Quantifying Interannual Variations in Groundwater Fluxes Within a Braidplain Aquifer Using Active‐Distributed Temperature Sensing

The problem

Recharge from braided rivers helps sustain groundwater resources and river ecosystems around the world (Coluccio and Morgan, 2019). In the Canterbury Plains of New Zealand, braided rivers are estimated to contribute more than half of the surface water flow (Canterbury Water, 2009).

Between a braided river and the deeper regional aquifer is a shallow layer of gravel, reworked by floods, that stores river water. Wilson et al. (2024) called this layer a braidplain aquifer (BPA). Flow between the river channels and the BPA is called parafluvial flow. Most methods for measuring exchange between a river and a shallow aquifer give an estimate at a single point, or over a longer distance at a single time (Coluccio and Morgan, 2019). Our goal in this study was to quantify how parafluvial flow changes over several years, as river flow and the shape of the riverbed change.

What we did

Our site is a 1 km reach of the upper Waikirikiri Selwyn River in Canterbury. Two 100 m fibre optic cables had been installed across the river by horizontal directional drilling (HDD) (Banks et al., 2022). One cable, HDD1, passes beneath the main channel, and the other, HDD2, beneath a smaller secondary channel. They are buried at depths of 0.56 to 5.55 m.

We heated copper conductors in each cable at 15 watts per metre of cable for 20 minutes. From late October 2021 we extended this to 1 hour. A distributed temperature sensing (DTS) instrument recorded the temperature along the optical fibre every 0.25 m. We then fitted a model of heat transfer by conduction and advection (transport by moving water) to the measured temperatures. This gave the specific discharge, the volume of groundwater crossing each square metre of aquifer per day, expressed in metres per day. The method gives the magnitude of the flux. We used only the saturated sections of each cable. An earlier study at the site put the uncertainty of a single survey at 6% (Sai Louie et al., 2024). Between November 2020 and August 2023 we ran 28 surveys, 14 on each cable, at roughly monthly intervals.

What we found

River flow at a gauge 9.5 km upstream ranged from 0.67 to 225 m³/s (cubic metres per second) over the study. A flood on 31 May 2021 peaked at 225 m³/s, about 90 times the median flow of 2.51 m³/s. The main channel moved south and became about 11 m wider. Several mature willow trees that grew on the bank were in the middle of the channel after the flood. Because surveys were not possible during floods, river flow at survey times varied by about one order of magnitude.

Averaged over the saturated part of each cable, specific discharge ranged from 3.5 to 5.9 m/day on HDD1 and from 2.4 to 4.5 m/day on HDD2. The average over both cables was 4.2 m/day. By our arithmetic, this is 4,200 litres of water crossing each square metre of aquifer each day. The interquartile range (the spread of the middle half of the surveys) was 20.4% of the median on HDD1 and 13.3% on HDD2. Specific discharge showed a negligible relationship with river flow on HDD1 and a moderate negative relationship on HDD2. The results suggest no strong seasonal influence, although we ran only two surveys in summer.

Changes to the riverbed after the May 2021 flood likely had the largest effect on specific discharge. On HDD1, the average after the flood was 30.5% higher than before. The first survey on HDD1 after the flood, about five weeks later, showed an increase under the left margin of the main channel. The next survey showed an increase under the right margin. Specific discharge beneath both margins remained above its pre-flood level for the rest of the study, although it fell in the last four or five surveys.

After the flood, the sediment above HDD1 was thinner at the margins of the main channel. The increases in specific discharge are likely related to this thinning. When the main channel moved, it also appears to have connected to a remnant braid. We suggest this contributed to the increase under the left margin. The secondary channel above HDD2 changed little in shape, yet on HDD2 the average fell by 17.6% after the flood. This fall could be due to the secondary channel switching from losing water to the aquifer to gaining water from it. However, we could not verify this. The flood destroyed the river monitoring equipment.

Limitations

Our method treats the flux as one-dimensional and does not resolve its direction. In reality, flow beneath the river is multidirectional. A numerical model of the site puts the horizontal flux at about an order of magnitude greater than the vertical flux (Di Ciacca et al., 2024). Our estimates are therefore likely dominated by horizontal flow. The parts inside the cable are not arranged symmetrically about its centre. This likely affects our estimates by an amount that we have not quantified. It may bias the absolute values of specific discharge, but the method can still identify changes along the cable and over time.

We measured at one site. At sites with a different geomorphology or flood history, the processes we identified may be stronger, weaker or absent. Horizontal cables beneath a river are costly to install. Vertical or steeply inclined cables would cost less and could help extend these findings to a whole river reach.

Glossary

A braided river has several active channels on a mobile bed of gravel.

A braidplain aquifer (BPA) is a shallow layer of flood-reworked gravel beneath and beside a braided river that stores river water.

An aquifer is a body of rock or sediment that stores groundwater and lets it flow.

Recharge is water that enters an aquifer, here from a river.

Parafluvial flow is the flow of water between a river channel and the BPA, through the sediment beside and beneath the channels.

Specific discharge is the volume of groundwater flowing through each square metre of aquifer per unit time, in metres per day.

Distributed temperature sensing (DTS) measures temperature at closely spaced points along a fibre optic cable.

Horizontal directional drilling is a method of drilling a near-horizontal hole underground, here beneath a river, to install a cable.

Conduction is the transfer of heat through a material, and advection is the transport of heat by moving water.

A saturated sediment has its pores completely filled with water.

The margins of a channel are the zones beneath its left and right banks, looking downstream.

A remnant braid is a former river channel that is no longer active.

A losing channel loses water to the aquifer beneath it, and a gaining channel receives water from it.

The median is the middle value of a set, and the interquartile range is the spread of the middle half of the values.

An order of magnitude is a factor of ten.