Rosemary. W. H. Carroll

The western United States faces two escalating, co-occurring hazards in the form of persistent water scarcity (drought) and increasingly frequent wildfires in places humans live and recreate. Drought adaptation has emphasized water savings from cropland retirement, while wildfire adaptation has emphasized the fire risks of urbanization. By overlooking what retired cropland becomes and what development replaces, neither has fully examined cropland’s role as a landscape-scale buffer to wildfire in the western US. We demonstrate this buffering capacity across 11 states and 39 years of data, showing that cropland is less likely to burn than other vegetation classes. Within 477 watersheds where cropland has interacted with wildfire in the western US, approximately 4.29 million hectares of cropland was lost between 1985 and 2024. Cropland loss in most watersheds (77%) was consistent with fallowing transitions to more burnable herbaceous and scrubland vegetation across the Pacific Northwest, Inland Northwest, Intermountain West, and Great Basin. In a smaller portion of watersheds (19%), losses occurred through cropland-to-developed transitions clustered around major metropolitan areas in California, Idaho, Oregon, Washington, Utah, and Colorado, where development simultaneously removes buffering capacity and adds ignition sources. The remaining watersheds saw cropland transitions to wetlands and forests. Results highlight emerging tradeoffs between drought adaptation and wildfire risk in the western US, with retirement-driven fallowing and development-driven conversion reshaping where wildfires stop at landscape-scales. Integrated planning tools that evaluate water use savings and wildfire risks jointly, and expanded agricultural conservation programs that incorporate fire-buffering value, can help manage tradeoffs.

Near‐surface air temperature grids are important environmental model forcings. In narrow mountain valleys, most gridded products are too coarse to resolve variability across slopes, so regional studies often map temperature across terrain following locally determined lapse rates. However, lapse rate approaches are complicated by patterns such as cold‐air pooling (CAP), which decouples the valley floor from upper slopes. We used a network of temperature sensors distributed across Colorado's East River watershed to characterize local CAP using empirical orthogonal function decomposition and a topographic algorithm. We then compared three approaches to map observed temperatures across terrain: a two‐station lapse rate, a lapse rate fit to a dense station network, and a CAP‐aware scheme which treated upper slopes separately from the valley floor. The CAP‐aware scheme reduced pooled root mean square error (RMSE) in predicted daily‐minimum temperatures () to 0.9°C, compared to 5.6°C and 1.6°C for the other two approaches, and an application with only a few local weather stations retained skill despite low sensor density (1.1°C). When used to construct 800 m grids, the CAP‐aware approach reduced pooled RMSE compared to PRISM (1.1°C vs. 3.8°C), even when only a few stations were used (1.9°C). When the best and worst performing approaches were mapped to a 100 m grid to force a regional hydrologic model, early summer cumulative basin snow water equivalent differed by 111% between runs, indicating strong sensitivity to melt season temperature input. Finally, analysis across Colorado state suggests CAP effects are widespread among SNOTEL observations.

The cascading effects of meteorological droughts on river flows and water quality in mountainous regions are poorly understood. Using a data-driven framework, we studied drought impacts on streamflow, water temperature (WT), specific conductance (SC) in the Upper Colorado River Basin (1998-2022). Severe meteorological droughts in 2001-2002, 2012, 2018, 2020-2021 reduced annual streamflows by 15%, increased WT by 9% and SC by 5% relative to normal adjacent years. Declines were greatest in spring and summer, with reduced peak flows. Streamflow and SC recovered until 2018, after which successive droughts prolonged impacts, whereas WT effects were transient. Catchments with higher freshwater withdrawals, deciduous forest cover, air temperatures, residence times, finer soils had greater streamflow declines, while higher subsurface flows and infiltration enabled resilience. WT increases were associated with higher elevation, slopes, precipitation and lower agriculture, shrub cover. Open water and reservoirs mitigated SC increases. Thus, catchment attributes influence drought vulnerability in mountainous regions.

Atmospheric warming is reducing snowpack, with uncertain effects on mountainous streamflow, a crucial water resource. Despite limited historical observations of groundwater–streamflow interactions above 2,500 m, new measurements in the Upper Colorado River headwaters indicate declining groundwater storage that is dated decades to millennia old. Here we use integrated hydrologic modelling spanning water years 2015–2021 to determine whether the loss of old-age groundwater buffers streamflow during low-snow years and whether that loss is exacerbated with warming. Results show that old-groundwater contributions to streams remain relatively steady through time, unlike the more variable contributions from young groundwater. Numerical experiments of increased surface air temperatures (+2.5 °C and +4 °C) increase rain–snow fractions and evapotranspiration and decrease runoff ratio by 2–3% per degree Celsius increase. As streamflow declines with warming, the age of groundwater supporting it gets older, in part owing to intermediate-aged (1–3 year) groundwater declining twice as fast. Simulations show that water table depths at higher elevations (>3,700 m) decline disproportionately and fail to recover even during wet years. These findings suggest altered groundwater–streamflow interactions with warming and snow loss, with implications for water resources.

Climate change is altering snow accumulation and ablation dynamics in snow-dependent regions worldwide, reshaping runoff timing and water availability for people, ecosystems, and agriculture. As the world’s largest consumer of freshwater, irrigated agriculture is particularly exposed, and these hydrologic changes impose substantial economic costs. Here, we demonstrate how hydrology and economics can be combined to assign costs to snow-driven hydrologic change, using irrigated agriculture in the western United States as a large, snow-dependent test case. By integrating reduced-form economic models with climate model projections spanning approximately 2–5 °C of warming, we estimate that irrigated cropland could decline by 27–46% by the end of the century, while agricultural profitability—proxied using land rental rates—declines by 11–26%, corresponding to annual losses of approximately $8.2–$14.7 billion. Against this economic backdrop, scientists have an opportunity to leverage expanding data and modeling capabilities to provide actionable information about adaptation strategies that can reduce damages.However, there remains a persistent gap between the scales at which snow loss research is conducted and the scales at which land and water management decisions are made. To address this challenge, we propose that archetypes—drawn from social-ecological systems research—could help accelerate matching adaptation strategies to specific decision-making contexts by explicitly accounting for governance capacity and behavioral dynamics. This approach has not been widely explored in agricultural adaptation to snow hydrology but could enable rapid, locally-relevant guidance.Yet rapid adaptation without integration across hazards risks unintended consequences. Through analysis of four decades of fire perimeter data (1984-present) and aerial imagery, we show that cropland is 2x less likely to be on the inside of a fire perimeter than any other land-cover type, suggesting an important landscape-scale buffering effect. Using drought and wildfire as an example, this finding demonstrates how cropland abandonment—a strategy that may enhance drought resilience—could amplify fire risk if poorly coordinated, illustrating how rational responses to one hazard can inadvertently increase exposure to others. Results underscore the complexity of adaptation under compounding climate risks and the importance of working in partnership with decision-makers to leverage ever expanding data and modeling capabilities for locally-relevant solutions.

Mountain snow is a globally important source of water for downstream users and ecosystem sustainability, but it is uncertain when and where traditional snow metrics alone provide reliable indicators of water supply. Here, we synthesize large-sample regional analysis with a high-resolution integrated hydrologic model of a headwater basin of the Colorado River (East River, Colorado, 750 km2) to demonstrate the tight coupling between land surface and subsurface process on mountain streamflow generation. Our analysis of nearly 4,700 western US mountain watersheds indicates that streamflow predictability depends not only on snow accumulation magnitude, but on whether snow represents the dominant water reservoir. Snow storage-dominated mountain basins exhibit a tight coupling between peak snow water equivalent and runoff, while mixed storage systems, such as the East River, depend on interactions among snow, rainfall, and groundwater. Recent declines in runoff efficiency and the degradation of low-flow metrics in the East River coincide with persistent subsurface storage deficits due to overlapping climate extremes, signaling a shift toward subsurface storage-limited behavior. Modeled scenarios of persistent and prolonged warming indicate enhanced vegetation water use reduces recharge, drives substantial groundwater storage loss that disproportionately affects dry years and limits recovery even during wet periods, ultimately reducing annual flows and increasing stream intermittency. Sensitivity experiments further show that the depth and porosity of active bedrock circulation strongly modulate drought response. Deeper, higher-porosity groundwater systems are able to buffer low flows during multi-year drought conditions but experience prolonged post-drought recovery. Collectively, these findings highlight the tight coupling among climate, vegetation, snow, and groundwater, and demonstrate that explicit representation of subsurface storage dynamics is essential for forecasting mountain water supply, ecosystem vulnerability, and drought response under future climate conditions.

Accurate near-surface air temperature grids are critical for environmental modeling. Standard lapse rate methods use linear regression between observations to distribute temperature by elevation, but typically ignore cold-air pooling (CAP), a major driver of night-time variability. We used a network of temperature sensors distributed across Colorado’s East River watershed to characterize local CAP using Empirical Orthogonal Function (EOF) decomposition and a topographic algorithm. We then compared three lapse rate schemes for mapping temperatures onto a 100 m grid: (1) a direct 2-station lapse rate, (2) a linear regression across the full network, and (3) an "organized network" scheme excluding CAP sites from the regression but reinserting the mean valley-floor temperature in CAP-prone terrain. This organized approach reduced mean RMSE in minimum temperature (Tmin) from 4.4°C (method 1) and 1.2°C (method 2) to 0.6°C, and outperformed a commonly used gridded dataset (PRISM; mean RMSE = 3.9°C). When two of these methods forced a hydrologic model, early summer cumulative basin snow water equivalent differed by 109% between model runs. Statewide analysis further revealed about half of Colorado’s SNOTEL sites exhibit CAP, pointing to widespread influences in observed Tmin across valley-floor weather stations. Finally, a sparse application of the organized scheme using only three CAP-free SNOTEL stations and one valley floor weather station retained skill despite low sensor density (mean RMSE = 1.1°C). These results highlight the issue of mountain weather station representativeness and demonstrate the value of CAP-aware lapse rate methods.

The cascading effects of meteorological drought on water availability and quality in mountainous regions are poorly understood. The Upper Colorado River Basin (UCRB) is a vital water source for the American Southwest, primarily fed by snowpack in head- water catchments. Reduced precipitation and rising temperatures have caused droughts in the basin since 2000. We used a modular data-driven approach to study streamflow and water quality changes during severe droughts from 1998 to 2022 in the UCRB. Drought events in 2001-2002, 2012, 2018, and 2020-2021 resulted in annual median streamflows declines of approximately 15% (N=202 sites), water temperature increases by 9% (N=36 sites), and specific conductance increases by 5% (N=23 sites) relative to normal adjacent years. The greatest declines for all variables occurred in summer months, with notable reductions in peak flows. The impacts are transient, recovering to normal levels post-drought. Catchments with higher freshwater withdrawals, deciduous forest cover, mean air temperatures, bedrock residence times, and silt or clay content had greater declines in flows, while those with higher baseflow index, soil water retention and infiltration capacity exhibited greater resilience. These findings highlight the importance of catchment attributes and human activities in determining their vulnerability to droughts in mountainous regions

Mountainous headwater streams are sustained by both snowmelt‐driven streamflow and groundwater discharge in the Upper Colorado River Basin. However, predicting headwater stream discharge magnitude and peak flow timing is challenging in mountainous terrains, where snowmelt rates vary with vegetation type and elevation, and heterogeneous subsurface physical properties influence groundwater storage and its release. We used a model‐data integration approach to investigate the roles of snowmelt and subsurface structure in stream discharge and groundwater level. We ran an ensemble of 100 integrated surface‐subsurface hydrologic models for a mountainous headwater catchment near Crested Butte, Colorado, USA. We also evaluated and calibrated these models against observed data sets, including snow depth measurements using distributed temperature probes, stream discharge, and groundwater levels. Calibration with multiple data sources using neural density estimators has further constrained uncertainty in subsurface properties and snowmelt rates. Results indicated that observed slower snowmelt rates in evergreen forests delayed the peak flow and baseflow onset. In upstream areas with lower subsurface permeability, water was stored within the subsurface but was not released as interflow or shallow groundwater flow, and thereby not contributing to downstream streamflow during recession limb periods. Double peaks in groundwater occurred in areas with spatial subsurface heterogeneity, in our case due to the contrast between granodiorite and Mancos shale. These process‐based insights into groundwater and snowmelt dynamics in mountainous headwaters will help improve predictions of headwater hydrology.

Mountainous headwater streams are sustained by both snowmelt-driven streamflow and groundwater discharge in the Upper Colorado River Basin. However, predicting headwater stream discharge magnitude and peak flow timing is challenging in mountainous terrains, where snowmelt rates vary with vegetation type and elevation, and heterogeneous subsurface physical properties influence groundwater storage and its release. We used a model-data integration approach to investigate the roles of snowmelt and subsurface structure in stream discharge and groundwater level. We ran an ensemble of 100 integrated surface-subsurface hydrologic models for a mountainous headwater catchment near Crested Butte, Colorado, USA. We also evaluated and calibrated these models against observed datasets, including snow depth measurements using distributed temperature probes, stream discharge, and groundwater level. Calibration with multiple data sources using neural density estimators has further constrained uncertainty in subsurface properties and snowmelt rates. Results indicated that observed slower snowmelt rates in evergreen forests delayed the peak flow and baseflow onset. In upstream areas with lower subsurface permeability, water was stored within the subsurface but was not released as interflow or shallow groundwater flow, and thereby not contributing to downstream streamflow during recession limb periods. Double peaks in groundwater occurred in areas with spatial subsurface heterogeneity, in our case due to the contrast between granodiorite and Mancos shale. These process-based insights into groundwater and snowmelt dynamics in mountainous headwaters will help improve predictions of headwater hydrology.