Update README.md

README.md

@@ -4,26 +4,39 @@ license: cc-by-4.0
 
 # PeakWeather 🌦️🌤️⛈️
 
-PeakWeather is a dataset of surface weather measurements for spatiotemporal deep learning 
-with data collected every 10 minutes over the course of more than 8 years, from January 2017 until March 2025.
-It includes a diverse set of meteorological variables obtained from of SwissMetNet, 
-the network of automatic weather stations operated by
-the Swiss Federal Office of Meteorology and Climatology (MeteoSwiss). The network counts over 300 station locations
-distributed across Switzerland's complex terrain.
-To provide geographical context, it is enriched with topographical features derived from a digital elevation model.
-Additionally, ensemble forecasts from the currently operational high-resolution NWP model ICON-CH1-EPS are included
-as a baseline for evaluating new forecasting approaches.
+PeakWeather is a high-resolution, quality-controlled dataset of surface weather 
+observations from 302 stations across Switzerland, collected by the Swiss 
+Federal Office for Meteorology and Climatology (MeteoSwiss) over more than 
+eight years. Designed to support research in spatiotemporal modeling, it 
+includes dense time series data, topographical features, and numerical weather 
+forecasts from an operational model at MeteoSwiss.
 
 <div style="display: flex; justify-content: center; gap: 20px;">
   <img src="https://raw.githubusercontent.com/MeteoSwiss/PeakWeather/main/figures/stations.png" alt="Description" width="450"/>
 </div>
 
+## What's Inside
+
+- 302 weather stations across complex Swiss topography;
+- 10-minute sampling from January 1st, 2017 to March 31th, 2025 (433,728 time steps);
+- 8 meteorological variables: air temperature, relative humidity, air pressure, precipitation amount, wind speed/direction, wind gust, sunshine duration;
+- Topographic features from a 50m-resolution digital height model;
+- Numerical weather forecasts (ICON-CH1-EPS) for benchmarking;
+- Metadata for stations and variables, as well as sensor relocations information;
+- Missing values and resampling utilities.
+
 ## How to use PeakWeather
 
-To load the data, refer to the dataset implementation and instructions provided [on GitHub](https://github.com/MeteoSwiss/PeakWeather) 📦.
+Use the companion [GitHub library](https://github.com/MeteoSwiss/PeakWeather) 📦 to:
+
+- Download and cache the data locally;
+- Load time series and static data into ready-to-use formats;
+- Extract sliding windows of a size W and horizon H from the time series;
+- Handle missing data via built-in imputation tools and (un)availability masks;
+- Downsample or resample as needed for your modeling task.
 
-The dataset is distributed as a lightweight, framework-agnostic Python library that exposes the data in a way that can
-easily integrate with modern deep learning frameworks while keeping dependencies to a minimum.
+For more details refer to the dataset implementation and instructions provided
+[on GitHub](https://github.com/MeteoSwiss/PeakWeather).
 
 **License**: CC-BY-4.0