the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Jan 2025
Combining empirical and mechanistic understanding of spruce bark beetle outbreak dynamics in the LPJ-GUESS (v4.1, r13130) vegetation model
Abstract. For evaluating the forests’ performance in a future with changing climate for different management alternatives, dynamic vegetation models are important tools. One of the functions in such models that has a big influence on the results is tree mortality. Bark beetles are important for the pattern of mortality in forest, especially for needle leaved forest in the temperate and boreal zones. The European spruce bark beetle (SBB, Ips typographus) has in the most recent years replaced wind as the most important disturbance agent in European forests. Historically, SBB damage is typically triggered by wind storms as they create breeding material with no defences to overcome for the beetles. Drought can contribute to increased damage and prolonged outbreaks by lowering the defence of the trees, but has been the main driver of some of the European forest damage in the last decade. In this study we implemented a SBB damage module in a dynamic vegetation model (LPJ-GUESS) that includes representation of wind damage and forest management. The module was calibrated against observations of storm and SBB damage in Sweden, Switzerland, Austria and France. An index of the SBB population size that changed over time driven by phenology, drought, storm felled spruce trees and density of the beetle population, was used to scale modelled damage. The model was able to catch the start and duration of outbreaks triggered by storm damage reasonably well but there was a large variability that partly can be related to salvage logging of storm felled forest and sanitary cutting of infested trees. The model showed increased damage in most recent years with warm and dry conditions, although below the level reported, which may suggest that the drought stress response of spruce in LPJ-GUESS is underestimated. The new model forms a basis to explore vulnerability of European forests to spruce bark beetle infestations.
- Preprint
(2436 KB) - Metadata XML
-
Supplement
(1103 KB) - BibTeX
- EndNote
Status: open (extended)
-
CC1: 'Comment on gmd-2024-239', Tomáš Hlásny, 03 Mar 2025
reply
General comments:
This work aimed to develop a module for the simulation of spruce bark beetle impact in the vegetation model LPJ-GUESS. The developed model was calibrated and tested based on data from several European countries, producing plausible and meaningfully accurate results. While the developed model was proved applicable, I found a number of aspects that require clarification. My general comments are:
- The used language is overly technical, and some paragraphs were not possible to understand (I listed some formal comments below). Revision is required to make the text accessible also to readers without deep technical understanding of the presented concept.
- The overall objective of this model development and its intended use were not clearly formulated, which resulted in combinations of simple empirical components with mechanistic components involving bark beetle population dynamics, phenology, etc. It was difficult to comprehend this logic – some implementation steps looked like workarounds helping to resolve technical problems.
- Some aspects of the development (e.g., salvaging implementation, wind impact on growing stock, etc., as in my comments below) seems to be rather arbitrary, lacking robust testing.
- The discussion is vague, not addressing the limitations of the proposed model or critically confronting this development with other works.
Specific comments:
- Bark beetle sub-models have already been implemented in different models; however, the overview of these implementations has not been provided (or provided only partly). I suggest shortening parts of the introduction on disturbance development in Europe and go directly to the current situation in bark beetle modelling and limitations, which this study aims to address.
- The link between bark beetle ecology and population dynamics and implementation of these processes in the developed model was insufficiently documented, which hampers the assessment of this implementation. Bark beetle dynamics are driven by a number of processes (inciting, predisposing, amplifying, terminating the outbreak, etc.), which need to have their adequate counterparts (thought simplified) in the model (in L55 this complexity was somewhat narrowed to phenology and forest conditions). In the current text, the information about beetle`s phenology; effects of windthrows, drought, salvage logging, etc. are rather scattered across the text, not providing a consistent framework to follow and understand the implementation of these processes. A section on Predisposing, triggering and contributing factors was placed on the end of the Results. However, this was done in a very inconsistent way, and the section consists of a single sentence only(?); therefore it is not very helpful.
- The objectives (L85) say that the point is to … „catch(?) outbreak dynamics, utilize empirical relationships where available but make use of suitable mechanistic knowledge … “. This process resolution is very unclear, and it should be clearly defined with regard to the overall objective of this development. If the purpose is to simulate flexibly bark beetle dynamics under different management and climates, including, for example, the recent Europe-wide transitions from wind- to drought-driven dynamics, a high-level of process detail may be needed. If the point is to reproduce past dynamics and use the model in the range of past conditions, more empirical implementation could be OK. However, this objective and decisions about desired process complexity seem to be arbitrary (or insufficiently documented) in the paper. For example, Eq. 3 models bark beetle impact increase rate by the empirical model that used temperature, precipitation windfelled amounts, etc. as predictors. However, this empirical relationship is modified by the population-level processes such as negative feedback from denser beetle populations, swarming-induced competition. It would be very helpful to clarify in the beginning the degree of process complexity this model aims at and, if possible, keep it consistent.
- Does this approach have any justification? „To enable that an outbreak can also be sustained at the highest population levels, the lowest possible total negative feedback from population size is just below the highest possible positive feedback from water stress and phenology (L….)“. Does not it prevent outbreak termination due to internal regulation mechanisms? This approach sounds like a workaround rather than ecologically grounded solutions (but I admit I had a difficulty to understand this part).
- I have a similar concern as in the previous point concerning the implementation of salvage logging (L215-220). This description sounds like a set of arbitrary decisions that technically allowed to include the effects of salvage logging into the model. However, no correspondence with a real dampening effect of salvage logging on outbreaks was presented.
- I am not sure if positive effect on water stress on beetle population growth can be termed as „positive feedback“ (L…). Positive feedback typically has a different meaning. Concerning the term „positive feedback from phenology“ – probably positive effect of temperature of bark beetle population growth (through altering phenology)?
- What is the difference in R in Eq. 3 and R in Eq. 4? I suppose they represent different variables (Eq. 3 increase rate of forest volume loss, Eq. 4 forest damage by bark beetles?; as I inferred from the text), therefore, they should be represented differently. I could not understand how the two Rs from Eq. 3 and Eq. 4 interact.
- L185, Fig. 1 caption – it sounds a bit weird to use term “components” and “ranges of components”. Cannot it be factors, variables, predictors, drivers or so?
- „Shape of the functions for the components of the increase rate“ Should not it be: Response of water stress coefficient driving bark beetle population growth to mean water stress? I generally found the used formulations very unnatural, making it difficult to understand the text .
- It is really extremely difficult to understand the logic of such statement: „As storms in Europe mainly occur in autumn and winter, and as the vegetation and bark beetle effect will be the same for a storm event in October or in February the next year, the storm damage statistics for a specific year were compiled for a storm season of July specified year until June the following year.“ Unfortunately, such formulations are frequent.
- Section 2.4. I could not understand paragraph in L235 describing data availability. L239-240 – does this mean that there was only a single volume value for France and Austria (from 2008) and this single value was used to generate annual time series spanning 1961-2010? I m not convinced if this approach can be considered reliable. In the case of such data limitation, would not it be better to focus on countries with better data coverage?
- L245 – the SDI concept from Patacca and its use in the current study would need to be much more elaborated. How is it to be used for evaluating the shift from wind to drought driven outbreaks in the developed simulation framework, as stated in L242?
- L255, wind implementation. As the text is whole overly complicated, I suggest simplifying this paragraph. It could be written directly that the wind impacts were prescribed to match the observed pattern, without elaborating on the experiment that did not work (driving the damage by real wind series data).
- L260-270. As far as I understand this part, the authors fixed a poor match of simulated wind damage with observations by introducing a correction by latitude, which correlated with productivity in Sweden (unpublished data?), and because there is higher damage in more productive sites, it should help simulate wind damage better. If this is correct, it looks like rather artificial solution for improving model performance. If the wind module is coupled with the vegetation model, should not the productivity and subsequently wind susceptibility be simulated as an emergent property? Without a need to imprint there this pattern externally.
- L 289 and paragraph 295: This calibration procedure does not look rigorous and reproducible. First, the parameters were adjusted based on expert judgement. Second, only 8 out of fourteen parameters were calibrated because „we wanted to keep the range of response function …… as the original Marini model“. I have doubts about such a sequence of arbitrary decisions.
- L400 The switch between stand-alone Mtlab implementation and the „calibrated main base run“ is confusing. Sounds more like developers’ jargon than the text aimed at a broader audience
- L403 – the 70 % overestimation for Austria was because of a single grid cell with missing negative feedback. The problem was fixed by removing this cell. In my opinion, this indicates a broader problem in the implementation, which should be fixed. Currently, it remains unclear how this missing negative feedback affected simulations in other remaining cells, where this effect could have been less pronounced than in the single Austrian cell.
- L410 – It is surprising that a 2° warming could cause such a severe spruce biomass reduction that it exerted a strong dampening effect on bark beetle damage (compensating for the amplifying effect on bark beetle activity). If this was the case, it would require exploring this vegetation feedback in greater detail. This issue was not addressed in Discussion; it just repeated the results.
- Concerning the Discussion – the text much better written and clearer than the previous sections. However, the presented model implementation, its limitations and confrontations with other models were addressed only marginally. The discussion mostly described general aspects of bark beetle dynamics and modelling.Technical corrections
Abstract requires revision. The introductory part on spruce bark beetle is overly long, while motivations for the presented development and the need for this solution are missing. Recommend avoiding terms salvage and sanitary felling in abstract, as their effect on bark beetle dynamics may not be clear to readers without forestry background. I did not notice in the results that the high variability of simulations was due to the variable effect of salvage logging (but I may have misunderstood this part).
L55-60 not only empirical approaches exist, see, for example, implementation in iLand but also in other models
L160-165 negative feedback. The paragraph is not possible to understand, requires revision
L170 – this paragraph seems essential, but I did not manage to understand it. Suggest revising this entire section to make it understandable (and reproducible) also for a reader without deep technical understanding of this framework.
L198 - “that goes from zero to total shutdown of photosynthesis” – consider revising the language
The authors operate across the text (already in abstract) with terms salvage logging and sanitary logging, and effect of these operations on bark beetle outbreaks. This concept can be unclear for the readers as these effects are not properly explained. Moreover, the definition of salvage (removal of windfelled trees) and sanitary (preventative removal of infested trees) is possible, but it is far from generally accepted and used definition.
L214 infected trees. Probably infested.
L322 The sentence is not possible to understand: “To test the robustness of the approach to test the model for different parameter combinations with structure and Lmort prescribed from an LPJ-GUESS simulation with default parameters, LPJ-GUESS was finally run with the optimized parameter set, with feedback of the damage associated with that setting to the simulated vegetation”. Unfortunately, such cases are frequent across the text.
Unclear citation in L141 (Pugh et al. Manuscript)
L331 “common model with the main base-run optimization” Consider please that readers main not have a deep technical understanding of this procedure
L333 The same as above – “It should be noted that the calibration always was based on data from all countries, also when the optimum model for the regions countries was selected, which explains why there is a difference for S Sweden and NE France between Table 3 and Table S1a and between Table S1b and Table S1c.“ It is necessary to find a language that makes these results accessible to and reproducible by a broader community.Citation: https://doi.org/10.5194/gmd-2024-239-CC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 125 | 38 | 9 | 172 | 20 | 8 | 7 |
- HTML: 125
- PDF: 38
- XML: 9
- Total: 172
- Supplement: 20
- BibTeX: 8
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1