Abstract
Groundwater is critical for many ecosystems, yet its role in supporting ecosystems is rarely acknowledged in global
conservation, climate, and sustainability goals. As groundwater depletion intensifies and spreads globally due to
climate change and high human water demands, it is imperative that groundwater-dependent ecosystems (GDEs) be
identified and considered in policy, programmatic, and management decisions. However, progress has been hindered
by significant data and ecohydrologic knowledge gaps. To advance the protection of GDEs globally, this dissertation
leverages publicly available datasets, and advancements in remote sensing, cloud computing, and machine learning
to evaluate ecosystem groundwater needs and map GDEs across global drylands. First, normalized difference
vegetation index (NDVI) derived from Sentinel-2 satellite imagery were coupled with field-based groundwater level
data to assess groundwater-dependent vegetation responses to groundwater depth differences across seasons and
streamflow regimes in California. A diminished reliance on groundwater for vegetation along anthropogenically
altered streams suggests that many riparian woodlands in California are subsidized by water management practices,
thus undermining their resilience to natural hydrologic variation. Second, NDVI data derived from Landsat satellite
imagery were standardized using Z-scores and used to identify groundwater thresholds corresponding to reductions
in vegetation greenness across a wide range of biomes and local conditions across California. While absolute
groundwater depths vary across locations and plant communities, our results showed that rooting depths for
dominant species inferred from vegetation maps can be used as threshold proxies for groundwater impact analyses.
Furthermore, potential drought refugia were mapped across California utilizing ZNDVI scores, which can help
practitioners to prioritize limited financial and natural resources to protect these critical habitats. Finally, GDEs were
mapped across global drylands by employing a Random Forest machine learning model along with publicly
available climate and satellite imagery data. Our map highlights the growing need to protect GDEs from the threat of
groundwater depletion. Overall, the cumulative findings from this dissertation provide critically needed technical
guidance for practitioners to identify and consider GDEs across various management scales globally so that global
conservation, climate, and sustainability goals can be achieved.