A Quantitative Precipitation Estimation Method Based on 3D Radar Reflectivity Inputs

Yanqin Wen,Jun Zhang,Di Wang,Xianming Peng,Ping Wang

DOI: https://doi.org/10.3390/sym16050555

2024-01-01

Symmetry

Abstract:Quantitative precipitation estimation (QPE) by radar observation data is a crucial aspect of meteorological forecasting operations. Accurate QPE plays a significant role in mitigating the impact of severe convective weather. Traditional QPE methods mainly employ an exponential Z–R relationship to map the radar reflectivity to precipitation intensity on a point-to-point basis. However, this isolated point-to-point transformation lacks an effective representation of convective systems. Deep learning-based methods can learn the evolution patterns of convective systems from rich historical data. However, current models often rely on 2 km-height CAPPI images, which struggle to capture the complex vertical motions within convective systems. To address this, we propose a novel QPE model: combining the classic extrapolation model ConvLSTM with Unet for an encoder-decoder module assembly. Meanwhile, we utilize three-dimensional radar echo images as inputs and introduce the convolutional block attention module (CBAM) to guide the model to focus on individual cells most likely to trigger intense precipitation, which is symmetrically built on both channel and spatial attention modules. We also employ asymmetry in training using weighted mean squared error to make the model concentrate more on heavy precipitation events which are prone to severe disasters. We conduct experiments using radar data from North China and Eastern China. For precipitation above 1 mm, the proposed model achieves 0.6769 and 0.7910 for CSI and HSS, respectively. The results indicate that compared to other methods, our model significantly enhances precipitation prediction accuracy, with a more pronounced improvement in forecasting accuracy for heavy precipitation events.