Identifying dissolved oxygen variability and stress in tidal freshwater streams of northern New Zealand

Abstract

Tidal streams are ecologically important components of lotic network, and we identify dissolved oxygen (DO) depletion as a potentially important stressor in freshwater tidal streams of northern New Zealand. Other studies have examined temporal DO variability within rivers and we build on this by examining variability between streams as a basis for regional-scale predictors of risk for DO stress. Diel DO variability in these streams was driven by: (1) photosynthesis by aquatic plants and community respiration which produced DO maxima in the afternoon and minima early morning (range, 0.6–4.7 g/m³) as a product of the solar cycle and (2) tidal variability as a product of the lunar cycle, including saline intrusions with variable DO concentrations plus a small residual effect on freshwater DO for low-velocity streams. The lowest DO concentrations were observed during March (early autumn) when water temperatures and macrophyte biomass were high. Spatial comparisons indicated that low-gradient tidal streams were at greater risk of DO depletions harmful to aquatic life. Tidal influence was stronger in low-gradient streams, which typically drain more developed catchments, have lower reaeration potential and offer conditions more suitable for aquatic plant proliferation. Combined, these characteristics supported a simple method based on the extent of low-gradient channel for identifying coastal streams at risk of DO depletion. High-risk streams can then be targeted for riparian planting, nutrient limits and water allocation controls to reduce potential ecological stress.

Appendix

WAIORA model

WAIORA uses the ‘approximate delta method’ of McBride and Chapra (2005), which is adapted from the delta method of Chapra and Di Toro (1991). The model enables the reaeration coefficient, primary production rate, and respiration rate to be calculated from single-site oxygen monitoring over a diurnal cycle (Wilcock et al. 1998). The rate of change of DO can then be calculated as:

$$ {\text{dDO}}/{\text{dt}} = k\left( {{\text{D}}{{\text{O}}_{\text{s}}} - {\text{ DO}}} \right) + P{ }-{ }R $$

where, DO is dissolved oxygen, t is time, k is the reaeration coefficient, DO_s is saturation DO, P and R are instantaneous rates of photosynthetic oxygen production and respiration, respectively.

A typical diurnal cycle was produced for calibration of the WAIORA model (Table 4) by averaging measured DO data for each time of day (data loggers were set to record at the same times each day). The effect of tidal variation was removed prior to calculating these diurnal averages by omitting periods of saline intrusion (as indicated by a sudden change in conductivity or temperature). The observed reaeration coefficient is calculated by WAIORA using the time lag between solar noon and the DO maxima (see McBride and Chapra 2005 for methods). Photosynthesis and respiration parameters are also derived from DO measurements (see McBride and Chapra 2005 for methods). For the purposes of modelling, photosynthesis and respiration were assumed to change proportional to flow (1) as follows:

Table 4 Parameters produced in calibrating WAIORA for each site to model the change in DO with flow

$$ {\text{Respiratio}}{{\text{n}}_i} = {\text{survey respiration }} \times {\text{ survey flow}}/{\text{flo}}{{\text{w}}_i}. $$

(1)

Sensitivity analysis indicated superior predictive performance when using flow compared to depth.

WAIORA models the effect of change in flow on DO by adjusting the reaeration coefficient (k) with average depth and velocity for the reach as follows:

$$ k{\text{ at flow }}i = {\text{survey }}k{ } \times {\left( {{\text{velocit}}{{\text{y}}_i}} \right)^{{0.{5}}}}/{\left( {{\text{dept}}{{\text{h}}_i}} \right)^{{{1}.{5}}}} $$

(2)

The necessary depth and velocity data were generated using the computer programme RHYHABSIM (River HYdraulic HABitat SIMulation; Jowett 1989), for which depth, velocity and offset were measured for 15 cross-sections (for survey details, see Wilding 2007). Repeat water-level measurements for each cross-section, and repeat flow measurements at one cross-section, were carried out over a range of flows in order to calibrate the rating curves for the RHYHABSIM model (typically within 6 weeks of the cross-section survey). Flow was measured at a more uniform cross-section, with velocity and depth measured at >20 offsets (more for complex profiles). Velocities were measured using a FLO-MATE Model 2000 and wading rod, typically at 0.6 times depth. The rating curve describes the relationship between water level and flow, and the model can predict depth, velocity and width for a given flow using water level and the channel profile. The RHYHABSIM model assumes gradually varying flow and a static water-level control.

The cross-section surveys were undertaken in lowland reaches immediately upstream of tidal influence, and hence upstream of the oxygen monitoring site (site separation, 0.25–3.0 km; median, 0.6 km). In a flowing stream, the DO concentration at any given point is a product of upstream conditions (reaeration and respiration). The lowland site therefore provided an approximation of average depth and velocity in the tidal reach at low tide and a direct representation of reaeration conditions for water prior to flowing through the tidal reach. Calibration of the WAIORA model was not possible for two sites (Whareroa, Waikanae) because the small diurnal peak in DO that did occur was observed prior to solar noon, rather than post-noon as required for the approximate delta method of McBride and Chapra (2005).

Flow requirements were standardised across sites using the predicted flow required to maintain a DO guideline of 4 g/m³, with flow expressed as a percent of MALF. The DO value of 4 g/m³ (average diurnal minima) was selected for use from various sources, including USEPA (1986) criteria: 7-day mean minimum for non-salmonid waters (see Wilding 2006 for more detail). This guideline is expected to provide adequate protection for more tolerant taxa in lowland and midland reaches. The average observed temperature was used to calibrate the model (Table 2), but predictions were standardised to an average water temperature of 20°C to improve resolution of tidal effects. In addition to the seven sites included in this study, flow requirements for DO were modelled for two additional tidal sites in the Coromandel (Opitonui and Whangamororo Rivers), as described in Wilding (2007).

WAIORA is a single-station model that assumes spatially uniform photosynthesis and temporally uniform reaeration (Jowett et al. 2004). Such assumptions are not likely to hold in tidal streams, so the validity of the WAIORA model was tested using repeat monitoring of the Whenuakite River. Four DO data logger deployments were carried out between 2004 and 2007 (summer/autumn), under contrasting flow and temperature conditions (Table 3). The model was calibrated using March 2005 observations and the WAIORA predictions based on this data were compared to observed daily minima for all four monitoring periods (Fig. 8). The predictions provide an acceptable approximation of observed DO concentrations for three of the four monitoring periods. The outlier (March 2007) represented an unusually dry period with few floods and the lowest mean spring flows on record (n = 32 years). The March 2007 outlier was followed by a flood (Tairua River recorded a 1.2 year return event with daily maximum at 33× median flow), and subsequent monitoring in April 2007 saw a return to predicted DO concentrations. The one outlier therefore demonstrates that WAIORA predictions were not robust against reduced flood frequency, which is expected to be a limitation of all DO models that do not incorporate inter-year macrophyte dynamics. This also emphasises the importance of disturbance regimes for stream ecosystems (see Sousa 1984; Poff et al. 1997; Riis and Biggs 2003).

Fig. 8

The predicted response of DO (average diurnal minima) to flow in the Whenuakite River was modelled using WAIORA, the lines representing predictions at three temperatures (15°C, 17.2°C and 20°C). For comparison, observed diurnal minima are plotted as points that represent four monitoring periods (Table 3)

Successful prediction for the other three monitoring periods indicates that the WAIORA predictions are robust against tidal influence and spatial heterogeneity, despite assumptions of uniformity for this single-station method. The WAIORA model is sufficiently robust to standardise DO concentrations for observed temperature and flow conditions during each monitoring period, and we therefore believe our use of the WAIORA model is valid for improving comparisons between streams.