[AHW37-P05] Deep learning for seasonal to sub-seasonal rainfall prediction in Chaophraya river basin
キーワード：ディープラーニング、季節予報、チャオプラヤ川
In Thailand 1-2 month ahead rainfall prediction is essential for flood mitigation of Chao Phraya river basin. Especially it is needed to decide dam operation for both purpose of preventing drought and flood. Such a seasonal to sub-seasonal scale prediction still remains as difficult problem since numerical prediction contains large uncertainty in this time scale. On the other hand, observed data has been collected for years and statistical analysis skill appeared as the possible solution. One of the most powerful statistical modeling methodology is deep leaning in terms of detecting hidden relationship beneath the complex combination of atmosphere, ocean and land conditions. In the field of rainfall prediction, however the number of observed data is still limited and not enough to train the deep learning model. It requires hundreds of data which enables us to use many variables. To overcome lacks of data, this research applied some data expanding methodology and verified its influence on modeling.