Expansion and increase of human pressures on global land ecosystems between 1990 and 2020

Ramm, Katharina; Brown, Calum; Arneth, Almut; Rounsevell, Mark

doi:10.64898/2026.04.16.718867

Preprints

https://doi.org/10.64898/2026.04.16.718867

Preprints

02 Jul 2026

| 02 Jul 2026

Expansion and increase of human pressures on global land ecosystems between 1990 and 2020

Katharina Ramm, Calum Brown, Almut Arneth, and Mark Rounsevell

Abstract. We present a spatially explicit, global-scale index to assess the effects of the five direct anthropogenic drivers of biodiversity loss identified by the IPBES: land use change, natural resource extraction, climate change, pollution, and invasive alien species. The Biodiversity Pressure Index (BPI) covers 30 years (1990–2020) with an annual time-step and a spatial resolution of 0.1°. We find that the coverage of drivers in available data varies and we highlight the key uncertainties that result from this. Using the best available data, we show that large parts of the terrestrial biosphere (approximately 89 %, including Antarctica and Greenland) are under medium or high human pressure and that almost all areas (approximately 96 %) have experienced an increase in pressure over the past three decades. The BPI shows varied spatial and temporal patterns across world regions and biomes, but many of these areas are dominated by pressures associated with rising temperatures and trade flows. Tropical and subtropical areas are subject to particularly rapidly-growing pressures, while wetlands consistently show the highest pressure levels across biomes. In revealing these and other patterns, the BPI provides a basis for improved understanding and management of biodiversity impacts in the future.

Received: 18 May 2026 – Discussion started: 02 Jul 2026

Katharina Ramm, Calum Brown, Almut Arneth, and Mark Rounsevell

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-2863', Anonymous Referee #1, 23 Aug 2026
The paper focusses on human threats to biodiversity and assesses their extent and expression at global scale with high resolution. This is an important topic given that human threats haven been increasing and differ across regions. The authors use a set of indicators, merge these to construct a global pressure index and discuss its dynamics during the past 30 years. Available data have limitations in scope, spatial and temporal resolution which is acknowledged by the authors. The value of this pressure index for better understanding threats to biodiversity however cannot easily be evaluated based on the current text. This is due to methodological shortcomings e.g. missing explanation of how the selected indicators threaten biodiversity, lacking interactions and limited discussion of implications of the chosen method and related biases. Improving the methodology would help provide a clearer picture of strengths and weaknesses of the calculated index. This would encourage to include more specific aspects of how the index can improve “understanding and management of biodiversity impacts in the future” as claimed in the abstract. I list more details in the below comments that may be useful for revising the paper.
Specific comments
The title speaks about “ecosystems” while the text focusses on “biodiversity pressure”. This is not the same and should be consistent. What aspects of biodiversity do you consider (structural, species communities etc)? This is important to clarify for better understanding what the index reveals.

Intro (line 30-47): IPBES has identified the main drivers of biodiversity loss in an early report. But presenting them loosely as if they were independent largely omits insights from IPBES Nexus and Transformative Change Assessments. Interactions are important and may reinforce effects on biodiversity even for drivers which are seemingly ‘less’ important when considered in isolation. A critical reflection on this point and its implication is necessary early on and should be made explicit in the data/methods section.

L 48: reference is needed

Intro: GBF is mentioned here but not taken into account anymore in the remaining text. You could either make its role more explicit throughout the text or delete.

L 58 “… understand their extent and magnitude (Bowler et al., 2020)” and interactions with biodiversity (https://doi.org/10.1016/j.oneear.2025.101416).

L 61 “analysis of combined pressures” is presented here in the simplest possible way – by summing the indicator values. This could be made clear early in the text.

L 73-81: this part could be moved up to become the second-last paragraph. This would help argue for the need to construct a new index.

L 80: very good.

L 66-72: This part could be better structured to clearly show how the data reflect the main drivers.

L 85: why 0.05? It would be useful to shortly explain it here (not only in the methods)

1: Legends for part b and d are missing.

Legend of Fig. 1: how is “change in pressure” calculated? Short explanation would be good.

L 106: “reduced land use intensity” seems surprising. Some hints or reference would be needed.

L 120-124: But Colombia and some other parts of S-America have the highest percentile of change. They should be presented here.

L 126-128: But wetland data are very limited covering only some wetlands! Better exclude these here and add to the discussion of limitations/uncertainties.

Fig 2 bottom: x-axis needs adjustment to 1990-2020.

L 143: ‘Trade’ needs to be explained. Analysis of trade impacts on species could be useful starting point to argue (https://doi.org/10.1038/s41559-016-0023)

3 is missing

L 174: mining data are not available as time series. This needs to be clearly presented here.

4: Perhaps better exclude mining because no time series available (figure is misleading).

L 237 – 239: The sentences ”The magnitudes … trade impacts” is unclear and even seems contradictory. Please rephrase.

L 240: This can be expected, right? Maybe chose other more novel aspects to present here.

L 260: also supported by 26% of wild land (https://doi.org/10.1093/nsr/nwz136)

L 275: but this is due to the method of summing indicator values. Is there any other aspect that you could present here?

L 279: trade threats on species are well studied (https://doi.org/10.1038/s41559-016-0023). This could be better reflected.

L 291: “…according to the study of (Theobald et al., 2020)” which corresponds with strong degradation of biodiversity (https://doi.org/10.1016/j.oneear.2025.101416).

L 296: But note that the low value results from the merging Europe with central Asia. These regions show a large variance in BPI. This could be better presented and discussed here before comparing with other studies.

L 306: careful discussion of limited wetland coverage needed here.

L 317: Better say “Trends in” individual pressures (or sometime like this)

L 321: invasive species are a limited perspective on trade, better draw attention to remote footprints that may exacerbate other pressures

L 335: “relatively stable”: seems to contradict Fuchs et al (https://doi.org/10.5194/esd-9-441-2018)

L 336: This section presents mainly limitations in data. The title could better reflect this by naming it ”Data limitations”

L 341: but if more data are added, min/max ranges will change and results may not be comparable in meaningful ways (of course based on 0-1 standardisation, but what does the max value then really mean?)

L 370-380: This is disconnected from previous part and could better serve as a concluding sub-section 3.5.

4: Adding table with all indicators and explanation of how they threaten biodiversity and which aspect of it would be good. Explanations of how each indicator threatens biodiversity needs to be added in the text. This link is missing.

L 405: how can you integrate mining data if not available as time series? This is doubtful.

L 406: sure but this needs to be explained for each indicator.

L 407-408: if input data are shared, then biases (if not even errors) occur. Please specify this more precisely here and for related indicators.

L 408-410: Diana Bowler and colleagues used cluster analysis to identify patterns of threats. So the threats do appear together (no likelihood here). This is a good example to discuss your method of summing indicator values with other methods (e.g. pattern analysis) to show strengths and weaknesses of different approaches. This would be needed to enhance the methods section (or discussion section).

L 412 “… generally occurring in the same place”: This is unclear. Conversion and use of fertilizers/pesticides have quite different impacts on biodiversity. Please specify.

4.2 – 4.9: Please explain through which processes the chosen indicators threaten biodiversity and which aspect of it. This is needed to understand BPI and its potential role in informing management of biodiversity effects.

L 423: These thresholds need to be explained and supported. My problem is also that they suggest homogenous impacts for cropland/pasture (due to single value). But this blurs existing differences and is inappropriate. Managed forests have at least a range, though it is unclear how this is translated in the analysis.

L 438 “number of transitions”: Is this really the important factor? When land is abandoned, biodiversity can decrease (f. ex. abandonment of high-nature value farmland) or increase (when rewilding of agric. areas) depending on taxonomic groups, propagule sources etc. It’s not only the number of switches, this seems a strong simplification.

4.3: It is odd to include an indicator with only one time slot. This disturbs the whole analysis! Can you not find a better replacement? Or remove it.

4.4: national level data introduce at least a spatial bias. This needs to be discussed. But back casting by regression assumes that trade mechanisms affecting biodiversity would not have changed between 1990-2000. This is hardly realistic. A compelling reasoning is needed here. So far it is not convincing.

L 477-478: it is far from reality to assume that all croplands per country receive the same amount of pesticides. Please distribute more realistically (e.g. based on groups of crop types).

L 485: why did you not use HYDE to indicate land use? This would have been more consistent.

L 499: this is an example for the need to explain the underpinning process (or hypothesis) to use “weight and value” of goods. But both dimensions have different implications. This needs clarification.

L 501: isn’t non-linearity expected?

L 503: the focus in invasive species is important but not sufficient to capture trade impacts. Biodiversity impacts in the exporting regions are severe (e.g S America, SE Asia).

5: summing and min-max normalization mean that outliers have large impacts. A z-scaling could have been more meaningful. Please explain.

L 516: “… BPI itself was normalised again”: is this done over all years? But then results of future analysis will not be comparable anymore. Even if done per year, problems of comparability appear. Please clarify.

L 518-520: “We did not …”: This is unclear. The equal weighing is a weighing decision. The inclusion/exclusion of indicators is also an implicit decision of weighing. Please revise.

L 523-524: This sentence doesn’t make sense and can be deleted.

L 525-532: the thresholds seem arbitrary and need a better justification. Using thresholds based on data distribution may be quite ‘artificial’. Accounting for the underpinning processes to be described would be a more solid approach to define thresholds (e.g. 20% of (semi-)natural habitats in agricultural land to support sustainable agriculture is derived from empirical evidence and functional relationships, not gaps in data; https://doi.org/10.1111/conl.12773).
Citation: https://doi.org/10.5194/egusphere-2026-2863-RC1
RC2:
'Comment on egusphere-2026-2863', Anonymous Referee #2, 28 Oct 2026
The manuscript “Expansion and increase of human pressures on global land ecosystems between 1990 and 2020” presents the Biodiversity Pressure Index (BPI), a spatially explicit metric that aggregates the five direct anthropogenic drivers of biodiversity loss (defined by IPBES) to map pressures on terrestrial biodiversity. Using publicly available datasets, the BPI is mapped annually between 1990 to 2020 to assess temporal trends and regional patterns. Results reveal a persistent global increase in pressure, with the highest recent BPI values observed in East Asia‑Pacific and rising pressures in South Asia. The study concludes that the BPI provides a useful tool for monitoring combined human pressures and guiding more targeted conservation policies.
The manuscript presents a novel indicator, is well written, and does a nice job situating the study in the current literature as well as addressing the limitations of the proposed BPI. However, several assumptions taken in order to operationalise the BPI mapping are (overly) strong and simplifying, which renders several limitations as critical and affect the usability of the BPI. I hence cannot recommend the manuscript in its current form and suggest major revisions. Please find my detailed comments below.
MAJOR
My main concern relates to the choice of indicators and how they are conceptualised to proxy pressures on biodiversity.
Land-use pressure values. The use of simple pressure values per land-use type risks oversimplifying biodiversity impacts and ignoring regional variation. For example, by assuming that cropland always has stronger negative effects than pasture, this neglects regional context, management intensity, or habitat quality. Also, it is doubtful at best that all croplands worldwide exert the same pressure on biodiversity, even if managed at the same level of intensity, given local landscape context. I agree that the intensity indicators used address the intensity gradients, but they do that for croplands only, which is a major drawback given that pasture comprise ~80% of all farmland globally. I would welcome more spatially nuanced pressure values related to land use, for example using empirical data such as the PREDICTS database.

Land-use transitions as a proxy for disturbance. The number of transitions alone does not adequately represent biodiversity pressure, as transitions differ in their ecological consequences and spatial context. For example, the conversion of forest to pasture to cropland (a typical sequence in deforestation frontiers) may exert greater pressure on biodiversity than repeated crop–fallow cycles, which can have higher transitions frequencies. Conversely, stable but intensively managed landscapes may exert substantial pressure on biodiversity, despite no transitions occurring between land-use classes.

Within-country homogeneity of selected indicators. This arguably neglects subnational variability and results in border artefacts of the BPI. Three out of eight input variables being available at country-level only does not really support the strong claims of the BPI being spatially explicit (which is true technically, but not thematically). Would it be an option to use additional layers (e.g., ports, airports, road networks) to create inverse distance layers to approximate pressure propagation from trade hubs? For pesticides, what was the reason for not adopting a similar workflow as Tian et al. (2022) to disaggregate country-level pesticide data to the pixel level? Another option could have been to use global crop type maps (e.g., https://esa-worldcereal.org/en/products/global-maps or Tang et al. (2024) https://doi.org/10.1038/s41597-024-03247-7) in combination with data on crop-specific pesticide application rates (e.g. from Maggi et al (2019) https://doi.org/10.1038/s41597-019-0169-4).

Related to comment 1), I would have expected some sort of sensitivity analysis or robustness test with different pressure responses (e.g., for land use) and/or indicators (if available).

While I agree with the authors that avoiding circularity by using independent datasets is important, excluding indicators such as livestock density (for which high-quality data is available, e.g., Du et al (2025) https://essd.copernicus.org/articles/17/5543/2025) can introduce substantial thematic biases. Would it be an option to compare two approaches: one prioritising data independence and another using the best available data, while explicitly addressing potential circularity?

Calculation of the composite BPI. While I understand (and agree with) the min-max transformation of the individual indicators representing the five IPBES drivers, I do not understand the min-max transformation of the BPI. After the indicator-specific normalisations and their averaging, the resulting BPI should already lie between 0 and 1. Applying a second min–max transformation is therefore unnecessary and rescales the BPI relative to its observed minimum and maximum, potentially complicating its interpretation and comparability across regions or time periods. For example, a score of 0.8 in one year might not correspond to the same absolute pressure level as 0.8 in another year, which is problematic.

Lastly, this is a purely descriptive manuscript. While I understand the justification of the authors in the discussion to not assess realised impacts on biodiversity, I feel that exactly this would be necessary to assess whether the BPI can actually identify areas likely to experience adverse biodiversity outcomes. Such “validation” would strengthen the ecological relevance of the index. You could simply link BPI patterns and trends to available data on richness or abundance trends of species groups, changes in ecosystem functions, or protected areas (to evaluate their potential “protection” against drivers).

MINOR
To identify regions of disproportionate pressure on biodiversity, I was wondering why the authors refrained from running a formal hotspot analysis, similar to, for example, Halpern et al. (2022) https://www.nature.com/articles/s41893-022-00965-x.

The format of the panels in Figures 2 and 4 (which should be 3) is not matching. I suggest to use proper plotting functions, e.g. ggplot in R, to produce high-quality figures.

I do not understand the approach of converting absolute to relative pressures (lines 525-532). How do the authors arrive at a threshold of 0.05 of the BPI, representing 5% of the pressure variables, if the threshold suggested by Williams et al. should be 8%. Is this choice then only based on the data distribution? If yes, why motivate with Williams? Moreover, a sensitivity analysis would have been needed here (see major comment 2), as the definition of this cut-off value strongly influences results.

DETAILS
Line 13: Perhaps add a (short) sentence motivating the study to start the abstract with?

Line 21: Perhaps use the IPBES drivers, not the proxy variables (as they have not been defined yet)?

Line 23: …and for spatially target decision making?

Line 33: “pollution” instead of “poisoning”?

Line 40: Surface water as well, I think.

Line 46: Consider changing terminology to low/middle/high income countries.

Lines 48-52: Please try to avoid two-sentence paragraphs and expand, linking to literature on polycrisis, driver interactions, cascading/compounding risks.

Lines 50-51: Perhaps point to the relatively recent IPBES Nexus Assessment?

Lines 53-62: Merge paragraphs?

Line 74: or the HPI by Lecomte et al. (2026)

https://www.sciencedirect.com/science/article/pii/S1470160X26001962

Line 82: I suggest to end the Introduction with clearly stated research objectives/questions.

Lines 99-131: A large part of the results reported in this section cannot be taken from Figure 1 alone, but also from Figure 2. Please revise your figure references in this section.

Lines 140-191: I suggest to have two main sections here -> 2.2 individual drivers (without sub-sections) and 2.3 dominant drivers.

Line 238: What is meant with “cryptic”?

Line 244: It should be Table 1, correct?

Line 355: Meyfroidt (2015) would fit here as a reference

http://dx.doi.org/10.1080/1747423X.2015.1117530
Citation: https://doi.org/10.5194/egusphere-2026-2863-RC2

Katharina Ramm, Calum Brown, Almut Arneth, and Mark Rounsevell

Data sets

Global Biodiversity Pressure Index (BPI) between 1990 and 2020 K. Ramm et al. https://doi.org/10.1594/PANGAEA.969567

Katharina Ramm, Calum Brown, Almut Arneth, and Mark Rounsevell

Viewed

Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.

Total article views: 171 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
167	0	4	171	0	0

HTML: 167
PDF: 0
XML: 4
Total: 171
BibTeX: 0
EndNote: 0

Views and downloads (calculated since 02 Jul 2026)

Month	HTML	PDF	XML
Jul 2026	48	1	49
Aug 2026	20	0	20
Sep 2026	38	1	39
Oct 2026	54	2	56
Nov 2026	7	0	7

Cumulative views and downloads (calculated since 02 Jul 2026)

Month	HTML	PDF	XML
Jul 2026	48	1	49
Aug 2026	20	0	20
Sep 2026	38	1	39
Oct 2026	54	2	56
Nov 2026	7	0	7

Viewed (geographical distribution)

Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.

Total article views: 151 (including HTML, PDF, and XML) Thereof 151 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 02 Nov 2026

Short summary

Human activities are having negative impacts on biodiversity worldwide. We have developed the Biodiversity Pressure Index to track the five key threats — land use change, resource extraction, climate change, pollution, and invasive species — across the entire planet from 1990 to 2020. Results show nearly 90 % of Earth's land is under medium or high human pressure, and 96 % has worsened over 30 years. Among other pressures, rising temperatures and global trade drive much of this damage.


Total:	0
HTML:	0
PDF:	0
XML:	0