Jaeho Choi

Jaeho Choi, Yura Kim, Kwang-Ho Kim et al.

The precipitation nowcasting methods have been elaborated over the centuries because rain has a crucial impact on human life. Not only quantitative precipitation forecast (QPF) models and convolutional long short-term memory (ConvLSTM), but also various sophisticated methods such as the latest MetNet-2 are emerging. In this paper, we propose a paired complementary temporal cycle-consistent adversarial networks (PCT-CycleGAN) for radar-based precipitation nowcasting, inspired by cycle-consistent adversarial networks (CycleGAN), which shows strong performance in image-to-image translation. PCT-CycleGAN generates temporal causality using two generator networks with forward and backward temporal dynamics in paired complementary cycles. Each generator network learns a huge number of one-to-one mappings about time-dependent radar-based precipitation data to approximate a mapping function representing the temporal dynamics in each direction. To create robust temporal causality between paired complementary cycles, novel connection loss is proposed. And torrential loss to cover exceptional heavy rain events is also proposed. The generator network learning forward temporal dynamics in PCT-CycleGAN generates radar-based precipitation data 10 minutes from the current time. Also, it provides a reliable prediction of up to 2 hours with iterative forecasting. The superiority of PCT-CycleGAN is demonstrated through qualitative and quantitative comparisons with several previous methods.

Jaeho Choi, Hyeri Kim, Kwang-Ho Kim et al.

Accurate precipitation forecasting is becoming increasingly important in the context of climate change. In response, machine learning-based approaches have recently gained attention as an emerging alternative to traditional methods such as numerical weather prediction and climate models. Nonetheless, many recent approaches still rely on off-the-shelf loss functions, and even the more advanced ones merely involve optimization processes based on the critical success index (CSI). The problem, however, is that CSI may become ineffective during extended dry periods when precipitation remains below the threshold, rendering it less than ideal as a criterion for optimization. To address this limitation, we introduce a simple penalty expression and reinterpret it as a quadratic unconstrained binary optimization (QUBO) formulation. Ultimately, the resulting QUBO formulation is relaxed into a differentiable advanced torrential (AT) loss function through an approximation process. The proposed AT loss demonstrates its superiority through the Lipschitz constant, forecast performance evaluations, consistency experiments, and ablation studies with the operational model.