pith. sign in

arxiv: 2605.04206 · v1 · submitted 2026-05-05 · 💻 cs.LG

Climate-based Pre-screening of Self-sustaining Regreening Opportunities in Drylands: A Case Study for Saudi Arabia

Katja Froehlich , Jonathan Klein , Ibrahim S. Elbasyoni , Julian D. Hunt , Yoshihide Wada , Dominik L. Michels This is my paper

Pith reviewed 2026-05-08 17:19 UTC · model grok-4.3

classification 💻 cs.LG
keywords climate suitability scoredryland restorationmachine learningSaudi Arabiaregreeningvegetation persistenceERA5-Landremote sensing
0
0 comments X

The pith

Machine learning models trained on expert reference sites can map climate suitability to pre-screen Saudi drylands for self-sustaining vegetation restoration without irrigation.

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

The paper develops a pre-screening method that combines climate records with machine learning to find dryland locations where native plants can establish and persist on their own. It trains models on carefully chosen successful sites to create a Climate Suitability Score that accounts for complex interactions among temperature, rainfall, and other variables. Applied to multi-year climate data across Saudi Arabia, the score produces national maps that are then overlaid with satellite vegetation measurements to highlight places where conditions look promising yet current cover remains low. This filtering step narrows the focus to thirteen priority sites, using intact ecosystems in similar climates as benchmarks for what a realistic increase in vegetation density might look like. The result is a lower-cost way to target restoration efforts in water-scarce regions where long-term irrigation is impractical.

Core claim

A Climate Suitability Score (CSS), derived from machine learning models trained on expert-curated reference sites, captures complex climatic dependencies on vegetation persistence. Using multi-year ERA5-Land data for Saudi Arabia, national-scale prediction maps are generated and combined with vegetation indices to identify areas where climate is favorable, but vegetation remains underdeveloped. Multi-criteria screening reduces candidates to thirteen priority locations. Climatically analogous intact ecosystems provide benchmarks for restoration targets and indicate that an average 2.5 fold increase in vegetation coverage is a realistic target for restoration efforts.

What carries the argument

The Climate Suitability Score (CSS) generated by machine learning models trained on expert-curated reference sites to predict where climatic conditions support persistent native vegetation without irrigation.

If this is right

  • National-scale climate suitability maps become feasible for any dryland region with comparable climate reanalysis data.
  • Overlaying CSS maps with vegetation indices reduces large search areas to a short list of high-potential restoration candidates.
  • Intact ecosystems in matching climates supply quantitative benchmarks, such as a 2.5-fold vegetation increase, for setting realistic project goals.
  • The workflow lowers the number of sites requiring expensive ground surveys, freeing resources for actual implementation.
  • Restoration planning in water-limited areas can prioritize locations already aligned with natural climatic niches.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same reference-site training approach could be adapted to other arid countries that lack detailed vegetation surveys but have access to ERA5-type climate grids.
  • Soil or topography layers could be added later to further refine the thirteen sites without changing the core climate-first filter.
  • Successful restoration at the priority locations would supply real-world test cases for whether the 2.5-fold target holds under active management.
  • If the CSS proves transferable, global dryland restoration programs could adopt similar pre-screening to avoid irrigation-dependent projects.

Load-bearing premise

The expert-curated reference sites accurately represent the full range of climatic conditions required for self-sustaining native vegetation to persist without irrigation across the Saudi landscape.

What would settle it

Long-term field monitoring of one or more high-CSS priority sites showing that native vegetation fails to establish or maintain cover without supplemental water.

Figures

Figures reproduced from arXiv: 2605.04206 by Dominik L. Michels, Ibrahim S. Elbasyoni, Jonathan Klein, Julian D. Hunt, Katja Froehlich, Yoshihide Wada.

Figure 1
Figure 1. Figure 1: Map of Saudi Arabia. The 230 sample locations and their classes are shown according to climatic suitability for vegetation and prevalence of vegetation. We apply our framework to Saudi Arabia as a model for dryland systems. Here, annual declines of 1.2–1.5% in shrub and tree cover within rangelands have been reported13,19. The Saudi Green Initiative pledges to restore 74 million ha of land by planting 10 b… view at source ↗
Figure 2
Figure 2. Figure 2: Model training results. Training results for BLUP and neural network for different input sizes. Average results and error bars are shown for the 10 training repetitions of each size. The input size corresponds to “number of Fourier coefficients per variable” for BLUP and “autoencoder latent vector size” for neural network. comparison. Neural networks are well known as powerful tools for identifying complex… view at source ↗
Figure 3
Figure 3. Figure 3: Correlation between sample reclassification of two categories and NDVI. view at source ↗
Figure 4
Figure 4. Figure 4: Climate Suitability Score (CSS) prediction maps. view at source ↗
Figure 5
Figure 5. Figure 5: Comparison between CSS prediction and NDVI. view at source ↗
Figure 6
Figure 6. Figure 6: Analysis of the 25 selected candidate land restoration locations. view at source ↗
Figure 7
Figure 7. Figure 7: Native ecosystem selection process. (a) Climate distance map. (b) NDVI map. (c) Climate distance map × NDVI overlay. (d) True color image (modified from Google Earth Airbus, 12-10-2023 for land restoration candidate location and 05-03-2023 for native ecosystem). (e) NDVI map (contains modified Copernicus Sentinel data [04-07-2024]) for land restoration candidate location and native ecosystem. All images co… view at source ↗
Figure 8
Figure 8. Figure 8: NDVI comparison for land restoration candidate locations and corresponding view at source ↗
read the original abstract

Large-scale restoration in drylands is widely promoted to address land degradation and biodiversity loss, yet many efforts rely on long-term irrigation, limiting sustainability in water-scarce regions. A key challenge is identifying locations where native vegetation can persist without intensive management while minimizing costly field campaigns. A scalable pre-screening framework is presented that integrates climate and remote sensing data to enable cost-efficient site selection in arid environments using Saudi Arabia as a case study. A Climate Suitability Score (CSS), derived from machine learning models trained on expert-curated reference sites, captures complex climatic dependencies on vegetation persistence. Using multi-year ERA5-Land data for Saudi Arabia, national-scale prediction maps are generated and combined with vegetation indices to identify areas where climate is favorable, but vegetation remains underdeveloped. Multi-criteria screening reduces candidates to thirteen priority locations. Climatically analogous intact ecosystems provide benchmarks for restoration targets and indicate that an average 2.5 fold increase in vegetation coverage is a realistic target for restoration efforts. Overall, this approach narrows the search space, reduces costs, and supports resilient ecosystem recovery planning in water-limited regions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript presents a scalable pre-screening framework for identifying self-sustaining regreening opportunities in drylands, using Saudi Arabia as a case study. It trains machine learning models on expert-curated reference sites to produce a Climate Suitability Score (CSS) that models complex climatic controls on vegetation persistence. National-scale CSS maps are generated from multi-year ERA5-Land data, then combined with vegetation indices via multi-criteria screening to reduce the search space to thirteen priority locations. Climatically analogous intact ecosystems are used to benchmark restoration targets, indicating that an average 2.5-fold increase in vegetation coverage is realistic for these sites.

Significance. If the CSS generalizes reliably, the work provides a practical, data-driven method to prioritize cost-effective restoration sites in water-limited regions without relying on long-term irrigation. It integrates ML for capturing non-linear climatic dependencies with remote-sensing overlays, narrowing candidates from national scale to a manageable set of thirteen locations. Credit is due for grounding the CSS in externally curated reference sites and for proposing a falsifiable vegetation-increase target derived from analogous ecosystems. The approach could reduce field-campaign expenses and support resilient planning in arid zones.

major comments (2)
  1. [Abstract] Abstract: the claim that an average 2.5-fold increase in vegetation coverage is a 'realistic target' is presented without any description of the analogous-site selection process, quantitative comparison metrics, or statistical derivation; this target is load-bearing for the restoration-planning claim yet cannot be evaluated from the given information.
  2. [Abstract] Abstract / Methods (inferred from pipeline description): no model performance metrics, cross-validation results, error bars, or details on validation splits and feature engineering are supplied for the supervised ML models producing the CSS; without these, the reliability of national-scale generalization from the expert-curated sites cannot be assessed and remains central to the pre-screening framework.
minor comments (1)
  1. [Abstract] The abstract would be clearer if it named the specific ML algorithms (e.g., random forest, gradient boosting) employed for the CSS and the exact vegetation indices used in the multi-criteria step.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback and recommendation for major revision. We address each major comment point by point below, providing the strongest honest defense of the manuscript while agreeing to revisions that improve clarity without misrepresenting the work.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that an average 2.5-fold increase in vegetation coverage is a 'realistic target' is presented without any description of the analogous-site selection process, quantitative comparison metrics, or statistical derivation; this target is load-bearing for the restoration-planning claim yet cannot be evaluated from the given information.

    Authors: We agree that the abstract, as a high-level summary, omits the specific details of analogous-site selection and the quantitative derivation of the 2.5-fold target. The full manuscript describes this process in the Methods, where climatically analogous intact ecosystems are selected via similarity in normalized ERA5-Land climate features (using Euclidean distance after PCA), and the target is computed as the average ratio of vegetation indices between these benchmarks and the priority sites, with associated statistics. We will revise the abstract to include a concise description of the selection process and derivation to make the claim evaluable. revision: yes

  2. Referee: [Abstract] Abstract / Methods (inferred from pipeline description): no model performance metrics, cross-validation results, error bars, or details on validation splits and feature engineering are supplied for the supervised ML models producing the CSS; without these, the reliability of national-scale generalization from the expert-curated sites cannot be assessed and remains central to the pre-screening framework.

    Authors: The abstract is intentionally brief and does not contain full methodological specifications. The manuscript's Methods section details the supervised ML models, including 5-fold cross-validation on the expert-curated reference sites, performance metrics, feature engineering on ERA5-Land variables, and error bars on the resulting CSS maps. To enable assessment of generalization from the abstract alone, we will add a short summary of the validation approach and key metrics. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper derives the Climate Suitability Score via supervised machine learning trained on independent expert-curated reference sites, then applies the model to ERA5-Land data for out-of-sample national predictions before combining with vegetation indices. This is a standard predictive workflow with no self-definitional loops, no fitted parameters renamed as predictions, and no load-bearing self-citations or imported uniqueness theorems. The abstract and described method remain self-contained without reducing any central claim to its own inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that reference sites provide unbiased training data for climatic suitability and that the resulting model generalizes nationally; no new physical entities are introduced.

free parameters (1)
  • Machine learning model parameters for CSS
    The Climate Suitability Score is generated by models trained on reference site data, so all internal weights and decision boundaries are fitted parameters.
axioms (1)
  • domain assumption Expert-curated reference sites accurately capture the climatic conditions required for long-term self-sustaining vegetation persistence
    The entire CSS training process depends on this premise that the selected sites represent viable non-irrigated conditions.

pith-pipeline@v0.9.0 · 5523 in / 1540 out tokens · 95220 ms · 2026-05-08T17:19:53.982472+00:00 · methodology

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

47 extracted references · 47 canonical work pages

  1. [1]

    Secretariat of the United Nations Convention to Combat Desertification, 1999

    United Nations Convention to Combat Desertification (Secretariat).United Nations Con- vention to Combat Desertification in Those Countries Experiencing Serious Drought And/or Desertification, Particularly in Africa. Secretariat of the United Nations Convention to Combat Desertification, 1999

    work page 1999

  2. [2]

    Vicente-Serrano et al

    Sergio M. Vicente-Serrano et al. The United Nations convention to combat desertification report on rising aridity trends globally and associated biological and agricultural implications. Technical report, UNCCD, 2024

    work page 2024

  3. [3]

    Increasing numbers of global change stressors reduce soil carbon worldwide.Nature Climate Change, 14:740–745, 2024

    Tomás Sáez-Sandino et al. Increasing numbers of global change stressors reduce soil carbon worldwide.Nature Climate Change, 14:740–745, 2024

    work page 2024

  4. [4]

    Jaime Martínez-Valderrama, Emilio Guirado, and Fernando T. Maestre. Desertifying deserts. Nature Sustainability, 3:572–575, 2020

    work page 2020

  5. [5]

    Forest landscape restoration—what generates failure and success? Forests, 11:938, 2020

    Melanie Höhl et al. Forest landscape restoration—what generates failure and success? Forests, 11:938, 2020

    work page 2020

  6. [6]

    Anzhong Deng, Xianjun Hao, and John J. Qu. A preliminary assessment of land restoration progress in the great green wall initiative region using satellite remote sensing measurements. Remote Sensing, 16:4461, 2024

    work page 2024

  7. [7]

    Parr, Michelle Te Beest, and Nicola Stevens

    Catherine L. Parr, Michelle Te Beest, and Nicola Stevens. Conflation of reforestation with restoration is widespread.Science, 383:698–701, 2024

    work page 2024

  8. [8]

    Taking root.Science, 2021

    Rachel Cernansky. Taking root.Science, 2021

    work page 2021

  9. [9]

    Turner et al

    Matthew D. Turner et al. Great green walls: hype, myth, and science.Annual Review of Environment and Resources, 48:263–287, 2023

    work page 2023

  10. [10]

    Determination of land restoration potentials in the semi-arid areas of Chad using systematic monitoring and mapping techniques.Agroforestry Systems, 97:1289–1305, 2023

    Bertin Takoutsing, Leigh Anne Winowiecki, Alicia Bargues-Tobella, and Tor-Gunnar Vågen. Determination of land restoration potentials in the semi-arid areas of Chad using systematic monitoring and mapping techniques.Agroforestry Systems, 97:1289–1305, 2023

    work page 2023

  11. [11]

    Eshetie, Berhanu K

    Gebrehiwot G. Eshetie, Berhanu K. Alemie, and Abebe M. Wubie. Assessing spatio- temporal dynamics of land degradation neutrality using Google Earth Engine in the Alawuha Watershed of North Wello Zone, Ethiopia.Environmental Challenges, 20:101202, 2025

    work page 2025

  12. [12]

    Landscape structure and suitable habitat analysis for effective restoration planning in semi-arid mountain forests.Ecological Processes, 10:17, 2021

    Hossein Piri Sahragard, Majid Ajorlo, and Pejman Karami. Landscape structure and suitable habitat analysis for effective restoration planning in semi-arid mountain forests.Ecological Processes, 10:17, 2021

    work page 2021

  13. [13]

    A computational sustain- ability framework for vegetation degradation and desertification assessment in arid lands in Saudi Arabia.Sustainability, 18:641, 2026

    Abdalrahman AlAmri, Mohammed Alshehri, and Osama Alharbi. A computational sustain- ability framework for vegetation degradation and desertification assessment in arid lands in Saudi Arabia.Sustainability, 18:641, 2026. 28

    work page 2026

  14. [14]

    Alqurashi, and Ahmed Saud Fahil

    Syed Sumaed Hasan, Omar Abdullah Alharbi, Atiqur F. Alqurashi, and Ahmed Saud Fahil. Assessment of desertification dynamics in arid coastal areas by integrating remote sensing data and statistical techniques.Sustainability, 16:4527, 2024

    work page 2024

  15. [15]

    Supervised NDVI composite thresholding for arid region vegetation mapping.Engineering, Technology & Applied Science Research, 14:14420–14427, 2024

    Razi Khalil, Muhammad Saqib Khan, Yahya Hasan, Nourelhouda Nacer, and Shahid Khan. Supervised NDVI composite thresholding for arid region vegetation mapping.Engineering, Technology & Applied Science Research, 14:14420–14427, 2024

    work page 2024

  16. [16]

    Saco, and Jose F

    Reem Almalki, Mehdi Khaki, Patricia M. Saco, and Jose F. Rodriguez. Monitoring and mappingvegetationcoverchangesinaridandsemi-aridareasusingremotesensingtechnology: A review.Remote Sensing, 14:5143, 2022

    work page 2022

  17. [17]

    Kirschner, Trevor T

    Guido K. Kirschner, Trevor T. Xiao, and Ikram Blilou. Rooting in the desert: A develop- mental overview on desert plants.Genes, 12:709, 2021

    work page 2021

  18. [18]

    ERA5-Land: A state-of-the-art global reanalysis dataset for land applications.Earth System Science Data, 13:4349–4383, 2021

    Joaquín Muñoz-Sabater et al. ERA5-Land: A state-of-the-art global reanalysis dataset for land applications.Earth System Science Data, 13:4349–4383, 2021. doi: 10.5194/ essd-13-4349-2021

    work page 2021

  19. [19]

    Annual report 2023

    Ministry of Environment, Water and Agriculture. Annual report 2023. Technical Report Report No. 1030, Government of Saudi Arabia, Riyadh, Saudi Arabia, 2023

    work page 2023

  20. [20]

    SGI targets — Greening Saudi.https://www.sgi

    Saudi and Middle East Green Initiative. SGI targets — Greening Saudi.https://www.sgi. gov.sa/about-sgi/sgi-targets/greening-saudi/, 2024

    work page 2024

  21. [21]

    Vegetation and condition of arid rangeland ecosystem in Central Saudi Arabia.Saudi Journal of Biological Sciences, 25:1022–1026, 2018

    Saud Al-Rowaily, Abdulaziz Assaeed, Salih Al-Khateeb, Ali Al-Qarawi, and Fahd Al Arifi. Vegetation and condition of arid rangeland ecosystem in Central Saudi Arabia.Saudi Journal of Biological Sciences, 25:1022–1026, 2018

    work page 2018

  22. [22]

    Hassaballa, and Elnazir Ganawa

    Abubaker Salih, Abdalhaleem A. Hassaballa, and Elnazir Ganawa. Mapping desertification degree and assessing its severity in Al-Ahsa Oasis, Saudi Arabia, using remote sensing-based indicators.Arabian Journal of Geosciences, 14:192, 2021

    work page 2021

  23. [23]

    Fernando, and Jack Dekkers

    David Habier, Rohan L. Fernando, and Jack Dekkers. The impact of genetic relationship information on genome-assisted breeding values.Genetics, 177:2389–2397, 2007

    work page 2007

  24. [24]

    Validation of cross-progeny variance genomic prediction using simulations and experimental data in winter elite bread wheat.Theoretical and Applied Genetics, 137:226, 2024

    Clément Oget-Ebrad et al. Validation of cross-progeny variance genomic prediction using simulations and experimental data in winter elite bread wheat.Theoretical and Applied Genetics, 137:226, 2024

    work page 2024

  25. [25]

    D’souza, Po-Yao Huang, and Fang-Cheng Yeh

    Reuben N. D’souza, Po-Yao Huang, and Fang-Cheng Yeh. Structural analysis and optimiza- tion of convolutional neural networks with a small sample size.Scientific Reports, 10:834, 2020

    work page 2020

  26. [26]

    Selective ensembles for consistent predictions, 2021

    Emily Black, Klas Leino, and Matt Fredrikson. Selective ensembles for consistent predictions, 2021

    work page 2021

  27. [27]

    Nehal, and Mikhail Moshkov

    Mohammad Azad, Taha H. Nehal, and Mikhail Moshkov. A novel ensemble learning method using majority based voting of multiple selective decision trees.Computing, 107:42, 2025. 29

    work page 2025

  28. [28]

    Ahmed M. F. Ghazal. Vegetation patterns and plant communities distribution along an altitudinal gradient at Asir Mountain, southwest Saudi Arabia.Pakistan Journal of Botany, 47:1377–1389, 2015

    work page 2015

  29. [29]

    Alharthi, Mohammed A

    Sultan T. Alharthi, Mohammed A. El-Shiekh, and Ahmed A. Alfarhan. Alien plant invasions of the natural habitat in the Western Region of Saudi Arabia: Floristic Diversity and Vegetation Structure.Diversity, 15:309, 2023

    work page 2023

  30. [30]

    Al-Harbi, Afaf A

    Hamad F. Al-Harbi, Afaf A. Alhuqail, Zahirul Islam, and Habes Ghrefat. Vegetation trends and dynamics in Shada Mountain, Saudi Arabia (1984–2023): insights from Google Earth Engine and R analysis.Frontiers in Environmental Science, 12:1397825, 2024

    work page 1984

  31. [31]

    Impact of anthropogenic activities on natural vegetation cover of Aseer Region Saudi Arabia.Egyptian Journal of Environmental Change, 13:33–50, 2021

    Ahmad Moatamed. Impact of anthropogenic activities on natural vegetation cover of Aseer Region Saudi Arabia.Egyptian Journal of Environmental Change, 13:33–50, 2021

    work page 2021

  32. [32]

    Nazrul Islam, and Philip Jones

    Mansour Almazroui, Rodeano Dambul, M. Nazrul Islam, and Philip Jones. Principal components-based regionalization of the Saudi Arabian climate.International Journal of Climatology, 35, 2015

    work page 2015

  33. [33]

    Rainfall trends and extremes in Saudi Arabia in recent decades

    Mansour Almazroui. Rainfall trends and extremes in Saudi Arabia in recent decades. Atmosphere, 11:964, 2020

    work page 2020

  34. [34]

    Nazrul Islam, Hussain Athar, Philip Jones, and Mohammed A

    Mansour Almazroui, M. Nazrul Islam, Hussain Athar, Philip Jones, and Mohammed A. Rahman. Recent climate change in the Arabian Peninsula: annual rainfall and temperature analysis of Saudi Arabia for 1978–2009.International Journal of Climatology, 32:953–966, 2012

    work page 1978

  35. [35]

    Regional Cli- mate Modelling Outputs for Saudi Arabia: Key Findings

    United Nations Economic Social Commission for Western Asia. Regional Cli- mate Modelling Outputs for Saudi Arabia: Key Findings. Technical Report E/ESCWA/CL1.CCS/2023/RICCAR/Technical Report, RICCAR, 2023

    work page 2023

  36. [36]

    Integratinghydrologicalimpactsforcost-effectivedrylandecologicalrestoration

    FengFuetal. Integratinghydrologicalimpactsforcost-effectivedrylandecologicalrestoration. Communications Earth & Environment, 6:667, 2025

    work page 2025

  37. [37]

    Roebroek et al

    Caspar T. Roebroek et al. Potential tree cover under current and future climate scenarios. Scientific Data, 12:564, 2025

    work page 2025

  38. [38]

    Dakhil, Reham F

    Mohammed A. Dakhil, Reham F. El-Barougy, Ali El-Keblawy, and Emad A. Farahat. Clay and climatic variability explain the global potential distribution ofjuniperus phoenicea toward restoration planning.Scientific Reports, 12:13199, 2022

    work page 2022

  39. [39]

    MacKenzie and Colin R

    William H. MacKenzie and Colin R. Mahony. An ecological approach to climate change- informed tree species selection for reforestation.Forest Ecology and Management, 481: 118705, 2021

    work page 2021

  40. [40]

    Luke A. N. Goodall, Adrianna C. Shannon, and Robert M. Scheller. Modeling forest restoration potential in the Scottish Highlands using multiple machine learning approaches. New Forests, 57:10, 2026. 30

    work page 2026

  41. [41]

    Elliott, and Prasit Wangpakapattanawong

    Pimonrat Tiansawat, Stephen D. Elliott, and Prasit Wangpakapattanawong. Climate niche modelling for mapping potential distributions of four framework tree species: implications for planning forest restoration in tropical and subtropical Asia.Forests, 13:993, 2022

    work page 2022

  42. [42]

    The coastal fog and ecological balance for plants in the Jizan region, Saudi Arabia.Saudi Journal of Biological Sciences, 30:103494, 2023

    Aleksandar Valjarevíć et al. The coastal fog and ecological balance for plants in the Jizan region, Saudi Arabia.Saudi Journal of Biological Sciences, 30:103494, 2023

    work page 2023

  43. [43]

    Almalki, Mohammed S

    Khalid A. Almalki, Mohammed S. Al Mosallam, Turki Z. Aldaajani, and Ali A. Al-Namazi. Landforms characterization of Saudi Arabia: Towards a geomorphological map.International Journal of Applied Earth Observation and Geoinformation, 112:102945, 2022

    work page 2022

  44. [44]

    Diego F. M. Burbano, Sonia H. Theodoro, Ana M. X. de Carvalho, and Claudete G. Ramos. Crushed volcanic rock as soil remineralizer: a strategy to overcome the global fertilizer crisis. Natural Resources Research, 31:2197–2210, 2022

    work page 2022

  45. [45]

    Stepwise ecological restoration: A framework for improving restoration outcomes.Geography and Sustainability, 5:160–166, 2024

    Jingjing Liu, Yuehan Dou, and Haoming Chen. Stepwise ecological restoration: A framework for improving restoration outcomes.Geography and Sustainability, 5:160–166, 2024

    work page 2024

  46. [46]

    Global-scale ERA5 product precipitation and temperature evaluation.Ecolog- ical Indicators, 166:112481, 2024

    Rui Liu et al. Global-scale ERA5 product precipitation and temperature evaluation.Ecolog- ical Indicators, 166:112481, 2024

    work page 2024

  47. [47]

    NOAA climate data record (CDR) of normalized difference vegetation index (NDVI), version 4

    Eric Vermote et al. NOAA climate data record (CDR) of normalized difference vegetation index (NDVI), version 4. NOAA National Climate Data Center, 2014. 31

    work page 2014