pith. machine review for the scientific record. sign in

arxiv: 2601.00857 · v1 · submitted 2025-12-30 · 💻 cs.LG

Recognition: 2 theorem links

· Lean Theorem

Harvesting AlphaEarth: Benchmarking the Geospatial Foundation Model for Agricultural Downstream Tasks

Yuchi Ma , Yawen Shen , Anu Swatantran , David B. Lobell
Authors on Pith no claims yet

Pith reviewed 2026-05-16 19:25 UTC · model grok-4.3

classification 💻 cs.LG
keywords geospatial foundation modelsAlphaEarthcrop yield predictiontillage mappingcover crop mappingremote sensingagricultural monitoringembeddings
0
0 comments X

The pith

AlphaEarth Foundation embeddings deliver competitive results with specialized remote sensing models for U.S. crop yield prediction and county-level tillage mapping when trained on local data.

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

This paper tests a new geospatial foundation model, AlphaEarth Foundation, on practical agricultural problems in the United States. Models built from its embeddings perform well on predicting crop yields, mapping tillage, and identifying cover crops. They match the accuracy of models designed specifically for remote sensing data in yield forecasting and county-scale tillage detection when given local training examples. The evaluation draws on datasets from public and private sources to cover different scales and places. The authors also identify shortcomings in how well the embeddings transfer to new locations, how easy they are to understand, and how responsive they are to changes over time.

Core claim

The central claim is that AlphaEarth Foundation embeddings, pre-trained on multi-source Earth observations, can be directly applied to agricultural downstream tasks such as crop yield prediction, tillage mapping, and cover crop mapping, generally showing strong performance and being competitive with purpose-built remote sensing models in yield prediction and county-level tillage mapping when trained on local data.

What carries the argument

AlphaEarth Foundation (AEF) embeddings, which are annual global embeddings derived from a geospatial foundation model pre-trained on continuous multi-source Earth observation data.

If this is right

  • AEF-based models can serve as a ready-to-use alternative for yield prediction without developing new remote sensing pipelines.
  • Local training data allows AEF to reach performance levels comparable to specialized models in tillage mapping at the county level.
  • Cover crop mapping benefits from AEF but may require additional adjustments for optimal results.
  • The observed limitations in spatial transferability indicate that AEF may need retraining or adaptation for use in new regions.
  • Low interpretability and time sensitivity suggest that AEF is less suitable for applications requiring detailed temporal analysis or explainable decisions.

Where Pith is reading between the lines

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

  • Foundation models like AEF could streamline agricultural monitoring by reducing the need for custom feature engineering in remote sensing applications.
  • Improving the time sensitivity of embeddings might enable better tracking of seasonal changes in crop conditions.
  • Hybrid approaches combining AEF with traditional RS models could address the transferability gaps identified.
  • Testing on global datasets beyond the U.S. would reveal if the competitiveness holds in diverse agricultural systems.

Load-bearing premise

The compiled datasets from public and private sources are representative and sufficient to evaluate generalizability across different scales and locations without significant biases in data quality or coverage.

What would settle it

A large-scale test on independent data from a new U.S. region or different year where AEF models show substantially lower accuracy than RS models in yield prediction or tillage mapping would falsify the competitiveness claim.

read the original abstract

Geospatial foundation models (GFMs) have emerged as a promising approach to overcoming the limitations in existing featurization methods. More recently, Google DeepMind has introduced AlphaEarth Foundation (AEF), a GFM pre-trained using multi-source EOs across continuous time. An annual and global embedding dataset is produced using AEF that is ready for analysis and modeling. The internal experiments show that AEF embeddings have outperformed operational models in 15 EO tasks without re-training. However, those experiments are mostly about land cover and land use classification. Applying AEF and other GFMs to agricultural monitoring require an in-depth evaluation in critical agricultural downstream tasks. There is also a lack of comprehensive comparison between the AEF-based models and traditional remote sensing (RS)-based models under different scenarios, which could offer valuable guidance for researchers and practitioners. This study addresses some of these gaps by evaluating AEF embeddings in three agricultural downstream tasks in the U.S., including crop yield prediction, tillage mapping, and cover crop mapping. Datasets are compiled from both public and private sources to comprehensively evaluate AEF embeddings across tasks at different scales and locations, and RS-based models are trained as comparison models. AEF-based models generally exhibit strong performance on all tasks and are competitive with purpose-built RS-based models in yield prediction and county-level tillage mapping when trained on local data. However, we also find several limitations in current AEF embeddings, such as limited spatial transferability compared to RS-based models, low interpretability, and limited time sensitivity. These limitations recommend caution when applying AEF embeddings in agriculture, where time sensitivity, generalizability, and interpretability is important.

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 / 3 minor

Summary. The manuscript evaluates AlphaEarth Foundation (AEF) geospatial embeddings on three U.S. agricultural downstream tasks—crop yield prediction, tillage mapping, and cover crop mapping—using compiled public and private datasets. It trains AEF-based models and compares them to purpose-built remote sensing (RS) models under local-training and spatial-transfer scenarios, reporting that AEF models show strong performance overall and are competitive with RS models for yield prediction and county-level tillage mapping when trained locally, while identifying limitations in spatial transferability, interpretability, and temporal sensitivity.

Significance. If the empirical comparisons hold, the work supplies a useful benchmark for geospatial foundation models in agriculture, quantifying where AEF embeddings match or fall short of specialized RS pipelines and thereby guiding both practitioners and future GFM development toward better temporal modeling and cross-location robustness.

major comments (2)
  1. [§4 (Datasets)] §4 (Datasets): the claim of comprehensive evaluation across scales and locations rests on compiled public and private sources, yet the manuscript provides only high-level descriptions of spatial coverage, temporal range, and class-balance handling; without these details it is impossible to determine whether the reported limited spatial transferability reflects model properties or dataset artifacts.
  2. [§5.2 (Yield prediction results)] §5.2 (Yield prediction results): competitiveness is asserted via point estimates of R² or RMSE, but the absence of statistical significance tests or confidence intervals on the AEF-versus-RS differences leaves the central performance claim vulnerable to sampling variability.
minor comments (3)
  1. [§3 (Methods)] The distinction between 'local' and 'transfer' training regimes should be formalized with an explicit protocol (e.g., county-level hold-out) in the methods section for reproducibility.
  2. [Figures 3–5] Figure captions for the tillage and cover-crop maps would benefit from explicit mention of the spatial resolution and the exact AEF embedding dimension used.
  3. [Abstract] The abstract states that internal AEF experiments outperformed operational models on 15 EO tasks; a brief citation or table reference to those tasks would help readers contextualize the agricultural results.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help improve the clarity and rigor of our manuscript. We address each major comment below and will revise the paper accordingly.

read point-by-point responses

  1. Referee: §4 (Datasets): the claim of comprehensive evaluation across scales and locations rests on compiled public and private sources, yet the manuscript provides only high-level descriptions of spatial coverage, temporal range, and class-balance handling; without these details it is impossible to determine whether the reported limited spatial transferability reflects model properties or dataset artifacts.

    Authors: We agree that more granular dataset details are necessary for full transparency. In the revised manuscript, we will expand §4 with additional specifics on spatial coverage (e.g., exact number of counties and states per task), temporal ranges (e.g., year spans for each dataset), and class-balance handling (e.g., explicit description of weighting or sampling strategies). These additions will help readers distinguish model limitations from potential dataset effects. revision: yes

  2. Referee: §5.2 (Yield prediction results): competitiveness is asserted via point estimates of R² or RMSE, but the absence of statistical significance tests or confidence intervals on the AEF-versus-RS differences leaves the central performance claim vulnerable to sampling variability.

    Authors: We acknowledge the value of statistical rigor for the performance comparisons. In the revised §5.2, we will incorporate bootstrap confidence intervals and paired significance tests (e.g., t-tests) on the differences in R² and RMSE between AEF-based and RS-based models to quantify uncertainty and strengthen the competitiveness claims. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical benchmarking of embeddings

full rationale

The paper conducts direct empirical comparisons of AEF embeddings against purpose-built RS models on compiled public and private agricultural datasets for yield prediction, tillage mapping, and cover crop mapping. No mathematical derivations, fitted parameters renamed as predictions, or self-citation chains underpin the central claims; performance metrics are computed on held-out test data independent of the embedding generation process. The study is self-contained against external benchmarks with no load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical benchmarking study with no new mathematical derivations, free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5613 in / 1062 out tokens · 40758 ms · 2026-05-16T19:25:09.959209+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Earth Embeddings Reveal Diverse Urban Signals from Space

    cs.LG 2026-04 unverdicted novelty 5.0

    Earth embeddings from satellite images predict neighborhood-level urban indicators with higher accuracy for built-environment outcomes than for behavior-driven ones, showing city-specific variation but year-to-year stability.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    Introduction Earth observation (EO), grounded in remote sensing imagery, has enabled scalable and timely monitoring of dynamic Earth system processes, with agriculture being one of its most widely applied domains. Since the launch of the first Landsat satellite in 1972, there were early studies that explored the association Under Review 2 between satellit...

    work page 1972

  2. [2]

    or contrastive learning (Wang et al., 2022). This pre-training process requires no task-specific labels and enables the model to extract generic features from the input remote sensing data that can be adapted to diverse downstream tasks through fine-tuning or zero-shot learning. Early GFMs were trained based on a single modality and targeted at a few rela...

    work page 2022

  3. [3]

    was pre-trained by jointly learning representations from optical images at both low and high scales, achieving improved accuracy in 8 land-use classification datasets. In addition to encoding spatial information, NASA and IBM released a multi-temporal GFM named Prithvi-EO-2.0 (Szwarcman et al., 2024), which explicitly leverages transformer attention acros...

    work page 2024

  4. [4]

    Although an increasing number of GFMs has been developed and released, several limitations remain in their application to practical downstream tasks

    integrates data from 11 satellite platforms and was pretrained progressively to learn both general representations and semantically enriched representations. Although an increasing number of GFMs has been developed and released, several limitations remain in their application to practical downstream tasks. Even though most GFM studies have released source...

    work page 2025

  5. [5]

    The corresponding agricultural data were collected as ground truth labels for model training and evaluation

    Experimental tasks and ground truth data We evaluated the potential of AEF embeddings for crop yield prediction, tillage classification, and cover crop mapping at both the regional scale (county level) and the field scale. The corresponding agricultural data were collected as ground truth labels for model training and evaluation. In particular, the county...

    work page 2021

  6. [6]

    The proportion of cropland under low-intensity tillage within each county were calculated as the ground truth data, and there are overall 2,001 county-year records

    We collected the data and combined conservation tillage and no-till as low-intensity tillage. The proportion of cropland under low-intensity tillage within each county were calculated as the ground truth data, and there are overall 2,001 county-year records. Field-level Spring tillage data were collected from farmers by Corteva Agriscience. The Corteva ti...

    work page 2016

  7. [7]

    Therefore, we dropped year 2017 and did experiments for 2018–2024, leading to a total number of 47,709 field-year samples

    Since the planting dates of the cover crops can happen in the fall after the harvest, or in the spring before the planting, we concatenated two-year AEF embeddings as the input predictors. Therefore, we dropped year 2017 and did experiments for 2018–2024, leading to a total number of 47,709 field-year samples. We did not conduct county-level cover crop ma...

    work page 2017

  8. [8]

    Table 1 A summary of the experimental settings. Task Level Year Crop # Samples Yield Prediction County 2017–2024 Corn 6,325 Soybean 6,024 Winter Wheat 3,020 Field 2017–2018 Corn 89,938 Soybean 73,492 2017–2022 Winter Wheat 20,401 Tillage Mapping County 2017&2022 N/A 2,001 Field 2017–2023 N/A 24,514 Cover Crop Mapping Field 2018–2024 N/A 47,709

    work page 2017

  9. [9]

    #"$% (1) GDD!=$$(max(0,min (T&−T'(),T'*+−T'()))),-&$%#

    Materials 3.1 Alpha Earth Foundation model The AEF model was trained on ~3 billions of observations across optical (Landsat 8/9, Sentinel-2), radar (Sentinel-1), LiDAR (GEDI), Climate (ERA5-Land), gravity fields (GRACE), Elevation (GLO-30), and text sources (Wikipedia) (Brown et al., 2025). A space–time encoder and a teacher–student framework was employed...

    work page 2025

  10. [10]

    Next, observations on the remaining pixels were aggregated to the county (field) level by calculating the mean values within each county (field) boundary

    to keep pixels on specific crop types and mask out non-cropland pixels. Next, observations on the remaining pixels were aggregated to the county (field) level by calculating the mean values within each county (field) boundary. Finally, the aggregated observations were downloaded from GEE, including AEF embeddings, time-series satellite remote sensing data...

    work page 2003

  11. [11]

    to each spectral bands and VIs based on all available observations during the growing season (Eq. (5)). 𝑦(𝑡)=𝑐+𝑎%cos(2𝜋𝑡)+𝑏%sin(2𝜋𝑡)+ 𝑎,cos(4𝜋𝑡)+𝑏,sin(4𝜋𝑡) (5) where t represents the date of the observation; c denotes the intercept coefficient; 𝑎% and 𝑏% (𝑎, and 𝑏,) represent the first-order (second-order) cosine and sine coefficients, respectively. The f...

    work page 2021

  12. [12]

    In total, 80 harmonic features were generated for each county or field, comprising 10 features for each of the 6 spectral bands and 2 VIs. Meanwhile, monthly GDD and PPT were used as predictor variables in the U.S., which added 10 additional predictors for corn and soybean (May to Sep) and 12 additional predictors for winter wheat (Jan to Jun). Figure 2 A...

    work page 2023

  13. [13]

    For regression tasks, coefficient of determination (R2) and root mean squared error (RMSE) are calculated as the evaluation metrics (Eq

    Each experiment is repeated five times under different random seeds, and the mean evaluation results are presented. For regression tasks, coefficient of determination (R2) and root mean squared error (RMSE) are calculated as the evaluation metrics (Eq. (9) – (10)). For classification tasks, the evaluation metrics include the overall accuracy (Accuracy), F...

    work page 2017

  14. [14]

    Considering that the XGB models slightly outperform the RF models in most cases, the comparison is presented using the results from the XGB models

    indicate that both AEF-based and RS-based models align well with the reported yields from USDA-NASS. Considering that the XGB models slightly outperform the RF models in most cases, the comparison is presented using the results from the XGB models. RS-based models showed notable underestimation of winter wheat yields in high-yield ranges (Figure 3(c2)), p...

    work page 2017

  15. [15]

    Generally, RF models had better performance than XGB models

    or with cover crops (class 1). Generally, RF models had better performance than XGB models. The degraded performance by XGB is potentially due to the noise in the input feature sets, since there were fewer satellite observations during winter and early spring because of high cloud coverage and the presence of snow and ice. (Fig. S1), leading to more noise...

    work page 2014

  16. [16]

    Global Change Biology 25, 2530–2543

    A critical review of the impacts of cover crops on nitrogen leaching, net greenhouse gas balance and crop productivity. Global Change Biology 25, 2530–2543. https://doi.org/10.1111/gcb.14644 Alonso-Ayuso, M., Gabriel, J.L., García-González, I., Del Monte, J.P ., Quemada, M.,

    work page doi:10.1111/gcb.14644

  17. [17]

    Crop Protection 112, 103–111

    Weed density and diversity in a long-term cover crop experiment background. Crop Protection 112, 103–111. https://doi.org/10.1016/j.cropro.2018.04.012 Archuleta, C.-A.M., Constance, E.W., Arundel, S.T., Lowe, A.J., Mantey, K.S., Phillips, L.A.,

    work page doi:10.1016/j.cropro.2018.04.012 2018

  18. [18]

    Global Food Security 23, 173–181

    The GEOGLAM crop monitor for AMIS: Assessing crop conditions in the context of global markets. Global Food Security 23, 173–181. https://doi.org/10.1016/j.gfs.2019.04.010 Boryan, C., Yang, Z., Mueller, R., Craig, M.,

    work page doi:10.1016/j.gfs.2019.04.010 2019

  19. [19]

    Geocarto International 26, 341–358

    Monitoring US agriculture: the US Department of Agriculture, National Agricultural Statistics Service, Cropland Data Layer Program. Geocarto International 26, 341–358. https://doi.org/10.1080/10106049.2011.562309 Breiman, L.,

    work page doi:10.1080/10106049.2011.562309 2011

  20. [20]

    Machine Learning 45, 5–32

    Random Forests. Machine Learning 45, 5–32. https://doi.org/10.1023/A:1010933404324 Brown, C.F ., Kazmierski, M.R., Pasquarella, V .J., Rucklidge, W.J., Zhang, C., Shelhamer, E., Lahera, E., Wiles, O., Ilyushchenko, S., Zhang, L.L., Alj, S., Schechter, E., Askay, S., Guinan, O., Moore, R., Boukouvalas, A., Kohli, P .,

    work page doi:10.1023/a:1010933404324

  21. [21]

    American Journal of Agricultural Economics 99, 592–613

    Impacts of Federal Crop Insurance on Land Use and Environmental Quality. American Journal of Agricultural Economics 99, 592–613. https://doi.org/10.1093/ajae/aaw075 Cong, Y ., Khanna, S., Meng, C., Liu, P ., Rozi, E., He, Y ., Burke, M., Lobell, D., Ermon, S.,

    work page doi:10.1093/ajae/aaw075

  22. [22]

    Remote Sensing of Environment 253, 112174

    A million kernels of truth: Insights into scalable satellite maize yield mapping and yield gap analysis from an extensive ground dataset in the US Corn Belt. Remote Sensing of Environment 253, 112174. https://doi.org/10.1016/j.rse.2020.112174 Eskandari, I., Navid, H., Rangzan, K.,

    work page doi:10.1016/j.rse.2020.112174 2020

  23. [23]

    International Soil and Water Conservation Research 4, 93–98

    Evaluating spectral indices for determining conservation and conventional tillage systems in a vetch-wheat rotation. International Soil and Water Conservation Research 4, 93–98. https://doi.org/10.1016/j.iswcr.2016.04.002 Under Review 22 Fawagreh, K., Gaber, M.M., Elyan, E.,

    work page doi:10.1016/j.iswcr.2016.04.002 2016

  24. [24]

    Systems Science & Control Engineering 2, 602–609

    Random forests: from early developments to recent advancements. Systems Science & Control Engineering 2, 602–609. https://doi.org/10.1080/21642583.2014.956265 Fendrich, A.N., Matthews, F ., Van Eynde, E., Carozzi, M., Li, Z., d’Andrimont, R., Lugato, E., Martin, P ., Ciais, P ., Panagos, P .,

    work page doi:10.1080/21642583.2014.956265 2014

  25. [25]

    Science of The Total Environment 873, 162300

    From regional to parcel scale: A high-resolution map of cover crops across Europe combining satellite data with statistical surveys. Science of The Total Environment 873, 162300. https://doi.org/10.1016/j.scitotenv.2023.162300 Gitelson, A.A., Viña, A., Arkebauer, T.J., Rundquist, D.C., Keydan, G., Leavitt, B.,

    work page doi:10.1016/j.scitotenv.2023.162300 2023

  26. [26]

    Earth-Science Reviews 243, 104462

    A scalable framework for quantifying field-level agricultural carbon outcomes. Earth-Science Reviews 243, 104462. https://doi.org/10.1016/j.earscirev.2023.104462 He, K., Chen, X., Xie, S., Li, Y ., Dollar, P ., Girshick, R.,

    work page doi:10.1016/j.earscirev.2023.104462 2023

  27. [27]

    In: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR), pp

    Masked Autoencoders Are Scalable Vision Learners, in: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Presented at the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, New Orleans, LA, USA, pp. 15979–15988. https://doi.org/10.1109/CVPR52688.2022.01553 Hollmann, N., Müller, S., Purucker, L., Kris...

    work page doi:10.1109/cvpr52688.2022.01553 2022

  28. [28]

    Nature 637, 319–326

    Accurate predictions on small data with a tabular foundation model. Nature 637, 319–326. https://doi.org/10.1038/s41586-024-08328-6 Kauth, R.J., Thomas, G.S.,

    work page doi:10.1038/s41586-024-08328-6

  29. [29]

    International Journal of Applied Earth Observation and Geoinformation 23, 192–203

    Winter wheat yield forecasting in Ukraine based on Earth observation, meteorological data and biophysical models. International Journal of Applied Earth Observation and Geoinformation 23, 192–203. https://doi.org/10.1016/j.jag.2013.01.002 Koudahe, K., Allen, S.C., Djaman, K.,

    work page doi:10.1016/j.jag.2013.01.002 2013

  30. [30]

    International Soil and Water Conservation Research 10, 343–354

    Critical review of the impact of cover crops on soil properties. International Soil and Water Conservation Research 10, 343–354. https://doi.org/10.1016/j.iswcr.2022.03.003 Lobell, D.B., Di Tommaso, S., Zhou, Q., Ma, Y ., Specht, J., Guan, K.,

    work page doi:10.1016/j.iswcr.2022.03.003 2022

  31. [31]

    Nat Sustain 1–9

    The mixed ehects of recent cover crop adoption on US cropland productivity. Nat Sustain 1–9. https://doi.org/10.1038/s41893-025-01599-5 Lu, C., Yu, Z., Hennessy, D.A., Feng, H., Tian, H., Hui, D.,

    work page doi:10.1038/s41893-025-01599-5

  32. [32]

    Nat Food 3, 266–274

    Emerging weed resistance increases tillage intensity and greenhouse gas emissions in the US corn–soybean cropping system. Nat Food 3, 266–274. https://doi.org/10.1038/s43016-022-00488-w Under Review 23 Luo, D., Zhang, H.K., Houborg, R., Ndekelu, L.M.N., Maimaitijiang, M., Tran, K.H., McMaine, J.,

    work page doi:10.1038/s43016-022-00488-w

  33. [33]

    Science of Remote Sensing 7, 100085

    Utility of daily 3 m Planet Fusion Surface Reflectance data for tillage practice mapping with deep learning. Science of Remote Sensing 7, 100085. https://doi.org/10.1016/j.srs.2023.100085 Ma, Y ., Chen, S., Ermon, S., Lobell, D.B., 2024a. Transfer learning in environmental remote sensing. Remote Sensing of Environment 301, 113924. https://doi.org/10.1016/j...

    work page doi:10.1016/j.srs.2023.100085 2023

  34. [34]

    Earth System Science Data 13, 4349–4383

    ERA5-Land: a state-of-the-art global reanalysis dataset for land applications. Earth System Science Data 13, 4349–4383. https://doi.org/10.5194/essd-13-4349-2021 Nakalembe, C., Becker-Reshef, I., Bonifacio, R., Hu, G., Humber, M.L., Justice, C.J., Keniston, J., Mwangi, K., Rembold, F ., Shukla, S., Urbano, F ., Whitcraft, A.K., Li, Y ., Zappacosta, M., Ja...

    work page doi:10.5194/essd-13-4349-2021 2021

  35. [35]

    Global Food Security 29, 100543

    A review of satellite-based global agricultural monitoring systems available for Africa. Global Food Security 29, 100543. https://doi.org/10.1016/j.gfs.2021.100543 Pedregosa, F ., Varoquaux, G., Gramfort, A., Michel, V ., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P ., Weiss, R., Dubourg, V ., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, ...

    work page doi:10.1016/j.gfs.2021.100543 2021

  36. [36]

    Renewable Agriculture and Food Systems 35, 38–48

    Cover crops use in Midwestern US agriculture: perceived benefits and net returns. Renewable Agriculture and Food Systems 35, 38–48. https://doi.org/10.1017/S1742170518000194 Reed, C.J., Gupta, R., Li, S., Brockman, S., Funk, C., Clipp, B., Keutzer, K., Candido, S., Uyttendaele, M., Darrell, T.,

    work page doi:10.1017/s1742170518000194

  37. [37]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp

    Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning, in: 2023 IEEE/CVF International Conference on Computer Vision (ICCV). Presented at the 2023 IEEE/CVF International Conference on Computer Vision (ICCV), IEEE, Paris, France, pp. 4065–4076. https://doi.org/10.1109/ICCV51070.2023.00378 Ritchie, J.T.,

    work page doi:10.1109/iccv51070.2023.00378 2023

  38. [38]

    John Wiley & Sons, Ltd, pp

    Wheat Phasic Development, in: Modeling Plant and Soil Systems. John Wiley & Sons, Ltd, pp. 31–54. https://doi.org/10.2134/agronmonogr31.c3 Sullivan, D.G., Truman, C.C., Schomberg, H.H., Endale, D.M., Strickland, T.C.,

    work page doi:10.2134/agronmonogr31.c3

  39. [39]

    Agronomy Journal 79, 53–60

    Estimating Corn Growth, Yield, and Grain Moisture from Air Growing Degree Days and Residue Cover1. Agronomy Journal 79, 53–60. https://doi.org/10.2134/agronj1987.00021962007900010012x Szwarcman, D., Roy, S., Fraccaro, P ., Gíslason, Þ.E., Blumenstiel, B., Ghosal, R., Oliveira, P .H. de, Almeida, J.L. de S., Sedona, R., Kang, Y ., Chakraborty, S., Wang, S....

    work page doi:10.2134/agronj1987.00021962007900010012x

  40. [40]

    https://doi.org/10.48550/arXiv.2412.02732 Tong, X.-Y ., Wang, S.,

    Prithvi-EO-2.0: A Versatile Multi-Temporal Foundation Model for Earth Observation Applications. https://doi.org/10.48550/arXiv.2412.02732 Tong, X.-Y ., Wang, S.,

    work page doi:10.48550/arxiv.2412.02732

  41. [41]

    Invariant Features for Global Crop Type Classification

    Invariant Features for Global Crop Type Classification. https://doi.org/10.48550/arXiv.2509.03497 USDA-NASS,

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2509.03497

  42. [42]

    IEEE Geoscience and Remote Sensing Magazine 10, 213–247

    Self-Supervised Learning in Remote Sensing: A review. IEEE Geoscience and Remote Sensing Magazine 10, 213–247. https://doi.org/10.1109/MGRS.2022.3198244 Wilson, B.T., Knight, J.F ., McRoberts, R.E.,

    work page doi:10.1109/mgrs.2022.3198244 2022

  43. [43]

    ISPRS Journal of Photogrammetry and Remote Sensing 137, 29–46

    Harmonic regression of Landsat time series for modeling attributes from national forest inventory data. ISPRS Journal of Photogrammetry and Remote Sensing 137, 29–46. https://doi.org/10.1016/j.isprsjprs.2018.01.006 Wu, K., Zhang, Yingying, Ru, L., Dang, B., Lao, J., Yu, L., Luo, J., Zhu, Z., Sun, Y ., Zhang, J., Zhu, Q., Wang, J., Yang, M., Chen, J., Zhan...

    work page doi:10.1016/j.isprsjprs.2018.01.006 2018

  44. [44]

    Nat Mach Intell 7, 1235–1249

    A semantic-enhanced multi-modal remote sensing foundation model for Earth observation. Nat Mach Intell 7, 1235–1249. https://doi.org/10.1038/s42256-025-01078-8 Xiong, X., Zhong, R., Jiang, H., Athanasiadis, I., Yang, Y ., Zhu, L., Lin, T.,

    work page doi:10.1038/s42256-025-01078-8

  45. [45]

    ISPRS Journal of Photogrammetry and Remote Sensing 231, 101–118

    Corn yield estimation under extreme climate stress with knowledge-encoded deep learning. ISPRS Journal of Photogrammetry and Remote Sensing 231, 101–118. https://doi.org/10.1016/j.isprsjprs.2025.10.020 Xiong, Z., Wang, Y ., Zhang, F ., Stewart, A.J., Hanna, J., Borth, D., Papoutsis, I., Saux, B.L., Camps-Valls, G., Zhu, X.X.,

    work page doi:10.1016/j.isprsjprs.2025.10.020 2025

  46. [46]

    Neural Plasticity-Inspired Multimodal Foundation Model for Earth Observation. https://doi.org/10.48550/arXiv.2403.15356 Zhang, C., Kerner, H., Wang, S., Hao, P ., Li, Z., Hunt, K.A., Abernethy, J., Zhao, H., Gao, F ., Di, L., Guo, C., Liu, Z., Yang, Z., Mueller, R., Boryan, C., Chen, Q., Beeson, P .C., Zhang, H.K., Shen, Y .,

    work page doi:10.48550/arxiv.2403.15356

  47. [47]

    Remote Sensing of Environment 330, 114995

    Remote sensing for crop mapping: A perspective on current and future crop-specific land cover data products. Remote Sensing of Environment 330, 114995. https://doi.org/10.1016/j.rse.2025.114995 Zhou, Q., Guan, K., Wang, Sheng, Jiang, C., Huang, Y ., Peng, B., Chen, Z., Wang, Sibo, Hipple, J., Schaefer, D., Qin, Z., Stroebel, S., Coppess, J., Khanna, M., Cai, Y .,

    work page doi:10.1016/j.rse.2025.114995 2025

  48. [48]

    Midwest Detected by Under Review 25 Fusing Multi-Source Satellite Data

    Recent Rapid Increase of Cover Crop Adoption Across the U.S. Midwest Detected by Under Review 25 Fusing Multi-Source Satellite Data. Geophysical Research Letters 49, e2022GL100249. https://doi.org/10.1029/2022GL100249 Zulauf, C., Schnitkey, G., Paulson, N., Coppess, and J.,

    work page doi:10.1029/2022gl100249

  49. [49]

    farmdoc daily

    Cover Crops and Covered Cropland, 2022 US Census of Agriculture. farmdoc daily

    work page 2022