Yuanhong Liao

Yuanhong Liao

Auto-generated profile, based on publicly available information

Yuanhong Liao

Department of Earth System Science, Institute for Global Change Studies, Ministry of Education Ecological Field Station for East Asian Migratory Birds, Tsinghua University, Beijing 100084, China

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 February 2026 in EGUsphere
Verified
Preprint
Open access
PDF file
SCShuang Chen
Shuang Chen
JWJie Wang
Jie Wang
SYShuai Yuan
Shuai Yuan
JLJiayang Li
Jiayang Li
YXYu Xia
Yu Xia
Peng GongPeng Gong
Peng Gong
The rapid evolution of satellite-borne Earth Observation (EO) systems has fundamentally revolutionized terrestrial monitoring, yielding comprehensive petabyte-scale archives. However, the immense computational resources and storage volumes required for global-scale analysis often preclude widespread use by many research teams, hindering broader scientific adoption and the execution of planetary-scale studies. To address these barriers, we present the Embedded Seamless Data (ESD), an ultra-lightweight, 30-m global Earth embedding database spanning the 25-year period from 2000 to 2024. By transforming high-dimensional, multi-sensor observations from the Landsat series (5, 7, 8, and 9) and MODIS Terra into information-dense, quantized latent vectors, ESD distils essential geophysical and semantic features into a unified latent space. Utilizing the ESDNet architecture and Finite Scalar Quantization (FSQ), the dataset achieves a transformative ~340-fold reduction in data volume compared to raw daily archives. This compression allows the entire global land surface for a single year to be encapsulated within approximately 2.4 TB, enabling decadal-scale global analysis on standard local workstations. Rigorous validation demonstrates that ESD maintains high reconstructive fidelity to the original reflectance values across the spectral dimension, achieving a Mean Absolute Error (MAE) of 0.0130 (averaged over six spectral bands, including Blue, Green, Red, NIR, SWIR1, and SWIR2), a Root Mean Square Error (RMSE) of 0.0179, and a Correlation Coefficient (CC) of 0.8543. By condensing the annual phenological cycle into 12 temporal latent steps, the embeddings provide inherent denoising effects and a semantically organized latent space that outperforms raw reflectance data in downstream land-cover classification tasks, achieving a comparable and even higher overall accuracy of 79.74 % than the 76.92 % obtained using raw sensor fusion data on globally distributed land cover sample sets. With robust few-shot learning capabilities and longitudinal consistency across 25 years, the ESD product provides a versatile foundation for democratizing planetary-scale Earth system research and advancing next-generation geospatial artificial intelligence.

16 February 2026 in ISPRS Journal of Photogrammetry and Remote Sensing
Verified
Journal Article
Closed access
YLYuanhong Liao
Yuanhong Liao
Yuqi BaiYuqi Bai
Yuqi Bai
YYYanping You
Yanping You
JWJie Wang
Jie Wang
SCShuang Chen
Shuang Chen
PGPeng Gong
Peng Gong

24 July 2026 in Earth System Science Data
Verified
Journal Article
Open access
PDF file
SCShuang Chen
Shuang Chen
JWJie Wang
Jie Wang
SYShuai Yuan
Shuai Yuan
JLJiayang Li
Jiayang Li
YXYu Xia
Yu Xia
Peng GongPeng Gong
Peng Gong
The rapid evolution of satellite-borne Earth Observation (EO) systems has fundamentally revolutionized terrestrial monitoring, yielding comprehensive petabyte-scale archives. However, the immense computational resources and storage volumes required for global-scale analysis often preclude widespread use by many research teams, hindering broader scientific adoption and the execution of planetary-scale studies. To address these barriers, we present the Embedded Seamless Data (ESD), an ultra-lightweight, 30 m global Earth embedding database spanning the 25-year period from 2000 to 2024. By transforming high-dimensional, multi-sensor observations from the Landsat series (5, 7, 8, and 9) and MODIS Terra into information-dense, quantized latent vectors, ESD distils essential geophysical and semantic features into a unified latent space. Utilizing the ESDNet architecture and Finite Scalar Quantization (FSQ), the dataset achieves a transformative ∼ 340-fold reduction in data volume compared to raw daily archives. This compression allows the entire global land surface for a single year to be encapsulated within approximately 2.4 TB, enabling decadal-scale global analysis on standard local workstations. Rigorous validation demonstrates that ESD maintains high reconstructive fidelity to the original reflectance values across the spectral dimension, achieving a Mean Absolute Error (MAE) of 0.0130 (averaged over six spectral bands, including Blue, Green, Red, NIR, SWIR1, and SWIR2), a Root Mean Square Error (RMSE) of 0.0179, and a Correlation Coefficient (CC) of 0.8543. By condensing the annual phenological cycle into 12 temporal latent steps, the embeddings provide inherent denoising effects and a semantically organized latent space that outperforms raw reflectance data in downstream land-cover classification tasks, achieving a comparable and even higher overall accuracy of 79.74 % than the 76.92 % obtained using raw sensor fusion data on globally distributed land cover sample sets. With robust few-shot learning capabilities and longitudinal consistency across 25 years, the ESD product provides a versatile foundation for democratizing planetary-scale Earth system research and advancing next-generation geospatial artificial intelligence. The ESD dataset is freely available at https://doi.org/10.12436/iEarth.0000.20251229.000064.v1 (Chen, 2025).

27 February 2026 in EGUsphere
Verified
Preprint
Open access
PDF file
YLYuanhong Liao
Yuanhong Liao
SCShuang Chen
Shuang Chen
YBYuqi Bai
Yuqi Bai
JWJie Wang
Jie Wang
Peng GongPeng Gong
Peng Gong
Accurate quantification of the structural evolution of global agricultural systems is critical for assessing food security and monitoring planetary boundaries. However, current agricultural monitoring is hindered by a baseline that reflects the continuous, annual nature of agricultural management and distinguishes active cultivation cycles from permanent structural changes. Existing products typically rely on fragmented snapshots, multi-year aggregation, or generalized and inconsistent definitions (introducing systematic bias and noise into trend analysis). To bridge this gap, we generated the first Global 30-m Annual Cropland Extent Dynamics (GACED30) dataset (2000–2024). Our continuous mapping framework integrates the gap-free SDC30 to capture the intra-annual phenological transitions distinctive of active cultivation, employs a spectral-semantic sample alignment strategy to resolve the inconsistencies across different cropland sample sets, and applies a rule-based processing to mitigate the spectral ambiguity between active fallow and natural bareland. Furthermore, we derived grid-based Structural Evolution Indicators via continuous statistical trend analysis, utilizing the full 25-year time-series density to rigorously quantify expansion and abandonment trajectories. Comprehensive assessment demonstrates the reliability of GACED30, which achieves a high overall accuracy of 96.5 % and significantly outperforms existing global products (e.g., GLAD Cropland, ESA World Cover and GLC_FCS30D) in capturing temporal stability. Crucially, GACED30 exhibits strong agreement with FAO national statistics, achieving a high correlation in cropland area (R2:0.95) and an 81.1 % consistency in production trends. Based on this consistent baseline, we analyzed the structural evolution of global agriculture, estimating the 2024 cropland area at 1488.5 Mha with a net change rate of 1.2 Mha/year. Our trend analysis reveals a distinct global divergence: structural expansion is heavily concentrated in the Global South (e.g., Africa and South America), driven by commodity frontiers, whereas the Global North is characterized by widespread stability or policy-driven contraction. GACED30 thus provides a reliable evidence base for monitoring the changing footprint of global agriculture. The dataset is publicly available at https://doi.org/10.5281/zenodo.18199675 (Chen et al., 2025).

11 August 2026 in Journal of Environmental Management
Verified
Journal Article
Closed access
ZWZimeng Wang
Zimeng Wang
WLWenrui Li
Wenrui Li
LWLingyu Wang
Lingyu Wang
YLYuanhong Liao
Yuanhong Liao
SLShuhan Lou
Shuhan Lou
YBYuqi Bai
Yuqi Bai

15 April 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
ZWZimeng Wang
Zimeng Wang
YBYuqi Bai
Yuqi Bai
WLWenrui Li
Wenrui Li
LWLingyu Wang
Lingyu Wang
YLYuanhong Liao
Yuanhong Liao
MZMingrun Zang
Mingrun Zang
The wildland–urban interface (WUI) is a critical zone where structures intermingle with combustible vegetation, frequently resulting in catastrophic wildfire losses. Traditional WUI mapping methods based on static empirical thresholds ignore the fine-scale heterogeneity of individual building wildfire exposure, leading to limited accuracy in high-precision risk management. Here, we propose a novel method to delineate the WUI based on building-level wildfire exposure. We developed two random forest models to quantify both the exposure status and exposure level of structures. Key variables and thresholds were identified using Shapley Additive Explanations (SHAP) and Partial Dependence Plots (PDP) to characterize high exposed buildings and surrounding vegetation. Validation against historical wildfire records in Northern Alberta demonstrated that, compared with the standard U.S. Federal Register method, our approach achieved significantly higher spatial consistency. Specifically, the proposed method achieved a 2.14 time higher burned land overlap, 1.69-fold increase in the number of captured ignition points, and a 31.3% higher proportion of exposed buildings compared with the U.S. Federal Register method. These findings indicate that identifying WUI areas based on exposure potential of buildings offers a more accurate tool for identifying areas with high exposure potential, thereby facilitating precision wildfire risk management and strategic policy development.