Pith. sign in

REVIEW 2 major objections 6 minor 53 references

WorldTensor puts climate, land, ocean, infrastructure, hazards, and socioeconomic data on one 0.25° annual grid for Earth-system foundation models.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-12 03:25 UTC pith:E4FZE47Q

load-bearing objection Solid data-release paper: a usable, documented multimodal Earth-system corpus with real engineering and validation, not a new model or causal claim. the 2 major comments →

arxiv 2607.03298 v1 pith:E4FZE47Q submitted 2026-07-03 cs.LG cs.AIecon.GNphysics.ao-phq-fin.EC

A harmonised dataset for Earth system foundation models

Carlos Rodriguez-Pardo , Massimo Tavoni This is my paper
classification cs.LG cs.AIecon.GNphysics.ao-phq-fin.EC
keywords Earth system foundation modelsmultimodal geospatial dataharmonised gridded datasetclimate-society coupling0.25 degree gridNetCDFhuman-environment systemsWorldTensor
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Earth-system foundation models have mostly been trained on physical climate and weather alone, leaving out the human systems that drive emissions, land conversion, infrastructure, and vulnerability. The paper argues that the barrier is data fragmentation: reanalyses, remote sensing, emissions inventories, land-use reconstructions, hazards, and socioeconomic indicators sit on incompatible grids, projections, and time steps, so joint machine learning is impractical. WorldTensor answers that by aligning hundreds of environmental and human-system variables onto a shared 0.25° latitude–longitude grid and an annual temporal framework, packaged as self-describing NetCDF files with CF metadata and PyTorch loaders. The claim is that co-locating these domains gives models the raw material to learn coupled human–environment dynamics at planetary scale rather than physical climate in isolation. A sympathetic reader cares because climate risk, impact assessment, and policy increasingly require models that reason across both systems jointly.

Core claim

The authors present WorldTensor as a reproducible, modular global dataset that harmonises hundreds of climate, land, ocean, cryosphere, emissions, infrastructure, hazard, and socioeconomic variables onto a common 0.25° grid and annual time unit, distributed as CF-convention NetCDF files designed for machine-learning workflows, thereby enabling foundation models that learn coupled environmental and human-system dynamics rather than physical climate alone.

What carries the argument

WorldTensor’s canonical data model: a fixed 0.25° lat–lon grid with annual layers (or static covariates), built by regridding continuous fields, rasterising points and vectors into density or proximity surfaces, aggregating or interpolating heterogeneous time series, and packaging everything as modular NetCDF with shared coordinates and CF metadata so domains can be stacked into multimodal tensors.

Load-bearing premise

That regridding, rasterising sparse points and vectors, annual aggregation, and interpolating between sparse socioeconomic years preserve each source’s main scientific signal well enough that joint models will not be misled by artefacts or spurious cross-domain correlations.

What would settle it

Train the same multimodal foundation model on WorldTensor versus carefully aligned native-resolution sources for a held-out coupled task (e.g., predicting socioeconomic impact given climate extremes) and check whether WorldTensor yields systematically worse accuracy or invents spatially coherent false couplings that disappear when native grids are used.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The manuscript introduces WorldTensor, a harmonised global dataset that co-locates hundreds of environmental and socioeconomic variables on a shared 0.25° latitude–longitude grid with an annual temporal framework, packaged as CF-convention NetCDF for machine-learning workflows. It documents source selection, spatial regridding and rasterisation of point/line/polygon data, temporal aggregation and inter-anchor interpolation, domain-specific pipelines, and a modular release of ~53k files (~46 GB) spanning climate, extremes, emissions, land use, vegetation, hydrology, cryosphere, ocean, agriculture, energy, human systems, hazards/conflict, and static context. Technical validation covers physical bounds, land-budget consistency, recovery of selected historical events, semivariograms, unsupervised ICA structure, and a geospatial embedding probe (Fourier vs RCF). Code, PyTorch loaders, and Zenodo data are released.

Significance. If the resource is as described, this is a useful piece of research infrastructure for multimodal Earth-system and geospatial foundation models that currently lack a common human–environment training grid. Strengths that should be credited explicitly include: (i) full public pipeline code and MIT-licensed processing; (ii) transparent Tables 1–3 mapping sources, harmonisation methods, and uncertainty treatment; (iii) multi-layer validation beyond format checks (land budget, event recovery, spatial structure, ICA, RCF); and (iv) practical PyTorch year/patch interfaces with finite-value masks. The contribution is a curated, reproducible co-location resource rather than a new physical law or a trained foundation model; that is an appropriate and valuable scope for a dataset paper.

major comments (2)
  1. Background & Technical Validation (Temporal signal fidelity): The central design choice of annual resolution is justified for interannual/coupled structure, but only 3/5 historical events are recovered at |z|>1, with Pinatubo and COVID attenuated as expected under annual aggregation. This is disclosed in Usage Notes, yet the framing in Background still emphasises human systems that “drive and respond” to change, some of which are sub-annual. Please tighten the scope statement (Abstract/Background/Usage Notes) so that WorldTensor is clearly positioned for interannual and multi-decadal coupling, not event-level or within-year human responses, and state which validation events are diagnostic of the intended use case versus known limitations of the temporal unit.
  2. Methods (Dataset design principles; Spatial/Temporal harmonisation) and Usage Notes: The paper correctly flags that co-location can induce spurious cross-domain correlations and that 0.25° regridding/rasterisation can average fine structure or invent apparent precision for sparse points. These are load-bearing caveats for the claim that the resource is ready for joint multimodal pretraining. Please add a short, concrete user-facing diagnostic or protocol (e.g., recommended coverage intersection rules, finite-mask usage, and a simple null/shuffle or domain-holdout check) so that the disclosed risk is operationalised rather than only warned about. The existing ICA/RCF analyses help but do not substitute for guidance on avoiding artefact-driven couplings.
minor comments (6)
  1. Table 1 / Data Records: Coverage spans differ sharply by domain (e.g., ocean 2010–2023 vs land use 1900–2024). A compact per-domain coverage matrix or machine-readable manifest would make subsetting safer than relying on Figure 1 alone.
  2. Technical Validation (Spatial structure): PM2.5 is flagged with an anomalously long fitted range. Briefly discuss whether this is a regridding artefact, source property, or probe limitation so users know how to treat air-quality layers.
  3. Table 3 / Quality flags and uncertainty: Uncertainty layers are not propagated. Consider a short “recommended reliability tiers” note (reanalysis vs interpolated socioeconomic vs sparse point rasterisations) for pretraining variable selection.
  4. Figure 2 caption and gallery: Units and colour scales are dense; ensure colourbars remain legible in print and that log-scaled panels (e.g., CH4, PM2.5, GDP) are explicitly labelled as such in the caption.
  5. Code availability: Confirm that the Zenodo DOIs and GitHub URL in the manuscript match the final release tags, and that examples/torch/ run against a documented minimal subset so reviewers can smoke-test ingestion without downloading 46 GB.
  6. Land-use processing: Residual coastal/island land-budget mismatches (MAD 0.015) are noted; a one-sentence pointer to whether residual maps or masks are released would help land-budget-sensitive users.

Circularity Check

0 steps flagged

No circularity: data-release paper with external sources and external validation; no derivation reduces to its inputs by construction.

full rationale

WorldTensor is a harmonisation and packaging paper, not a first-principles or predictive derivation. Load-bearing content is (i) selection and regridding/rasterisation of third-party products (ERA5, EDGAR, LUH3, MODIS, GPW, etc.), (ii) documented spatial/temporal rules, and (iii) validation against independent benchmarks (WMO temperature, ERA5 precipitation docs, EDGAR CH4 totals, historical events, ICA/semivariogram/RCF probes). None of these steps define a quantity in terms of itself, fit a parameter then re-label it as a prediction, or rest on a uniqueness theorem or ansatz from the same authors. The only self-references are the Zenodo deposit and code archive of this same release (Data/Code availability; ref. 59), which is standard data-product citation and is not used to justify any scientific claim. Heterogeneous coverage, annual aggregation, and 0.25° regridding are disclosed design trade-offs, not circular reductions. Score 0 is therefore appropriate.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 1 invented entities

As a dataset paper, load-bearing choices are design decisions and domain assumptions about source fidelity and resampling, not free parameters fitted to prove a physical law. The central claim rests on those choices plus the premise that co-located annual fields are useful for foundation-model pretraining.

free parameters (4)
  • target_grid_resolution = 0.25 degrees
    0.25° chosen to match ERA5 and balance detail vs sparsity/storage; not derived from a uniqueness theorem.
  • annual_aggregation_statistics = variable-specific (mean/sum + std/min/max)
    Per-variable choice of mean/sum/std/min/max and incomplete-year exclusion rules are design choices that define the released signal.
  • inter_anchor_interpolation_rules = dataset-level linear/year-assignment rules
    Population, GDP, settlement, and similar sparse products use explicit interpolation between anchor years; the interpolation scheme is a free design choice.
  • release_start_cap_year = 1900
    Long historical sources capped at 1900 for release consistency.
axioms (5)
  • domain assumption Source reanalyses, remote-sensing products, inventories, and socioeconomic grids are sufficiently accurate that their central estimates, after masking, are useful training targets.
    Invoked throughout Source dataset selection and Quality flags sections; uncertainty layers are not propagated.
  • domain assumption Bilinear/area-weighted regridding and nearest-neighbour categorical assignment preserve the main scientific signal of continuous and discrete fields on the target grid.
    Spatial harmonisation and Table 2; validated only partially via variograms and ICA.
  • domain assumption Rasterising points/lines into counts, densities, and distance-to-nearest fields yields spatially meaningful fields comparable to gridded environmental variables.
    Rasterisation of point, line, and polygon datasets section; energy and hazards workflows.
  • ad hoc to paper Annual resolution is an appropriate shared temporal unit for learning coupled human–environment structure (as opposed to sub-daily weather dynamics).
    Dataset design principles and Usage Notes explicitly scope the product to annual tasks.
  • domain assumption Standard CF/NetCDF conventions and float32 packaging are adequate for ML interoperability.
    Metadata standardisation and file packaging.
invented entities (1)
  • WorldTensor multimodal tensor schema independent evidence
    purpose: Canonical domain/variable/year NetCDF layout and shared lat/lon/time coordinates for joint ML ingestion.
    The release abstraction is the paper’s product; independent evidence is the public Zenodo corpus and code, not a new physical object.

pith-pipeline@v1.1.0-grok45 · 23356 in / 3215 out tokens · 28904 ms · 2026-07-12T03:25:49.006091+00:00 · methodology

0 comments
read the original abstract

Foundation models for Earth systems have so far been trained primarily on physical climate and weather data, with limited representation of the human systems that both drive and respond to environmental change. The lack of a unified global training resource that combines climate, land, ocean, cryosphere, infrastructure, hazards, and socioeconomic data on a common grid hinders progress toward truly multimodal Earth system foundation models. We present WorldTensor, a harmonised global dataset that aligns hundreds of environmental and socioeconomic variables to a standardised 0.25$^\circ$ spatial grid and annual temporal framework. WorldTensor integrates reanalysis products, remote sensing, emissions inventories, land use reconstructions, hydrological observations, infrastructure and hazard datasets, and socioeconomic indicators within a single representation designed for machine learning workflows. To build the dataset, we regridded inputs across heterogeneous native resolutions and projections, rasterised point and vector datasets into spatially meaningful gridded fields, and reconciled temporal coverages ranging from daily observations to sparse multiyear socioeconomic snapshots. All outputs are distributed as NetCDF files with standardised coordinates, variable metadata, and a common CF metadata convention. WorldTensor provides a reproducible resource for training and evaluating foundation models that learn coupled dynamics across environmental and human systems at planetary scale.

Figures

Figures reproduced from arXiv: 2607.03298 by Carlos Rodriguez-Pardo, Massimo Tavoni.

Figure 1
Figure 1. Figure 1: Temporal density of WorldTensor, 1900–2025. Each column represents one year; colour encodes the total number of variable families with data available in that year, from roughly 270 in the early record to over 750 after 2000. The progressive saturation reflects the staggered onset of source datasets across domains. as a fixed covariate across annual series. Examples include bathymetric elevation and ocean d… view at source ↗
Figure 2
Figure 2. Figure 2: Representative maps from WorldTensor for the year 2015. The gallery illustrates the spatial diversity of the dataset across physical climate fields, biogeochemical and land surface indicators, socioeconomic variables, infrastructure, and static contextual layers, all on the common 0.25◦ grid. Internal consistency Land-use state fractions from LUH3 (12 variables: primary and secondary forest, crops, pasture… view at source ↗
Figure 3
Figure 3. Figure 3: Composite temporal volatility across all WorldTensor time-varying variables, computed as the cross-variable mean of normalised standard deviation, mean absolute year-to-year change, and terminal shift (|late mean−early mean|). Bright regions indicate grid cells where multiple variables changed substantially over the observational record. Hotspots include the Arctic (amplified warming, sea-ice loss), East a… view at source ↗
Figure 4
Figure 4. Figure 4: Spatial autocorrelation summary for 15 representative WorldTensor variables. Left: effective range (km) of the fitted semivariogram, indicating the distance at which spatial correlation decays. Right: nugget-to-sill ratio, measuring the fraction of total variance attributable to sub-grid or measurement noise. Multivariate coherence via ICA As a holistic check, we applied Independent Component Analysis (ICA… view at source ↗
Figure 5
Figure 5. Figure 5: Global composite of the first three ICA components derived from all non-static WorldTensor variables (2015), mapped to RGB. Distinct colours correspond to coherent eco-climatic regimes, confirming that cross-domain variables jointly capture geographically meaningful structure without supervision. Geospatial embedding analysis To assess whether WorldTensor preserves meaningful spatial structure beyond simpl… view at source ↗
Figure 6
Figure 6. Figure 6: Fourier positional encoding versus RCF spatial features (torchgeo) for the 19 WorldTensor targets eligible for RCF evaluation; the 7 variables used as RCF input channels are excluded to avoid information leakage. Points above the diagonal indicate variables where local spatial texture (captured by RCF) is more informative than geographic coordinates alone. Marker colour indicates domain. Usage Notes Machin… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 12 canonical work pages

  1. [1]

    Nguyen, T., Brandstetter, J., Kapoor, A., Gupta, J. K. & Grover, A. Climax: A foundation model for weather and climate. InInternational Conference on Machine Learning, 25904–25938 (PMLR, 2023). 3.Bodnar, C.et al.A foundation model for the earth system.Nature641, 1180–1187 (2025). 4.Epicoco, I.et al.Medformer: a data-driven model for forecasting the medite...

    2023

  2. [2]

    6.Ou, Y .et al.Artificial intelligence to support cross-disciplinary climate change research.Nature Climate Change(2026)

    Agarwal, M.et al.General geospatial inference with a population dynamics foundation model.arXiv preprint arXiv:2411.07207(2024). 6.Ou, Y .et al.Artificial intelligence to support cross-disciplinary climate change research.Nature Climate Change(2026). In press

    Pith/arXiv arXiv 2024

  3. [3]

    Hersbach, H., Bell, B., Berrisford, P.et al.The ERA5 global reanalysis.Quarterly Journal of the Royal Meteorological Society146, 1999–2049, DOI: 10.1002/qj.3803 (2020). 13/19

    doi:10.1002/qj.3803 1999

  4. [4]

    M., Beguería, S

    Vicente-Serrano, S. M., Beguería, S. & López-Moreno, J. I. A multiscalar drought index sensitive to global warming: The Standardized Precipitation Evapotranspiration Index.Journal of Climate23, 1696–1718, DOI: 10.1175/2009JCLI2909.1 (2010)

    work page doi:10.1175/2009jcli2909.1 2010

  5. [5]

    Dunn, R. J. H., Alexander, L. V ., Donat, M. G.et al.Development of an updated global land in situ-based data set of temperature and precipitation extremes: HadEX3.Journal of Geophysical Research: Atmospheres125, e2019JD032263, DOI: 10.1029/2019JD032263 (2020)

    work page doi:10.1029/2019jd032263 2020

  6. [6]

    J., Alexander, L

    Hobday, A. J., Alexander, L. V ., Perkins, S. E.et al.A hierarchical approach to defining marine heatwaves.Progress in Oceanography141, 227–238 (2016)

    2016

  7. [7]

    Inness, A.et al.The cams reanalysis of atmospheric composition.Atmospheric Chemistry and Physics19, 3515–3556 (2019)

    2019

  8. [8]

    0).Earth System Science Data16, 2811–2830 (2024)

    Crippa, M.et al.Insights into the spatial distribution of global, national, and subnational greenhouse gas emissions in the emissions database for global atmospheric research (edgar v8. 0).Earth System Science Data16, 2811–2830 (2024)

    2024

  9. [9]

    Hoesly, R. M.et al.Historical (1750–2014) anthropogenic emissions of reactive gases and aerosols from the community emissions data system (ceds).Geoscientific Model Development11, 369–408 (2018)

    2014

  10. [10]

    & Andres, R

    Oda, T., Maksyutov, S. & Andres, R. J. The open-source data inventory for anthropogenic co 2, version 2016 (odiac2016): a global monthly fossil fuel co 2 gridded emissions data product for tracer transport simulations and surface flux inversions. Earth System Science Data10, 87–107 (2018)

    2016

  11. [11]

    C.et al.Harmonization of global land-use change and management for the period 850–2100 (luh2) for cmip6

    Hurtt, G. C.et al.Harmonization of global land-use change and management for the period 850–2100 (luh2) for cmip6. Geoscientific Model Development Discussions2020, 1–65 (2020)

    2020

  12. [12]

    Huete, A.et al.Overview of the radiometric and biophysical performance of the modis vegetation indices.Remote sensing of environment83, 195–213 (2002)

    2002

  13. [13]

    P., Humber, M

    Giglio, L., Boschetti, L., Roy, D. P., Humber, M. L. & Justice, C. O. The collection 6 modis burned area mapping algorithm and product.Remote sensing of environment217, 72–85 (2018)

    2018

  14. [14]

    Moesinger, L.et al.The global long-term microwave vegetation optical depth climate archive (vodca).Earth System Science Data12, 177–196 (2020)

    2020

  15. [15]

    & Tian, F

    Qin, X., Wu, B., Zeng, H., Zhang, M. & Tian, F. Global gridded crop production dataset at 10 km resolution from 2010 to 2020.Scientific Data11, 1377 (2024). 20.Coello, F.et al.Global crop-specific fertilization dataset from 1961–2019.Scientific data12, 40 (2025)

    2010

  16. [16]

    Du, Z.et al.Annual global gridded livestock mapping from 1961 to 2021.Earth System Science Data Discussions2025, 1–20 (2025)

    1961

  17. [17]

    D., Bettadpur, S., Watkins, M

    Tapley, B. D., Bettadpur, S., Watkins, M. & Reigber, C. The gravity recovery and climate experiment: Mission overview and early results.Geophysical research letters31(2004)

    2004

  18. [18]

    W.et al.Extending the global mass change data record: Grace follow-on instrument and science data performance.Geophysical Research Letters47, e2020GL088306 (2020)

    Landerer, F. W.et al.Extending the global mass change data record: Grace follow-on instrument and science data performance.Geophysical Research Letters47, e2020GL088306 (2020)

    2020

  19. [19]

    Rodell, M.et al.The global land data assimilation system.Bulletin of the American Meteorological society85, 381–394 (2004)

    2004

  20. [20]

    26.Luojus, K.et al.Globsnow v3

    Zhang, Z.et al.Development of the global dataset of wetland area and dynamics for methane modeling (wad2m).Earth System Science Data13, 2001–2023 (2021). 26.Luojus, K.et al.Globsnow v3. 0 northern hemisphere snow water equivalent dataset.Scientific Data8, 163 (2021)

    2001

  21. [21]

    Obu, J.et al.Northern hemisphere permafrost map based on ttop modelling for 2000–2016 at 1 km2 scale.Earth-Science Reviews193, 299–316 (2019)

    2000

  22. [22]

    Dussaillant, I.et al.Annual mass change of the world’s glaciers from 1976 to 2024 by temporal downscaling of satellite data with in situ observations.Earth System Science Data17, 1977–2006 (2025)

    1976

  23. [23]

    Aqua MODIS level-3 global mapped chlorophyll data, version 2022.0, DOI: 10.5067/AQUA/MODIS/L3M/CHL/2022 (2022)

    NASA Ocean Biology Processing Group. Aqua MODIS level-3 global mapped chlorophyll data, version 2022.0, DOI: 10.5067/AQUA/MODIS/L3M/CHL/2022 (2022)

    work page doi:10.5067/aqua/modis/l3m/chl/2022 2022

  24. [24]

    Gridded population of the world, version 4 (GPWv4), revision 11, DOI: 10.7927/ H4JW8BX5 (2018)

    CIESIN, Columbia University. Gridded population of the world, version 4 (GPWv4), revision 11, DOI: 10.7927/ H4JW8BX5 (2018). 14/19

    2018

  25. [25]

    World population prospects 2024

    United Nations, Department of Economic and Social Affairs, Population Division. World population prospects 2024. https://population.un.org/wpp/ (2024)

    2024

  26. [26]

    & Masoumzadeh Sayyar, S

    Kummu, M., Kosonen, M. & Masoumzadeh Sayyar, S. Downscaled gridded global dataset for gross domestic product (gdp) per capita ppp over 1990–2022.Scientific Data12, 178 (2025)

    1990

  27. [27]

    & Sun, F

    Wang, T. & Sun, F. Global gridded GDP data set consistent with the shared socioeconomic pathways.Scientific Data9, 221, DOI: 10.1038/s41597-022-01300-x (2022)

    work page doi:10.1038/s41597-022-01300-x 2022

  28. [28]

    & Watanabe, M

    Shoji, T., Yamazaki, D., Kita, Y . & Watanabe, M. Global spatially-distributed sectoral gdp map for disaster risk analysis. Earth System Science Data Discussions2024, 1–18 (2024)

    2024

  29. [29]

    Chrisendo, D.et al.Rising income inequality across half of global population and socioecological implications.Nature sustainability1–13 (2025)

    2025

  30. [30]

    & Guillaume, J

    Kummu, M., Taka, M. & Guillaume, J. H. Gridded global datasets for gross domestic product and human development index over 1990–2015.Scientific data5, 180004 (2018)

    1990

  31. [31]

    & Zhao, X

    Li, X., Zhou, Y ., Zhao, M. & Zhao, X. A harmonized global nighttime light dataset 1992–2018.Scientific data7, 168 (2020)

    1992

  32. [32]

    Zhao, M.et al.A global dataset of annual urban extents (1992–2020) from harmonized nighttime lights.Earth System Science Data Discussions2021, 1–25 (2021)

    1992

  33. [33]

    Pesaresi, M.et al.Advances on the global human settlement layer by joint assessment of earth observation and population survey data.International Journal of Digital Earth17, 2390454 (2024)

    2024

  34. [34]

    41.Marconcini, M.et al.Outlining where humans live, the world settlement footprint 2015.Scientific Data7, 242 (2020)

    Huang, X.et al.30 m global impervious surface area dynamics and urban expansion pattern observed by landsat satellites: From 1972 to 2019.Science China Earth Sciences64, 1922–1933 (2021). 41.Marconcini, M.et al.Outlining where humans live, the world settlement footprint 2015.Scientific Data7, 242 (2020)

    1972

  35. [35]

    Mu, H.et al.A global record of annual terrestrial human footprint dataset from 2000 to 2018.Scientific Data9, 176 (2022)

    2000

  36. [36]

    M.et al.Global extent and change in human modification of terrestrial ecosystems from 1990 to 2022

    Theobald, D. M.et al.Global extent and change in human modification of terrestrial ecosystems from 1990 to 2022. Scientific Data12, 606 (2025)

    1990

  37. [37]

    Global integrated power tracker

    Global Energy Monitor. Global integrated power tracker. https://globalenergymonitor.org/projects/ global-integrated-power-tracker/ (2026). March 2026 release. Accessed 2026-03-19

    2026

  38. [38]

    & Melander, E

    Sundberg, R. & Melander, E. Introducing the UCDP Georeferenced Event Dataset.Journal of Peace Research50, 523–532, DOI: 10.1177/0022343313484347 (2013)

    work page doi:10.1177/0022343313484347 2013

  39. [39]

    Geological Survey

    U.S. Geological Survey. ANSS comprehensive catalog of earthquake events and products, DOI: 10.5066/F7MS3QZH (2017)

    work page doi:10.5066/f7ms3qzh 2017

  40. [40]

    R., Kruk, M

    Knapp, K. R., Kruk, M. C., Levinson, D. H., Diamond, H. J. & Neumann, C. J. The international best track archive for climate stewardship (ibtracs) unifying tropical cyclone data.Bulletin of the American Meteorological Society91, 363–376 (2010)

    2010

  41. [41]

    NCEI/WDS Global Significant V olcanic Eruptions Database, DOI: 10.7289/V5JW8BSH (2014)

    National Geophysical Data Center / World Data Service. NCEI/WDS Global Significant V olcanic Eruptions Database, DOI: 10.7289/V5JW8BSH (2014). 49.Rosvold, E. L. & Buhaug, H. Gdis, a global dataset of geocoded disaster locations.Scientific data8, 61 (2021)

    work page doi:10.7289/v5jw8bsh 2014

  42. [42]

    Danielson, J. J. & Gesch, D. B. Global multi-resolution terrain elevation data 2010 (GMTED2010). Open-File Report 2011–1073, U.S. Geological Survey (2011). DOI: 10.3133/ofr20111073

    work page doi:10.3133/ofr20111073 2010

  43. [43]

    ETOPO 2022 15 arc-second global relief model, DOI: 10.25921/ fd45-gt74 (2022)

    NOAA National Centers for Environmental Information. ETOPO 2022 15 arc-second global relief model, DOI: 10.25921/ fd45-gt74 (2022)

    2022

  44. [44]

    M., Batjes, N

    Poggio, L., de Sousa, L. M., Batjes, N. H.et al.SoilGrids 2.0: producing soil information for the globe with quantified spatial uncertainty.SOIL7, 217–240, DOI: 10.5194/soil-7-217-2021 (2021)

    doi:10.5194/soil-7-217-2021 2021

  45. [45]

    A., Jackson, T

    Reynolds, C. A., Jackson, T. J. & Rawls, W. J. Estimating soil water-holding capacities by linking the FAO soil map with global pedon databases and continuous pedotransfer functions.Water Resources Research36, 3653–3662, DOI: 10.1029/2000WR900130 (2000)

    work page doi:10.1029/2000wr900130 2000

  46. [46]

    McNally, A., Arsenault, K., Kumar, S.et al.A land data assimilation system for sub-Saharan Africa.Scientific Data4, 170012, DOI: 10.1038/sdata.2017.12 (2017). 15/19

    work page doi:10.1038/sdata.2017.12 2017

  47. [47]

    Distance to the nearest coastline: 0.01-degree grid

    NASA Ocean Biology Processing Group. Distance to the nearest coastline: 0.01-degree grid. https://oceancolor.gsfc.nasa. gov/resources/docs/distfromcoast/ (2009)

    2009

  48. [48]

    & Grill, G

    Lehner, B. & Grill, G. Global river hydrography and network routing.Hydrological Processes27, 2171–2186, DOI: 10.1002/hyp.9740 (2013)

    doi:10.1002/hyp.9740 2013

  49. [49]

    J., Nelson, A., Gibson, H

    Weiss, D. J., Nelson, A., Gibson, H. S.et al.A global map of travel time to cities to assess inequalities in accessibility in 2015.Nature553, 333–336, DOI: 10.1038/nature25181 (2018)

    doi:10.1038/nature25181 2015

  50. [50]

    Costanza, R.et al.Changes in the global value of ecosystem services.Global environmental change26, 152–158 (2014)

    2014

  51. [51]

    & Tavoni, M

    Rodriguez-Pardo, C. & Tavoni, M. WorldTensor: a harmonised dataset for Earth system foundation models, DOI: 10.5281/zenodo.19047618 (2026)

    work page doi:10.5281/zenodo.19047618 2026

  52. [52]

    Rolf, E.et al.A generalizable and accessible approach to machine learning with global satellite imagery.Nature Communications12, 4392, DOI: 10.1038/s41467-021-24638-z (2021)

    work page doi:10.1038/s41467-021-24638-z 2021

  53. [53]

    J.et al.Torchgeo: Deep learning with geospatial data

    Stewart, A. J.et al.Torchgeo: Deep learning with geospatial data. InProceedings of the 30th International Conference on Advances in Geographic Information Systems, DOI: 10.1145/3557915.3560953 (2022). Acknowledgements Authors acknowledge support from the European Research Council, ERC grant agreement number 101044703 (EUNICE) CUP D87G22000340006. Contribu...

    doi:10.1145/3557915.3560953 2022