Impacts of stormwater infiltration on downslope soil moisture and tree water use

The study was conducted around a large, vegetated infiltration basin, with mature, deep-rooted forest upslope and downslope of it, using a combination of monitoring techniques to understand interactions between the infiltrated stormwater and the surrounding trees. As presented conceptually in figure 1, this monitoring comprised:

• (a)  
  Continuous hydrometric monitoring of depth to water table, conducted in three zones: 1. within the infiltration basin (herein referred to as 'basin'), 2. upslope, to act as a reference, as it is unaffected by the basin, and 3. downslope, which acts as a treatment, hypothesized to be affected by the basin.
• (b)  
  Sampling for isotopic composition in 1. shallow groundwater bores, 2. sapwood cores taken from trees upslope and downslope, 3. soil cores within 2 × DBH (diameter at breast height) of the sampled trees, 4. infiltration basin water and 5. rainfall. Analysis of isotopic composition aimed to provide insights into the sources of water being used by the trees (i.e. rainwater, soil water or groundwater). The isotopic composition (²H and ¹⁸O) sampling was conducted over three separate sampling campaigns: 29 November 2018 ('spring'), 21 February 2019 ('summer') and 16 May 2019 ('autumn'), with the exception of rainfall that was sampled monthly.
• (c)  
  Continuous measurement of sap flow in the upslope and downslope trees was undertaken from 1 June 2018 to 31 May 2019 to characterize the temporal patterns of tree water use. Sap flow provides a direct measurement of the movement of water within a tree, and thus provides the most direct estimate of tree evapotranspiration. In doing so, it provides an indirect measure of tree water availability.

Zoom In Zoom Out Reset image size

Combined, this monitoring programme allowed us to deduce the extent to which stormwater infiltrated from the basin became available to downslope vegetation and how that availability varied seasonally. All data used to produce this article are publicly available on the Open Science Framework at https://osf.io/4hysc/.

The study site and basin have been fully described [20], including its water balance (infiltration) and water quality treatment performance. Wicks Reserve Infiltration Basin is located near Melbourne, in South-Eastern Australia (Lat-37.857°, Long 145.321°). It was constructed in 2011 and has a surface area of 1800 m², receiving stormwater from 5 ha of impervious surfaces, making up 15% of the total catchment area. Stormwater is filtered as it percolates down the 0.8 m deep filter media of the basin. The basin is densely planted with a mix of swamp grasses, sedges, reeds, rushes and shrubs. The basin's infiltrated water flows through the subsurface towards Dobsons Creek, a peri-urban stream ∼80 m downslope of the basin. The dominant trees upslope and downslope of the basin are mature Messmate (Eucalyptus obliqua) and Swamp Gum (Eucalyptus ovata), with a dense understory vegetation comprised of native shrubs and ferns.

Rainfall depth was measured continuously using a Davis Instruments Tipping Bucket Rain Collector. Groundwater levels were monitored every 6 min using capacitance probes (Odyssey Capacitance Water Level recorder). Short periods of missing data due to battery failure were recorded. Probes were calibrated before and after deployment, with typical accuracy under field conditions of ±20 mm.

To monitor infiltrated stormwater and its interaction with the shallow water table, a network of monitoring bores was established in and around the basin (figure 1). A well within the infiltration basin (Basin) allowed water level measurement and water sampling. Downslope monitoring bores formed a transect 0 m (DS0), 7 m (DS7), 17 m (DS17) and 24 m (DS24) downslope (DS) from the edge of the basin. Several bores were also installed upslope of the basin, as a reference, unaffected by the basin, but remained dry, precluding water level monitoring or sampling for isotopic composition. Monitoring bores were 1.5–4 m deep and did not reach the full depth of underlying regolith, whose depth is not known, but soil coring to 5 m indicated very low hydraulic conductivity at this depth.

The stable isotopes of water (²H and ¹⁸O) are well suited to trace the movement of water between storages or through its environs [21]. Water from the basin and its surroundings was therefore characterized using isotopic composition. Samples were collected from soil, surrounding trees (sapwood), rain, groundwater, and water within the filter media of the basin. Rainfall was sampled monthly over (1 June 2018–31 May 2019) in accordance with methods in the Global Network of Isotopes in Precipitation [22].

All water, soil, and plant samples were collected in 12 ml exetainer vials (Labco Limited, Product No. 737W). Water samples were filled completely to avoid air column whilst soil and plant samples were filled only up to ¾ of the vial following sampling protocols from the Global Institute for Water Security (GIWS). Sample vials were sealed immediately with parafilm to prevent leakage and evaporation. Duplicate samples were collected as part of the quality assurance and quality control procedure to assess the precision of field samples. The samples were transported to the laboratory in a cool box and stored at 4 °C before being shipped to the GIWS at the University of Saskatchewan, Canada, for ¹⁸O and ²H analysis, where all isotopic analysis occurred. The GIWS uses Los Gatos Research liquid and vapour water off-axis integrated-cavity output spectroscopy machines, with an accuracy of <±1.0‰ for δ²H and ±0.2‰ for δ¹⁸O. Isotopic compositions of are expressed in standard delta (δ) notation, in units of per-mille (‰), as the differences relative to Vienna Standard Mean Ocean Water [23].

Prior to groundwater sampling, each bore was pumped continuously to purge at least three casing-volumes of water. Samples were collected using a single-use syringe rinsed three times with sample water, filtered through a 45 µm syringe filter and stored in 12 ml exetainer vials. Samples were collected from the filter, bores DS0, DS7, DS17, DS24 and the rainfall collector.

Soil cores were collected from depths of 0.1 m, 0.3 m, 0.5 m, 0.7 m and 0.9 m from upslope and downslope of the basin. At each location, soil cores were taken around the base of the three trees instrumented with sap flow sensors, each tree representing a different size range (a) small (26–35 cm), (b) medium (36–45 cm), (c) Large (46–60 cm), which are representative of the composition of the forest. Individual trees within each category were selected randomly. Soil water was first extracted using Cryogenic Vacuum Extraction. All soil samples were weighed prior to extraction, samples (in vials) were frozen in liquid nitrogen for 20–25 s, and then evacuated down to 0.6 mbar. Extraction was conducted at 180 °C for 15 min. Post extraction, all samples were re-weighed and then placed in an oven for 24 h at 100 °C. Samples were then re-weighed and any weight loss was calculated for extraction efficiency. Soil water was then analysed for δ²H and δ¹⁸O.

Sapwood cores were collected at breast height from six trees upslope and six trees downslope of the basin. This included the three trees instrumented with sap flow sensors at each site, and an additional three individual trees at each site. The cores were initially collected using a 300 mm Haglof three threaded increment borer (more suited to softwood) and a core extractor. Subsequent cores were collected using an 11 mm hollow punch coring tool. Several cores were required to obtain a sufficient volume per tree. Water was extracted using cryogenic vacuum extraction following the same protocol as for soil samples. Plant water was then analysed for δ²H and δ¹⁸O.

Two SFM1 sap flow sensors (ICT International) were installed on the north and south side of each of six trees (three upslope and three downslope) approximately 2–3 m from the ground. DBH of these trees were 30 cm (small), 42 cm (medium) and 53 cm (large) for the upslope trees, and 36 cm (small), 43 cm (medium) and 57 cm (large) for the downslope trees. Sap flow sensors utilized the heat ratio method [24] which calculates sap flux density (rate of movement of sap through the tree xylem, cm³ cm⁻² h⁻¹) from heat velocity (cm h⁻¹). Heat velocity was calculated as the average ratio of temperature change between thermistors inserted in the tree xylem 5 mm above and below a central heater, following the emission of a heat pulse. To ascertain moisture content, sapwood was cored adjacent to the sensor and sapwood separated from the core by identifying the boundary between sapwood and heartwood with methyl orange dye. Sapwood was then weighed fresh, dried in an oven at 80 °C for three days and weighed dry. To correct for potential probe misalignment, climate data from a nearby Bureau of Meteorology climate station (Scoresby Station ID:086104) were filtered to identify conditions where sap flux density should equal zero [25] and data were linearly offset accordingly.

Due to failure of sensors installed on the north side of several trees, we used data from sensors installed on the south side of trees only. Periods of less than four hours with missing data were gap-filled using a linear interpolation. Days where at least one sensor had more than four hours of missing data were excluded. Two periods of missing sap flux data (August–September 2018 and April–May 2019) were due to failure of power supply to the sensor network. Daily sap flux density (J_S) was calculated as the sum of hourly values. To compare the response of trees to seasonal changes in water availability, we further calculated a normalized sap flux as the proportion of maximum daily sap flux (J_Smax) of each tree (J_S /J_Smax, %). Seasonal means were calculated for winter (June–August 2018), spring (September–November 2018), summer (December 2018–February 2019), and autumn (March–May 2019) to assess statistical differences among trees upslope and downslope from the basin in each season.

All statistical tests were conducted in MINITAB 17. To assess the effect of season and location on tree δ¹⁸O isotopes, sap flux, or normalized sap flux, both single and two-factor analyses of variance (ANOVA) were performed. Assumptions of normality were evaluated and determined to be satisfied after natural log transformation was applied to sap flux. To evaluate the nature of the differences between season and location for tree isotopes, statistically significant ANOVA was followed-up with Fisher's least significant difference (LSD) post-hoc tests at the 95% confidence interval. Despite a slight skew in the standardized residual plot for tree δ¹⁸O isotopes, Fisher's LSD results on raw data and square-root-transformed data were near-identical; we thus report only results based on the raw data.

This study revealed a significant impact of stormwater from a constructed infiltration basin on downstream water fluxes and on the water use of downstream vegetation. The trees upslope from the basin used shallow soil water during spring (November) 2018, following recent substantial rainfall and when the landscape was starting to dry. However, these trees moved to using deep soil water in February 2019 under dry conditions, before demonstrating a use of shallow soil water in May 2019, following a strong decline in PET and recent autumn (March–June) rainfalls (i.e. reduced water demand).

The downslope tree water use follows a distinctly different pattern. In spring (September–November), the depth of water use is unclear because there is little vertical change in isotopic composition. In February 2019, during much drier conditions, upslope trees are clearly accessing deeper water than downslope trees and upslope trees are accessing water at about 0.5 m. In May 2019, as water use reduces, the downslope trees use deeper water while upslope trees have changed back to using shallower water. A critical difference here is that the water table has started to rise downslope, suggesting water was likely more available at depth than at the surface, despite recent rain.

Together, these data suggest the trees are using the most available water under drier conditions. Both upslope and downslope of the basin, the residual deeper water is being used in February 2019's dry conditions. Recently replenished shallow soil water is used at the upslope site in the wetter May 2019 period, while lateral flow replenishing the water table from the infiltration basin provides water for use by downslope trees. Vertical water fluxes (i.e. from direct rainfall) are sufficient to explain the upslope behaviour, but lateral flow (i.e. water from the infiltration basin) is required to explain the downslope behaviour characterized by higher observed sap flux per tree.

In terms of lateral movement of water, the isotopic signatures in the soil water also provide evidence of different behaviour upslope and downslope of the basin. There are clearer relationships with depth at the upslope sites and the isotopic signature is more depleted upslope. At the end of the cool, wet season, after significant periods of high water table, the relationship between isotopic composition and depth has been obliterated downslope of the basin and is close to the basin's isotopic composition. This is most likely due to lateral flow of water from the infiltration basin overprinting the locally (i.e. direct rainfall) infiltrated water during the high water table period through winter–spring (June–November) of 2018.

The infiltration basin allowed downslope vegetation to continue using shallow soil water during times when upslope soil water had become limiting. This is evidenced by contrasting water use by trees upslope and downslope of the basin. The sap flux data shows upslope trees are using less water (figure 2(D)) and, when normalized by the maximum observed sap flux per tree (figure 2(E)), the temporal pattern also suggests that low water availability likely controls tree water use for long periods at the upslope sites. This is evident both through winter when water tables were deep at the upslope sites and shallow downslope, and in summer–autumn (December–May) when water tables deepened, but were still present downslope of the basin (figure 2(B)). Together with the isotope data, this suggests that trees use water from deeper layers in the profile downslope of the basin (figure 4), confirming that the trees are accessing significantly more water due to lateral contributions from the infiltration basin.

Overall, these results paint a picture of both the infiltration basin being a significant source of water for lateral flow and the trees being a significant sink of water coming from the basin, particularly in summer (December–February). Water from the infiltration basin supports higher transpiration by downslope vegetation than would otherwise have occurred, thereby contributing to the maintenance of urban vegetation in a water-limited environment [26]. While this was not an objective in the original infiltration basin site selection, it nevertheless suggests that infiltration remote from streams could contribute significantly to such an objective, with subsequent benefits in terms of provision of shade [27], amelioration of the urban heat island [28] and the range of other social and psychological benefits provided by urban vegetation [29].

Conversely if the purpose of the infiltration basin were to contribute to stream baseflows, the hydrologic, isotopic and tree physiological evidence presented in this study suggests that the influence of vegetation will likely impede that objective, drawing on subsurface water originating from the infiltration basin. Indeed, water use of mature trees can be very large in such situations, where water availability is non-limiting. Tree water use will be substantially greater than the water use of grass or other low-level vegetation [30]. In this context, siting stormwater infiltration closer to the receiving stream, or where there were relatively few trees between the infiltration basin and the stream, would be more effective, at the expense of the broader landscape-scale and urban amenity benefits.