Abstract
This study evaluates satellite precipitation products (CHIRPS, PERSIANN-CDR, TAMSAT, ARC2, and ERA5) against in-situ rainfall data from Djibouti-Airport station (1983–2017) using multiple statistical metrics: Correlation Coefficient, RMSE, Percentage Bias, Nash-Sutcliffe Efficiency (NSE), and Relative Standard Deviation. ERA5 demonstrated the best overall performance with the highest correlation (0.75) and NSE (0.56), while TAMSAT and ARC2 showed poorer accuracy (R² ≈ 0.09). Seasonal analysis revealed ERA5 performed exceptionally during JF (R²=0.81, COR=0.90), while PERSIANN-CDR excelled during MAM and OND.
Principal Component Analysis (PCA) of 14 rainfall stations identified two distinct precipitation regimes: eastern coastal stations (influenced by the Red Sea/Gulf of Aden) and western/interior stations (altitude-dependent). The first two principal components explained 70.3% of total variance, with PC1 capturing interannual variability (92.1%) and PC2 capturing annual cycle variability (40.75%). Four imputation methods (PPCA, AMELIA II, MICE, missMDA) were applied to reconstruct missing data from 1946–2017.
Trend analysis indicated significant decreasing rainfall trends in MAM (Airport: -3.15 mm/season, p=0.009; CHIRPS: -0.71 mm/season, p=0.010), with mixed seasonal patterns across products. These findings establish ERA5 as the most reliable satellite product for Djibouti's rainfall estimation and confirm the country's spatially heterogeneous precipitation regime shaped by both maritime influence and orographic factors.
Keywords ERA5, CHIRPS, Nash-Sutcliffe, PCA, Djibouti.
Abdi-Basid Adan – abdi-basid@outlook.com – 2018-2019. https://www.impactio.com/publication-attachments/2139/954225551.pdf
© 2026 The Abdi-Basid Courses Institute (tABCi). All rights reserved.
馃實 The EcoClimate Hub | Focus: Climate Science & Environment.
Contact : Abdi-Basid ADAN [abdi-basid@outlook.com]
Click below to view the full Course Catalogue :
*The Abdi-Basid Courses Institute*
© 2017-2026 The Abdi-Basid Courses Institute (tABCi). All rights reserved.
