As climate change leads to more frequent and intense extreme precipitation events, accurately predicting rainfall during the flood season has become increasingly critical.
A recent study has employed machine learning (ML) algorithms to address the nonlinear challenges faced by traditional models in predicting flood season rainfall, resulting in significant improvements in accuracy. The findings were published in Advances in Atmospheric Sciences .
Current predictions for flood season rainfall rely largely on outputs from climate system numerical models, which often contain systematic biases. To correct these outputs and reduce errors, researchers traditionally combine historical observational data with statistical methods.
This approach, known as the dynamical-statistical method, has its limitations. Prediction errors from numerical models tend to grow nonlinearly over time, and traditional correction methods, which primarily rely on linear approaches, struggle to effectively address these errors.
Recognizing ML's strength in managing nonlinear relationships, the study applied the LightGBM algorithm to enhance the dynamical-statistical correction method. In trials conducted from 2019 to 2022, the predictions improved significantly, with the prediction score (PS) increasing from 68.6 to 74—an improvement of 7.87%. This represents a 6.63% enhancement over traditional dynamical-statistical methods, substantially boosting the accuracy of flood season rainfall predictions.
Many data-driven ML methods used for climate prediction often lack sufficient physical interpretability. To address this issue, the researchers carefully selected meteorological factors with clear physical connections to rainfall and integrated them into the climate system model. The team also quantified the contribution of each forecasting factor, offering a clearer understanding of the physical significance of the predictors used.
The study emphasizes a key point: relying solely on either physical models or ML models to improve predictions of flood season rainfall has inherent limitations. This study explores a climate prediction method that effectively integrates ML with physical models.
The rapidly evolving fields of artificial intelligence and big data offer new opportunities to optimize and refine model outputs, addressing nonlinear and complex challenges that traditional dynamic-statistical methods cannot resolve.
This study proposed a feasible approach to developing the traditional dynamic-statistical method into a dynamic-ML method.
Despite the progress, challenges remain.
"Our next steps will focus on extracting pre-existing and real-time signals from research on flood season precipitation formation mechanisms to develop dynamic-ML method with stronger physical interpretability," said Dr. YU Haipeng, the corresponding author from Northwest Institute of Eco-Environment and Resources of the Chinese Academy of Sciences.
This research marks a significant step forward in precipitation prediction and offers valuable insights for developing future meteorological methods that integrate artificial intelligence and big data.
"Our ultimate goal is to create an efficient, stable, and interpretable system combining climate system models and ML techniques for predicting flood season rainfall, helping to mitigate the impacts of extreme precipitation and related disasters," said Dr. YU.
As technology continues to advance, integrating physical mechanisms with ML-based prediction methods holds great potential for addressing the challenges posed by climate change.
