US-UrbanLST: A Large-Scale High-Resolution Dataset to Advance Research on Land Surface Temperature

GitHub • HuggingFace Dataset • Visualization

KDD 2026 Datasets & Benchmarks Track Submission

A large-scale open-source benchmark dataset for predicting monthly Land Surface Temperature (LST) at 30m spatial resolution across 124 U.S. metropolitan areas, with Earthformer and CNN+LSTM baselines and an interactive web application.

---

Overview

Land Surface Temperature (LST) serves as a critical indicator for quantifying urban heat islands and informing climate-resilient urban planning, particularly for vulnerable communities. However, the lack of open-source, large-scale, spatio-temporal datasets poses significant challenges to research at the national scale across the United States.

This repository presents:

• A benchmark dataset spanning 124 U.S. cities from 2013-2025 (~1.4 million 128x128 tiles at 30m resolution)
• Earthformer and CNN+LSTM baselines with reproducible training pipelines
• An interactive web application for visualizing LST predictions over San Antonio, TX

---

Key Contributions

1. Cross-city prediction: 124 cities spanning diverse climates, enabling models that generalize beyond single-city training
2. Neighborhood-scale resolution: 30m spatial resolution, sufficient to distinguish thermal differences between adjacent blocks
3. Reproducible baselines: Open code, data, and model weights; Earthformer achieves 13.93°F RMSE, a ~26% improvement over CNN+LSTM
4. LST Web Application: Integrates LST predictions into mapping software through an LLM interface for planning guidance

---

Dataset

Download

The dataset is available on HuggingFace: JesseGuerrero/US-UrbanLST

huggingface-cli download JesseGuerrero/US-UrbanLST --local-dir ./Data/ML --repo-type dataset

Data Components

Feature	Description	Resolution
LST	Land Surface Temperature (°F)	30m, Monthly
NDVI	Normalized Difference Vegetation Index	30m, Monthly
NDWI	Normalized Difference Water Index	30m, Monthly
NDBI	Normalized Difference Built-up Index	30m, Monthly
Albedo	Surface reflectance ratio	30m, Monthly
RGB	Red, Green, Blue channels	30m, Monthly
DEM	Digital Elevation Model (NASADEM)	30m, Static
LCZ	Local Climate Zones (CONUS-wide)	100m, Annual

Dataset Statistics

Attribute	Value
Spatial Resolution	30 meters
Tile Size	128 x 128 pixels (3.84 km²)
Input Channels	9 (DEM, LST, R, G, B, NDVI, NDWI, NDBI, Albedo)
Output Channel	1 (LST)
Input Sequence Length	12 months
Output Sequence Length	1 month
Total Rasters	~1.4 million tiles
Total Training Sequences	~200,000
Storage Size	~150 GB

Temporal Splits

Split	Years	Duration
Training	2013-2021	9 years
Validation	2022-2023	2 years
Testing	2024-2025	2 years

Data Distribution

LCZ class distribution across 124 cities and LST temperature distribution across ~22 billion pixels. Mean temperature: 85.4°F, dominated by LCZ Class 6 (Open low-rise, 57.9%).

LST temperature distribution by Local Climate Zone class. Built-up areas (LCZ 8: Large low-rise, 94°F mean) and bare surfaces (LCZ 16: Bare soil, 104°F mean) show the highest temperatures, while water (LCZ 17, 65°F mean) and dense trees (LCZ 11, 67°F mean) are coolest.

Build Your Own (STAC Scraper)

Alternatively, collect raw Landsat imagery using the STAC API via Microsoft Planetary Computer with stac_scrapper.ipynb. This allows you to customize the geographic extent, extend the temporal range, or modify processing parameters.

pip install pystac-client planetary-computer odc-stac rioxarray

---

Installation

Requirements

• Python 3.8+
• PyTorch 2.0+
• CUDA 11.8+ (for GPU training)

Setup

git clone https://github.com/JesseGuerrero/US-UrbanLST.git
cd US-UrbanLST

python -m venv venv
source venv/bin/activate  # Linux/Mac
# or: venv\Scripts\activate  # Windows

pip install torch pytorch-lightning rasterio numpy pandas wandb tqdm scikit-learn matplotlib earthformer

---

Quick Start

1. Download Dataset

huggingface-cli download JesseGuerrero/US-UrbanLST --local-dir ./Data/ML --repo-type dataset

2. Setup Data Cache

python setup_data.py \
    --dataset_root "./Data/ML" \
    --cluster "all" \
    --input_length 12 \
    --output_length 1 \
    --train_years 2013 2014 2015 2016 2017 2018 2019 2020 2021 \
    --val_years 2022 2023 \
    --test_years 2024 2025

3. Train Model

python train_with_cache.py \
    --dataset_root "./Data/ML" \
    --cluster "all" \
    --input_length 12 \
    --output_length 1 \
    --model_size "earthnet" \
    --batch_size 32 \
    --max_epochs 200 \
    --learning_rate 0.0001 \
    --train_years 2013 2014 2015 2016 2017 2018 2019 2020 2021 \
    --val_years 2022 2023 \
    --test_years 2024 2025 \
    --gpus 2

---

Model Architectures

Model	Description
earthnet	Earthformer (CuboidTransformer) - recommended
lstm	CNN+LSTM baseline (DMVSTNet)
tiny / small / medium / large	Transformer variants at different scales

Training with Land Cover Clusters

# Cluster 1: Dense urban (LCZ 1-3)
python train_with_cache.py --cluster "1" --model_size "earthnet"

# Cluster 2: Suburban (LCZ 4-6)
python train_with_cache.py --cluster "2" --model_size "earthnet"

# All data
python train_with_cache.py --cluster "all" --model_size "earthnet"

Channel Ablation

# RGB only (remove spectral indices, DEM, and historical LST)
python setup_data.py --remove_channels DEM ndvi ndwi ndbi albedo LST
python train_with_cache.py --remove_channels DEM ndvi ndwi ndbi albedo LST

---

Benchmark Results

Performance by Land Cover Cluster (Test RMSE °F)

Land Clusters	LCZ	Description	Earthformer	CNN+LSTM
All	0-17	All Lands	13.93	18.75
1	1-3	Dense buildings, sparse green space	22.82	15.52
2	4-6	Less dense buildings, more green space	15.14	19.48
3	7-10	Remaining urban classes	12.59	14.42
4	11-17	Natural landscapes	11.84	17.61

Ablation Study (Test RMSE °F)

Feature Set	Earthformer	CNN+LSTM
All Features	13.93	18.75
LST Only	14.67	19.96
Spectral (no LST)	13.89	22.44
RGB Only	15.62	26.68

---

Web Application

The interactive visualization demonstrates LST predictions for downtown San Antonio, TX using a 3D ArcGIS map with:

• Monthly LST overlay with time slider (2025-2026 predictions)
• Pre-rendered PNG map tiles at zoom levels 14-17
• Chat interface powered by an LLM for planning guidance

Run locally:

cd web-app
conda env create -f environment.yml
conda activate earthformer
python _inference_city.py   # generate tiles and temperature grids
python -m http.server 3000  # preview at http://localhost:3000

---

Repository Structure

US-UrbanLST/
├── dataset.py              # PyTorch dataset with interpolation and caching
├── model.py                # Earthformer and CNN+LSTM (DMVSTNet) models
├── setup_data.py           # Data preprocessing and sequence cache builder
├── train_with_cache.py     # Training script with WandB logging
├── stac_scrapper.ipynb     # Landsat STAC data collection
├── preprocess.ipynb        # Data preprocessing notebook
├── main.ipynb              # Main experiment notebook
├── CONUS_LCZ.tif           # CONUS-wide Local Climate Zone raster
├── scripts/                # Shell scripts for training and ablation
├── test/                   # Test and evaluation scripts
├── analysis/               # Dataset analysis and visualization
│   └── out/                # Distribution plots and statistics
├── web-app/                # Interactive LST visualization app
│   ├── index.html          # ArcGIS 3D map with chat interface
│   ├── _inference_city.py  # City-wide inference pipeline
│   ├── model.py            # Model loading for inference
│   └── model_baseline.ckpt # Pre-trained Earthformer checkpoint (Git LFS)
└── Data/
    └── City_Shapes/        # City boundary shapefiles

---

Citation

If you use this dataset or code in your research, please cite:

@inproceedings{guerrero2026urbanlst,
  title={US-UrbanLST: A Large-Scale High-Resolution Dataset to Advance Research on Land Surface Temperature},
  author={Guerrero, Jesus and Najafirad, Leon and Corley, Isaac and Tabar, Maryam and Rad, Paul},
  booktitle={Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  year={2026},
  organization={ACM}
}

Acknowledgements

• Secure AI Autonomy Laboratory (SAAL) at the University of Texas at San Antonio
• UTSA High Performance Computing Platform
• Data: Landsat 8/9 (USGS/NASA), CONUS LCZ, Urban Footprints (Esri)

License

MIT