A holistic big data approach to understand and manage increasing pollen-induced respiratory allergies under global change

Nearly one-third of the world's population suffers from pollen-induced respiratory allergies—and the number is growing. These include hay fever, rhinitis, asthma, allergic sinusitis, allergic conjunctivitis, and serious complications for those with chronic inflammatory lung disease. For those affected, the situation is becoming more stressful every year: Global warming and rising atmospheric CO₂ levels are causing longer pollen seasons with earlier onset and higher pollen concentrations (Anderegg et al., 2021; Ziska et al., 2019). This is projected to worsen in the future (Sapkota et al., 2020). These phenomena form a complex interface between human health and global change, yet the critical information needed to unravel it remains fragmented (Zhu et al., 2024). The need for a transdisciplinary approach and close collaboration between Earth system science and human health experts to address these pressing challenges has been recognized (Asrar et al., 2020; Davies et al., 2015).

For many researchers living and working in southern Australia, the wake-up call was the world's most catastrophic epidemic thunderstorm asthma event in Melbourne in November 2016, when allergenic ryegrass pollen broke up into small toxic fragments, leading to thousands of acute respiratory presentations to emergency departments and was associated with 10 deaths. These events forced us to think: What knowledge would have enabled us to provide better early warning information to the public about respiratory health risks? Because periods of high pollen concentrations are dynamic and can be associated with long-range transport under certain weather conditions, it is not enough to know pollen concentrations at a few locations. What we really need is spatially explicit information about which species grow where, when do they start releasing pollen, and how both are being reshaped by climate change and urbanization (Zhu et al., 2024).

Global warming is known to significantly alter pollen phenology (Sapkota et al., 2020). However, few studies have examined other variables, particularly the landscape-scale ecological factors. When a piece of land is converted from native vegetation or agricultural land to an urban area, the local pollen aerobiology is completely altered. Non-native species are planted in gardens and parks, requiring additional management and resources, such as mowing and watering.

For example, a study in Germany showed an increase in the abundance and diversity of pollen allergens, due to an increase in allergenic non-native plants in urban parks and gardens (Bernard-Verdier et al., 2022). In Japan, the planting of Japanese cedar for timber production has led to an explosion of hay fever (Saito, 2014). In Australia, ryegrass, introduced as a pasture species by European settlers, has become dominant in peri-urban and rural landscapes, triggering asthma attacks each spring (Silver et al., 2018).

The highly dynamic urban–rural gradient resulting from rapid population growth creates a more complex aerobiological environment that, when combined with climate change factors, can trigger or exacerbate respiratory health risks (Li et al., 2017; Li, Hao, et al., 2022; Li, Li, et al., 2022). To better attribute and manage these health risks, we need knowledge of local to global trends in pollen phenology, and the relative importance of rising CO₂ levels, changes in climate (e.g., warming, changing precipitation, and extreme weather events), and landscape conditions (e.g., land cover and land use change) in driving these trends (Ma et al., 2022). To date, we have only studied some of these factors independently, or in some limited combinations, but not holistically. This limits our ability to make reliable predictions and understand future conditions.

There is at least one piece of good news: Advances in the collection and processing of large environmental datasets are now providing the basis for analyzing and predicting global trends in pollen seasonality and its drivers, providing deeper insights for research and management (Figure 1).

For example, Earth observation missions from space agencies and commercial companies are providing meter-scale observations that, when combined with in situ pollen samplers and phenocams, can generate invaluable data sets with sufficient detail to resolve highly heterogeneous plant species phenology and distribution from urban to rural areas globally (Ma et al., 2022; Zhu et al., 2024). These data, along with multi-decadal historical archives such as MODIS and Landsat, are now available to understand the confounding effects of factors that contribute to spatial and temporal changes in pollen and its association with respiratory allergies (Devadas et al., 2018). Liu et al. (2024) correlated grass phenology derived from Sentinel-2 and phenocam observations for grass pollen seasons to improve understanding of the spatial distribution and inter-seasonal variation of grass pollen sources in Australian urban landscapes. High temporal and spatial resolution satellite data, such as VENμS, PlanetScope, and HLS (Harmonized Landsat and Sentinel-2), also enable near-real-time and short-term prediction of phenology, which is critical for early warning of allergic pollen outbreaks (Gao & Zhang, 2021).

In situ pollen monitoring networks are expanding, and traditional samplers are being replaced by automated pollen counters that can monitor pollen concentrations in real time and provide more information about plant species. Meanwhile, environmental DNA—small amounts of genetic material captured in air samplers—now allows pollen to be classified and quantified from family and genus levels to finer species levels (Clare et al., 2022; Van Haeften et al., 2024). Mobile apps for species identification using artificial intelligence (AI) and computer vision technologies are becoming available and can provide critical information on urban vegetation species' distributions and phenology (Rzanny et al., 2024). Human behavior data are also widely available through social media and the Internet of Things. Local pollen apps are increasingly available to provide daily pollen forecasts to protect those with respiratory allergies, and a positive “early warning” association has been found between the frequency of Google searches for terms such as “hay fever,” “allergies,” and “runny nose” and local pollen concentrations (Kang et al., 2015).

While these technical advances provide a solid foundation of data, innovative models informed by observations are needed to derive insights and useful knowledge for effective interpretation, forecasting, and management practices. Traditional site-based regression models for pollen forecasting, which rely heavily on local expert knowledge and meteorology, are being replaced by more advanced hybrid approaches that combine pollen networks and physical process models with the versatility of data-driven machine learning to fully exploit the spatio-temporal contextual information. Emerging technologies such as AI-based approaches can be used to integrate the complex processes of pollen production and dispersal, effectively improving the accuracy of pollen concentration and seasonal predictions. In addition, it is also important for scientific mitigation and adaptation strategies to consider human population dynamics in pollen studies to measure population-weighted exposure to pollen risks. Ultimately, the key is to establish a dynamic data-knowledge feedback loop with data-driven knowledge distillation in one direction and knowledge-guided data collection in the other (Figure 1).

Unraveling the complex interface between climate change and urbanization and its impact on human health is more than a technical challenge. A holistic big data approach, with close collaboration between Earth system scientists and health experts, is needed to gain meaningful insights that will advance our understanding and management of increasing pollen-induced respiratory allergies under global change.

Xuanlong Ma: Conceptualization; funding acquisition; visualization; writing – original draft; writing – review and editing. Alfredo Huete: Conceptualization; funding acquisition; writing – original draft; writing – review and editing. Yuxia Liu: Writing – original draft; writing – review and editing. Xiaoyu Zhu: Writing – original draft; writing – review and editing. Ha Nguyen: Writing – original draft; writing – review and editing. Tomoaki Miura: Writing – original draft; writing – review and editing. Min Chen: Writing – original draft; writing – review and editing. Xuecao Li: Writing – original draft; writing – review and editing. Ghassem Asrar: Conceptualization; writing – original draft; writing – review and editing.

The authors declare that they have no conflict of interest.