Research
PollenGeo: Transforming Botanical Collections into Digital Treasures
The Smithsonian Institution is home to several important palynological collections, now being digitized and standardized through the PollenGeo project. Among the most significant is the Graham Palynological Collection, established in 1954 and donated by Dr. Alan Graham in 2008. With approximately 15,000 species primarily from the Neotropics, it is the largest collection of its kind in the region. The collection consists of 25,000 index cards containing detailed taxonomic information, each with an associated microscope slide.
In addition to the Graham Collection, several other noteworthy collections are part of this effort. The Joan Nowicke Collection includes 5,229 slides representing 3,269 taxa, mainly from temperate North America. The Barro Colorado Collection served as the basis for the 1991 catalogue Pollen and Spores from Barro Colorado Island, documenting 1,269 taxa. Another key resource is the Amazon Collection, which features 1,215 taxa commonly found in the Holocene pollen record of the Amazon region.
One of the main challenges faced by the project was digitizing the 25,000 handwritten index cards from the Graham Collection. This task was addressed through the Smithsonian Transcription Center, where nearly 100 volunteers worked over a period of six months to photograph and transcribe the entire collection. The data from the cards, including taxonomy, locality, and collection details, were then organized into a structured database, with each record linked to its corresponding image.
As part of the digitization process, a taxonomic revision was also carried out. Scientific names were reviewed and updated based on current standards from the World Flora Online (WFO). Each species was assigned a unique identifier consistent with WFO’s taxonomy, ensuring alignment with global biodiversity data frameworks.
The digitization and taxonomic standardization of these collections not only preserve decades of scientific work, but also greatly enhance access and usability. Researchers, educators, and the general public are now able to explore these resources more easily, helping to bridge historical data with modern scientific inquiry.
Selecting Materials for Digitization
To accurately represent the diversity of pollen grains, the project digitized 20 grains per taxon, capturing variation in orientation, size, and morphology. Because of inconsistent labelling and differing preservation conditions, each slide had to be inspected manually. This close examination revealed that more than 30% of the slides were unusable due to deterioration—an expected issue in collections that are over three decades old. For each species, the best-preserved slide was selected, and only specimens identified at the species level were included in the digitization process.
Grain coordinates were recorded using a Nikon Eclipse Ni brightfield microscope. These positions were then converted by the PollenGeo system to be compatible across various microscopes used during digitization. This careful and methodical approach ensures not only the accuracy of the data but also the long-term preservation and usability of these valuable scientific records.
Capturing Botanical Details with BF and DIC Imaging
To reveal the structural complexity of pollen grains, each specimen was imaged using both brightfield (BF) and differential interference contrast (DIC) microscopy. Using Nikon and Zeiss microscopes, we applied complementary imaging techniques that captured both surface features and internal structures. Typically, 20 grains were photographed per slide; for fossil specimens, up to 50 grains were imaged to account for their greater morphological variation.
The process began with the import of grain coordinates from the PollenGeo database. Imaging parameters—such as magnification and Z-plane depth—were optimized for each slide. Imaging was then performed automatically in both BF and DIC modes, with each slide taking approximately 15 to 20 minutes. To compensate for occasional technical issues like vibrations or focus drift, additional grains were photographed when necessary.
All resulting images underwent a manual review process. Microscopists checked each image for clarity, focus, and proper orientation, applying light adjustments when needed to ensure uniform quality across the dataset.
Once approved, image files were saved, renamed systematically, and linked back to the PollenGeo database. The final images were stored on a 1,000-terabyte Smithsonian storage system, with automated backups in place to ensure data security. This workflow enabled the generation of up to 1,000 high-resolution images per day, creating a valuable resource for palynologists worldwide.
Zeiss Airyscan: A New Standard in Pollen Digitization
Zeiss Airyscan confocal microscopy was used to produce high-resolution, three-dimensional images of pollen grains, capturing both internal and external structures in remarkable detail. This method offers resolution comparable to electron microscopy, but without the need for additional sample preparation—making it especially valuable for preserving delicate botanical specimens.
To balance detail with efficiency, five grains per species—previously imaged using brightfield and DIC—were selected for confocal imaging. Due to the longer acquisition time required for this method, careful selection and preparation were essential.
Laser settings were customized for each grain, using 405, 488, and 561 nm wavelengths depending on the specimen’s fluorescence response and the mounting medium. Parameters were fine-tuned to avoid image saturation or loss of detail in low-intensity areas.
Different sampling modes were applied based on grain size and depth. SR-4Y was the default mode for grains between 20 and 50 microns, while SR-8Y and CO-8Y were used for larger or more complex structures. Averaging levels were adjusted accordingly—higher values (8x or 16x) for small grains to enhance resolution, and lower values for larger ones to balance quality and acquisition time. Images were collected in Z-stacks, capturing the grain layer by layer, then integrated into a single composite using the Airyscan module.
Following acquisition, images were renamed, linked to the PollenGeo database, and stored in the .czi format, which preserves all associated metadata. This process generated approximately 90 gigabytes of data per day, securely archived with backup protocols in place.
Through this approach, PollenGeo is setting a new standard for morphological detail in pollen digitization—advancing both preservation and scientific research.
Quality Control and Cropping in PollenGeo
Once multiple pollen grains are captured in a single image frame, the next step is to isolate and crop each grain individually. This process not only reduces file size but also removes debris or irrelevant background elements. It takes place on a dedicated workstation, separate from the microscope, and serves as the final quality control checkpoint before images are added to the PollenGeo database.
Each image is carefully reviewed for resolution, clarity, and the presence of any digital noise. Microscopists also verify the accuracy of the taxonomic identification and assess overall image quality. If any issues are found—such as contamination, incomplete grains, or subpar image quality—the system flags the image and notifies the original microscopist to retake the photograph.
Once approved, grains are manually cropped in both the X-Y and Z planes, ensuring only informative sections are retained. This step is critical for optimizing file size without compromising data quality. Each cropped grain is saved as a separate file, with a structured name that includes the slide code, species name, original image number, imaging technique, and the label “SG” (Single Grain).
Metadata from the original image is extracted and linked to the new single-grain files. Nikon images (.nd2) are processed using NIS software, while Zeiss images (.czi) are handled with Zen. After cropping, the original composite images are discarded, and the renamed files are integrated into the PollenGeo database.
This method ensures that each digital specimen is high-quality, properly labelled, and ready for use in palynological research, contributing to a well-organized and reliable dataset.
Interactive Tools for Exploring Pollen Morphology
The digitized PollenGeo collection offers a variety of ways for users to explore and interact with the data. To make the collection more accessible, small-file GIFs of each pollen grain have been generated. These animated files allow users to view the grain in motion across all Z-planes, effectively mimicking the experience of focusing through a microscope without needing to download the full-resolution image. For users requiring more detail, the high-resolution original files are also available for download in their native formats.
All images, including those of both extant and fossil taxa, have been uploaded to a custom-built public morphological database developed using MySQL 8.2 and PHP 8. This platform houses thousands of detailed morphological descriptions, covering most species in the digitized collection. Users are able to search by taxon name or filter results based on specific traits such as aperture type, ornamentation, or a combination of morphological features.
To enhance research capabilities, the database has been designed to support integration with R, allowing users to run complex queries and extract data for statistical or comparative analyses directly from the platform.
Additionally, a user-friendly web interface has been developed to connect with the OMERO system. This interface allows users to browse and visualize pollen grains as though they were viewing them under a microscope. Each image is accompanied by rich metadata, including details on the grain’s origin, slide information, position coordinates, imaging technique, resolution, Z-stack depth, and file size.
These tools are designed to ensure that the PollenGeo collection is not only accessible, but also practical and powerful for scientific research—bringing decades of palynological work into the hands of researchers, educators, and the public in a dynamic, interactive format.