Abstract:
General Circulation Models (GCMs) play a vital role in forecasting future climate changes;
however, their rainfall outputs frequently exhibit significant biases due to coarse resolution and
oversimplified processes. Consequently, it is essential to apply bias correction methods to this data
before utilizing it for climate change initiatives. This study introduces a context-aware framework
that integrates multiple statistical bias correction methods specifically tailored to the climatic zones
and rainfall characteristics of Sri Lanka. Rainfall data from 68 monitoring stations across the island
were utilized alongside outputs from the CNRM-CM6-1 GCM. The study was conducted under
moderate and high climate change scenarios, Shared Socioeconomic Pathways (SSP) (SSP2-4.5
and SSP5-8.5) from the CMIP6 program. Bias correction techniques, including Linear Scaling (LS),
Empirical Quantile Mapping, Delta Change, Power Transformation, and Local Intensity Scaling,
were employed to evaluate the performance of each method. The framework dynamically identifies
the most effective method based on metrics such as RMSE, R², and Taylor diagrams. The results
indicate that LS significantly enhances rainfall accuracy, particularly for monthly precipitation,
outperforming other methods under SSP2-4.5. However, LS has limitations, particularly regarding
its inability to account for topographic variation and spatial biases, highlighting the necessity for
complementary approaches. Spatial analysis revealed regional disparities, especially in northern
areas, attributed to local climate dynamics. Furthermore, the corrected data demonstrate improved
accuracy and effectively capture seasonal trends, including intensified extremes projected under
SSP5-8.5. This research underscores the importance of adaptive, regionally tailored correction
strategies to enhance the reliability of rainfall data in climate impact assessments and water resource
management.
