Finding where groundwater actually enters a stream

Between storms, a mountain stream is mostly groundwater. That flow — base flow
— is what keeps the water cool in August, keeps the channel connected so fish
can move, and carries whatever chemistry the rock it came through gave it.

It does not arrive evenly. Groundwater reaches a stream through **preferential
discharge** at particular seeps, springs, riparian wetlands and stretches of
streambed, and elsewhere the stream loses water back to the ground, leaving
disconnected pools at low flow.

Measuring that at the scale of a whole headwater network is the problem the
USGS Next Generation Water Observing System set out to solve in the Neversink
Reservoir watershed, in the headwaters of the Delaware River Basin.

Four methods, four scales

The streambank

The USGS has long measured the hydraulic gradient from groundwater to stream
at some Super Gage sites. NGWOS added heat as a tracer to find where the
discharge actually is along the West Branch Neversink (site 01434498) and the
mainstem (site 01435000), then instrumented those seeps and springs directly
— sensors on live connections rather than on a regular grid.

The network

51 temporary monitoring stations installed in autumn 2020, each pairing
streambank air temperature against stream water temperature. Water samples in
spring and autumn went for stable water isotopes and dissolved radon.

The temperature pairing is the elegant part. Without groundwater, stream
temperature tracks the air. Where groundwater comes in, it does not — so the
slope of stream temperature against air temperature is a map of groundwater
influence, made with two cheap sensors.

The rock underneath

Soil thickness routes the flow paths, and the route determines both the
temperature and the chemistry of what emerges. From 2019 to 2022 the USGS took
passive seismic measurements at over 80 locations along stream corridors
and hillslopes to find depth to bedrock — which turned out to differ between
the East and West Branches, and improved the near-surface geologic maps.

The map itself

Surficial sediment and bedrock exposure were mapped at 1:24,000 or finer. To
do it, a deep-learning image segmentation model was trained to pick out
exposed and covered bedrock from a 1-metre lidar elevation model, and a map of
minimum sediment thickness was derived from it.

Why bother with all four

Because a model of groundwater discharge is only as good as what calibrates
it, and the things that calibrate it are of different kinds: tracers of
groundwater age, which tell you the flow path and therefore the water
quality and how long the base flow will last; direct measurement at the seep;
temperature across the network; and the geology that routes it all.

The strategy is being repeated. NGWOS is running it at several more Delaware
River Basin watersheds across different hydrogeology, and in the upper
Colorado River Basin. All the data is public.

Sources

Written from the U.S. Geological Survey Fact Sheet *A multiscale approach for
monitoring groundwater discharge to headwater streams by the U.S. Geological
Survey Next Generation Water Observing System*, a work of the United States
government in the public domain. Preferential discharge from Briggs and Hare
(2018); ecosystem dependence from Miller and others (2016); depth-to-bedrock
methods from Glas and others (2021); the lidar segmentation model from Odom
and Doctor (2022).