Multi-decadal high-resolution dynamic downscaling over the CORDEX Southeast Asia domain were performed using the regional model WRF forced by the ERA-40 reanalysis dataset as well as the present climate and future projections (RCP4.5, RCP8.5) of the Met Office Hadley Centre's HadGEM2-AO. The downscaled low-level circulation and precipitation in the hindcast experiment, historical run, near future changes, and far future changes were analyzed and compared to observations or corresponding large-scale forcing.

In the hindcast experiment, the added values of dynamic downscaling over land exceeded those over ocean owing to the precisely described underlying surface over land considered in the WRF model. The added values over land increased when the small islands were neglected. The poor performance of the WRF model over small islands degraded the general skill over the entire land region. Although the hindcast experiment showed biases in low-level circulation fields, which can probably be attributed to the air-sea interaction being neglected, the WRF model has reliable skills for the big lands when it was driven by the “perfect” lateral boundary forcing from the reanalysis dataset.

The historical downscaling enhanced the low-level circulation biases in the driving fields from HadGEM2-AO because the regional model was unable to reduce the uncertainties in the lateral boundary forcing. Therefore, the biases in the historical downscaling were larger than those in the hindcast downscaling forced by the reanalysis dataset. However, the precipitation biases over land in the HadGEM2-AO model were reduced in the historical downscaling largely because the dominant forcing from the underlying surface over land was represented better in the fine-resolution model. The historical downscaling showed poor added value over the ocean.

Over all, the uncertainty in the RCP4.5 projection for JJA was the highest among all the cases. The mean precipitation change in the HadGEM2-AO model was more likely enhanced by the WRF model in the far future than in the near future. The uncertainty in precipitation variability projection is larger than that in the mean precipitation projection.

1. INTRODUCTION

2. MODEL AND DATA

3. EXPERIMENTAL DESIGN

4. RESULTS 40

4.1 Hindcast experiment

4.2 Historical experiment

4.3 Future projection

5. CONCLUSION