The Infoplaza Model: A Unique Approach

The meteorologists at Infoplaza have access to numerous weather models, but their proprietary Infoplaza model is the most important. The Infoplaza model is also available on Imweather.com. It uniquely differs from well-known models such as the European ECMWF, the American GFS, the German ICON, and the Dutch KNMI Harmonie model. While these traditional models simulate the atmosphere through physical and mathematical calculations, the Infoplaza model takes a different approach.

How Does the Infoplaza Weather Model Work?

Conventional weather models calculate forecasts by processing observations within a grid, using varying resolutions and time steps for physical computations. The Infoplaza model approaches this differently: instead of relying solely on atmospheric data, it integrates multiple weather models. These models are 'blended' using smart algorithms, resulting in Infoplaza's distinctive output.

You might think: “If Model A predicts 18°C and Model B predicts 14°C, the Infoplaza model will simply average to 16°C.” However, it’s far more sophisticated than that. The Infoplaza model’s blending process goes beyond calculating averages.

Step 1: Continuous Verification and Model Weighting

The process begins with verification: all weather models are continuously monitored at thousands of locations. Through this ongoing verification, Infoplaza gains insight into each model's accuracy, based on location and weather element. For instance, one model might excel in temperature prediction but underperform for wind forecasts. This data informs the model weighting process, where models are given more or less influence depending on their performance.

Step 2: Data Formatting and Grid Filtering  

Raw data from various weather sources arrives in diverse formats. The first step is to standardize this data using a grid filter. At this stage, adjustments are made for elevation differences and the specific characteristics of land and sea locations.

Step 3: Model Mixing and the Smart Grid

For as many as 20,000 locations, the Infoplaza model generates a weather forecast. For each location, it determines which model carries the most weight and which carries less. Beyond this, the model calculates forecasts for an additional 400,000 points within a 'smart grid' that spans much of the North Sea and Europe. Unlike the fixed grids used in conventional models, the smart grid only calculates relevant points. Data density is higher in coastal or mountainous areas, while it’s lower over open seas. This intelligent system ensures efficient computational power and enhanced precision at critical locations.

Step 4: Data Calibration and Assimilation

After the initial model blending, a base forecast is ready. This forecast undergoes further refinement through a data calibration module. For 5,000 of the 20,000 locations, historical observations are used to make corrections. For example, a location consistently predicted half a degree too cold can be adjusted. For the remaining 15,000 locations, data assimilation is applied: calibrated data points are used to fine-tune neighboring uncalibrated points, leading to an even more precise model.

Step 5: Final Finetuning

After all calculations, the Infoplaza model is nearly complete. For the Netherlands, the first few hours of the forecast are fine-tuned with the most recent observations and, when necessary, adjusted manually by a meteorologist. The entire process takes about 45 to 60 minutes, ensuring the most accurate model is available for Infoplaza’s meteorologists and clients.

From Infoplaza's I'm Weather

Innovation and the Future: The Cloud and Artificial Intelligence

Noud Brasjen has been the driving force behind Infoplaza’s model blending for over a decade. During this time, both computational power and the volume of available data have significantly increased. A major milestone is planned for 2025: migrating to a cloud-based weather computer. This move will provide greater capacity, speed, and efficiency, ultimately enabling a global smart grid to make the best model available everywhere.

Additionally, the WXtech team, which Noud is part of, is exploring how AI can enhance model blending. Initially, the focus is on the calibration process. According to Noud, there is potential to shift from static to dynamic calibration, which would represent a significant leap in forecast accuracy.

With the Infoplaza model, Infoplaza envisions a future where the best weather forecasts are available worldwide—always tailored to specific locations and weather conditions.

Try I'm Weather for free

I'm Weather is the weather tool for the true weather enthusiast! You can try it yourself. And when you become a Imweather Pro user you get access to more features in our Labs environment like Quickview, Particles and Probabilities.