Certain wind-pollinating plants produce massive amounts of pollen as part of their reproductive process. These grains, which usually look like a fine, yellow powder, are usually 15-90 micrometers in size, but some can be as small as 2 micrometers.
Pollen tracking is very difficult, so it’s no wonder we frequently get asked how we provide such granular pollen monitoring with daily up-to-date and forecast information.
In this guide, we explain the science behind our pollen forecasting model and how we help millions around the world manage their exposure to pollen on a daily basis.
Multiple studies suggest a strong link between human-induced climate change and worsening pollen seasons across the globe. This means individuals with strong sensitivities, and even those with only moderate reactions may begin experiencing more extreme symptoms.
Experts suggest annual pollen seasons are now producing more grains and impacting more people:
Longer Seasons: Researchers estimate climate change has increased the length of North American pollen seasons by roughly 20 days over the past two decades.
Higher Concentrations: Researchers estimate climate change has increased North American pollen concentrations by 21% over the past two decades. Climate change could also increase Northwestern Europe’s grass pollen season severity by up to 60% - based on a simulation that assumes double the global average atmospheric CO2 concentrations over two decades.
More Ragweed: Researchers also project ragweed pollen allergy presence in Europe will more than double over the next two to four decades (2041-2060). By 2100, high allergy risk areas in Europe for common ragweed species are projected to increase in size by up to 100%.
Greater Distribution: As the climate warms, new habitats develop for allergenic plant species, often closer to populated areas. The consequence of this is that more people become exposed to new forms of pollen, triggering potentially more severe reactions.
Traditional approaches to pollen analysis involve rudimentary techniques that offer only limited reporting options. Many sources are typically unable to provide personalized, granular, and actionable information in a timely fashion, due to a number of limitations:
In order to deliver accurate and timely pollen forecasts, BreezoMeter moves beyond monitoring station information and regional models by leveraging sophisticated AI-driven data analysis and prediction methods.
In addition to station information, we layer this information with data about pollen emission schedules, vegetation land covers, and climate and weather conditions - all of which serve to inform our pollen forecasting models.
Using AI and machine learning algorithms, our model can forecast pollen dispersion and report hyperlocal threat levels:
We combine government land cover databases and satellite imagery to calculate the percentage of pollen-producing plants at a 100-meter grid resolution. We then reference that data against local weather and climate to calculate pollen production schedules and total annual production for local plant types.
After establishing the beginning and end of pollen seasons and total annual productions, we forecast local pollen emissions for 3 consecutive days. We continue referencing dynamic weather changes, as well as data from pollen monitoring stations across the world to adjust daily results.
The calculated emissions data is then fed into our AI-driven pollen dispersion model to estimate grains per cubic meter. These numbers are translated into BreezoMeter’s Pollen Index to provide a 5-level threat severity assessment based on scientific literature and continuous research.
Hyperlocal and highly accurate 3-day pollen forecasts at a 1km resolution. Businesses can easily integrate data via our Pollen API to provide users with daily risk alerts, personalized actionable recommendations, and even color-coded pollen heatmaps visualizing location-based threat levels.
Most common pollen indexes vary in how they calculate pollen concentrations. Most indexes only categorize general pollen threat levels based on pollen counts without indicating risk for specific plant sensitivities.
Individuals often experience different allergic reactions to different pollen types, and lack of personalization creates some serious limitations:
Different ‘High’ and ‘Low’ Standards
A ‘low’ count in one pollen index can register as ‘medium’ or even ‘high’ in others, due to the use of different scales. Allergy sufferers that need to take medication in advance can become vulnerable to high pollen exposure considered ‘low’ by some indexes.
Many pollen indexes base their levels on average measurements in a wide geographical area, not daily variations based on granular location. In addition, they often don’t categorize measured pollen counts into plant species or types, which pose different risks based on individual sensitivities.
Common pollen indexes often don’t account for the impact on respiratory health, and reported levels don’t reflect the threat of causing or aggravating different allergy symptoms. Individuals with pollen sensitivities can’t refer to them to estimate personal risk.
To achieve a high level of accuracy, BreezoMeter adopts a multi-data layer approach and uses a geographical grid to model daily variations in pollen levels based on the user’s location. But what measures do we take to ensure the information we provide is reliable?
Validating Collected Information
Utilizing multiple data sources enables BreezoMeter to maintain better information authentication. For example, we compare satellite imagery with land cover databases, thus enabling us to verify the relevant local flora population.
Comparison Studies with Pollen Monitoring Stations
In 2019, we compared BreezoMeter’s pollen forecast information to the Italian POLLnet monitoring stations and found our model agreed with the stations’ threat risk assessment over 95% of the time for a period of 6 months (April 1st, 2019 - Sept 29th, 2019). This demonstrates that BreezoMeter’s pollen model reported the correct pollen risk category (in terms of ‘Low’, ‘Medium’, ‘High’) almost 100% of the time.
Constant Assessment & Improvement
We constantly strive to improve our models by taking new research into account, by comparing our model results to measurement stations where available, and based on customer and user feedback.