Institution profile

China Meteorological Administration

Academic institutionasia · cn
Official website
Research library7linked papers
Opportunities0open roles
Selected work

Representative Papers

Robustness of Deep Learning Models for PV Power Forecasting under NWP Forecast Errors: A Spatiotemporal and Physically Interpretable Analysis

Jul 14, 2026

This study addresses the limitations of existing evaluations that overlook the temporal correlation, state dependence, and physical coupling inherent in numerical weather prediction (NWP) errors. To this end, it proposes a simulation-based, physics-constrained robustness evaluation framework that incorporates clear-sky modulated heteroscedastic perturbations and an Erbs-based solar irradiance reconstruction mechanism to generate dynamic, heteroscedastic, and radiometrically consistent NWP perturbations. By isolating input uncertainty propagation through synthetic photovoltaic power generation, the framework systematically benchmarks models including PatchTST, GRU, N-HiTS, and LightGBM. Results demonstrate that sequence-based models outperform tabular models under moderate to high perturbation levels, exhibiting superior noise filtering and temporal resilience. Interpretability analysis further reveals a shift in feature reliance—from future forecasts toward historical observations and physical priors—as perturbations intensify. Combined with Pareto analysis, these insights offer principled guidance for model selection across trade-offs among accuracy, robustness, and latency.

0 citationsRead paper

Enhancing the Forecasting Capability of Multi-Model Blending Algorithms for Extreme Precipitation via Joint Use of Station and Gridded Observations

Jul 06, 2026

This study addresses the persistent challenges in numerical weather prediction (NWP) models—namely, the underestimation of extreme precipitation intensity and spatial displacement errors—which are often exacerbated by conventional multi-model blending that diffuses rainfall areas and smooths peak values. To overcome these limitations, the authors propose a novel two-stage fusion framework based on U-Net architecture: first performing probabilistic classification of precipitation occurrence, followed by quantitative reconstruction of rainfall amounts. The approach integrates six leading NWP models and innovatively embeds observations from 2,411 Chinese national meteorological stations directly into the loss function, enabling joint supervision over both grid points and station locations to simultaneously constrain spatial structure and peak intensity. Evaluated on independent data from the 2025 flood season, the method improves the Threat Score (TS) for heavy rain (≥50 mm) by 38.4% over the best individual model and achieves a TS exceeding 0.1 for extreme precipitation (≥100 mm)—a notable milestone—demonstrating substantially enhanced forecast accuracy and practical utility.

0 citationsRead paper

SIMBA: ABidirectional Retrieval Forward Simulation Framework for Modeling FY-4A GIIRS Hyperspectral Infrared Radiances Toward NWP Applications

Jun 18, 2026

This study addresses the limitation of existing deep learning approaches, which typically focus solely on the one-way inversion from radiance observations to atmospheric profiles while neglecting forward radiative transfer simulation and the physical consistency between observations and atmospheric states. To overcome this, the authors propose a unified bidirectional framework that jointly performs atmospheric profile inversion and radiative transfer simulation, incorporating a cycle-consistency constraint to enforce physical coupling. A novel bidirectional Mamba state-space module is designed to capture long-range dependencies across pressure levels. Trained on collocated FY-4A GIIRS observations and ERA5 reanalysis data, the model significantly outperforms state-of-the-art deep learning baselines in both temperature and humidity profile retrieval and shortwave–longwave radiative flux reconstruction, demonstrating the efficacy of the bidirectional architecture and cycle-consistency mechanism.

0 citationsRead paper
Recent publications

Latest Papers

Robustness of Deep Learning Models for PV Power Forecasting under NWP Forecast Errors: A Spatiotemporal and Physically Interpretable Analysis

Jul 14, 2026

This study addresses the limitations of existing evaluations that overlook the temporal correlation, state dependence, and physical coupling inherent in numerical weather prediction (NWP) errors. To this end, it proposes a simulation-based, physics-constrained robustness evaluation framework that incorporates clear-sky modulated heteroscedastic perturbations and an Erbs-based solar irradiance reconstruction mechanism to generate dynamic, heteroscedastic, and radiometrically consistent NWP perturbations. By isolating input uncertainty propagation through synthetic photovoltaic power generation, the framework systematically benchmarks models including PatchTST, GRU, N-HiTS, and LightGBM. Results demonstrate that sequence-based models outperform tabular models under moderate to high perturbation levels, exhibiting superior noise filtering and temporal resilience. Interpretability analysis further reveals a shift in feature reliance—from future forecasts toward historical observations and physical priors—as perturbations intensify. Combined with Pareto analysis, these insights offer principled guidance for model selection across trade-offs among accuracy, robustness, and latency.

0 citationsRead paper

Enhancing the Forecasting Capability of Multi-Model Blending Algorithms for Extreme Precipitation via Joint Use of Station and Gridded Observations

Jul 06, 2026

This study addresses the persistent challenges in numerical weather prediction (NWP) models—namely, the underestimation of extreme precipitation intensity and spatial displacement errors—which are often exacerbated by conventional multi-model blending that diffuses rainfall areas and smooths peak values. To overcome these limitations, the authors propose a novel two-stage fusion framework based on U-Net architecture: first performing probabilistic classification of precipitation occurrence, followed by quantitative reconstruction of rainfall amounts. The approach integrates six leading NWP models and innovatively embeds observations from 2,411 Chinese national meteorological stations directly into the loss function, enabling joint supervision over both grid points and station locations to simultaneously constrain spatial structure and peak intensity. Evaluated on independent data from the 2025 flood season, the method improves the Threat Score (TS) for heavy rain (≥50 mm) by 38.4% over the best individual model and achieves a TS exceeding 0.1 for extreme precipitation (≥100 mm)—a notable milestone—demonstrating substantially enhanced forecast accuracy and practical utility.

0 citationsRead paper

SIMBA: ABidirectional Retrieval Forward Simulation Framework for Modeling FY-4A GIIRS Hyperspectral Infrared Radiances Toward NWP Applications

Jun 18, 2026

This study addresses the limitation of existing deep learning approaches, which typically focus solely on the one-way inversion from radiance observations to atmospheric profiles while neglecting forward radiative transfer simulation and the physical consistency between observations and atmospheric states. To overcome this, the authors propose a unified bidirectional framework that jointly performs atmospheric profile inversion and radiative transfer simulation, incorporating a cycle-consistency constraint to enforce physical coupling. A novel bidirectional Mamba state-space module is designed to capture long-range dependencies across pressure levels. Trained on collocated FY-4A GIIRS observations and ERA5 reanalysis data, the model significantly outperforms state-of-the-art deep learning baselines in both temperature and humidity profile retrieval and shortwave–longwave radiative flux reconstruction, demonstrating the efficacy of the bidirectional architecture and cycle-consistency mechanism.

0 citationsRead paper