TL;DR
This study leverages 33 years of daily satellite rainfall data (ARC2) to analyze changes in the Sahel's rainfall regime, finding a general consistency with ground-based measurements but highlighting discrepancies in low-intensity rainfall events. The research reveals distinct trends between the western/central and eastern Sahel, with the former experiencing increased rainfall and longer wet seasons, while the latter shows more complex and sometimes negative trends.
What Problem Does This Paper Address?
The Sahel region, crucial for rain-fed agriculture and livestock, is characterized by high rainfall variability, making it vulnerable to droughts and impacting food security. Understanding these rainfall patterns and their changes is vital for adaptation strategies. However, the sparse and decreasing network of rain gauges in the region limits detailed analysis. Satellite-based rainfall estimates offer a promising alternative due to their wider coverage and temporal resolution, but their reliability and ability to capture the nuances of the rainfall regime, especially compared to ground data, require thorough investigation. This paper aims to bridge this gap by evaluating the ARC2 satellite dataset for analyzing spatio-temporal changes in the Sahelian rainfall regime and comparing it with ground-based data.
Methods
The study employs a multi-faceted approach to analyze rainfall data:
- Rainfall Regime Variable Definition: Key variables describing the rainfall regime were defined, including onset and cessation of the rainy season, length of the wet season, seasonal rainfall amount, number of rainy days, intensity, frequency of rainfall events, and dry spell characteristics. These definitions were adapted from existing literature to be suitable for both satellite and gauge data.
- Spatio-temporal Correlation Analysis: Linear regression and Pearson's correlation coefficients were used to assess the agreement between ARC2 satellite data and rain gauge data, both spatially (pixel-to-station) and temporally (time series analysis).
- Trend Analysis: Sen's slope estimator and the Mann-Kendall test were applied to identify significant trends in rainfall variables over time, accounting for serial correlation. This was done for various sub-periods to capture different temporal scales.
- Wavelet Analysis: Continuous wavelet power spectrum analysis was used to investigate inter-annual and multi-year variability in rainfall patterns.
- Standardized Rainfall Index: A standardized rainfall index was applied to quantify seasonal rainfall anomalies in relation to the long-term climatology, adapted for spatial variability.
Datasets
- ARC2 (African Rainfall Climatology Version 2): A daily gridded rainfall dataset with 0.1°× 0.1° spatial resolution, available from 1983 to the present. This dataset is a revision of ARC1, with recalibration for the 1983-2005 period.
- MSWEP (Multi-Source Weighted-Ensemble Precipitation): A global rainfall dataset with 3-hour temporal resolution and 0.25° spatial resolution, covering the period 1979-2015. This dataset merges various precipitation data sources.
- Rain Gauge Data: Derived from the Global Historical Climatology Network (GHCN-Daily). Approximately 30 stations with at least 80% data availability between 1983-2015 were selected for comparison, distributed across the western and central Sahel.
Key Insights and Ideas
- Satellite vs. Gauge Discrepancies: While ARC2 generally aligns well with gauge data for many rainfall variables, a significant bias emerges in the estimation of low-intensity rainfall events (<10 mm day⁻¹). This suggests that satellite algorithms may struggle to accurately capture the frequency and intensity of these smaller events, impacting derived variables like rainfall intensity and dry spell characteristics.
- Spatial vs. Temporal Reliability: The study indicates that ARC2 might capture inter-annual fluctuations of dry spells better than their spatial distribution, suggesting that temporal consistency is a stronger suit for certain variables.
- Regional Divergence in Trends: A striking finding is the distinct spatio-temporal trends observed between the western/central Sahel and the eastern Sahel. The western/central region shows a clear trend towards increased rainfall, longer wet seasons, and more high-intensity events, while the eastern Sahel exhibits more complex and sometimes negative trends, particularly for medium and high rainfall events.
- Data Quality Artifacts: The research points to potential data quality issues in the ARC2 product, particularly around rainfall stations, possibly due to calibration artifacts related to ground observations. This can lead to spurious patterns in trends for variables sensitive to daily rainfall event counts.
Key Results
- Correlation: Linear correlation coefficients (r) between ARC2 and gauge data for most rainfall variables ranged between 0.29 and 0.77, indicating a fair to good agreement. Dry spell variables showed weaker correlations (0.06-0.22).
- Trends:
- Western/Central Sahel: Positive trends were observed for rainy season length, seasonal rainfall amount, number of rainy days, and high rainfall events (>20 mm day⁻¹). This was associated with earlier onset and later cessation of the rainy season.
- Eastern Sahel: Trends were less significant and sometimes negative, especially for medium and high rainfall events, and seasonal rainfall and length of season in some areas.
- Seasonal Distribution: A shift towards later monsoon rainfall was observed in the central Sahel.
- Inter-annual Variability: Wavelet analysis revealed that inter-annual variability is dominated by shorter year-to-year fluctuations, with strong fluctuations observed in onset/cessation dates, seasonal rainfall amount, and number of rainy days, particularly in the last 15 years.
Limitations and Future Work
- Low-Intensity Rainfall Bias: The study acknowledges that the bias in estimating low-intensity rainfall events between satellite and gauge data makes direct comparisons of daily event-based variables less appropriate.
- Calibration Artifacts: The presence of artifacts around rain gauge stations in ARC2 data, especially during the dry season, suggests that improvements in gauge calibration could enhance the product's reliability.
- Future Work: The authors suggest that despite these limitations, temporal trends derived from satellite and gauge data are still valid for comparison. They also propose that the insights gained from disentangling seasonal rainfall variables could be applied to other dryland regions and linked with satellite-based vegetation productivity data to better understand changing vegetation patterns. The consistency of ARC2 and MSWEP data for most variables indicates their potential for rainfall regime analysis in other dryland areas.
Summary updated