Geology topics

Jacob Oram

Publications and source records attributed to Jacob Oram.

2 recordsLinked to original sources

USGS2025

Predicting bat roosts in bridges using Bayesian Additive Regression Trees

Human-built structures can provide important habitat for wildlife, but predicting which structures are most likely to be used remains challenging. To evaluate the predictive capabilities of data-driven ensemble modeling approaches, we conducted surveys for bats and signs of bat use, such as urine and guano staining, at bridges across the southwestern United States. We developed a bat roost discovery tool using Bayesian Additive Regression Trees (BART) and evaluated the predictive ability of this model against other commonly used approaches. We found that the lack of nearby water resources was associated with a lower predicted probability of bat presence or signs of bat use at bridges. While the presence of nearby water resources was associated with higher average predicted probability of bat presence or signs of bat use, high uncertainty surrounding these estimates indicates that other factors also play a role in determining which bridge roosts bats are more likely to use. As such, our model could be particularly useful for predicting which bridges can be excluded from survey efforts due to low probability of bat presence or signs of bat use. We extrapolated our model to unsurveyed bridges across the study region and provide an interactive dashboard application interface for the exploration of these results. Overall, this study demonstrates the application of BART as a predictive tool for prioritizing future bridge surveys for bats roosting in transportation structures.

Arizona, California, Nevada, New Mexico, Texas↗

USGS2025

Leveraging an observed-data likelihood improves the use of machine learning labels in a Bayesian hierarchical model for bioacoustic data

Classification of massive datasets by machine learning (ML) algorithms is promising for many scientific domains, especially wildlife monitoring programs that rely on passive acoustic surveys for detecting species. However, treating ML-predicted class labels (e.g., species identity) as truth biases inferences of focal parameters within common modeling frameworks. One solution is to model the misclassification process explicitly using human-validated true-class labels for a subset of observations. Validation by experts can present a substantial bottleneck in otherwise efficient workflows that use ML predictions. Bioacoustics practitioners seek guidance on both the quantity and process for selecting ML-labeled data to validate by an expert. We derive an alternative model formulation that jointly models human-validated and ML-predicted class labels with an observed-data likelihood (ODL) and use empirically informed simulations motivated by a real-data application to explore different probability designs for selecting class labels for validation. Simulation results suggest that with smaller validation sets the ODL formulation increases computational speed and reduces estimation error compared to a default MCMC data augmentation routine. Our methodology is transferable to applications that treat predictions from classification algorithms as the response variable of interest.

Annals of Applied Statistics↗