Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Location of the weather radar (red point), positioned in the eastern Amazon, and some municipalities within its coverage area.

More »

Fig 1 Expand

Fig 2.

Accumulated precipitation in the municipalities within the weather radar’s coverage area by month (a), and the mean daily precipitation of the municipalities (b), both for the year 2021.

More »

Fig 2 Expand

Fig 3.

Cumulative daily precipitation on December 31, 2021, obtained from weather radar data.

The Z-R relationship was used to estimate precipitation in mm.

More »

Fig 3 Expand

Fig 4.

Overview of the multiple-horizon approach for precipitation nowcasting and an example of horizon selection.

More »

Fig 4 Expand

Fig 5.

Overview of the U-Net architecture.

U-Net is a U-shaped convolutional neural network with an encoder-decoder structure capable of capturing spatio-temporal features.

More »

Fig 5 Expand

Fig 6.

Performance evaluation of multi-horizons through the relationship between RMSE and MAE scores.

More »

Fig 6 Expand

Fig 7.

Comparison of multi-horizon performance using categorical scores with the 20 dBZ threshold.

More »

Fig 7 Expand

Fig 8.

Continuous and categorical scores for each 5-minute interval within the 60-minute forecast horizon.

The evaluation includes the best (120-minute past horizon) and worst (60-minute past horizon) models for this forecast horizon, as well as the baseline models Extrapolation and STEPS, both with a 120-minute past horizon.

More »

Fig 8 Expand

Fig 9.

Comparison of precipitation forecasts over various time intervals for models U-Net-120-60, U-Net-60-60, and reference models Extrapolation and STEPS, both with 120-60 horizons.

More »

Fig 9 Expand