Assessing Risk and Prioritizing Safety Interventions in Human Settlements Affected by Large Wildfires

Round 1

Reviewer 1 Report

This is an important study which is extremely useful for many locations internationally.

I would like to see the introduction incorporated with the discussion of the 2017 wildfires to provide a clear roadmap indicating all the elements of the paper.

There is a major challenge in presenting the results visually. I recognise that the considerable efforts have been made but perhaps some attempt to further integrate the elements of Figure 3 so that exposure, vulnerability and risk can be more easily visualised simultaneously. Similarly, for Figure 4. Figure 7 (the heat map) is an important summary tool but also needs further work to make the findings more accessible. Perhaps information on the exposure and risk scenarios and key for the colour coding of course.

Can you provide some technical background for equation 1 (Line 210) to support it.

You introduce the term “pixel” in line 276 without a definition. Can you please elaborate?

Given the centrality of the fuel management scenarios I think there is a need to elaborate more on both the forest management activities (Mngt) and the fuel break network (FBN). I would like to have a greater understanding of what these fuel management scenarios actually look like – the type and extent of forest management and how fuel breaks actually work in the landscape.

The choice of fuel management strategy is raised in Line 494 and social, legal, economic and land ownership constraints are recognised but I wonder if it would be worth a sentence or two discussing these in the context of the constraints on fuel management initiatives.

Sheltering is also important in your analysis but I don’t feel as though I have a good understanding of the nature of shelters that are deemed to be adequate given obviously that the risk for many residents relates to the need to travel sometimes long distances to a shelter.

I feel that the manuscript could be improved and meanings clarified if English language could be edited.

Author Response

This is an important study which is extremely useful for many locations internationally.

I would like to see the introduction incorporated with the discussion of the 2017 wildfires to provide a clear roadmap indicating all the elements of the paper.

R: we appreciate the comments and suggestions of the reviewer. We have added some more information about the wildfires of 2017, please see L77-81, regarding the driving conditions, and L94-98, regarding the impacts in the study area.

There is a major challenge in presenting the results visually. I recognise that the considerable efforts have been made but perhaps some attempt to further integrate the elements of Figure 3 so that exposure, vulnerability and risk can be more easily visualised simultaneously. Similarly, for Figure 4. Figure 7 (the heat map) is an important summary tool but also needs further work to make the findings more accessible. Perhaps information on the exposure and risk scenarios and key for the colour coding of course.

R: We understand the issue raised, but we were not able to build a figure with a different format that could convey all the results like these ones. The superimposition of the 3 layers of fig. 3 (now Fig. 5), for example, would not allow the distinction of the ranking level of each layer/risk component, which are also relevant results. As suggested, we have the colour code in Fig. 7 (now Fig. 9). Regarding the information about the scenarios, we added few details on the caption and refer to Table 3 for further details, to avoid a very long caption that would be difficult to follow.

Can you provide some technical background for equation 1 (Line 210) to support it.

R: The equation was based on the technical guidelines of the National Forest Services, particularly the weighting of population and buildings (please see L229-230). This option is consistent with our purpose to provide results useful and coherent with operational activities in the country.

You introduce the term “pixel” in line 276 without a definition. Can you please elaborate? ok

R: we have changed the term to grid cell, to be more accurate and clearer.

Given the centrality of the fuel management scenarios I think there is a need to elaborate more on both the forest management activities (Mngt) and the fuel break network (FBN). I would like to have a greater understanding of what these fuel management scenarios actually look like – the type and extent of forest management and how fuel breaks actually work in the landscape.

R: We understand the issue raised and we have added the following:

- Text with further information of the different options, section 2.4, L341-369;

- A figure with the proposed location of the fuel-break network, in the 3 different priorities (now Fig.4, page 11).

These scenarios are explained in detail elsewhere, specifically in a technical report and an article to be published in the same issue (Refs. 49, 53), which resulted from the same project, therefore our first description was not detailed.

The choice of fuel management strategy is raised in Line 494 and social, legal, economic and land ownership constraints are recognised but I wonder if it would be worth a sentence or two discussing these in the context of the constraints on fuel management initiatives.

R: we have added further discussion on this matter, please see L555-559 for some of the challenges landowners need to face in that region and that affect land management

Sheltering is also important in your analysis but I don’t feel as though I have a good understanding of the nature of shelters that are deemed to be adequate given obviously that the risk for many residents relates to the need to travel sometimes long distances to a shelter.

R: We have clarified the selection of shelters and their required conditions; this is described in L281-290. The conditions for a shelter are defined in the programme “Safe villages, safe people” and these were followed to carry out the inventory through field work.

I feel that the manuscript could be improved and meanings clarified if English language could be edited.

R: We have made a general revision of the English language.

Author Response File: Author Response.docx

Reviewer 2 Report

GENERAL COMMENTS

This study evaluates wildfire risk and the capacity to cope with that risk in a region of central Portugal that was affected by large wildfires in 2017. It characterizes the fire hazard, exposure, vulnerability, and coping capacity of settlements throughout the study region. Then it evaluates how two management activities (fuel management and a fuel break network), with different levels of management intensity could reduce future wildfire risk. Overall, it was a comprehensive evaluation of the risk environment in this region.

The main limitation was that little information was provided on the specific activities included in the fuel management or fuel break networks or how the effectiveness of those treatments was evaluated. For example, did vegetation management involve thinning, clearcutting, the use of prescribed burning, or some combination of these treatments? How wide were the fuel breaks and what guided the placement of fuel breaks across the landscape? Although the management was implemented with different levels of intensity, I did not understand how the different treatment types (fuel management vs fuel break network) or the different levels of management affected the degree of exposure to wildfire.

Were the management activities incorporated into the FARSITE modeling, or were fuel breaks and fuel management always assumed to achieve the desired effects? Is there any information about how long the effects of the management strategies persist or the range of weather conditions under which the management strategies are or are not effective? For instance, line 486 acknowledges that fuel management might not be effective under severe fire weather. On lines 397-398, it is suggested that moderately increasing the managed forest area would reduce the proportion of exposed settlements by 32% compared to the business-as-usual scenario, but this should depend on underlying assumptions about the effectiveness of fuel management? Likewise, it is difficult to understand the impact of different forest management scenarios on wildfire risk (lines 408-414) without knowing how the effectiveness of the different management scenarios was evaluated.

The other main area for improvement would be to discuss the degree to which this analysis evaluates vulnerability to future fires, as opposed to characterizing the risk environment for recent fires (is there reason to believe large fires in the near future are likely to follow similar patterns to the recent large fires?). For example, the FUNC-SIM model was calibrated using an ignition probability density surface that was generated based on ignitions that recently led to very large wildfires (> 1,000 ha; line 189). It is not clear why this size threshold was used, but it seems like this approach could make the model seem very effective at recreating what has occurred in the recent past and less effective at predicting exposure to future wildfires. It is reasonable to exclude very small fires if those fires are not consistently represented in the data or if those fires are most likely to occur in places where they are rapidly suppressed. However, each fire that burns more than one hectare (or more than a few hectares) could potentially develop into a very large fire, and excluding these smaller fires when calibrating the probability surface of fire ignitions might make the model less effective at predicting future fires.

Related to the previous comment, is there any reason why fires would spread more easily across recently burned vegetation than vegetation that has remained long unburned in this region? If the vegetation that develops soon after fire is more flammable than long unburned vegetation, then it may be reasonable to conduct modeling based on the locations and patterns of recent large fires. However, if the vegetation that develops soon after fire is less flammable than long unburned vegetation, then areas affected by recent fire may be less fire prone than other parts of the landscape. Also, would the fuel management activities in the study area replicate the effects of wildfire or would they seek to generate vegetation much different from the vegetation that develops following wildfire?

ADDITIONAL COMMENTS

Line 128: When you refer to Alvares being located in a region of Portugal with “high fire frequency including several large fires,” it would be useful to give some quantitative data on fire frequency and fire sizes.

Lines 336-337: Delete these lines. It looks like it was copied from the guidelines for authors.

Line 510: In the cases where evacuations are not feasible and there is no structure available that could serve as a safe shelter on-site, the authors suggest that a new structure could be built specifically for that purpose. Is this a realistic option? What is the funding source for these new buildings that could be used as shelters from wildfire?

Author Response

GENERAL COMMENTS

This study evaluates wildfire risk and the capacity to cope with that risk in a region of central Portugal that was affected by large wildfires in 2017. It characterizes the fire hazard, exposure, vulnerability, and coping capacity of settlements throughout the study region. Then it evaluates how two management activities (fuel management and a fuel break network), with different levels of management intensity could reduce future wildfire risk. Overall, it was a comprehensive evaluation of the risk environment in this region.

The main limitation was that little information was provided on the specific activities included in the fuel management or fuel break networks or how the effectiveness of those treatments was evaluated. For example, did vegetation management involve thinning, clearcutting, the use of prescribed burning, or some combination of these treatments? How wide were the fuel breaks and what guided the placement of fuel breaks across the landscape? Although the management was implemented with different levels of intensity, I did not understand how the different treatment types (fuel management vs fuel break network) or the different levels of management affected the degree of exposure to wildfire.

R: We appreciate the thorough comments and suggestions of the reviewer.

R: We understand the comments raised regarding the lack of information about the fuel management and fuel break network scenarios. These scenarios are explained in detail elsewhere, specifically in a technical report and an article to be published in the same issue (Refs. 49, 53), which resulted from the same project, therefore our first description was not detailed. We understand that it is necessary to provide more information on this and, as such, we added the following to the manuscript:

• Text with further information of the different options, section 2.4, L341-369;
• A figure with the planned location of the fuel-break network, in the 3 different priorities (now Fig.4, page 11)

Were the management activities incorporated into the FARSITE modeling, or were fuel breaks and fuel management always assumed to achieve the desired effects? Is there any information about how long the effects of the management strategies persist or the range of weather conditions under which the management strategies are or are not effective? For instance, line 486 acknowledges that fuel management might not be effective under severe fire weather. On lines 397-398, it is suggested that moderately increasing the managed forest area would reduce the proportion of exposed settlements by 32% compared to the business-as-usual scenario, but this should depend on underlying assumptions about the effectiveness of fuel management? Likewise, it is difficult to understand the impact of different forest management scenarios on wildfire risk (lines 408-414) without knowing how the effectiveness of the different management scenarios was evaluated.

R: Overall, the management options and the fuel break network implemented were considered to bring changes in fuel distribution and loads. The simulations included different levels of fuel loads/density in their parameters, depending on the scenario considered. This changes the probability that a fire reaches the villages. The fuel breaks were defined with different purposes, according to the priority (please see added text and fig. 4 in the revised manuscript) and followed the technical guidelines defined by the National Forest Services/ICNF (at least 125 m large).  Fuel management for understory vegetation, should occur maximum every 5 years, regardless of the technique used; the simulations incorporate the density/arrangement of fuels for each scenario, assuming that, in the first year when fuels are managed, there is no burnable understory vegetation (assigning the respective fuel class according to the forest type, eucalyptus or pine), progressively increasing to low shrubs after 5 years, when the cycle restarts. This is based on the technical guidelines of ICNF and their prior experience (ICNF was a partner of the project). The weather conditions were integrated in the simulations as another variable and the effectiveness of the forest management activities were not assessed according to weather conditions, due to lack of appropriate data.

The other main area for improvement would be to discuss the degree to which this analysis evaluates vulnerability to future fires, as opposed to characterizing the risk environment for recent fires (is there reason to believe large fires in the near future are likely to follow similar patterns to the recent large fires?).

R: We did not assess how future climatic conditions would affect risk levels, since we focused on the characteristics of the landscape that can directly change with human intervention (fuel management). However, based on recent studies, it is reasonable to assume that future climatic conditions in the study area/region, will cause similar circumstances to those observed in 2017, probably occurring more frequently. We have added text, please see L77-81. We also added in the discussion that the simulations can be adjusted to incorporate weather parameters for future climatic scenarios (L531-532), although this was not our purpose in this project.

For example, the FUNC-SIM model was calibrated using an ignition probability density surface that was generated based on ignitions that recently led to very large wildfires (> 1,000 ha; line 189). It is not clear why this size threshold was used, but it seems like this approach could make the model seem very effective at recreating what has occurred in the recent past and less effective at predicting exposure to future wildfires. It is reasonable to exclude very small fires if those fires are not consistently represented in the data or if those fires are most likely to occur in places where they are rapidly suppressed. However, each fire that burns more than one hectare (or more than a few hectares) could potentially develop into a very large fire, and excluding these smaller fires when calibrating the probability surface of fire ignitions might make the model less effective at predicting future fires.

R: The fire size threshold was based on the fire history of the study area; 90% of the burned area since 1975 was caused by 10 fires which were above 1 000 ha. These are the fires that cause more concerns in the study area, therefore the simulations were adjusted to this reality.

Related to the previous comment, is there any reason why fires would spread more easily across recently burned vegetation than vegetation that has remained long unburned in this region? If the vegetation that develops soon after fire is more flammable than long unburned vegetation, then it may be reasonable to conduct modeling based on the locations and patterns of recent large fires. However, if the vegetation that develops soon after fire is less flammable than long unburned vegetation, then areas affected by recent fire may be less fire prone than other parts of the landscape. Also, would the fuel management activities in the study area replicate the effects of wildfire or would they seek to generate vegetation much different from the vegetation that develops following wildfire?

R: The simulations were calibrated with a fire history of nearly 40 years (1980-2017), to cover the diversity of weather, fuel and fire conditions. Fuel management activities here represent a mean to reduce fuel continuity and loads, rather than changing vegetation types. This is because the existing economic activities in the study area are based on pines and eucalyptus forests, and it is not reasonable to assume this will substantially change in the near future, although other economic activities could be explored (as mentioned in the conclusions). Nevertheless, landcover types such as agriculture and pastures are suggested possibilities for the fuel break network, but this is also represented as different fuel loads (or unburnable) in the simulations. On a general note, the biophysical and climatic conditions of the region foster the growth of shrubland-type vegetation, that rapidly colonizes burned areas and accumulates over time, increasing fuel loads. Eucalyptus also resprout easily, just after 2 weeks, as we have observed in Alvares, whereas pines (obligate seeders) take longer and depend on the available seedbank. The longer the time since last fire, the more fire-prone the area is considered, if no management has been carried out.

ADDITIONAL COMMENTS

Line 128: When you refer to Alvares being located in a region of Portugal with “high fire frequency including several large fires,” it would be useful to give some quantitative data on fire frequency and fire sizes.

R: We have added a figure (now Fig. 1, page 4) with the location of the study area within the region, as well as further data on fire frequency and burned area. Please see L137-140.

Lines 336-337: Delete these lines. It looks like it was copied from the guidelines for authors.

R: the lines were not part of the manuscript, they were kept my mistake and were removed.

Line 510: In the cases where evacuations are not feasible and there is no structure available that could serve as a safe shelter on-site, the authors suggest that a new structure could be built specifically for that purpose. Is this a realistic option? What is the funding source for these new buildings that could be used as shelters from wildfire?

R: The construction of a new building to be used as a fire shelter is one of the options presented by the National Authority of Civil Protection in the Programme “Safe Villages, Safe People”, which is mentioned in the introduction. This link with the programme was added, please see lines L288-289 and L573-575. The implementation of the shelter options and evacuation measures in each village, depends on the actions defined and carried out by the local authorities under their fire defense and emergency plans, and are budgeted accordingly. Our main purpose with this analysis was to assist in the process of diagnosing the availability/need for a shelter and the evacuation time, based on scientific knowledge and technically advanced GIS tools, that could be of use to the operational strategies. It is out of the scope of this work to define how these shelters can be funded, although we know that the safety measures fostered by the Programme are already being implemented in multiple villages across the country.

Author Response File: Author Response.docx