Pith. sign in

REVIEW 3 major objections 5 minor 57 references

HeatCast: A Benchmark for Neighborhood-Scale LST Forecasting across 124 U.S. Cities

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper introduces HeatCast, an open, reproducible benchmark for monthly land-surface temperature forecasting at 30 m resolution across 124 U.S. cities, and reports that Earthformer achieves 7.74 K RMSE versus 10.42 K for a CNN+LSTM.

desk verdict A genuinely useful, released LST forecasting benchmark whose '30m' claim overstates thermal resolution and whose baselines need error bars. read the letter →

arxiv 2608.07640 v1 pith:E6CGNOWW submitted 2026-08-07 cs.CV

Jesus Guerrero , Isaac Corley , Leon Najafirad , Maryam Tabar , Paul Rad This is my paper · ORCID
classification cs.CV
keywords landsurfacetemperatureurbanheatislandbenchmarkdatasetspatiotemporalforecastingremotesensingclimateresilienceEarthformerlocalzones
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper introduces HeatCast, an open benchmark for forecasting monthly land-surface temperature at 30 m resolution across 124 U.S. cities from 2013 through June 2025. It defines a fixed task: given twelve months of nine aligned channels (LST, elevation, RGB, three spectral indices, and albedo) on 128x128 tiles, predict the next month's LST, with a temporal split, Local Climate Zone stratified metrics, and a reference evaluator. On this protocol, the paper reports that Earthformer reaches 7.74 K aggregate RMSE versus 10.42 K for a CNN+LSTM, and that an Earthformer using only the eight non-LST channels reaches 7.72 K. If the benchmark is adopted, urban-heat model comparisons become reproducible and comparable across studies. The data, code, and weights are released under MIT.

What carries the argument

The carrying object is the benchmark protocol itself: monthly 30 m tiles organized as 128x128 patches with nine aligned channels, a fixed one-month-ahead task conditioned on twelve monthly observations, a temporal split (train 2013-2021, validation 2022-2023, test January 2024-June 2025), and an evaluation script that reports mean tile-level RMSE in Kelvin, stratified by Local Climate Zone super-clusters. Earthformer, a space-time transformer using cuboid attention, is the reference model that produces the headline numbers; the released fixed sample manifests and reference evaluator are what let the results be regenerated and compared.

What would settle it

Run the released reference evaluator on a persistence baseline that predicts the last observed monthly LST, and on a monthly climatology baseline; if either reaches 7.7 K or better, the headline baselines do not demonstrate skill beyond the seasonal cycle, and if the 7.74 K and 7.72 K numbers cannot be regenerated from the released manifests and checkpoints, the benchmark's reproducibility claim fails.

Watch

Extended reading notes

Core claim

The paper claims that urban LST forecasting lacks a shared benchmark and that HeatCast fills this gap. The central discovery, stated on the benchmark's own terms, is that Earthformer reaches 7.74 K aggregate test RMSE on the fixed next-month task, compared with 10.42 K for the CNN+LSTM, and that an Earthformer trained on only the eight non-LST channels reaches 7.72 K, against 8.15 K from historical LST alone and 8.68 K from RGB alone. It also reports that the CNN+LSTM performs better on compact urban pixels (8.62 K vs 12.68 K), while Earthformer dominates the open urban, other urban, and natural categories. The claim is that these numbers are reproducible from the released manifests, configurations, and checkpoints, giving the field a common evaluation standard for 30 m urban heat forecasting.

Load-bearing premise

The benchmark's 'neighborhood-scale' claim assumes the Landsat Collection 2 LST product, which is derived from the 100 m native TIRS band and only distributed on the 30 m optical grid, actually carries thermal information at 30 m; if the true thermal support is 100 m, the benchmark standardizes a regridded product rather than 30 m thermal measurements.

Editorial extensions

If this is right

  • Any future forecasting model can be scored against the same 124 cities, months, and pixels as the published baselines, making cross-study comparisons possible.
  • The auxiliary-only result (7.72 K) implies that elevation, spectral indices, and albedo carry strong predictive signal for next-month heat, so feature engineering is as important as architecture choice.
  • LCZ-stratified metrics show compact urban pixels remain the least accurate class for Earthformer (12.68 K vs 8.62 K for CNN+LSTM), pointing to class imbalance as a concrete target.
  • Because the split is temporal, the benchmark measures how well models forecast future months in known cities; it does not by itself measure transfer to unseen cities.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: A persistence baseline that predicts last month's LST, or a monthly climatology, would sharpen the claim: if it lands near 7.7 K, the benchmark's baselines would show skill mostly from the seasonal cycle rather than from learned dynamics.
  • Inference: Because the Landsat thermal band is natively 100 m, users should read '30 m' as grid spacing, not thermal resolution; a planning study that treats 30 m LST hotspots as thermal measurements would over-interpret the product.
  • Inference: The released artifacts make a natural held-out-city experiment possible, training on a subset of cities and testing on the rest, which the current temporal split does not address.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. HeatCast introduces a Landsat-based benchmark for monthly land surface temperature forecasting at a nominal 30 m grid resolution across 124 U.S. cities, covering mid-2013 through June 2025. The dataset provides nine aligned channels, Local Climate Zone labels, quality masks, fixed train/validation/test splits, and an evaluation harness reporting LCZ-stratified RMSE in Kelvin. The paper evaluates a CNN+LSTM and Earthformer on a 12-month-to-next-month task, reporting 10.42 K versus 7.74 K aggregate RMSE, and an auxiliary-only Earthformer variant at 7.72 K. The data, code, configurations, and checkpoints are released under an MIT license.

Significance. The contribution is potentially valuable: HeatCast appears to be the first open, reproducible benchmark for urban LST forecasting at a 30 m grid, with a clear task definition, frozen splits, explicit NoData masks, and released artifacts. Its geographic and temporal scope (124 cities, 12 years) is much larger than typical one-to-three-city LST studies, and its LCZ-stratified evaluation is a sensible response to class imbalance. The paper also benefits from transparent limitation statements in Section 9. However, the significance is conditional on a central qualification: the LST target is derived from a 100 m native thermal band and only gridded at 30 m, so the '30 m neighborhood-scale' claim needs to be reframed or empirically supported. The benchmark remains useful as a standardized forecasting task even after this reframing.

major comments (3)
  1. [3.1, Table 2, Section 9] The headline characterization of HeatCast as a 30 m neighborhood-scale benchmark is not supported by the paper's own source-product description. Table 2 states that the Landsat TIRS surface temperature product has '100 m→30 m' resolution, and Section 3.1 explains that USGS derives ST from the 100 m TIRS band and distributes it on the 30 m optical grid. Since the target channel carries thermal information at roughly 100 m native support, the 30 m grid contains resampled or interpolated thermal detail rather than 30 m measurements. The abstract, title, and Figures 1 and 5 foreground '30 m' and 'block-scale thermal gradients,' but the caveat appears only in Section 9. This is load-bearing for the central claim. I recommend either qualifying all resolution claims, for example by saying '30 m grid, 100 m native thermal support,' or adding an analysis that quantifies the impact, such as comparing forecasts and RMSE against an evaluation aggregated to 100 m or against a product with genuine 30 m thermal support. The benchmark itself can remain as released, but the characterization needs revision.
  2. [6.2, Table 8, Section 9] The feature-set ablation is presented as a headline result ('Earthformer achieves its lowest RMSE with auxiliary inputs', 7.72 K versus 7.74 K), yet Section 9 states that each baseline uses one training run and that the 0.02 K difference is not a stable ranking. The paper therefore both relies on and disclaims this difference. Because the feature-set ablation is one of the stated contributions, Table 8 and the corresponding text should either provide multiple-seed means and standard deviations or explicitly demote the 7.72 K versus 7.74 K comparison to an observation without a 'best' designation. Without error bars, the claim that non-LST channels are the strongest input is not supported at the reported precision.
  3. [6.2, Section 7, Section 9] The resolution issue also affects the interpretation of the ablation. Because the eight auxiliary channels are natively 30 m while the LST target is natively 100 m, the auxiliary-only Earthformer may be learning to predict the 30 m spatial pattern of the regridded target rather than thermal dynamics; the reported 0.43 K improvement over LST history could reflect the spatial resolution mismatch. At minimum, the paper should compare against a spatial persistence or monthly climatology baseline (which Section 7 lists as future work) and discuss whether the auxiliary advantage would survive an evaluation on 100 m aggregates. This comparison is needed before the auxiliary-only result is presented as evidence about input informativeness.
minor comments (5)
  1. [3.1, 3.5] The monthly lowest-cloud-scene selection does not account for differences in overpass time between Landsat 8 and 9 or across months, which can add non-climatic variance to the target; a sentence acknowledging this and any normalization would help.
  2. [4.3] Please clarify how LCZ-stratified RMSE is aggregated: whether pixels are grouped by their own LCZ label before averaging, and whether tile-level weighting applies within each stratum.
  3. [Table 3] The LST minimum of -123°C is listed among observed ranges; although the text says residual outliers are masked before evaluation, the table may confuse readers about the valid range, so consider reporting QC-passed ranges instead.
  4. [Figure 1] The caption asserts 10–15 K variation within a single neighborhood; if the LST product is resampled from 100 m native support, this statement should be tied to the actual effective resolution or softened.
  5. [Listing 1] The quickstart uses test_years=[2024, 2025], but the test set ends in June 2025; a comment noting the six-month cutoff would avoid ambiguity.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the benchmark is an empirical assembly of public satellite data and the reported results are measurements, with only minor non-load-bearing self-citations.

full rationale

HeatCast's central claims are (1) the release of a large, reproducible benchmark and (2) reference baseline performance numbers. Neither reduces to its own inputs by construction. The forecasting task uses a 12-month input window to predict next-month LST, and the reported RMSE values come from training and evaluating models on the fixed temporal split; the target month is never part of the input sequence, so the prediction is not definitionally forced. The ablation results are also measurements: the auxiliary-channel Earthformer is trained without the LST channel and evaluated against future LST, so the comparison is genuine rather than a fitted parameter renamed as a prediction. The paper does cite two prior works sharing authors with the present paper, [7] and [39], but only to justify the mosaicking/reprojection pipeline and to identify a related Landsat benchmark; these citations are not used to produce the reported RMSEs or to justify the central forecasting claim, so they are not load-bearing. The acknowledged limitation that the USGS Collection 2 LST product has 100 m native thermal support while being distributed on a 30 m grid is a physical-resolution caveat about the target variable, not a circular derivation; the benchmark still standardizes forecasting of the distributed 30 m-aligned product, and the reported errors remain valid measurements for that product. No equation in the paper defines a claimed result in terms of the result itself, and no fitted value is relabeled as a prediction. The derivation chain is therefore self-contained with respect to circularity concerns.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The benchmark rests on standard public geospatial data plus four hand-set protocol choices. No new physical entities are postulated. The most consequential unproven premise is that 30 m gridding of a 100 m native LST product supports neighborhood-scale thermal claims.

free parameters (4)
  • Lowest-cloud scene fallback limit = 20
    Chosen by hand in §3.1 acquisition; sets how aggressively a month's scene is replaced, which shapes coverage and per-month representativeness.
  • Max consecutive missing months for interpolation = 2
    Chosen by hand in §3.5 QC; tiles with longer gaps are excluded, changing the training and test distribution.
  • Max input no-data fraction per sample = 0.60
    Appears in the released loader in §4.5; controls which sequences are valid and affects the reported errors.
  • Temporal split boundaries = train 2013-2021; val 2022-2023; test 2024-2025
    Hand-selected split defining the benchmark; all reported RMSEs depend on it.
assumptions (5)
  • domain assumption Landsat Collection 2 Level-2 ST product accuracy of 1-2 K under cloud-free conditions transfers to the benchmark labels.
    Invoked in §3.2 and §9; benchmark targets inherit any product-level bias.
  • domain assumption A LST product with 100 m native TIRS support, regridded to 30 m, can represent neighborhood-scale thermal structure.
    Relied on by the central 30 m claim; disclosed in §9 but not tested.
  • domain assumption QA_PIXEL cloud, cloud-shadow, and invalid flags correctly identify pixels to mask before index computation and evaluation.
    Used in §3.5; masking errors propagate into both training and test errors.
  • domain assumption CONUS LCZ labels correctly assign each pixel to one of 17 classes for stratified evaluation.
    Used in §3.3 and §4.3; §9 acknowledges source-product errors.
  • domain assumption Choosing cities with Esri urban footprints larger than 90 mi2 yields a representative benchmark population.
    City selection in §3.1; §9 notes coverage favors large cloud-free cities.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HeatCast: A Benchmark for Neighborhood-Scale LST Forecasting across 124 U.S. Cities." pith.science (2026). https://pith.science/paper/E6CGNOWW

@misc{pith2026260807640,
  author       = {Pith},
  title        = {Pith review of: HeatCast: A Benchmark for Neighborhood-Scale LST Forecasting across 124 U.S. Cities},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E6CGNOWW}},
  note         = {Machine review of arXiv:2608.07640}
}
read the original abstract

Land Surface Temperature (LST) is a widely used satellite-derived measure of urban surface heat, but there is no shared benchmark for forecasting it at 30 m. Prior studies usually cover one to three cities, use kilometer-scale products, or do not release data and code. We introduce HeatCast, a Landsat-based benchmark for monthly LST forecasting across 124 U.S. cities from 2013 through June 2025. HeatCast contains 30 m monthly tiles with LST, elevation, surfacereflectance RGB, three spectral indices, broadband albedo, quality masks, and Local Climate Zone (LCZ) labels, together with a fixed temporal split, LCZ-stratified metrics, and a reference evaluation harness. We evaluate a CNN+LSTM and Earthformer on next-month forecasting, where Earthformer reaches 7.74 K RMSE against 10.42 K for the CNN+LSTM. Forecasting from the eight nonLST channels alone reaches 7.72 K, against 8.15 K from LST history and 8.68 K from RGB. The data, code, and weights are released under MIT at https://doi.org/10.57967/hf/9889.

Figures

Figures reproduced from arXiv: 2608.07640 by the authors.

Figure 1
Figure 1. Landsat-derived LST over downtown San Antonio, TX. Surface temperature varies by 10–15 K within a single [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. HeatCast covers 124 U.S. cities. Marker color denotes per-city mean LST, and marker size denotes the number of [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. San Antonio shows a strong annual LST cycle. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: All HeatCast channels share the same 30 m grid. The panels show LST, RGB, three spectral indices, broadband albedo, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Intra-city LST patterns appear across the benchmark. The panels show time-mean LST for eighteen cities sampled [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Temporal split overlaid on the benchmark-wide LST trace. The line is the median per-city scene-mean LST across the [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Each city is distributed as a Zarr v3 store. Channels [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Seasonal and spatial LST patterns differ across cities. Each Hovmöller diagram shows time horizontally, north-to-south [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Annual-mean LST trends vary across HeatCast cities. [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 30 canonical work pages

  1. [1]

    Hashem Akbari, Surabi Menon, and Arthur Rosenfeld. 2009. Global Cooling: Increasing World-wide Urban Albedos to Offset CO2.Climatic Change94, 3–4 (2009), 275–286. doi:10.1007/s10584-008-9515-9

    work page doi:10.1007/s10584-008-9515-9 2009

  2. [2]

    Benjamin Bechtel, Matthias Demuzere, Gerald Mills, Wenfeng Zhan, Panagiotis Sismanidis, Christopher Small, and James Voogt. 2019. SUHI analysis using Local Climate Zones—A comparison of 50 cities.Urban Climate28 (2019), 100451. doi:10.1016/j.uclim.2019.01.005 HeatCast: A Benchmark for Neighborhood-Scale LST Forecasting across 124 U.S. Cities SIGSPATIAL ’2...

    work page doi:10.1016/j.uclim.2019.01.005 2019

  3. [3]

    Sofiane Bouaziz, Adel Hafiane, Raphaël Canals, and Rachid Nedjai. 2026. Deep learning for spatio-temporal fusion in land surface temperature estimation: A comprehensive survey, experimental analysis, and future trends.Remote Sensing 18, 2 (2026), 289. doi:10.3390/rs18020289

    work page doi:10.3390/rs18020289 2026

  4. [4]

    Zachary D Calhoun, Frank Willard, Chenhao Ge, Claudia Rodriguez, Mike Bergin, and David Carlson. 2024. Estimating the effects of vegetation and increased albedo on the urban heat island effect with spatial causal inference.Scientific Reports14, 1 (2024), 540. doi:10.1038/s41598-023-50981-w

    work page doi:10.1038/s41598-023-50981-w 2024

  5. [5]

    Xiao-Ling Chen, Hong-Mei Zhao, Ping-Xiang Li, and Zhi-Yong Yin. 2006. Remote sensing image-based analysis of the relationship between urban heat island and land use/cover changes.Remote Sensing of Environment104, 2 (2006), 133–146. doi:10.1016/j.rse.2005.11.016

    work page doi:10.1016/j.rse.2005.11.016 2006

  6. [6]

    Yezhen Cong, Samar Khanna, Chenlin Meng, Patrick Liu, Erik Rozi, Yutong He, Marshall Burke, David Lobell, and Stefano Ermon. 2022. SatMAE: Pre-training transformers for temporal and multi-spectral satellite imagery. InAdvances in Neural Information Processing Systems, Vol. 35. Curran Associates, Inc., Red Hook, NY, USA, 197–211. arXiv:2207.08051

    arXiv 2022

  7. [7]

    Isaac Corley, Lakshay Sharma, and Ruth Crasto. 2025. Landsat-Bench: Datasets and Benchmarks for Landsat Foundation Models. arXiv preprint arXiv:2506.08780. arXiv:2506.08780 [cs.CV] doi:10.48550/arXiv.2506.08780

    work page Pith review arXiv doi:10.48550/arxiv.2506.08780 2025

  8. [8]

    Matthias Demuzere, Steve Hankey, Gerald Mills, Wenwen Zhang, Tianjun Lu, and Benjamin Bechtel. 2020. Combining expert and crowd-sourced training data to map urban form and functions for the continental US.Scientific Data7, 1 (2020), 264. doi:10.1038/s41597-020-00605-z

    work page doi:10.1038/s41597-020-00605-z 2020

Show all 57 references
  1. [9]

    Feng Gao, Jeffrey Masek, Matt Schwaller, and Forrest Hall. 2006. On the Blending of the Landsat and MODIS Surface Reflectance: Predicting Daily Landsat Surface Reflectance.IEEE Transactions on Geoscience and Remote Sensing44, 8 (2006), 2207–2218. doi:10.1109/TGRS.2006.872081

    2006

  2. [10]

    Zhihan Gao, Xingjian Shi, Hao Wang, Yi Zhu, Yuyang Bernie Wang, Mu Li, and Dit-Yan Yeung. 2022. Earthformer: Exploring space-time transformers for earth system forecasting. InAdvances in Neural Information Processing Systems, Vol. 35. Curran Associates, Inc., Red Hook, NY, USA...

    2022 arXiv

  3. [11]

    Zhangyang Gao, Cheng Tan, Lirong Wu, and Stan Z Li. 2022. SimVP: Simpler yet better video prediction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, New Orleans, LA, USA, 3160–3170. doi:10.1109/CVPR52688.2022.00317

    2022

  4. [12]

    Antonio Gasparrini, Yuming Guo, Masahiro Hashizume, Eric Lavigne, Antonella Zanobetti, Joel Schwartz, Aurelio Tobias, Shilu Tong, Joacim Rocklöv, Bertil Forsberg, Michela Leone, Manuela De Sario, Michelle L. Bell, Yue-Liang Leon Guo, Chang-fu Wu, Haidong Kan, Seung-Muk Yi, Mic...

    2015

  5. [13]

    Siavash Ghorbany, Ming Hu, Siyuan Yao, and Chaoli Wang. 2024. Towards a sustainable urban future: A comprehensive review of urban heat island research technologies and machine learning approaches.Sustainability16, 11 (2024), 4609. doi:10.3390/su16114609

    2024 doi

  6. [14]

    Jiahao Han, Shibo Fang, Qianchuan Mi, Xinyu Wang, Yanru Yu, Wen Zhuo, and Xiaofeng Peng. 2024. A time-continuous land surface temperature (LST) data fusion approach based on deep learning with microwave remote sensing and high-density ground truth observations.Science of The T...

    2024

  7. [15]

    Farzad Hashemi and Mahsa Adib. 2024. Examining thermal inequities: Land surface temperature, social vulnerability, and historical redlining in San Antonio, TX.Urban Climate55 (2024), 101960. doi:10.1016/j.uclim.2024.101960

    2024

  8. [16]

    Juelin He, Wei Zhao, Ainong Li, Fengping Wen, and Daijun Yu. 2019. The impact of the terrain effect on land surface temperature variation based on Landsat-8 observations in mountainous areas.International Journal of Remote Sensing40, 5-6 (2019), 1808–1827. doi:10.1080/01431161...

    2019

  9. [17]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural Computation9, 8 (1997), 1735–1780. doi:10.1162/neco.1997.9.8.1735

    1997 doi

  10. [18]

    Jeremy S Hoffman, Vivek Shandas, and Nicholas Pendleton. 2020. The effects of historical housing policies on resident exposure to intra-urban heat: a study of 108 US urban areas.Climate8, 1 (2020), 12. doi:10.3390/cli8010012

    2020 doi

  11. [19]

    Angel Hsu, Glenn Sheriff, Tirthankar Chakraborty, and Diego Manya. 2021. Disproportionate exposure to urban heat island intensity across major US cities. Nature Communications12, 1 (2021), 2721. doi:10.1038/s41467-021-22799-5

    2021 doi

  12. [20]

    Johannes Jakubik, Sujit Roy, C. E. Phillips, Paolo Fraccaro, Denys Godwin, Bianca Zadrozny, Daniela Szwarcman, Carlos Gomes, Gabby Nyirjesy, Blair Edwards, Daiki Kimura, Naomi Simumba, Linsong Chu, S. Karthik Mukkavilli, Devyani Lambhate, Kamal Das, Ranjini Bangalore, Dario Ol...

  13. [21]

    Serkan Kartal and Aliihsan Sekertekin. 2022. Prediction of MODIS land surface temperature using new hybrid models based on spatial interpolation techniques and deep learning models.Environmental Science and Pollution Research29, 44 (2022), 67115–67134. doi:10.1007/s11356-022-20572-9

    2022 doi

  14. [22]

    Dirk Lauwaet, Julie Berckmans, Hans Hooyberghs, Hendrik Wouters, Guy Driesen, Filip Lefebre, and Koen De Ridder. 2024. High resolution modelling of the urban heat island of 100 European cities.Urban Climate54 (March 2024), 101850. doi:10.1016/j.uclim.2024.101850

    2024

  15. [23]

    Shunlin Liang. 2001. Narrowband to broadband conversions of land surface albedo I: Algorithms.Remote Sensing of Environment76, 2 (2001), 213–238. doi:10.1016/S0034-4257(00)00205-4

    2001 doi

  16. [24]

    Shengjie Liu, Siqin Wang, and Lu Zhang. 2025. Daily Land Surface Temperature Reconstruction in Landsat Cross-Track Areas Using Deep Ensemble Learning With Uncertainty Quantification.IEEE Transactions on Geoscience and Remote Sensing63 (2025), 1–18. doi:10.1109/TGRS.2025.3643985

    2025

  17. [25]

    Ilya Loshchilov and Frank Hutter. 2019. Decoupled Weight Decay Regularization. InInternational Conference on Learning Representations (ICLR). OpenReview.net, New Orleans, LA, USA. arXiv:1711.05101 https://openreview.net/forum?id= Bkg6RiCqY7

    2019 arXiv

  18. [26]

    Rajesh Maddu, Abhishek Reddy Vanga, Jashwanth Kumar Sajja, Ghouse Basha, and Rehana Shaik. 2021. Prediction of land surface temperature of major coastal cities of India using bidirectional LSTM neural networks.Journal of Water and Climate Change12, 8 (2021), 3801–3819. doi:10....

    2021 doi

  19. [27]

    Malakar, Glynn C

    Nabin K. Malakar, Glynn C. Hulley, Simon J. Hook, Kelly Laraby, Monica Cook, and John R. Schott. 2018. An Operational Land Surface Temperature Product for Landsat Thermal Data: Methodology and Validation.IEEE Transactions on Geoscience and Remote Sensing56, 10 (2018), 5717–573...

    2018 doi

  20. [28]

    Stuart K McFeeters. 1996. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features.International Journal of Remote Sensing17, 7 (1996), 1425–1432. doi:10.1080/01431169608948714

    1996 doi

  21. [29]

    Yujia Ming, Yong Liu, Yingpeng Li, and Yongze Song. 2024. Unraveling nonlinear and spatial non-stationary effects of urban form on surface urban heat islands using explainable spatial machine learning.Computers, Environment and Urban Systems114 (Dec. 2024), 102200. doi:10.1016...

    2024

  22. [30]

    Juan Nathaniel, Yongquan Qu, Tung Nguyen, Sungduk Yu, Julius Busecke, Aditya Grover, and Pierre Gentine. 2024. ChaosBench: A multi-channel, physics-based benchmark for subseasonal-to-seasonal climate prediction. InAdvances in Neural Information Processing Systems (NeurIPS) Dat...

    2024 arXiv

  23. [31]

    Lu Niu, Zhong Peng, Ronglin Tang, and Zhengfeng Zhang. 2021. Development of a Long-Term Dataset of China Surface Urban Heat Island for Policy Making: Spatio-Temporal Characteristics. In2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS. IEEE, Brussels, Belg...

    2021

  24. [32]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Des- maison, Andreas Köpf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, L...

    2019 arXiv

  25. [33]

    Lutz Prechelt. 1998. Early Stopping-But When?. InNeural Networks: Tricks of the Trade, This Book is an Outgrowth of a 1996 NIPS Workshop. Springer-Verlag, Berlin, Heidelberg, 55–69. doi:10.1007/3-540-49430-8_3

    1998 doi

  26. [34]

    Stephan Rasp, Peter D Dueben, Sebastian Scher, Jonathan A Weyn, Soukayna Mouatadid, and Nils Thuerey. 2020. WeatherBench: A benchmark data set for data-driven weather forecasting.Journal of Advances in Modeling Earth Systems 12, 11 (2020), e2020MS002203. doi:10.1029/2020MS002203

    2020 doi

  27. [35]

    Christian Requena-Mesa, Vitus Benson, Markus Reichstein, Jakob Runge, and Joachim Denzler. 2021. EarthNet2021: A large-scale dataset and challenge for Earth surface forecasting as a guided video prediction task. InProceedings of the IEEE/CVF Conference on Computer Vision and P...

    2021

  28. [36]

    Bishal Roy and Ehsanul Bari. 2022. Examining the relationship between land surface temperature and landscape features using spectral indices with Google Earth Engine.Heliyon8, 9 (2022), e10668. doi:10.1016/j.heliyon.2022.e10668

    2022 doi

  29. [37]

    Xingjian Shi, Zhourong Chen, Hao Wang, Dit-Yan Yeung, Wai-Kin Wong, and Wang-chun Woo. 2015. Convolutional LSTM network: A machine learning ap- proach for precipitation nowcasting. InAdvances in Neural Information Processing Systems. Curran Associates, Inc., Red Hook, NY, USA,...

    2015 arXiv

  30. [38]

    Karen Simonyan and Andrew Zisserman. 2014. Two-Stream Convolutional Networks for Action Recognition in Videos. InAdvances in Neural Information Processing Systems (NeurIPS). Curran Associates, Inc., Red Hook, NY, USA, 568–

    2014

  31. [39]

    Adam Stewart, Nils Lehmann, Isaac Corley, Yi Wang, Yi-Chia Chang, Nassim Ait Ali Braham, Shradha Sehgal, Caleb Robinson, and Arindam Banerjee. 2023. SIGSPATIAL ’26, November 3–6, 2026, Riverside, CA, USA Jesus Guerrero, Isaac Corley, Leon Najafirad, Maryam Tabar, and Paul Rad ...

    2023 arXiv

  32. [40]

    Ian D Stewart and Tim R Oke. 2012. Local climate zones for urban temperature studies.Bulletin of the American Meteorological Society93, 12 (2012), 1879–1900. doi:10.1175/BAMS-D-11-00019.1

    2012 doi

  33. [41]

    Amir Tahooni, Ata Abdollahi Kakroodi, and Majid Kiavarz. 2023. Monitoring of land surface albedo and its impact on land surface temperature (LST) using time series of remote sensing data.Ecological Informatics75 (2023), 102118. doi:10.1016/j.ecoinf.2023.102118

    2023

  34. [42]

    Alireza Attarhay Tehrani, Omid Veisi, Kambiz Kia, Yasin Delavar, Sasan Bahrami, Saeideh Sobhaninia, and Asma Mehan. 2024. Predicting urban Heat Island in European cities: A comparative study of GRU, DNN, and ANN models using urban morphological variables.Urban Climate56 (July ...

    2024

  35. [43]

    Compton J. Tucker. 1979. Red and Photographic Infrared Linear Combinations for Monitoring Vegetation.Remote Sensing of Environment8, 2 (1979), 127–150. doi:10.1016/0034-4257(79)90013-0

    1979 doi

  36. [44]

    Sajid Ullah, Xiuchen Qiao, and Aqil Tariq. 2025. Impact assessment of planned and unplanned urbanization on land surface temperature in Afghanistan using machine learning algorithms: a path toward sustainability.Scientific Reports15, 1 (Jan. 2025), 3092. doi:10.1038/s41598-025-87234-x

    2025 doi

  37. [45]

    E Vermote, Jean-Claude Roger, Belen Franch, and Sergii Skakun. 2018. LaSRC (Land Surface Reflectance Code): Overview, application and validation using MODIS, VIIRS, LANDSAT and Sentinel-2 data’s. InIGARSS 2018 – 2018 IEEE International Geoscience and Remote Sensing Symposium. ...

    2018

  38. [46]

    Han Wang, Jiachuan Yang, Guangzhao Chen, Chao Ren, and Jize Zhang. 2023. Machine learning applications on air temperature prediction in the urban canopy layer: A critical review of 2011–2022.Urban Climate49 (May 2023), 101499. doi:10.1016/j.uclim.2023.101499

    2023

  39. [47]

    Yu, and Mingsheng Long

    Yunbo Wang, Haixu Wu, Jianjin Zhang, Zhifeng Gao, Jianmin Wang, Philip S. Yu, and Mingsheng Long. 2023. PredRNN: A Recurrent Neural Network for Spatiotemporal Predictive Learning.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 2 (2023), 2208–2225. doi:10.1109...

    2023

  40. [48]

    Zi Wang, Rui Zhou, and Yang Yu. 2025. The impact of urban morphology on land surface temperature under seasonal and diurnal variations: marginal and interaction effects.Building and Environment272 (2025), 112673. doi:10.1016/j. buildenv.2025.112673

    2025

  41. [49]

    Ding Wen, Li Wang, Qian Cao, Man Hong, Hao Wang, and Guojian Bian. 2024. A comparative study of the effects of urban morphology on land surface temperature in Chengdu and Chongqing, China.Scientific Reports14, 1 (2024), 25130. doi:10. 1038/s41598-024-77036-y

    2024

  42. [50]

    Qihao Weng, Dengsheng Lu, and Jacquelyn Schubring. 2004. Estimation of land surface temperature–vegetation abundance relationship for urban heat island studies.Remote Sensing of Environment89, 4 (2004), 467–483. doi:10.1016/j.rse. 2003.11.005

    2004 doi

  43. [51]

    Donghang Wu, Weiquan Liu, Bowen Fang, Linwei Chen, Yu Zang, Lei Zhao, Shenlong Wang, Cheng Wang, José Marcato, and Jonathan Li. 2022. Intracity Temperature Estimation by Physics Informed Neural Network Using Modeled Forcing Meteorology and Multispectral Satellite Imagery.IEEE ...

    2022

  44. [52]

    Shuo Xu, Dongdong Wang, Shunlin Liang, Aolin Jia, Ruohan Li, Zhihao Wang, and Yuling Liu. 2024. A novel approach to estimate land surface temperature from landsat top-of-atmosphere reflective and emissive data using transfer-learning neural network.Science of The Total Environ...

    2024

  45. [53]

    Qiquan Yang, Yi Xu, T. C. Chakraborty, Meng Du, Ting Hu, Ling Zhang, Yue Liu, Rui Yao, Jie Yang, Shurui Chen, Changjiang Xiao, Renrui Liu, Mingjie Zhang, and Rui Chen. 2024. A global urban heat island intensity dataset: Generation, comparison, and analysis.Remote Sensing of En...

    2024

  46. [54]

    Huaxiu Yao, Fei Wu, Jintao Ke, Xianfeng Tang, Yitian Jia, Siyu Lu, Pinghua Gong, Jieping Ye, and Zhenhui Li. 2018. Deep multi-view spatial-temporal network for taxi demand prediction. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 32. AAAI Press, Palo Al...

    2018

  47. [55]

    Yong Zha, Jay Gao, and Shaoxiang Ni. 2003. Use of normalized difference built-up index in automatically mapping urban areas from TM imagery.International Journal of Remote Sensing24, 3 (2003), 583–594. doi:10.1080/01431160304987

    2003 doi

  48. [56]

    Chunhong Zhao, Jennifer Jensen, Qihao Weng, and Russell Weaver. 2018. A Geographically Weighted Regression Analysis of the Underlying Factors Related to the Surface Urban Heat Island Phenomenon.Remote Sensing10, 9 (2018), 1428. doi:10.3390/rs10091428

    2018 doi

  49. [2023]

    arXiv preprint arXiv:2310.18660

    Foundation Models for Generalist Geospatial Artificial Intelligence. arXiv preprint arXiv:2310.18660. arXiv:2310.18660 [cs.CV] doi:10.48550/arXiv.2310. 18660

    doi arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.