T46 Critical Zone Science: Intersection of Processes Linked to Geomorphology, Ecology and Climate (Posters)

Sunday, 22 September 2024
8:00 AM - 5:30 PM
Hall D

FINDING MIMA: A DEEP LEARNING BASED INVENTORY MAP OF MIMA MOUND PRESENCE IN THE PACIFIC NORTHWEST

Paper number: T46-12
Program: Technical Sessions
Time: 8:00 AM – 5:30 PM

Authors

Abstract

Mima mounds are an abundant microtopographic feature in the western United States, particularly in the Pacific Northwest. These roughly circular domes, in places numbering in the millions, show considerable variation in morphology and spatial patterns. Despite over 150 years of investigation leading to various compelling hypotheses such as vegetation anchoring and fossorial rodents, a widely applicable explanation for the formation and variation of these mounds remains elusive. This mystery is compounded by the cumbersome task of manual mapping, which restricts research to small areas, and the lack of detailed regional inventories, which prevents effective comparison of site-specific studies to the broader population. These limitations obscure key patterns and associations with geologic, biologic, and climatic factors, thereby hindering explanation of the widespread and morphometrically varied population of Mima mounds. To address these limitations, we developed a method to automate Mima mound inventorying over large regions using widely available LiDAR data and a machine learning model. To build a comprehensive training database, we created a Python tool that streamlines the generation and labeling of lidar imagery with evidence of Mima mounds. Using tens of thousands of 1m digital elevation models, we developed and trained a convolutional neural network to label 200m x 200m square areas for the presence or absence of Mima mounds. We reserved 20% of the training dataset to assess our model’s performance, obtaining an accuracy of 0.91. The model showcased resilience in discerning mounded areas with a precision of 0.90 and non-mounded areas with a recall of 0.92. Furthermore, the area under the receiver operating characteristic curve was 0.98, providing additional evidence of its efficacy in this classification task. We deployed this model across an extensive ~100,000 km2 region of Oregon and Washington east of the Cascade Mountains, producing a heat map of likely mound presence. This map allowed us to explore broad associations of Mima mounds with geologic, topologic, biologic, and climatic factors. We observed a strong preference of mounds for shallow, sandy soils, predominantly situated on flat terrains. When mounds occur on slopes, they are often facing southward. The Mima mounds in Washington are frequently associated with exposures of Columbia River Basalt.

Geological Society of America Abstracts with Programs. Vol. 56, No. 5

doi: 10.1130/abs/2024AM-402180

© Copyright 2024 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.