Most nocturnal bird migration takes place beyond direct visual observation, but moving flocks leave measurable signatures in weather-radar records. BirdNet-AI was developed to turn these records into spatially continuous estimates of migration over the United States. NEXRAD scans were organized as multi-channel images and passed to a U-Net that separated likely bird echoes from precipitation and background returns. On the benchmark images, the model reached an F1-score of 85.1%, a small improvement over MistNet. The resulting station-level estimates were then analyzed with a graph neural network. Its prediction errors were 0.138 for MAE and 0.189 for RMSE, with wind speed and temperature improving performance over migration history alone. The mapped results were strongest along central and eastern flyways and showed the expected seasonal reversal between spring and fall. These findings suggest that radar-image segmentation and network-based forecasting can be combined for broad-scale migration monitoring, although the estimates remain dependent on radar coverage and labeling quality.

Zihan Zhao. BirdNet-AI: Spatio-Temporal Modeling of Nocturnal Bird Migration Using Weather Radar and Graph Neural Networks. Academic Journal of Computing & Information Science (2026), Vol. 9, Issue 7: 28-33. https://doi.org/10.25236/AJCIS.2026.090704.

[1] “The Migration Ecology of Birds,” ScienceDirect. https://www.sciencedirect.com/book/978 0125173674/the-migration-ecology-of-birds

[2] A. M. Dokter et al., “Seasonal abundance and survival of North America’s migratory avifauna determined by weather radar,” Nature Ecology & Evolution, vol. 2, no. 10, pp. 1603-1609, Sep. 2018, doi: https://doi.org/10.1038/s41559-018-0666-4.

[3] B. M. V. Doren and K. G. Horton, “A continental system for forecasting bird migration,” Science, vol. 361, no. 6407, pp. 1115-1118, Sep. 2018, doi: https://doi.org/10.1126/science.aat7526.

[4] National Centers for Environmental Information, “Next Generation Weather Radar,” National Centers for Environmental Information (NCEI), Sep. 22, 2020. https://www.ncei.noaa.gov/ products/radar/next-generation-weather-radar

[5] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” arXiv.org, May 18, 2015. https://arxiv.org/abs/1505.04597

[6] T. N. Kipf and M. Welling, “Semi-Supervised Classification with Graph Convolutional Networks,” arXiv:1609.02907 [cs, stat], Feb. 2017, Available: https://arxiv.org/abs/1609.02907

[7] “NOAA National Climatic Data Center (NCDC) | CAKE: Climate Adaptation Knowledge Exchange,” Cakex.org, 2026. https://www.cakex.org/community/directory/organizations/noaa-national-climatic-data-center-ncdc (accessed Mar. 05, 2026).

[8] T. Lin et al., “MistNet: Measuring historical bird migration in the US using archived weather radar data and convolutional neural networks,” Methods in Ecology and Evolution, vol. 10, no. 11, pp. 1908-1922, Aug. 2019, doi: https://doi.org/10.1111/2041-210x.13280.

[9] F. A. La Sorte, K. G. Horton, C. Nilsson, and A. M. Dokter, “Projected changes in wind assistance under climate change for nocturnally migrating bird populations,” Global Change Biology, Dec. 2018, doi: https://doi.org/10.1111/gcb.14531.

[10] K. G. Horton et al., “Phenology of nocturnal avian migration has shifted at the continental scale,” Nature Climate Change, vol. 10, no. 1, pp. 63-68, Jan. 2020, doi: https://doi.org/10.1038/s41558-019-0648-9.