E21 Quantitative Precipitation Nowcasting using PySteps and S-Band dual-pol Weather Radar for Squall Lines in Southern Brazil

Monday, 29 January 2024
Hall E (The Baltimore Convention Center)
Leonardo Calvetti, Universidade Federal de Pelotas, Pelotas, RS, Brazil; Universidade Federal de Pelotas, Pelotas, RS, Brazil; and K. Andrzejewski, C. Beneti, T. M. Buriol, and E. Brignol

Quantitative Precipitation Forecasting (QPF) continues to be a significant challenge in hydrometeorological research. The numerical weather prediction (NWP) still needs to improve accuracy for the first three hours of QPF compared with nowcast methods. For small and urban watersheds, the quick response of discharges needs fast and accurate QPF for short-range periods. Including operational environments, nowcasting methods can be a good alternative to complex numerical models, giving higher skill of QPF. Here, the authors present and compare the results of methods included in the PySteps package (pysteps.github.io) applied in data of dual-pol S-Band Weather radar of SIMEPAR Meteorological Service installed in Cascavel City, west of Parana.

The radar data from December 13, 2022, and February 26, 2023, were used for two cases of the Quasi-Linear Convective System (QLCS) that propagates toward the western region of the Paraná state where more than 20 mm/h was registered, causing flash-flooding in some cities. In order to use the algorithm, it was necessary to transform the reflectivity factor into a precipitation intensity, for which the classic Marshall and Palmer Z-R relationship was chosen (a=200 and b=1.6). It has been used the PySTEPS package (pysteps.github.io) to address the QPF, including the Lucas-Kanade (LK) optical flow method, a local tracking approach, Dynamic and Adaptive Radar Tracking of Storms (DARTS) method, which use a spectral approach for the optical flow based on the discrete Fourier transform (DFT) and a Semi-Lagrangian extrapolation that is based on motion field estimation, which here was performed using both the LK and DARTS approaches (PULKKINEN et al., 2019).

For the motion field prediction, in the December 13, 2022 case, both methods satisfactorily presented similar results, while for the February 26, 2023 case, there was a slight difference between the observed and predicted with the DARTS method doing a more realistic motion field.

Regarding extrapolating the rainfall field for February 26, 2023, the LK method presents a different result than the one observed, elongating the storm area and slightly underestimating precipitation intensity (Figure below). On the other hand, the DARTS approach presents a great result, very similar to what was observed.

In the extrapolation for the December 13, 2022 case, the LK method slightly reduces the storm area and again underestimates the precipitation intensity, while the DARTS method presents satisfactory results with a subtle difference in relation to what is observed; however, the disadvantage of this method is that a more extended sequence of radar fields is necessary to estimate the movement, while in the LK approach, the local features are tracked in a sequence of two or more radar images.

More testing is still needed with the algorithms provided by PySTEPS to find the best nowcasting method considering the peculiarities of Brazilian meteorological systems. In this way, it is expected to contribute to short-term operational forecasting and increase alerts and monitoring of storms that may cause damage to the local population.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner