PeakWeather: MeteoSwiss Weather Station Measurements for Spatiotemporal Deep Learning

TL;DR

This work introduces PeakWeather, a high-quality dataset of surface weather observations collected every 10 minutes over more than 8 years from the ground stations of the Federal Office of Meteorology and Climatology MeteoSwiss's measurement network, which supports a broad spectrum of spatiotemporal tasks.

Abstract

Accurate weather forecasts are essential for supporting a wide range of activities and decision-making processes, as well as mitigating the impacts of adverse weather events. While traditional numerical weather prediction (NWP) remains the cornerstone of operational forecasting, machine learning is emerging as a powerful alternative for fast, flexible, and scalable predictions. We introduce PeakWeather, a high-quality dataset of surface weather observations collected every 10 minutes over more than 8 years from the ground stations of the Federal Office of Meteorology and Climatology MeteoSwiss's measurement network. The dataset includes a diverse set of meteorological variables from 302 station locations distributed across Switzerland's complex topography and is complemented with topographical indices derived from digital height models for context. Ensemble forecasts from the currently operational high-resolution NWP model are provided as a baseline forecast against which to evaluate new approaches. The dataset's richness supports a broad spectrum of spatiotemporal tasks, including time series forecasting at various scales, graph structure learning, imputation, and virtual sensing. As such, PeakWeather serves as a real-world benchmark to advance both foundational machine learning research, meteorology, and sensor-based applications.

Figures (9)

• Figure 1: PeakWeather station locations across the Swiss territory.
• Figure 2: Performance comparison of deep learning models with ICON and persistence model baselines for wind forecasting. Results are averaged over five runs with different random seeds, except for ICON and Chronos-2. Shaded areas indicate $\pm$3 standard deviations. The persistence models, ICON and Chronos-2 require no training. The variability of the persistence models and ICON arises solely from Monte Carlo sampling during evaluation and the ensemble forecasts, respectively. Chronos-2 provides quantile predictions of the velocity, which are used to compute the speed and direction.
• Figure 3: Visualizations of the placement of PeakWeather stations. Panel \ref{['fig:stations']}) distribution of the stations across the Swiss territory; circles denote meteorological stations, while triangles denote rain gauges. Panel \ref{['fig:stations_graph']}) a graph obtained using the geographical distance to compute a similarity heuristic for the meteorological stations. Panel \ref{['fig:height_cov_latlon']}) vertical coverage of the stations.
• Figure 4: Visualization of the topographic descriptors.
• Figure 5: Number of stations with valid measurements of the meteorological variables included in PeakWeather.