As the cold winter recedes, springtime delights us with a splash of color, the sound of birds, warmer temperatures, increased daylight hours, and an overall vibrant energy.
But this freshness comes with its challenges, like pollen allergies. They are most prevalent during spring but also in summer because many plants, trees, grasses, and weeds pollinate at this time, releasing pollen into the air.
While pollen counts tend to be higher during warmer seasons, some plants pollinate throughout the year.
Then there’s climate change, which may also affect the pollen spread. By shifting temperatures, atmospheric carbon dioxide (CO2), and precipitation, climate change can impact the pollen season duration, pollen quantity, pollen allergenicity, and the risk of experiencing allergy symptoms.
Pollen is an airborne allergen that triggers various allergic reactions. These include allergic rhinitis or hay fever, which happens when pollen enters the body, the immune system mistakenly recognizes it as a threat, and then reacts.
Pollen allergy is actually pretty common, with 10% to 30 % of the global population estimated to be affected by it. In the US, about 7.8% of people aged 18 and over experience hay fever.
Exposure to pollen can also trigger symptoms of allergic conjunctivitis, an inflammation of the eye lining. About 30% of the normal population experience eye allergy, which involves red, watery, or itchy eyes, while a much higher number, 70% of those with allergic rhinitis, get these symptoms.
Those with respiratory conditions like asthma could even be more sensitive to pollen, and exposure to it can cause asthma attacks, respiratory issues, and reduced productivity at work and school.
There are also massive medical costs involved, with pollen-related medical expenses surpassing $3 billion annually. Half of these are prescription medication costs.
Addressing this major public health concern means accurately identifying pollen grains, which are tiny ‘seeds’ dispersed from flowering plants and trees. Their correct identification is important for understanding human-environment interactions and reconstructing landscapes and environments.
“With more detailed data on which tree species are most allergenic and when they release pollen, urban planners can make smarter decisions about what to plant and where.”
– Study co-author Behnaz Balmaki, an assistant professor of research in biology at UT Arlington
She further noted the importance of selection and placement in high-traffic areas like parks, schools, hospitals, and neighborhoods.
However, differentiating between pollen grains of conifer genera, particularly those from Abies (fir), Pinus (pine), and Picea (spruce), poses challenges with palynology, the study of pollen grains and spores, due to their morphological similarities.
High Similarities Make it Challenging to Distinguish Pollen
Researchers use pollen data to study historical and contemporary environmental analysis and to plan cities.
Analyzing pollen grains preserved in sediments and peat bogs actually allows paleoecologists to identify the types of vegetation that existed in a particular location at different times in the past. This helps them reconstruct past vegetation patterns and historical climate conditions, as plant distribution is closely linked to specific climate parameters like rainfall and temperature.
As such, researchers are able to understand changes in landscapes and trace the ecological impacts of climate fluctuations over centuries. This provides insights into just how ecosystems respond to changes in the environment, as well as helps make predictions about future ecological responses to climate change.
Pollen grains further help us identify the interactions between human activities and environmental factors that significantly shape these landscape patterns.
Here, coniferous genera are of importance due to being the representatives of specific ecological and climatic adaptations. For instance, pine trees are known for their resilience to environmental stresses like fire. Meanwhile, spruce trees are adapted to cold environments, and fir trees are susceptible to moisture changes.
Data on this plant species can help us gain an extensive understanding of fire regimes, climatic fluctuations, precipitation patterns, and historical humidity. It can also play a key role in allergen treatment and public health management.
In particular, pollen analysis can help pinpoint allergenic species and forecast related health issues, proving beneficial in allergy and health research.
Of course, the problem here arises in the morphological similarities of coniferous genera. Researchers rely on pollen grains’ morphological features, which include size, shape, symmetry, polarity, apertures, and ornamentation, to study pollen grains.
When it comes to pollen grains that are closely related, the morphological differences are very subtle, making it challenging to differentiate one species from another accurately and quickly.
For instance, using the microscope to identify pollen grains is a resource-intensive process. Not only is it expensive and time-consuming, but it is also dependent on subjective criteria, which results in error rates as high as 33%.
Pollen grains of conifer species, specifically, have been widely documented as rather difficult to identify due to their little morphological distinctness. All of the grains in the group share high similarities, having two air sacs with a central body. As such, even accurately recognizing these species under a microscope is challenging.
Researchers have been utilizing digital imaging techniques and graphical software to enhance the analysis. However, this is still largely dependent on human visual inspection, which is subject to classification errors, especially if novice palynologists are involved.
These limitations call for more objective, efficient, and precise techniques to identify pollen grains. This difficult task requires expert knowledge, high-resolution micrographs, and a substantial number of reference slides to make accurate comparisons and then identification.
“Even with high-resolution microscopes, the differences between pollens are very subtle.”
– Dr. Balmaki
AI Can Help Find the Pollen Responsible for Allergies
Artificial intelligence (AI) has been helping make advances in almost all industries. It utilizes vast amounts of data to learn from and then improve, allowing the technology to identify patterns and relationships that humans may miss.
As we recently shared, AI is enabling scientists to find the best solid-state electrolyte (SSE) candidates by drawing from a massive database of previous studies. This quick search through all the potential options accelerates the discovery of optimized SSE options to advance high-performance solid-state batteries (SSBs) for sustainable energy demands.
Now, researchers from the University of Texas at Arlington (UTA) are leveraging the tech to enhance pollen analysis1 by identifying species from a deep learning model trained on thousands of images.
Founding deep learning to be a great technique for this purpose, researchers noted that this approach can significantly improve pollen classification accuracy and dramatically reduce the time required for identification.
Traditional methods, such as manually identifying an individual pollen sample, can take hours, depending on the sample complexity and a person’s expertise.
In contrast, machine learning (ML) and deep learning (DL) models can process thousands of images in seconds once trained. The “exponential improvement in speed” makes DL particularly valuable for large-scale ecological and environmental studies.
This way, having a trained model can potentially improve species identification while reducing the need for considerable morphological training in palynology.
“Our study shows deep-learning tools can significantly enhance the speed and accuracy of pollen classification. That opens the door to large-scale environmental monitoring and more detailed reconstructions of ecological change. It also holds promise for improving allergen tracking by identifying exactly which species are releasing pollen and when.”
– Dr. Balmaki
To take a deeper dive into the technology used by researchers at UT in collaboration with the University of Nevada and Virginia Tech, they leveraged advanced deep learning techniques, specifically transfer learning models.
These models involve reusing a pre-trained model for a different but related task. This way, transfer learning prevents the need to start from scratch and helps reduce the time and resources needed to train new models, even with limited data.
According to the study, the transfer models are effective in recognizing any similarities in detailed features. They can actually help create models for the identification of difficult species, especially in conifer species classification, and enhance pollen grain analysis.
Researchers actually utilized nine transfer learning models — VGG16, VGG19, ResNet101, ResNet50, MobileNetV2, InceptionV3, EfficientNetV2S, DenseNet201, and Xception.
They trained and validated each model on a dataset of pollen grain images collected from samples preserved by the University of Nevada’s Museum of Natural History.
The models were also assessed on various performance metrics, including precision, accuracy, recall, and F1-score across training, testing, and validation phases. Based on the results for each of these models, ResNet101 was found to be outperforming all others. It achieved a test accuracy of 99% with equally high precision, recall, and F1-score. According to Balmaki:
“This shows that deep learning can successfully support and even exceed traditional identification methods in both speed and accuracy.”
While powerful, AI doesn’t remove the need for the expertise of trained palynologists, though. In fact, it “confirms how essential human expertise still is,” she added. After all, in order to create necessary datasets, we need “well-prepared samples and a strong understanding of ecological context. This isn’t just about machines—it’s a collaboration between technology and science.”
With this new AI system, researchers aim to provide city planners with a tool to make more informed tree selection decisions and build more health-conscious landscapes, which would be a big relief to allergy sufferers. Dr. Balmaki noted:
“Health services could also use this information to better time allergy alerts, public health messaging, and treatment recommendations during peak pollen seasons.”
Even farmers can benefit greatly from this study, as “pollen is a strong indicator of ecosystem health.” As Dr. Balmaki explained:
“Shifts in pollen composition can signal changes in vegetation, moisture levels and even past fire activity. Farmers could use this information to track long-term environmental trends that affect crop viability, soil conditions, or regional climate patterns.”
Furthermore, the study can be useful for the conservation of wildlife and pollinators.
Insects like bees and butterflies depend on specific plants for their food and habitat. By identifying the plant species present or declining in an area, we can better understand how such changes impact the entire food web and then take appropriate measures to protect critical relationships between plants and pollinators.
The research plans to expand to include a wider range of plant species. The researchers’ goal is to develop a comprehensive pollen identification system that can be applied across the US to better understand how plants may shift in response to extreme weather events.
Click here to learn if robotic pollinators play a key role in vertical farming.
Investing in Artificial Intelligence (AI)
In the realm of AI, Nvidia (NVDA +0.26%) is the biggest name. This semiconductor powerhouse has been powering many deep learning models with its GPU technologies and AI hardware. In fact, all the models in this study were trained and tested on an NVIDIA GeForce RTX 3060 with 12GB of memory using Python 3.10.6 and TensorFlow.
NVIDIA Corporation (NVDA +0.26%)
Nvidia’s GeForce RTX 3060 was introduced by Nvidia back in early January 2021 as the second generation of NVIDIA RTX™ to deliver up to 10x the ray-tracing performance of the GTX 1060 with support for NVIDIA DLSS.
It is powered by NVIDIA Ampere Architecture, which is created with 54 billion transistors and is the largest 7 nm chip ever built. The architecture features groundbreaking innovations including 3rd-generation Tensor Cores to accelerate and simplify AI adoption, multi-instance GPU (MIG) to allows workloads to share the GPU, GPU-to-GPU direct bandwidth of 600 GB/s, structural sparsity to improve the performance of model training, second-generation RT Cores to speedup workloads, and 2 TB/sec of memory bandwidth.
Currently, Nvidia’s Hopper (H100) GPU and next-gen Blackwell GPU architecture are stealing all the limelight, being the preferred chips of businesses that want to lead the AI innovation.
Through its advanced solutions and several top clients, Nvidia has become an AI darling stock that helped its prices go from sub-$30 exactly two years ago to its current level of $119. This 296% upside has been primarily driven by AI mania, which resulted in NVDA stocks hitting a peak at nearly $150 in Nov. 2024, just after the conclusion of the US presidential election.
Since then, NVDA shares have lost 20.66% of their value while recording a decline of 12.83% this year as tariffs and a trade war disrupt the stock market. Last month, Nvidia actually reported anticipating $5.5 bln in charges related to the export of H20 to China.
Amidst this, the leading AI chip maker announced its plan to invest hundreds of billions of dollars in the U.S. supply chain over the next four years. “Having the support of an administration that cares about the success of this industry and not allowing energy to be an obstacle is a phenomenal result for AI in the U.S.,” CEO Jensen Huan told the FT a couple of months ago.
The Trump administration is also prepared to rescind the “AI diffusion rule,” which will effectively prevent a set of AI chip controls from taking effect later this month. The rule organized countries into three tiers, with each one having different restrictions on the chips made by Nvidia and others that can be shipped there without a license.
Nvidia has been against the rule, with Huang saying earlier this week that being locked out of the Chinese AI market, which could be worth $50 bln, would be a “tremendous loss.”
When it comes to profitability, Nvidia has an EPS (TTM) of 2.94, a P/E (TTM) of 39.82, and an ROE (TTM) of 119.18%. Nvidia also pays a dividend yield, but it’s a mere 0.03%. Its market capitalization, meanwhile, is a whopping $2.85 trillion, which makes it the world’s third-largest company.
NVIDIA Corporation (NVDA +0.26%)
Now, for Q4, which ended January 26, 2025, Nvidia reported a record quarterly revenue of $39.3 billion, an increase of 12% and 78% from the previous quarter and a year ago, respectively.
For the full fiscal 2025, revenue was $130.5 billion, a massive 114% surge from a year ago. GAAP earnings per diluted share also jumped 147% to $2.94 while non-GAAP earnings per diluted share went up 130% to $2.99.
These strong financials were the result of Nvidia’s new GPU microarchitecture, Blackwell, which is designed for Gen AI and accelerated computing and is seeing “amazing” demand. While sharing the results in February, Huang said that they have been ramping up the extensive production of Blackwell AI supercomputers, and already in the first quarter, they have generated billions of dollars in sales.
Latest on NVIDIA Corporation
Conclusions
Deep learning techniques are known to boost efficiency and accuracy and reduce errors and manual effort in object detection, image classification, and task recognition. They have also been extremely effective in classifying pollen. The study used transfer learning, in particular, and found it to be cost-effective and less time-consuming while also addressing the challenges of data scarcity in aiding conifer species.
By leveraging AI, researchers have provided a tool to relieve allergy sufferers and help farmers and city planners. With scalable, fast, and accurate pollen identification, urban environments can be tailored to reduce allergen exposure and improve public health responses.
Moreover, given climate change’s impact on the environment, combining ecological science with AI will not only drive more health-conscious planning of our cities and landscapes but also preserve wildlife and pollinators and produce higher crop yields.
Click here to learn all about investing in artificial intelligence.
Studies Referenced:
1. Rostami, M. A., Kydd, L., Balmaki, B., Dyer, L. A., & Allen, J. M. (2025). Deep learning for accurate classification of conifer pollen grains: enhancing species identification in palynology. Frontiers in Big Data, 8, 1507036. https://doi.org/10.3389/fdata.2025.1507036