Research

Applied machine learning for scientific problems: fusing messy, multi-modal real-world data into predictions scientists can use.

Research

Senior thesis

Predicting Rangeland Herbaceous Biomass with Multi-Branch LSTMs

KDD Lab, New Mexico State University · Sep 2025 – Present

Read the write-up

Multi-branch LSTMs combining satellite biomass history, on-ranch rainfall, collar-derived grazing pressure and land cover to predict rangeland herbaceous biomass change. On held-out pixels from one 2025 season, adding rainfall and land cover raised R² from 0.36 to 0.93; grazing pressure did not improve it.

LSTM
multi-modal fusion
sequence modeling
remote sensing
time-series regression

Publications & Presentations

Neakrase, L. D. V.; Chanover, N.; Hartwick, V. L.; Huber, L.; Ajo-Montaño, S.; Dey, R.; Miller, N. The PDS Atmospheric Modeling Annex: Adapting to the Community Need for a Planetary Atmospheres Modeling Output Repository. 2026 AGU Annual Meeting (AGU26), San Francisco, CA, 2026.
poster · accepted
Neakrase, L. D. V.; Chanover, N.; Hartwick, V. L.; Huber, L.; Ajo-Montaño, S.; Dey, R.; Miller, N. The PDS Atmospheric Modeling Annex: A new repository for modeling output. 58th Annual Meeting of the AAS Division for Planetary Sciences (DPS 58), Spokane, WA, 2026.
poster · accepted

orcid.org/0009-0001-6194-4916