Review of essd-2025-393

General comments

Overall, this paper is well written and organized, and provides a valuable contribution for efforts to conserve biodiversity in Australia. More generally, discussing and working through a few aspects described below would strengthen this paper.

This approach is relevant to reporting on the Global Biodiversity Framework, as the authors point out. As a result, the alignment of the “pressures” with the stressor/threat taxonomy (https://www.iucnredlist.org/resources/threat-classification-scheme) should be clarified. This may be a semantic difference between the definition of a “pressure” vs. a stressor/threat – is this considered the same as a stressor/threat in the framework? There are 16 pressures mapped here (and in previous datasets fewer). What is the rationale for including these pressures, and not others? Is it lack of data or relevancy? What happens when additional datasets are discovered or created – are they included within an existing pressure or not? E.g,. why are major roads and minor roads a single pressure, when trails and railroads are distinct from roads (and from each other). Please clarify the relationship of the pressures to the stressors framework.

The arbitrary but explicit scoring of impact strength associated with each pressure is described (e.g., a value of 10 for built-up lands) and addressed briefly in the limitations/caveats section. But, work on mapping human pressures generally, particularly in the context of national (or smaller regional, local applications) as described here, would benefit from a more data-driven approach/method to develop the scores. There are many papers that have used the general scoring scheme that this work builds on, but national-level mapping provides an opportunity to further improve how these scores are estimated or assigned – at least mentioned in the limitations section. In particular, methods from decision science that can elicit expert information in more careful, robust ways. In addition, the mapping approach described here would be strengthened by briefly discussing other work on mapping human pressures – at global and national scales (more specifics below).

While human pressure mapping is often used as a surrogate and is clearly a practical approach to provide critical information for informing conservation, it would be valuable to briefly distinguish the difference with ecologic (or habitat) condition, and what the assumptions are and to what situations this applies, e.g., high pressure corresponds to habitat degradation. 

Specific comments

• The terms edge effect, intactness, integrity, fragmentation, and connectivity are used throughout, and part of the paper is on mapping intactness – but it is unclear if these are synonymous or are they different aspects? Rectifying with the biogeography and landscape ecology literature would be valuable, in particular, intactness (as mentioned in the paper) brings in the spatial context/configuration, but is also used to describe individual pixels that are lower than a threshold (e.g., 4). How is intactness different from connectivity (or is intactness a certain aspect of connectivity)? are often used interchangeably. Because the indicator used is intactness, then defining and using it consistently would help to clarify the contribution that it brings to pressure mapping would be more apparent and strengthen a central focus of this paper.

• The uncertainty analysis is helpful – but the map in Figure 4 seems to be generated by extrapolating (using IDW) from the validation points. Because the patterns of human land use can be quite abrupt, an assumption of simple distance decay is overly simplistic. Why not simply show the uncertainty for each pixel (perhaps at a reduced resolution such as 1 km).
• Please clarify what is meant by high or low resolution by specifying quantitatively what you mean (do this consistently throughout the paper).
• L67: Please clarify – perhaps there have been no other efforts carried out in Australia – but there are other efforts to map human pressures at a national level – please revise, or better, provide a sentence or two in the introduction that identifies some of these and how your approach is similar or different, strengths and weaknesses.
• L68-72. There are other global pressure maps that have addressed these pressures (e.g, Theobald et al. 2025), and other national level maps that do meet your criteria.
• L72-74 While additional pressures may have improved the results discussed in Arias-Patino et al., the logical conclusion of adding more and more pressures does not follow – most spatial science literature describes the opposite occurring, of compounding errors due to uncertainty. Please modify this statement to address this limitation.
• L92: “... edge effects from habitat fragmentation…” –  edge effects are typically thought of as a different ecological process but are not fragmentation per se. This is clear in the landscape and road ecology literature… also see Sisk, Fahrig work as examples.

Sisk, T.D., Haddad, N.M. and Ehrlich, P.R., 1997. Bird assemblages in patchy woodlands: modeling the effects of edge and matrix habitats. Ecological applications, 7(4), pp.1170-1180.

Fahrig, L., 2017. Ecological responses to habitat fragmentation per se. Annual review of ecology, evolution, and systematics, 48, pp.1-23.

• Cumulative – land use, mining, cropland, pastureland (so 4 of these are non-overlapping… then is the max value not 50??? 
• L94-96. Please clarify – the statement on how mutually exclusive pressures coexist was difficult to follow. If a pixel has attentive land use, then it can’t have mining cropland or pastureland… then why would they take the max value? I think your assumption/rule is that they are mutually exclusive and so to achieve this you found the maximum value of those 4 pressures. Why weren’t others considered? It seems no other land uses can occur in the middle of a highway?
• Please clarify the method(s) used to transform polygons and lines to raster cells. If each cell of 100 m is binary – if a road touches a pixel, it is assigned that value, correct? Or does it follow a fractional basis? If the former, then effectively pipelines, roads, etc. have a ~50 m buffer on them. What is the consequence on calculation of the area of impact? What if you downscaled to 30 m pixels, or compared it to 1000 m pixels – would the overall HIF be the same? I don’t think so.
• Please clarify, or better calculate the median year of each pressure to describe the “currency” of the map you produced using different years represented by the pressures.
• L119-120, Please clarify – I think that CLUMP was provided in a raster form to you – but it originally was created in vector (polygon) format? This is related to the conversion and representation of smaller feature by relatively large pixels.
• Related, CLUMP is an interesting dataset and a good example of the benefits of going to a national level where there is unique data available… please provide a brief overview of how they develop their polygons and classes – aerial photography? Land owernship/parcel data?
• It is surprising that nightlights information was not used – when it was used in past HIF maps (e.g., Venter et al. 2016). If more pressures (dataests) lead to higher accuracy (as argued earlier) then why not include it? Are there no oil and gas wells that can be readily detected by gas flaring? 
• L133 - Urban area is not a land use… remove – replace with residential, commercial, industrial, etc.
• L182 - Agreed that data on pastureland and livestock grazing are challenging, the Gridded Livestock of the World dataset(s) would be a valuable addition – please consider adding or describe limitations that precluded their incorporation.

Robinson, T. P., Wint, G. W., Conchedda, G., Van Boeckel, T. P., Ercoli, V., Palamara, E., and Gilbert, M.: Mapping the global distribution of livestock, PloS one, 9, e96084, https://doi.org/10.1371/journal.pone.0096084, 2014. 

• It would be valuable to understand what are the proportions of the dominant pressure nationally, and perhaps specific to each location? 
• Please discuss the limitations of the assumption a simple, uniform buffer used to represent “access” away from roads, for example. Traveling off-road depends greatly on the adjacent topography and land cover (not to mention land ownership). Please describe how the method used in the paper compares to methods based on travel time, etc., such as:

Nelson, A., Weiss, D.J., van Etten, J., Cattaneo, A., McMenomy, T.S. and Koo, J., 2019. A suite of global accessibility indicators. Scientific data, 6(1), p.266.

Weiss, D.J., Nelson, A., Vargas-Ruiz, C.A., Gligorić, K., Bavadekar, S., Gabrilovich, E., Bertozzi-Villa, A., Rozier, J., Gibson, H.S., Shekel, T. and Kamath, C., 2020. Global maps of travel time to healthcare facilities. Nature medicine, 26(12), pp.1835-1838.

• L216 - Please provide more specific guidance on the edge effect of roads – the citation used (Trombulak and Frissel 2000) is a seminal paper (but is 25 years old), more recent reviews (e.g., Rytwinski and Fahrig 2015) would be helpful to briefly summarize. Are there additional, more recent, more specific to Australia, citations to support the parameterization of the distance?

Rytwinski, T. and Fahrig, L., 2015. The impacts of roads and traffic on terrestrial animal populations. Handbook of road ecology, pp.237-246.

• L143: – “... assign a score of 10 for any pixel overlapping a building.” – to be clear, this is if the building footprint polygon touches any part of a pixel, yes? If this is common to all other pressures, then this should be stated, or if different, then detailed for each pressure. E.g, that is – a building footprint of 100 m2 would translate to a 10000 m2 pixel, correct? (Yes, you assume some modification around the building). Or, is it the centroid of the pixel that must intersect with the polygon?
• L145: clarify why there are commission errors – these would be errant footprints?
• L156: Agreed that degradation is associated with proximity (that are accessible) to human populations. But are maps of human populations (typically based on where residences are located, represented largely by building footprints) then constrained to the pixels they touch? 
• How large are large dams? Many of these visually are comparable to the farm dams. What is the specification, typically in terms of area at full or dam height?
• L210: farm dams – buffered by 500 m. Aren’t the impacts dominated downstream? Also, large dams are presumably polygons – why is their buffers so much lower than farm dams?
• Roads and trails are two different pressures, correct? In the discussion in the paper, it would be valuable to have them separated for clarity. 
• Trails – why 0.9? Why not 1.0 – seems scoring is ordinal 1-10!? Presumably just the pixels that touch the trail line. Also, please describe which data and key/attribute of OSM was used (what classes of roads, etc.) so that this work can be reproduced if needed.
• Be consistent with positional accuracy – you’ve got that in the table, so probably no need to mention it here.
• L235: Please correct – transmission lines are included in other human pressure datasets (e.g. Theobald et al. 2025).
• L243-244. Good point about service roads paralleling transmission and pipelines.
• L255. Because this is an open access publication, please provide the validation data (point locations and scoring from visual inspection) in the repo. 
• Please consider noting that if the stratification was based on the output HIF map, then the randomization was in part based on the resulting map and is not strictly independent.
• L269 - normalized to a 0-1 scale – please provide the formula used – Is this max-normalized? Please describe why the highest cumulative value would be the max value used (and why not the theoretical value of 73?). 
• L274 – it is not clear why you also report validation using a 20% threshold of being correct – why not just use the continuous value results?
• It seems that section 2.4 is duplicative of 2.3.1 – is there a way to combine these or distinguish them more?
• Please provide the statistical distribution of the cumulative pressure values (histogram or cumulative frequency distribution) to understand their distribution better. This would be helpful context to understand how a central tendency measure like the mean (or median, etc.) portrays the full range of values. Assuming that the distribution is highly skewed with many more low values (this is typical of spatial data generally), is a mean appropriate metric? Please describe, perhaps in the limitations section, the assumptions/intepretation of the addition of the pressures (although there are 4 that are mutually exclusive). 
• Interesting and helpful validation plot figure 2 (validation). Just curious, why are there horizontal patterns (or even boxes) of points… e.g., mapped values at 0.025.
• L312-315. Related to the comment above in the general comments section – intactness includes the spatial configuration. Would be very helpful to clarify this, and not call single pixels as intact or not. This has led to much confusion amongst scientists and policy/decision makers in the CBD Global Biodiversity Framework context.
• L330 - remove note on projection – you already described in above and more specifically.
• The data layers for pressures 7 and 13 were not in the Zenodo repo (forestry plantations, trails). Perhaps these are subsumed in other layers, but should be in separate datasets to maintain consistency.
• L369 – Not tracking these statements – the lower RMSEs could also be a function of resolution – not just increased number of threats, can you disentangle these? This would be valuable to know for sure, but it's not clear that this statement is justified by these initial (but limited) findings.
• L397-402 Remove, this has already been discussed.
• Perhaps it would be valuable in describing intactness relates to connectivity is to describe – briefly – how EII is similar to a connectivity measure such as Ferrier et al’s PARC index or Brennan et al’s 2020 circuitscape paper.

Brennan, A., Naidoo, R., Greenstreet, L., Mehrabi, Z., Ramankutty, N. and Kremen, C., 2022. Functional connectivity of the world’s protected areas. Science, 376(6597), pp.1101-1104.

• Please align the numbering/naming scheme of the data layers to the description in the text. E.g., human population is 03_ dataset but is 4th in description (2.2.4).

Technical corrections

L58: “Australis” spelling

L64: “gazettal” – is this a typo or an uncommon word?

L145: Microsoft (2022) ?? isn’t it Microsoft 2018?.

L162: WorldPop is at 10 m resolution (or area of 100 m2)? I think you mean 100 m (10000 m2)

L442: Great to see these limitations, caveats. 

SI2: is the ABARES dataset the same as CLUMP (in the paper)?

Datasets

Built-up - in an ad hoc viewing of the data layer, this data seems this covers a broad range of intensity. Also, consider aligning the file names for the pressures with the description in the text (e.g., 01_builtup = 2.2.1 Intensive land uses. E.g., the town of Lithgow (150.1527, -33.4815) and just north near Marrangaroo (150.11423, -33.44008) is a much lower intensity area (dominated by forest/shrub). It would be valuable to examine this more systematically to this occurs elsewhere, and address this perhaps by describing the range of land use intensity (perhaps better would be using built-up as a value that ranges from 1 to 10 rather than just 10 or 0, such as is done with human population – but not suggesting that this has to be re-done). Numerous small (5-25 pixels) areas in very rural areas (albeit farmsteads/ranches) occur as well. These seem to be categorically different from high-density residential/commercial in cities. Similarly, there is a fair amount of speckling (single/couple pixels with 0 values) in high density areas e.g., (151.2506, -33.91478). This might be related to the conversion of the polygonal CLUMP data to raster (the details of this are needed).

Farm ponds and reservoirs

Amazing to see the number of farm ponds! The buffering of the ponds (500 m?) resulting in ~118 pixels seems disproportionate to the un-buffered reservoirs, which presumably have a larger impact in general than farm ponds. For example, at 118.61132, -31.97881 the footprint of the ponds covers ~50% of the land, while many (most?) reservoirs are smaller than a single pixel, and represented by 5 pixels (except for very large reservoirs, eg. >100 pixels. The result seems counter-intuitive, while the intensity value of 8 vs. 5 is higher, the impact is much greater on farms ponds… so 118 x 5=590 vs 5 x 9=40. Please clarify.

Roads

If two datasets are used to represent the roads, can the same road be represented in both datasets if they don’t align spatially, are they double-counted? In a quick look, it didn’t appear that there were any, but would be valuable to describe how this was handled. Also, what were the attributes and values used to distinguish major from minor roads and trails?

Cumulative map

Seems there are counter-intuitive results, e.g., 141.125431, -17.840316 where a major road (National Highway 1) has nearly half the cumulative value (~10.4) than a nearby powerline (18.7). Another example is a series of lower values in the middle of a major road (value of 8, correct) compared to adjacent areas (e.g., 145.69361, -34.08512) vs. adjacent to the road with a value of 15 (because of other pressures, in this example cropland). Are these caused by the summation of the pressures or something else?

EII – more detail – even just a sentence or two of how you calculated EII would be valuable. For example, what is the normalization of HIF to 0-1, what was the radius, shape of the kernel used for EII, so that reader doesn’t have to go to the Beyer et al. paper for pertinent parameters.

This paper aims to develop the first national-specific human industrial footprint for terrestrial Australia. However, I struggle to understand the rationale behind this work. The production merely summarizes various pressure layers based on subjective scoring, how can the author claim it is 'accurate' or not? Additionally, it lacks practical significance for both biodiversity and ecology, as the article fails to demonstrate this through analysis or discussion. Therefore, I recommend rejecting this paper. Before submitting it elsewhere, the author(s) should reconsider the novelty and practicality of their work and extensively revise the manuscript.

Introduction

1. What are the drawbacks of the existing global-scale data? What academic contributions can be achieved by addressing these drawbacks, such as improving biodiversity prediction?
2. How do you define ‘pressure’? What is the relationship between different pressures and biodiversity? The authors should reconsider the correlation between the Human Impact Factor (HIF) and biodiversity.
3. The industrial footprint may be misleading, as I would expect to see some analysis on the trade-induced impacts of industrial sectors on biodiversity, commonly referred to as the ‘footprint.’

Methods

1. Do you believe your data layers can represent all pressures on biodiversity or ecology? Additionally, since the intensity of human activity can be represented by many proxies, why do you only consider population density while neglecting others, such as nighttime light?
2. The temporal inconsistency of data layers may introduce significant bias.
3. How do you determine the score and spatial scale for all the indirect impacts?
4. Why do you assign the pressure of a dam to the focal pixel rather than its downstream effects?
5. What does ‘HFI’ in line 272 refer to? It appears to be an abbreviation error.
6. The single score for cropland seems insufficient to represent the pressure on biodiversity, as there are distinct differences under various intensification levels, land-use strategies, and biochemical conditions.

Results

1. What is the ecological significance of the HIF value? I suspect that pixels with the same cumulative HIF value may experience different levels of pressure on ecology or biodiversity. Additionally, can I assert that a pixel with an HIF value of 40.0 experiences double the pressure of a pixel with an HIF value of 20.0?
2. There is a numerical inconsistency regarding the R² value in line 333 and figure 2. Why did you use R², which typically measures goodness of fit, instead of Pearson’s r?
3. I noticed a higher bias in regions with a high footprint. Why is that?
4. The validation was based on subjective scoring, which is insufficient to support the reliability of the data.
5. How can you claim that your production is more accurate solely based on low correlation with existing global-scale data? Furthermore, how do you define the accuracy of your work?
6. What is the purpose of calculating the Ecological Impact Index (EII)? It does not seem to indicate any practical significance of your findings.

Discussion

1. Please expand on the novelty, results, practical implications, and potential applications of your work.

Based on 16 pressure layers of national relevance, the authors have developed the Human Industrial Footprint (HIF) and Ecological Intactness Index (EII) with high spatial resolution for Australia. These two indices are of critical significance for guiding vegetation restoration initiatives and biodiversity conservation practices. Overall, the manuscript is well-structured, with clear logical flow and coherent writing. To further enhance its academic rigor and contribution, the following suggestions are proposed for potential revisions:​

1. Introduction (Lines 54–60): The current paragraph places excessive emphasis on the detailed background of the Global Biodiversity Framework (GBF). Given the focus of this study, an in-depth elaboration of the GBF is unnecessary and may divert attention from the core research context. Instead, the authors should systematically synthesize and present global advancements in pressure mapping research—a key foundation for justifying the novelty of this study. For instance,  studies such as Gassert et al. (2023) and Arias-Patino et al. (2024) should be integrated to clarify research gaps that the current HIF and EII aim to address.

2. Methods: While the Discussion section addresses uncertainties associated with data sources and methodological design, an important uncertainty remains unaccounted for: the influence of fire regimes. As a dominant disturbance in Australian ecosystems, fire exerts profound effects on both vegetation dynamics and intensive land use. For example, forestry operations across different regions exhibit varying levels of fire resistance, which may lead to divergent HIF/EII values even for the same ecosystem. The authors are advised to discuss whether fire regime variables were incorporated into the index development framework; if not, an additional analysis of fire-induced uncertainty should be added to strengthen the robustness of the methods.

3. Section 3.3 (Lines 365–370): The comparison between the proposed HIF/EII and existing Global Human Footprint datasets is currently insufficiently detailed. To fully demonstrate the advantages and limitations of the new indices, the authors should expand this section to include spatial comparative analyses by visualizing spatial patterns of discrepancies (e.g., via difference maps) to identify regions where the new indices diverge most significantly from global datasets.

4. The current Discussion section functions more as a Conclusion, as it primarily summarizes key findings rather than engaging in critical, in-depth synthesis. 

5. The Discussion section can explicitly outline targeted application scenarios for the two indices to enhance their relevance for policymakers and practitioners. For example: compare the suitability of HIF and EII for specific management objectives (e.g., Is HIF more effective for evaluating industrial disturbance risks, while EII better captures ecological integrity for biodiversity hotspots?