This work attempts to validate a new method for past burnt area reconstructions using pollen and charcoal data. The methodology involves complex statistical and geospatial analyses, which are applied for the first time for this purpose. The geographic context is the Iberian Peninsula (IP). While I certainly appreciate the important aim of this work and the large effort in data gathering and manipulation here presented, I think there are some major issues that need to be addressed before this manuscript is accepted.  I am providing details of some major (+ some minor) concerns here below.

Major issues:

• The calibration of area burnt using modern charcoal is not well explained. There is quite limited information about the modern charcoal samples, which seem to be core tops of unknown time coverage. This information if crucial to assess the validity of the approach. Also, at L99 it says that interpolation was used to extract present-day burnt area at each of the sites with modern charcoal records. However, the comparison of the locations of the modern charcoal samples with the GLM output suggests relatively low spatial coverage of the calibration (e.g. a large area burnt fraction was derived using the GLM in north-central Portugal, but there is only one modern charcoal sample in the region). Another figure showing the calibration of area burnt/modern charcoal needs to be presented, at the moment it is unclear how these two match.
• The area burnt fraction reconstruction shown in Figure 4 only ‘matches’ charcoal if we consider a long-term trend perspective. The individual wiggles are often anti-phase, which raise questions about the validity of the approach. Please reconsider the robustness of this validation.
• Using pollen data only to reconstruct area burnt can only work in limited conditions. The underlying assumption of the whole methodology is that vegetation (fuel) availability (derived by pollen assemblages) would vary through time only due to fire activity/spread. This assumption does not work in many regions on Earth and I am not quite sure it would work within some parts of the IP where there are other important controls and where fuel-limitation of fire activity is less dominant. The paper should clarify this limitation and improve its discussion.

Minor issues:

• L30: the assertion that pollen records are more abundant than charcoal records is not valid for many regions on Earth. Maybe this generalisation is true for the Iberian Peninsula, but this has to be clarified.
• LL32-40: the first paragraph of the introduction reads like a series of loosely connected statements. The rationale for this work needs to be apparent in this paragraph, but at the moment it’s quite confusing.
• L42: the Holocene is certainly a period when human agency was ‘pervasive’ in many regions. This is another generalisation that needs to be better expressed.
• L49: remove ‘qualitative’
• LL54-57: this section does not consider climate into the equation, assuming that fires are fuel-limited. To make it work as a general statement, this should include susceptibility to burn. Alternatively, if this is only referring to the Iberian Peninsula, where fuel availability plays a more important role, this needs to be specified. This is still probably a generalisation, but it works better to introduce the study region.
• LL54-64: this whole paragraph is quite jumpy and confusing (starts with fuel availability, then it goes to pollen assemblages as a method to reconstruct past climates)
• LL72-74: This section makes your previous inference about the importance of fuel availability less valid and the whole approach more confusing. I think there needs to be a section introducing the drivers of fires in the IP.
• L79: this is a general reference to the EPD, but a list of record with references needs to be provided in supporting information
• L84: list with references needed for the charcoal records too
• L270: unclear links between paragraphs
• L272: this is of course true, but fire is not the only way to achieve land clearance and this approach assumes pollen assemblages are only varying in response to area burnt
• LL287-289: unclear sentence, confusing how the scarce availability of charcoal records would have led to large-scale patterns (these normally require lots of records)

This is an interesting, if challenging, exercise in numerical data transformation. The results are worth reporting, even if the exercise does not appear to have been especially successful.  What should be removed is the claim in the conclusion/abstract that “this new method opens up the possibility of reconstructing changes in fire regimes quantitatively from pollen records” in regions where charcoal data are lacking. The pollen-burnt area relationship that they established for the Iberian peninsula is not transferrable to other geographical regions, even to adjacent regions such as France. For example, wildfire in most of Iberia is fuel-limited so that burning increases at times of wetter climate, when biomass increases. In contrast, in most of France, there is abundant plant biomass so that fires are caused by drought conditions when vegetation becomes more flammable. There are some specific plant taxa that are fire-sensitive or fire-tolerant such as Cistus monspeliensis, but these indicator species are the exception not the norm. In reality, what the authors have done is to use pollen and charcoal data in combination to fill spatial gaps in coverage within Iberia. They have not shown that charcoal can be replaced by pollen as a palaeo-fire proxy in other regions (e.g., Greece) where charcoal data are lacking. The analysis carried out for Iberia could, in theory, be scaled up to cover larger regions, but they are not transferable from one part of the planet to another. Sub-Saharan Africa, for example, is deficient in charcoal records although African savannahs account for almost half of all wildfires globally. This deficiency cannot be resolved by using the pollen-burnt area relationship in, for example, North America, and transferring it to African pollen records.

At a more fundamental level, trying to use pollen data as a fire proxy also means that pollen cannot then be used to test vegetation-wildfire dynamics and relationships, in the way that Connor et al (2019) did for Iberia during the Holocene. Pollen data, on their own, are not able to provide both cause and effect without falling into the trap of circular reasoning.

How successful was the Iberian test case? Not very successful, as far as I can see. The authors report that pollen data predict charcoal abundances through time “relatively well (R2 = 0.47)”. However, as Figure 4 in the original version of this manuscript shows (Fig. 5 in reply to reviewer comments) this is almost entirely due to a long-term trend during the Holocene towards increased burning.  Centennial or millennial scale peaks and troughs in this graph (no longer included in the paper, nor is the helpful flow chart of methodology – why?) are mostly mis-aligned, a point made already by reviewer 1. In their reply to her/him, the authors also report the results of CCA which shows that burnt area alone explains only 1% of the variability, while other factors explain a much higher share. My guess is that much of this is due to the fact that most of the 29 sites with coupled pollen-charcoal analyses come from two relatively small mountain areas of central Spain (see Figure 1 map in the current manuscript version), so that other regions and biotypes are under-represented in the “training set”.

In summary, with modified conclusions, this exercise is worth reporting, but largely because it highlights the difficulties and challenges of using pollen data on their own as a palaeo-fire proxy.