Semi-Arid Egypt Climate Dataset: A Gap-Filled Daily Meteorological Dataset (1990–2020)
Description
This repository contains a high-quality, gap-filled daily meteorological time series dataset for 13 meteorological stations across Egypt, covering the strict period from 1 January 1990 to 31 December 2020 (11,323 days per station). The dataset was developed as part of the study: "A novel framework for gap-filling meteorological time series data" (Agricultural Water Management, 2026).
The framework combines bias-corrected ERA5-Land reanalysis data, ground-based meteorological observations, and machine learning ensemble models to reconstruct missing records.
📂 REPOSITORY CONTENTS
1. SemiArid-Egypt-Climate-Dataset-1990-2020.zip
The final, gap-filled dataset. Contains 13 CSV files (one for each station). Each file contains exactly 11,323 rows with the following columns:
• Date: Strict ISO 8601 format (YYYY-MM-DD)
• Tmax: Maximum air temperature (°C)
• Tmin: Minimum air temperature (°C)
• Tmean: Mean air temperature (°C)
• Tdew: Dew point temperature (°C)
• WindSpeed: Wind speed (m/s)
• Pressure: Sea level pressure (kPa)
• is_imputed: Boolean flag (True = ML-reconstructed gap fill; False = original ground observation)
2. Ground-Based Observations.zip
The raw historical daily observations obtained from the Global Weather Historical Database for the 13 stations.
3. ERA5-Land Reanalysis Data.zip
The raw satellite-derived reanalysis variables extracted from the ECMWF ERA5-Land dataset using Google Earth Engine.
4. Corrected_ERA5-Land Reanalysis Data.zip
The ERA5-Land variables after applying statistical bias correction (Linear Scaling) against ground observations.
5. semiarid-egypt-climate-dataset-original-code.ipynb
The complete Python code used for data preprocessing, bias correction, and the machine learning ensemble gap-filling framework (Released under the MIT License).
6. Supplementary Tables (CSV)
• Table_S1_Baseline_Comparison.csv
• Table_S2_QC_Removal_Counts.csv
• Table_S3_Bias_Correction_All_Stations.csv
• Table_S4_Top_Three_Machine_Learning_Models_Selected.csv
• Table_S5_Uncertainty_All_Stations.csv
• Table_S6_Chronological_Validation_Dabaa.csv
📍 STUDY AREA
The dataset includes 13 meteorological stations distributed across diverse climatic regions of Egypt (coastal, inland, and desert environments): Aswan, Asyut, Baharia, Dabaa, Dakhla, El-Arish, El-Tor, Farafra, Ismailia, Kharga, Luxor, Minya, and Siwa.
⚙️ DATA PROCESSING WORKFLOW
Step 1 — Preprocessing: Missing value detection, outlier removal (IQR-based), physical consistency checks (e.g., removing temperatures < -5°C or > 55°C, wind > 25 m/s, pressure outside 85–110 kPa).
Step 2 — Bias Correction: ERA5-Land variables were bias-corrected using Linear Scaling (LS), Variance Scaling (VS), Empirical Quantile Mapping (EQM), and Theoretical Quantile Mapping (TQM). LS achieved the best performance.
Step 3 — Machine Learning Gap Filling: Seven ML models (RF, ET, GB, XGBoost, LR, Ridge, Lasso) were evaluated. A Voting Regressor Ensemble of the top 3 models per variable/station was used for final reconstruction.
📊 MODEL PERFORMANCE
• Temperature variables (Tmax, Tmin, Tmean): R² = 0.89 – 0.99
• Dew point (Tdew): R² = 0.85 – 0.96
• Sea level pressure: R² > 0.98
• Wind speed: R² = 0.47 – 0.79
📜 CITATION
If you use this dataset, please cite the associated paper:
El-Agha, D. E., Elgamal, M. H., Gado, T. A., & Atiea, M. A. (2026). A novel framework for gap-filling meteorological time series data. Agricultural Water Management, 331, 110454. https://doi.org/10.1016/j.agwat.2026.110454
Files
Corrected_ERA5-Land Reanalysis Data.zip
Files
(58.4 MB)
| Name | Size | |
|---|---|---|
|
md5:4881b820efca0c54702f58dbae6da868
|
17.3 MB | Preview Download |
|
md5:2992429480820dae229b937b4d4e221e
|
29.2 MB | Preview Download |
|
md5:e4d1eeb8269fc7c0769ed3b456e6dd97
|
8.9 MB | Preview Download |
|
md5:7978c7492568839a102140966b908bf8
|
2.9 MB | Preview Download |
|
md5:1b87054a6e16f4e6d8243bf80a78d68b
|
14.5 kB | Preview Download |
|
md5:dbb91b8ff2fd6923c4f7a855f9b2f1a9
|
1.2 kB | Preview Download |
|
md5:1c38aa469ff622287777338e44ccd6fb
|
375 Bytes | Preview Download |
|
md5:04cf5a8d9b16e2dcbd3650681163002a
|
8.1 kB | Preview Download |
|
md5:a90d429c00672cb93a3ef722ff86d3b2
|
4.0 kB | Preview Download |
|
md5:50e1bb5bdb1a4e65e34a258c0a3211fc
|
2.0 kB | Preview Download |
|
md5:4db0be8c83cc388c8701a439a1ccdefe
|
108 Bytes | Preview Download |
Additional details
References
- El-Agha, D. E., Elgamal, M. H., Gado, T. A., & Atiea, M. A. (2026). A novel framework for gap-filling meteorological time series data. Agricultural Water Management, 331, 110454. https://doi.org/10.1016/j.agwat.2026.110454