Future Climate of Colombo Downscaled with SDSM-Neural Network
by Singay Dorji, Srikantha Herath and Binaya Kumar Mishra
- Paper focuses on downscaling global climate model (GCM) data for climate impact studies.
- Compares artificial neural network (ANN) with regression-based statistical downscaling model (SDSM) for downscaling precipitation in Colombo, Sri Lanka.
- ANN outperforms SDSM in downscaling precipitation based on model biases and root mean squared error (RMSE).
- Proposes combining SDSM with neural networks for improved downscaling skills.
- Utilizes CanESM2, NCEP reanalysis data, and APHRODITE observational data.
- Considers RCP 8.5 scenario for future projections: temperature increase of 2.83 °C (SDSM) and 3.03 °C (TDNN), and rainfall increase of 33% (SDSM) and 63% (TDNN) by the 2080s.
