Automated Mapping
Overview
Geoscientists traditionally create maps during detailed study of a field site in person. But a personal visit is not possible for most places in the solar system. Automating the process of mapping rock units, surface types, or depositional facies could help improve our understanding of sites we have yet to encounter.
Projects
Cucomungo Mars analog alluvial fan
We integrated topographic and spectral data into a machine learning workflow to identify fan aqueous deposits (e.g., sheet, fluvial, and debris flows). We evaluated the performance of our model in the field at the Cucomungo Canyon alluvial fan Mars analog in Death Valley, CA. We found that cluster mapping techniques accurately guided our traverse to geologic contacts and outliers.
Jezero crater automated map
We evaluate a variety of automated mapping techniques including clustering (i.e., k-means clustering, gaussian mixture models, BIRCH, etc.) and segmentation methods on Mars orbital data of the Perseverance rover landing site in Jezero crater.
Big Maria Mountains field mapping
This EDMAP funded project will use ultra-high resolution drone imagery and machine learning to map the surface geology of a portion of the Big Maria Embayment (CA). The project has two objectives: 1. develop a protocol for semi-autonomous mapping of surface deposits in arid environments, and 2. interpret the Plio-Quaternary migration history of the lower Colorado River.
Related Works
Kerner, H. R. and Adler, J. B. (2022). Guiding Field Exploration on Earth and Mars with Outlier Detection. In Proceedings of the International Geoscience and Remote Sensing Symposium (IGARSS), https://doi.org/10.1109/IGARSS46834.2022.9884366.
Adler, J., Rivera-Hernández, F., and Kerner, H. (2023) Machine learning techniques aid in Mars analog surface characterization by guiding in-situ exploration to geologic contacts and outliers. Under review.