Yunue et al. (2026) Multiscale Meteorological Drought Spatial Reconstruction in North-Central Urban Core of Mexico City: An Explainable Deep Learning Approach

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water
• Year: 2026
• Date: 2026-05-12
• Authors: Garza-Pimentel Yunue, González-Olvera Marcos Angel, Santos-Reyes Jaime Reynaldo
• DOI: 10.3390/w18101165

Research Groups

Not specified in the provided text.

Short Summary

The study developed an explainable deep learning framework using LSTM networks to spatially reconstruct three drought indices (SPI, SPEI, and RDI) across various temporal scales in Mexico City.

Objective

• To develop an operational spatial monitoring system capable of reconstructing meteorological drought anomalies across multiple temporal scales to support water stress assessment in an urban environment.

Study Configuration

• Spatial Scale: Intra-urban (Mexico City), utilizing microsites with inter-station distances ranging from 1.8 to 6.7 km.
• Temporal Scale: Accumulation scales of 3, 6, 12, 18, and 24 months.

Methodology and Data

• Models used: Long Short-Term Memory (LSTM) networks and SHapley Additive exPlanations (SHAP) for feature importance.
• Data sources: Meteorological station data (microsites); SPI calculated via WMO Gamma distribution fitting, while SPEI and RDI used empirical monthly standardized anomalies.

Main Results

• The LSTM model achieved high predictive performance for long-term drought scales due to its capacity for deep sequential memory.
• Short-term reconstructions were more affected by the inherent noise of urban convective precipitation.
• The framework demonstrated reliable intra-urban spatial generalization using a leave-one-microsite-out validation strategy.

Contributions

• Introduces an explainable deep learning approach for the spatial reconstruction of drought indices in non-stationary urban climates, providing mathematical transparency through SHAP.

Funding

Not specified in the provided text.

Citation

@article{Yunue2026Multiscale,
  author = {Yunue, Garza-Pimentel and Angel, González-Olvera Marcos and Reynaldo, Santos-Reyes Jaime},
  title = {Multiscale Meteorological Drought Spatial Reconstruction in North-Central Urban Core of Mexico City: An Explainable Deep Learning Approach},
  journal = {Water},
  year = {2026},
  doi = {10.3390/w18101165},
  url = {https://doi.org/10.3390/w18101165}
}