are hard to distinguish from the green stable green regions. Consider revising color palette.

Figure 7. Are the changes shown with respect to the historical simulation using CESM1-CAM5?

Line 335: Typo “winners”

Line 452: Typo: associated with using

Author Response

Point 1: The four GCM’s were chosen to represent a range of projected change in annual mean temperature and precipitation. However, the annual mean temperature and precipitation are not highly relevant climate measures for ecosystems. The timing of precipitation and annual cycle in temperature are more influential on which vegetation types can be supported by a region. Therefore, projected changes in timing and variability should also be taken into account when selecting GCM’s to represent a range of possible futures. To support the choice of GCM’s used in this study I would suggest adding text describing how projected change in seasonality and variability can also influence vegetation distributions and adding two figures to the SI: 1) Reproduce figure 2 for each season and 2) reproduce figure 2 for each season showing the projected change in variability of temperature and precipitation.


Response 1: We agree that there are many alternative ways to characterize the qualities of GCMs relative to each other. However, in the Pacific Northwest region the climate has a strong Mediterranean pattern, where the seasonal curves for precipitation and temperature are clearly sinusoidal, with a single maximum per year. Therefore, versions of Figure 2 using only one season would produce plots that are nearly identical. In any case, we recognize that this is a valid issue for the reader to consider, so we inserted text in the section referring to Figure 2 to say that annual averages are similar to seasonal averages in this region, and that the figure only captures GCM traits broadly, and that individual organisms may respond to finer features of GCMs in different ways. See lines 178-184.

Point 2: It is not clear how the vegetation distributions simulated with each GCM were averaged across GCMs and differenced between time periods (line 196 & Figures 3/4/5). The

projected changes in potential natural vegetation type may not represent the change due to greenhouse gas concentrations, but may be confounded by climate model biases, and/or distorted by averaging.

Response 2: We understand that our description could be clearer. We clarified that we averaged the percent of the study area for each vegetation type across the four GCMs and two future time periods. We then compared the historical and averaged percent area for the futures. See lines 231-232. As suggested, we further highlight some of the confounding biases and issues in the discussion, see lines 490-491.

Point 3: For each time period, was the simulated fractional coverage of each vegetation type averaged over the four simulations, then the dominant vegetation type determined? If not, then how was averaging performed with discrete dominant vegetation types?

Response 3: Thanks, this question relates to Point 2 - we did not average the vegetation types across multiple GCMs. We attempted to clarify this in our Response 2.

Point 4: Were the historical simulations performed with observed historical climate or with the GCM simulated climate for the historical period?

Response 4: We apologize that this was not very clearly described. The results that we label “historical” (e.g., in Figures 3, 4 and 5) were simulated using PRISM data, which is observed historical climate (Daly et al. 2001[1]). We added this information and the citation to section 2.2: Climate Change Projections and 2.3: Vegetation Projections. PRISM also served as the reference dataset for the NEX-DCP30, the future climate dataset we used, so PRISM and the historical portions of the GCM simulated climate have good agreement. However, where we present projected changes in habitat (vegetation types), those were calculated by comparing the future projection against the historical portion of the same simulation (i.e., driven by the same GCM). We inserted text in Section 2.4 to make this point clear.

Point 5: Regardless of the answer to a and b, there are issues with comparing average future projections to historical simulations in Figures 3, 4, & 5. The changes in potential natural vegetation type for each GCM should be calculated with respect to the simulated vegetation type using the same GCM during the historical period, and then averaged. Taking the difference and then averaging isolates the change in vegetation distribution due to future climate change. Each GCM has biases in climate, by comparing the future climate simulated by one GCM to the average results from 4 GCMs historical climate the delta represents both the change due to greenhouse gas concentrations and the model bias. Model biases are often on the same order of magnitude as projected climate changes, particularly over small domains. Presently the way projected change in vegetation distribution is presented it is confounded by climate model biases (and also biases due to averaging over the historical period). To illustrate this point, the future climate projections with CANESM2 minus the average historical climate from all 4 models would likely not represent the “hot and wet” future scenario anymore, depending on the model biases in historical climate. The future vegetation distribution should be compared directly to the historical vegetation distribution

simulated using the same GCM, and then the average of the vegetation changes can be taken. Future vegetation distributions shown in Figures 3 and 4 should be paired with their respective historical simulation. Figure 5 should show the change in percent coverage for each model relative to the respective model simulation of the historical period.

Response 5: We agree with the Reviewer and we did use the GCM simulations of historical vegetation when calculating the percent change for future simulations (see Response 4 for more information). And we did not average future projections of vegetation types. We clarify this in the Methods.

Point 6: The metric used to indicate a directional change (Figure 6) does not account for the relative size of each vegetation type and can have misleading results. For example, by end-of century simulations with CESM1-CAM5 a species associated with moist coniferous forest (contracted ~8%) coniferous with moist coniferous forest (contracted ~8%), coniferous forest (expanded 1%), and dry coniferous forest (expanded 1%) would receive a score of +1 indicating a potential expansion of suitable habitat, but the percent of the study area covered by these vegetation types actually contracted by ~6%. Converting the expansions/contractions to binary distorts the results. It also distorts comparisons among ecosystem services in figure 6. An alternative would be to retain the % change in habitat, particularly since this study is not aiming to identify where specific changes are projected to occur.

Response 6: We recognize that our use of a binary metric is simplistic. However, we settled on this method after consultation with our stakeholders. Significantly modifying our methodological approach would compromise our stakeholder process. Additionally, some vegetation types have dramatic differences in their percent change in habitat between historical and future. For example, dry shrub steppe covers essentially 0% of the study area during the historical simulation and 7% in the future simulation, resulting in a 71,109,732,590% increase. Nevertheless, we added some text in the methodology to clarify our decision to use this method and to highlight the caveats with our binary metric approach. See lines 233-249.

Point 7: Mapping ecosystem services to MC2 vegetation types introduces considerable uncertainties which impact the results presented in this study. For example, huckleberries

may be associated with moist coniferous forests but not all moist coniferous forests provide suitable habitat for huckleberries. Expansion/contraction of moist coniferous forests that historically did not support huckleberries do not result in changes in suitable habitat. Can additional information be used to constrain the range of suitable habitat for particular species? Are soils data informative? Are there known temperature thresholds for specific species? If not, more text should be devoted to acknowledging these uncertainties and discussing the potential implications for results.

Response 7: Excellent points and we agree with you. We added text to clearly state these uncertainties and offer some future research opportunities, as suggested. See lines 506-511.

Point 8: Please clarify how uncertainty was leveraged in this study. Variability among model projections was used to examine several plausible future scenarios but this does not leverage

uncertainty. I would suggest rewording to communicate that model-to-model variability can be a tool to explore and communicate a range of plausible futures (lines 21/32/99).

Response 8: We agree with your comment and have changed our use of the word, “leveraged” to “explored”.

Point 9: “Suitable habitat” may be too precise of a term to use when referring to expansion/contraction of potential natural vegetation type. Suitable habitat implies that a specific area can sustain a species, but the method used in this study relates species occurrence to potential natural vegetation type rather than rigorously determining an area as suitable habitat.

Response 9: Good point and we replaced the term “suitable habitat” with potential habitat throughout the manuscript.

Point 10: The discussion could benefit from a section on how anthropogenic influence may affect results with respect to changes in potential natural vegetation.

Response 10: Updated as suggested, see lines 502-503.

Point 11: Table 1 only includes examples of the species associated with each MC2 vegetation type. The results cannot be fully reproduced without a table showing which vegetation types each species in Table B1 was associated with.

Response 11: We have included a new table in the appendix with individual species associated with each MC2 vegetation type, as suggested. See “AppendixB_TableB1.xlsx”.

Point 12: For bitterbrush, areas lower than 700m are excluded from the definition of suitable habitat. Is this elevation limit is due to a temperature threshold; and would the elevation limit be lower under future climate conditions? Is potential suitable habitat excluded by setting the elevation limit?

Response 12: The elevation limit for bitterbrush is a reflection of precipitation thresholds, not temperature thresholds. It is not known how climate change would affect these precipitation patterns and therefore we decided to hold this limitation constant for future simulations. We clarified this in the manuscript, see lines 280-282.

Point 13: Please clarify if MC2 simulations show changes in dominant vegetation type or density, and whether more than one vegetation type can occupy a grid cell (no co-occurrence).

Response 13: We clarified this by inserting the following text into Section 2.3: Only a single dominant tree or shrub vegetation type may occupy a cell, with grass in the understory [Bachelet et al. 2001]. MC2 simulates per area carbon stocks, but not plant density.

Point 14: Line 21: Typo: emphasize

Response 14: Corrected.

Point 15: Line 30: Typo: inherent uncertainty associated with using model Output

Response 15: Corrected.

Point 16: Line 83: Ref. 20 does not explicitly address climate impacts on huckleberry and bitterbrush so the following sentence beginning on line 83 should either provide a reference (perhaps ref. 51/52) or be reworded to identify that this is the first study to investigate

the climate risk to huckleberry and bitterbrush.

Response 15: We rewrote this section in question in response to a previous comment, please see 86-97.

Point 17: Line 99: Typo: emphasize

Response 17: Corrected.

Point 18: Line 112: Section 2.1 Since this is intended to be a reproducible methodology, it may be worth adding more description of the stakeholder engagement process. Were the methods of ref. 34 explicitly followed? Were specific methods or consultation formats more useful? Did any aspects of the consultation process present challenges? How were differing perspectives reconciled? Did stakeholders provide feedback on the process or the outcomes? Acknowledging the complexity associated with stakeholder engagement would make this study a more valuable methodological guide when applied in other regions.

Response 18: We added text and a relevant appendix to better describe our stakeholder process, see lines 125-129 and Appendix A. Although we agree that detailed descriptions of the process and issues that arose during the stakeholder consultation, we feel it’s out of the scope of for this manuscript.

Point 19: Line 112: Section 2.1 Stakeholder viewpoints are generally referred to as a monolith throughout this section. Changing the language to acknowledge the diversity of input and perspectives would better describe the depth of breadth of knowledges

incorporated into the study.

Response 19: Good point, we added some text to this section as suggested, see lines 125-130.

Point 20: Line 138: Missing references after 34: Could not review references from this point on.

Response 20: We apologize for this error and have updated the references accordingly.

Point 21: Line 147: Typo: four GCMs

Response 21: Corrected.

Point 22: Line 147: Reword for accuracy. The four GCM’s do not span the range of future climate space, only a range in annual mean temperature and precipitation climatology. For example, seasonality, modes of variability, and trends were not taken into account.

Response 22: We clarified this section to indicate that our selection of GCMs represents the future range of annual mean temperature and precipitation climate space represented by the full ensemble of GCMs, see lines 160-162.

Point 23: Line 153: Section 2.2 – Were stakeholders involved in these decisions? If so please describe, if not please clarify. unless otherwise stated decisions? If so please describe, if not please clarify.

Response 23: We engaged the stakeholders in part of this process – for example the GCM selection. We clarified this, see line 176.

Point 24: Line 167: Edit “the range” to “a range”

Response 24: Corrected.

Point 25: Line 200: Typo projected

Response 25: Corrected.

Point 26: Line 201: Typo projected

Response 26: Corrected.

Point 27: Line 213: Typo: medicinal purposes

Response 27: Corrected.

Point 28: Line 243: typo: identified

Response 28: Revised as suggested.

Point 29: Line 251: Suggested edit for accuracy: Potential natural vegetation types

Response 29: Revised as suggested.

Point 30: Line 264: Figures 3 & 4: Please clarify what “historical” represents. Are these simulations performed under historical observed climate conditions or is it the ensemble mean of simulations under GCM historical climate?

Response 30: Please see our response to Point 4 above.

Point 31: Table 2. An estimate of the uncertainty associated with projected changes in suitable habitat for huckleberry and bitterbrush can be added by repeating the analysis with the 3 other GCM’s. This would add considerable credibility to the magnitude and

direction of projected changes.

Response 31: We agree and have included a series of maps for both bitterbrush and huckleberry showing their future simulations under difference GCMs for both 2050 and 2100, see Appendix D.

Point 32: Figure 7. The blue regions of expansion are hard to distinguish from the green stable green regions. Consider revising color palette.

Response 32: Revised as suggested.

Point 33: Figure 7. Are the changes shown with respect to the historical simulation using CESM1-CAM5?

Response 33: Yes.

Point 34: Line 335: Typo “winners”

Response 34: Corrected.

Point 35: Line 452: Typo: associated with using

Response 35: Corrected.

[1] Daly, C.; Taylor, G.H.; Gibson, W.P.; Parzybok, T.W.; Johnson, G.L.; Pasteris, P.A. 2001. High-quality spatial climate data sets for the United States and beyond. Transactions of the American Society of Agricultural Engineers. 43:1957–1962.