Surangi W. Punyasena

This paper uses deep learning to analyze fossil grass pollen morphology and reconstruct 25,000 years of ecological change in East African grasslands, showing reduced diversity and shifts in C3:C4 ratios during the last ice age.

Fusing a small-grain expert detector, fine-tuning across microscopes, human-in-the-loop continual learning and zero-shot SAM-2 segmentation improve detection and isolation of fossil pollen on scanned slides.

This study uses deep learning to analyze pollen morphology and links it to environmental factors and phylogenetic history in Podocarpus species.

The paper evaluates niche conservatism in two Amherstieae clades by comparing their climatic niches using geographic and environmental space analyses.

CuticleTrace automates epidermal cell outline measurement in leaf cuticle images, achieving ~93% agreement with expert tracings and improving efficiency for paleoecology and paleoclimatology studies.

This paper uses deep learning to place fossil pollen specimens in a phylogeny by transforming morphological features into predicted phylogenetic distances.