Watershed Structure Mapped Through Remote Sensing and Geophysics

Several concepts recur throughout the findings below. Foresummer drought sensitivity describes how much peak plant productivity drops in years with a dry early growing season (May–June). Aspect (the compass direction a slope faces) and microtopography (small bumps and depressions) create local differences in sun exposure and water pooling. Functional zonation refers to dividing a watershed into zones with similar snow, soil, and plant behavior so that scientists can model large landscapes tractably. Finally, machine learning methods — including Long Short-Term Memory (LSTM) neural networks that learn from time series data — are increasingly used to fuse these diverse measurements into predictions of snow water equivalent, evapotranspiration, and ecosystem response.

Foundational work

Early studies in the East River Watershed near Gothic established that microtopography and aspect exert powerful, fine-scale control on mountain ecosystems. Falco and colleagues showed that fusing LiDAR with surface geophysics could predict plant community distributions at submeter resolution, finding a strong inverse relationship between soil electrical conductivity and slope and demonstrating that soil moisture — driven by microtopography — determines where species like veratrum and sagebrush establish (Falco et al., 2019). In parallel, Tran and colleagues used the Community Land Model to quantify evapotranspiration across the same watershed, showing that 55% of annual precipitation is returned to the atmosphere through evapotranspiration and that transpiration alone accounts for roughly half of that flux (Tran et al., 2019).

These studies set the template for an integrated approach: combine satellite and airborne remote sensing with ground-based geophysical imaging and physically based models to characterize watershed structure and function across scales.

Key findings

A central finding across this body of work is that watershed heterogeneity is not random — it is organized by topography, aspect, and bedrock structure, and these controls can be mapped systematically. Wainwright and colleagues demonstrated that hillslopes are a natural unit for capturing watershed-scale heterogeneity, and that unsupervised clustering of LiDAR, hyperspectral, and airborne electromagnetic data can divide a watershed into zones with characteristic snow, plant, and subsurface signatures (Wainwright et al., 2022). Uhlemann and colleagues then showed that above-ground features visible to remote sensing (vegetation patterns, microtopography) covary with below-ground bedrock properties measured by geophysics, meaning that surface observations can be used to infer hidden subsurface structure across entire watersheds (Uhlemann et al., 2022).

Drought response emerged as a second major theme. Using nearly two decades of Landsat NDVI data, Wainwright and colleagues found that foresummer drought sensitivity is spatially heterogeneous and is positive across 94% of vegetated area, meaning earlier snowmelt and drier early summers consistently reduce peak plant productivity, especially at lower elevations and in grasslands (Wainwright et al., 2020). Subsequent monitoring by Dafflon and colleagues confirmed the mechanism at the hillslope scale: heavy snow years deliver more infiltration and shallower groundwater, supporting higher greenness, while low-snow years with early bare ground produce muted growing seasons (Dafflon et al., 2023). Worsham and colleagues extended this snow-centered view to forests, showing that peak snow water equivalent and snow disappearance rate are the dominant abiotic controls on subalpine conifer forest structure across the watershed (Worsham et al., 2025).

A third thread is the role of machine learning in turning sparse measurements into watershed-scale predictions. LSTM neural networks trained on SNOTEL records produced highly accurate forecasts of snow water equivalent (Meyal et al., 2020), and a hybrid deep-learning approach successfully estimated evapotranspiration and ecosystem respiration in data-sparse mountain settings using only readily available inputs like air temperature, precipitation, and NDVI (Chen et al., 2021). Hybrid data-model approaches have also mapped soil thickness at half-meter resolution, revealing that northeast-facing hillslopes carry deeper soils than southwest-facing ones (Yan et al., 2021).

Current frontier

Early work from 2019–2020 established the feasibility of fusing remote sensing and geophysics in the East River. Recent studies since 2022 have shifted toward finer-resolution monitoring, predictive frameworks, and explicit uncertainty quantification. New distributed temperature profiling systems now resolve soil and snow temperature profiles at unprecedented vertical density and accuracy (Dafflon et al., 2022), while electrical resistivity monitoring combined with hydrological modeling has revealed that subsurface bedrock and vegetation gradients create starkly different hillslope flow paths over distances of tens of meters — shallow lateral flow on thin-soil upper slopes versus vertical infiltration and groundwater dynamics on colluvial lower slopes (Uhlemann et al., 2024). Researchers are also addressing previously overlooked physics: Feldman and colleagues showed that terrain-emitted longwave radiation contributes roughly 22% of surface longwave flux in the Upper Colorado River Basin, an effect commonly omitted from atmospheric models (Feldman et al., 2022).

The frontier is increasingly about using these integrated datasets to guide the next round of measurements. Wainwright and colleagues recently developed a machine-learning workflow that uses remote sensing and hydrological model outputs to identify optimal locations for ecohydrological monitoring, finding that soil moisture variability outperforms static variables like elevation as a predictor of drought sensitivity (Wainwright et al., 2024). Probabilistic hypothesis testing is also entering the field, allowing researchers to formally evaluate competing geological interpretations against geophysical data (Miltenberger et al., 2021).

Open questions

Many important questions remain. How will the tight coupling between snowpack, soil moisture, and plant productivity change as snow disappears earlier and rain replaces snow at mid-elevations? How transferable are watershed zonation schemes developed in the East River to other basins with different bedrock, vegetation, and climate? Can subsurface properties — fracture density, soil thickness, bedrock weathering — be predicted reliably from surface remote sensing alone, or will geophysical ground-truthing always be required? How can three-dimensional radiation effects, lateral subsurface flow, and root water uptake be efficiently represented in the large-scale models used for water resource planning? Addressing these questions over the next decade will require sustained, co-located measurements of snow, soil, vegetation, and subsurface properties, paired with machine-learning frameworks that honestly quantify uncertainty when extrapolating beyond the conditions in which they were trained.

References

Chen, J., Dafflon, B., Tran, A. P., Falco, N., & Hubbard, S. S. (2021). A deep learning hybrid predictive modeling (HPM) approach for estimating evapotranspiration and ecosystem respiration. Hydrology and Earth System Sciences. →

Dafflon, B., et al. (2023). Advanced monitoring of soil-vegetation co-dynamics reveals the successive controls of snowmelt on soil moisture and on plant seasonal dynamics in a mountainous watershed. Frontiers in Earth Science. →

Dafflon, B., Wielandt, S., Lamb, J., et al. (2022). A distributed temperature profiling system for vertically and laterally dense acquisition of soil and snow temperature. The Cryosphere. →

Falco, N., Wainwright, H., Dafflon, B., et al. (2019). Investigating Microtopographic and Soil Controls on a Mountainous Meadow Plant Community Using High-Resolution Remote Sensing and Surface Geophysical Data. Journal of Geophysical Research: Biogeosciences. →

Feldman, A. F., et al. (2022). Three-Dimensional Surface Downwelling Longwave Radiation Clear-Sky Effects in the Upper Colorado River Basin. Geophysical Research Letters. →

Meyal, A. Y., Versteeg, R., et al. (2020). Automated Cloud Based Long Short-Term Memory Neural Network Based SWE Prediction. Frontiers in Water. →

Miltenberger, A., et al. (2021). Probabilistic Evaluation of Geoscientific Hypotheses With Geophysical Data. Application to Electrical Resistivity Imaging of a Fractured Bedrock Zone. Journal of Geophysical Research. →

Tran, A. P., Rungee, J., Faybishenko, B., Dafflon, B., & Hubbard, S. S. (2019). Assessment of Spatiotemporal Variability of Evapotranspiration and Its Governing Factors in a Mountainous Watershed. Water. →

Uhlemann, S., Dafflon, B., Wainwright, H. M., et al. (2022). Surface parameters and bedrock properties covary across a mountainous watershed: Insights from machine learning and geophysics. Science Advances. →

Uhlemann, S., et al. (2024). Variations in bedrock and vegetation cover modulate subsurface water flow dynamics of a mountainous hillslope. Water Resources Research. →

Wainwright, H. M., et al. (2024). Model and remote-sensing-guided experimental design and hypothesis generation for monitoring snow-soil-plant interactions. Frontiers in Water. →

Wainwright, H. M., Steefel, C., Trutner, S. D., et al. (2020). Satellite-derived foresummer drought sensitivity of plant productivity in Rocky Mountain headwater catchments: spatial heterogeneity and geological-geomorphological control. Environmental Research Letters. →

Wainwright, H. M., Uhlemann, S., Franklin, M., et al. (2022). Watershed zonation through hillslope clustering for tractably quantifying above- and below-ground watershed heterogeneity and functions. Hydrology and Earth System Sciences. →

Worsham, M., et al. (2025). Abiotic influences on continuous conifer forest structure across a subalpine watershed. Remote Sensing of Environment. →

Yan, Q., et al. (2021). A hybrid data-model approach to map soil thickness in mountain hillslopes. Earth Surface Dynamics. →