# Enhanced forecasting of bird nocturnal migration intensity in relation to previous days and synoptic weather patterns

## TL;DR

This paper improves bird migration forecasts by considering past and large-scale weather patterns, helping predict migration intensity more accurately.

## Contribution

The study introduces a novel method that integrates local, synoptic, and previous weather conditions to enhance bird migration forecasts.

## Key findings

- The model explains 1.3 and 2.25 times more variance in migration intensity than local weather-based models in spring and autumn.
- Previous weather conditions are crucial for predicting high-intensity migration events, as they reflect bird accumulation due to unfavorable conditions.
- The framework is interpretable and transferable, offering insights into ecological processes and aiding conservation efforts.

## Abstract

Operational bird migration forecast models have recently offered promising perspectives for mitigating the impacts of human activities on avifauna. These models improve on simple phenological expectations by harnessing the intricate relationship between bird movements and weather conditions to forecast migration fluxes days in advance. However, state-of-the-art models face limitations as bird fluxes are often simply modelled as a response to local and instantaneous weather without accounting for previous and synoptic weather patterns. This study focuses on enhancing bird migration forecasts by evaluating the contributions of weather dynamics at various spatial and temporal scales. We use bird vertical density data from 9 French weather radars over 6 years and employ gradient-boosted regression trees for predictions. Dimension reduction tools are used to describe local and continental-scale weather conditions from the previous three days. We also explore the contributions of the different meteorological metrics considered using explainable regression trees tools. Our model improved phenology models by explaining about 1.3 and 2.25 times more additional variance than approaches based on local and instantaneous weather conditions in spring and autumn, respectively. Local and instantaneous weather metrics contributed the most, but they mainly helped identifying nights with low migration. In contrast, weather metrics for previous 3 days were crucial to forecast highest intensity migration events, as they enabled to account for bird accumulation in relation to unfavorable weather locally and remotely. This study enhanced forecast accuracy and contributed to a deeper understanding of the factors influencing bird migration. It enabled the identification local and synoptic weather patterns related to important migration events without a priori knowledge. It is therefore easy to interpret, easy to transfer to other ecological systems, and promising for the accurate forecast of migration peaks. Forecasted peaks can guide conservation efforts, for example by dimming lights for birds at night or by shutting down wind turbines.

The online version contains supplementary material available at 10.1007/s00484-025-02917-4.

Nocturnally migrating birds undergo large distances and their motivation to migrate is linked to weather conditions. We here relate the local, synoptic (i.e. continental scale), and preceding weather conditions to forecast peaks and lows in migration intensity. This methodological framework enhanced accuracy of our forecasts and deepened our understanding of the underlying ecological process. It is particularly valuable for the design of relevant bird migration forecast tools to mitigate the impacts of human activities on avifauna.

The online version contains supplementary material available at 10.1007/s00484-025-02917-4.