DaYu: Data-Driven Model for Geostationary Satellite Observed Cloud Images Forecasting

TL;DR

This paper proposed an advanced data-driven thermal infrared cloud images forecasting model, DaYu, which is specifically designed for geostationary satellite observations, and based on a large-scale transformer architecture, which enables it to capture fine-grained cloud structures and learn fast-changing spatio-temporal evolution features effectively.

Abstract

In the past few years, Artificial Intelligence (AI)-based weather forecasting methods have widely demonstrated strong competitiveness among the weather forecasting systems. However, these methods are insufficient for high-spatial-resolution short-term nowcasting within 6 hours, which is crucial for warning short-duration, mesoscale and small-scale weather events. Geostationary satellite remote sensing provides detailed, high spatio-temporal and all-day observations, which can address the above limitations of existing methods. Therefore, this paper proposed an advanced data-driven thermal infrared cloud images forecasting model, "DaYu." Unlike existing data-driven weather forecasting models, DaYu is specifically designed for geostationary satellite observations, with a temporal resolution of 0.5 hours and a spatial resolution of ${0.05}^\circ$ $\times$ ${0.05}^\circ$. DaYu is based on a large-scale transformer architecture, which enables it to capture fine-grained cloud structures and learn fast-changing spatio-temporal evolution features effectively. Moreover, its attention mechanism design achieves a balance in computational complexity, making it practical for applications. DaYu not only achieves accurate forecasts up to 3 hours with a correlation coefficient higher than 0.9, 6 hours higher than 0.8, and 12 hours higher than 0.7, but also detects short-duration, mesoscale, and small-scale weather events with enhanced detail, effectively addressing the shortcomings of existing methods in providing detailed short-term nowcasting within 6 hours. Furthermore, DaYu has significant potential in short-term climate disaster prevention and mitigation.

Figures (6)

• Figure 1: Overview of the DaYu Architecture. DaYu cascades the parameters of Stage 1 and Stage 2 models. Taking continuous cloud images from two moments $x^{-1}$, $x^0$ as inputs, DaYu Stage 1 autoregressively forecasts 12 cloud images for the 0-6 hour period, and then uses $x^{11}$, $x^{12}$ as inputs. Stage 2 forecasts 12 cloud images for the 6-12 hour period as the same pattern. Red-bordered boxes indicate the initial inputs for the stage models. Spatio-Temporal Encoder extracts feature embeddings from inputs. Transformer layers are then used to learn the global relationships of high-level abstract features. Finally, Spatio-Temporal Decoder generates the predicted cloud image for the next moment. In this figure, $i$ ranges from 0 to 23.
• Figure 2: PCC and RMSE of DaYu forecasting the brightness temperature states in 2023. In each subplot, the x-axis represents the lead time, with intervals of 0.5 hours over a 12-hour lead time. The y-axis represents the PCC and RMSE as defined in the Eq. \ref{['pcc_eq']} and Eq. \ref{['rmse_eq']}.
• Figure 3: Visualization of forecast results for the observational wavelength at 6.9$\mu m$. The visualization of cloud image forecasts for 3 hours (first row), 6 hours (second row) and 12 hours (third row). For each row, the results shown are: DaYu forecast (left), AHI observation (middle), and the difference between DaYu forecast and AHI observation (right). For all cases, the initial time is 12:00 UTC on October 12, 2023.
• Figure 4: Visualization of forecast results for the observational wavelength at 11.2$\mu m$. The visualization of cloud image forecasts for 3 hours (first row), 6 hours (second row) and 12 hours (third row). For each row, the results shown are: DaYu forecast (left), AHI observation (middle), and the difference between DaYu forecast and AHI observation (right). For all cases, the initial time is 12:00 UTC on October 12, 2023.
• Figure 5: Visualization of forecast results for Super Typhoon "Bolaven". The first two rows show the cloud images from DaYu forecasts at different lead times, while the last two rows display the AHI observed cloud images. For all cases, the initial time is 12:00 UTC on October 12, 2023.