Hydrologic and topographic analyses of bog turtle (Glyptemys muhlenbergii) wetlands

Abstract

Bog turtles (Glyptemys muhlenbergii), a federally threatened species, depend on mountain wetland seeps to support their habitat needs. Bog turtles have been found to predominantly use saturated soil areas, making them highly sensitive to small changes in wetland hydrology. Recent research determined that water level elevations across these wetland systems are spatially variable and not equilibrated, even over short (ca. 5-10 m) distances, highlighting knowledge gaps regarding spatial variation in hydrologic regimes and associated habitat features. Building from that research, we integrated monitoring well data, delineated habitat patches, and computed terrain-based variables from 28 bog turtle-inhabited wetlands spanning both their northern and southern ranges. Using a previous classification system of water level dynamics, we reconstructed a principal component analysis (PCA) to categorize hydrologic regimes of 22 individual site wells. These individual wells were installed in predicted core nesting habitat of each site, as determined by observations of perennial saturation and emergent herbaceous vegetation (palustrine emergent wetland; PEM). However, the PCA and associated cluster analysis revealed that most of the core nesting habitat wells aligned more closely with previously characterized drier, intermittently saturated regimes. Further, across-site comparisons highlight high variability among sites, suggesting that either some of the wells were not installed in the same hydrologic regimes or, and more likely, that wetlands differ in their groundwater inputs which influences site wetness. To explore relationships between water level dynamics and wetland spatial features, we conducted regression analyses using water level metrics and water temperature regimes (indicators of groundwater seeps) as predictors of the extent of total wetland area and percent PEM areas. We also included a topographic wetness metric (TWI) and ditch presence as additional predictors. However, very few significant relationships emerged, suggesting the limitations of using a single well to infer spatial hydrologic and habitat patterns. To improve spatial predictions, we utilized LiDAR-derived topographic variables (e.g. topographic wetness index, deviation from mean elevation, roughness index) to classify wetland habitats within the study sites at a 1-m resolution. A random forest model effectively classified the three main habitat classes– emergent (PEM1), scrub-shrub (PSS1), and forested (PFO1) – with an overall accuracy of 84.6%, and classified the most suitable bog turtle habitat class (PEM1) with 97.5% accuracy. This research highlights the hydrologic variability across sites, demonstrates the limitations of a single well, and provides an effective approach for predicting habitat patches using terrain-based modeling. Collectively, these findings contribute to the broader knowledge of hydrologic and topographic controls on bog turtle wetlands and offer insights to inform future conservation and management efforts.