Historical rainfall reconstruction in the period of 1900–2010 for extreme climate event analysis (case study in Java Island, Indonesia)

Abstract

Extreme weather events and climate anomalies are increasing in both frequency and intensity due to climate change, posing greater risks to both human and natural systems. Analyzing hazards like hydrometeorological events is therefore crucial for managing these risks, particularly by utilizing long-term precipitation data spanning up to 100 years. This study aims to develop a method for reconstructing historical rainfall data for the 1900–2010 period in Indonesia, using the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-20C dataset. The primary methodology involves the retrospective use of the constructed analogues statistical downscaling (CA-SD) method. This approach employs ERA-20C's zonal (U) and meridional (V) wind parameters at 850 hPa, which are converted into scalar field variables as predictors for rainfall reconstruction. The results show that the reconstructed climatological mean is overestimated compared to Multi-Source Weighted-Ensemble Precipitation (MSWEP) data but underestimated compared to station observations. A more comprehensive evaluation, using several deterministic and probabilistic metrics at various rainfall thresholds, was conducted. The ensemble mean correlations improved with temporal aggregation. According to probabilistic metrics, the reliability of estimated 5 mm and 20 mm rainfall events was better than that of rain days (>0.5 mm), while the reliability for 50 mm and 100 mm events was lower. This study produced reconstructed rainfall data for the 1900–2010 period, which could be valuable for climatological and hydrological research. Specifically, the severe drought in Java between 1960 and 1970 can be examined in greater detail; this is important due to its potential reoccurrence in the future.