Here we compare the four systems’ budget results for residence time, balance of catchment and oceanic nutrient loading, NEM and net denitrification. The NEM and net denitrification results are discussed in context of a global dataset of LOICZ mass balance budget values and other literature. We discuss the implications of contrasting catchment loading across the systems and seasons for ecosystem stressors associated with NEM state (including lowered O2, pH) and denitrification (eutrophication) and conclude by considering the utility of the budgeting results for coastal management.
Tasman Bay had annual mean residence time of 44 days (range: 13–77 days) which was consistent with Harris (1990) who used a method based on salinity and river flow, for an estimate of 2 months. There are no published residence times for Golden Bay. For the Firth, the methods of Plew et al. (2018), based on the modified tidal prism modelling of Luketina (1998), were used to estimate residence time, yielding a time of 12 days. This is a shorter time than we estimated (annual averages of 24 day in 2000–2001 and 21 days in 2012–2013). The shorter time was explicable, however, in that the Firth water body definition used by Plew et al. (2018) placed the Firth boundary considerably inshore of our boundary and yielded a considerably smaller estuary volume (5 km3) than in our study (16 km3). There are no published residence time values for the Hauraki Gulf. Our residence times determined by mass balance were thus consistent with available published values obtained using other methods. Furthermore, the published values confirmed the validity of our selection of cases to exclude from further analysis, based on their unrealistically short calculated residence times.
The Balance of Catchment and Oceanic Nutrient Loading to the Systems
River DIN loadings to Golden and Tasman Bays were much less than to the Firth, while offshore loading of DIN to Golden and Tasman Bays was much greater than to the Firth. The latter was a function of the greater DIN levels in the Cook Strait source waters for the bays (~ 2.6 μmol) than for the marine source waters for the Firth (i.e. the Hauraki Gulf: ~ 0.6 μmol). Averaged exchange rates (Vx) for Golden and Tasman Bays with their shelf waters were also greater than between the Firth and the Gulf (about 1.9-fold) (Table 2). Thus, Cook Strait shelf waters seaward of Golden and Tasman Bays dominated supply of N to the bays, in setting their nutrient stock levels. Similar strong shelf exchanges were found for Hauraki Gulf, although on occasion (winter 2013) the Firth exported large proportions of loading to the Gulf (Table 2). This probably arose because of winter light limitation of N uptake by Firth primary producers (see below), combined with high river loading (Figs. 3 and 5). Golden and Tasman Bays and Hauraki Gulf were found to export DON on average, consistent with their often autotrophic metabolism.
For the Firth, the 2000–2001 and 2012–2013 budgets showed averages of ~ 85% of its DIN contributions were catchment-derived (all sources), with remainders arising from mixing with the adjacent Hauraki Gulf. The Firth consumed PON on a net basis, to balance its denitrification sink. When PON was included in the Firth N fluxes, the average N contributions to Firth system fluxes from the Firth catchment were ~ 51%.
In the following sections, we consider the consequences of these contrasting nutrient source balances for the key biogeochemical functions of NEM and denitrification.
NEM, Seasonality and Land Use History
Golden and Tasman Bays usually had positive or near neutral (p − r), indicating net autotrophy or near-balanced NEM. The seasonal (p − r) values for the two bays were similar and were near the modal value of (p − r) estimates derived from ~ 150 budgets in the global LOICZ budget database (Fig. 8a) compiled from data in Smith et al. (2010). The (p − r) values for Hauraki Gulf also lay near the modal value.
The approximately balanced metabolisms of Golden/Tasman Bays and Hauraki Gulf contrasted with the Firth (Fig. 8a), which was often net heterotrophic. Unlike Golden and Tasman Bays and Hauraki Gulf, DIN loading to the Firth was dominated by catchment input (Table 2). Catchment DON loading to the Firth was also much greater (3- to 5-fold) than from Golden/Tasman Bays and Hauraki Gulf catchments. In addition, it apparently received substantial input of PON from the Gulf, contributing to its heterotrophy.
In addition to the oxidation of imported organic material, the loading of substantial DIN to the Firth suggests that this was also involved in the net heterotrophy observed. The mechanism by which this occurs was indicated by the seasonality of NEM in the Firth (Fig. 7) and consideration of the primary production cycle in the Hauraki Gulf/Firth system. Net primary production (NPP) measured at the NIWA outer Firth monitoring site (Fig. 1) increases in spring and is maximal in summer (Chang et al. 2003; Zeldis and Willis 2015), but with onset of summer water column stratification (see Fig. 9c) and decreased river loads (Fig. 3), nutrients are reduced at rates faster than their re-supply. By autumn and early winter, nutrient limitation becomes intense, slowing production in the upper water column. Thus, while production and respiration are continuous processes the year around, their seasonal balance changes, becoming increasingly net heterotrophic later in the production season.
Zeldis et al. (2015) presented moored chl-a fluorometry data from the upper and lower water columns at the NIWA Firth monitoring site (Fig. 1) that showed deepening of the phytoplankton distribution in autumn, as the cells senesce and sink under nutrient stress. Zeldis et al. (2015) also showed that the inner Firth supports the highest chl-a and dissolved organic matter concentrations, and the most intense seasonal DIC production (in autumn) of the whole Firth system. This area also has a higher residence time than the outer Firth, which is exposed to the Hauraki Gulf (Ferreira et al. 2005). Consequently, as well as having high exposure to riverine DIN and organic N loading, the inner Firth supports the greatest potential for retention of the resulting phytoplankton and the respiration of its breakdown products.
That succession of net production and net respiration that occurs in the Firth is shown by time series of O2 data (Fig. 9a) at the long-term NIWA monitoring site (Fig. 1). This 15-year, three-monthly time series showed that the whole water column was generally well-oxygenated (> 90% saturation) in winter and spring but that starting in late summer and autumn (when temperature was maximal: Fig. 9b), there was usually DO depression in the lower 20 m of the water column (often between 70 and 60% saturation although with extreme events as low as < 40%). The water column tended to be density-stratified at these low DO times (Fig. 9c). The changing seasonal balance between production and respiration described above therefore coincides with cycles in O2 (Fig. 9) and DIC (Table 2) concentration, as also described by Sunda and Cai (2012) and Wallace et al. (2014).
The seasonal O2 and stratification patterns (Fig. 9) resemble those seen in a number of eastern USA estuaries including Chesapeake Bay (Kemp et al. 2005; Kemp et al. 1992; Scully 2016) and Narragansett Bay (Wallace et al. 2014) where seasonal stratification is a necessary (but not sufficient) driver of low O2. The onset of seasonal hypoxia in Chesapeake Bay is set by the timing of seasonal physical stratification, but the severity and extent of hypoxia are more closely related to the variability in the production cycle including eutrophication effects (Hagy et al. 2004). Although oxygen depression is more severe in these very heavily loaded estuaries than in the Firth, the systems share this sensitivity to stratification and the primary production cycle.
It was notable that while the winter 2012–2013 budget showed strong heterotrophy in the Firth, there was typically little expression of O2 depletion in winter in the long-term time series (Fig. 9a). This is explicable in that in winter, the Firth water column is actively physically mixed (Fig. 9c), equilibrating its O2 with the atmosphere. This was similar to dynamics described by Wallace et al. (2014) and Ianson et al. (2016) who described stoichiometric de-coupling between DIC concentration (which we index using P in the budgeting) and O2 saturation, driven by physical mixing.
The history of the Firth’s catchment agricultural development helps explain its accentuated late season heterotrophy. For the developed catchments of the Firth (the Hauraki Plains: Fig. 1), point and diffuse (agricultural) human sources now contribute about 8 and 70%, respectively, of TN load to its major rivers, with ‘natural’ sources the remainder (Vant 2013). Thus, while the Firth is now heavily catchment N loaded (Fig. 6), prior to its land use intensification, ocean-side loading was likely to have contributed a larger percentage to a much lower overall nutrient load. This is consistent with other works showing dominant effects of agricultural intensification on nutrient export from NZ landscapes (Cooper and Thomsen 1988; Quinn and Stroud 2002; Snelder et al. 2017). Although land use intensification has continued in the Hauraki Plains in the last 20 years, its development was largely complete by the 1990s (Kelly et al. 2017). Intensification has increased nutrient leaching rates to the major Hauraki Plains’ rivers (Waihou and Piako Rivers), making them among the most heavily N-loaded rivers in NZ (Snelder et al. 2017). The findings that ca. 78% of Hauraki Plains’ loading is anthropogenic, combined with the heavy catchment-side N loading shown by our budgeting (51% for TN, 85% for DIN), lead to the conclusion that historical land use intensification has substantially increased the N loads to the Firth (by 42 and 66%, for TN and DIN, respectively). Snelder et al. (2017) reported anthropogenic increases of Waikato region river TN yields of 82% over natural levels. Given that about half the total Firth TN load is from its catchment, our estimate of 42% is consistent with this.
The catchments of Golden and Tasman Bays, on the other hand, have experienced much less historical increase in TN load (38%) in their rivers, starting from a lower natural level (2.3 vs 14.9 Gg year−1 for Firth rivers) (Snelder et al. 2017). The lack of large catchment inputs for Golden and Tasman Bays probably allows the relatively neutral (p −r ) they exhibit, which is instead fuelled by mixing of externally mineralised inorganic nutrients into the bays from offshore. The Hauraki Gulf is similar, with relatively neutral (p − r) and ocean-side nutrient supply usually dominant. This conclusion is consistent with the observation by Wallace et al. (2014) that the intensity of DIC elevation across systems was a function of their catchment loading rates.
Physical water exchange appears to have a role in conditioning susceptibility to strong (p − r) effects across the systems. Golden and Tasman Bays and Hauraki Gulf all had higher physical exchange rates (Vx: m3 year−1) than Firth of Thames (Table 2). This suggests that the fact that the Firth is a relatively enclosed water body rather than a more open system further accentuates its potential for trophic impact (Ferreira et al. 2005), especially toward the head of the estuary. However, in some seasonal cases, the differences in physical exchange between the more open systems and the Firth were not large (e.g. for Tasman Bay, and for Hauraki Gulf in 2000–2001—order of 25%). This suggests that other factors, including the absolute size of the load and its inorganic/organic makeup, are also influential (Caffrey 2004). The Firth has been substantially enriched with increased catchment loading, and its present day expression of low O2 and elevated DIC is consistent with the responses of estuaries elsewhere that have undergone cultural eutrophication, as described by Provoost et al. (2010) for European estuaries and Wallace et al. (2014) in eastern USA.
Denitrification and Trends in Water Quality
In general, the systems all exhibited net denitrification (Table 2), with areal (m−2) rates consistent with near modal values in the LOICZ meta-analysis literature (Fig. 8b). Our estimated rates (0.1 to 1.5, mean 0.7 mmol N2 m−2 day−1) were also within the ranges for San Francisco Bay (0.6–1.0 mmol N2 m−2 d−1) and 20 other coastal systems listed by Cornwell et al. (2014), which were largely between 0 and 4 mmol N2 m−2 day−1.
Our measures of denitrification as a proportion of total N exports for the four coastal systems ranged from ~ 0.1 to 0.7, with mean of 0.51, indicating the critical importance of this regulating ecosystem service (MacDiarmid et al. 2013) in terms of ameliorating eutrophication. A mean estimate of 39% of denitrified proportions was reported by Seitzinger (1988) across 10 coastal systems.
An exception to the findings of uniform net denitrification was in winter in Golden and Tasman Bays (Table 2), which showed strong net N fixation. We consider this an unreliable result for the following reasons. It was driven by strong net export of DON in winter, accompanied by weak fluxes of DOP (Table 2). This imbalance may be explained by research showing that organic P is more readily mineralised to its inorganic form than is organic N (Álvarez-Salgado et al. 1997; Garber 1984; Hopkinson et al. 2002; Lønborg et al. 2009). When residence times are short, this can result greater hydrographic export of DON relative to DOP (Álvarez-Salgado et al. 1997). In our study, the short winter water residence times in both Golden and Tasman Bays (Table 2) were consistent with this explanation. Strong N fixation was further discounted by findings from phytoplankton samples taken during our surveys that showed no N fixing species (e.g. Trichodesmium spp.) present in either bay (J. Zeldis, unpubl. data). Seagrass beds can fix N (Welsh 2000), and Golden Bay supports seagrass (Zostera muelleri) beds at Farewell Spit (Fig. 1). However, their areal extent is only 1.6% of the area of Golden Bay (Battley et al. (2005) and only 0.02% of Tasman Bay (Robertson and Stevens 2012). Mean areal N fixation rate of temperate Zostera species (Welsh 2000) was 77 mmol m−2 year−1, which is less than the denitrification values we budgeted for the bays in the other seasons (over much larger areas: Table 2). These results showed that bay-wide net N fixation was not occurring in winter in Golden and Tasman Bays.
For the Firth, denitrification averaged over 2012–2013 was only 42% of the average rate in 2000–2001 (Table 2). All seasonal budgets in common to the two surveys (spring, summer and autumn) showed lower values in 2012–2013 than in 2001–2002. These decreased rates coincided with a period of significant increases in N and phytoplankton concentration in the Firth (Fig. 10). Nutrient and phytoplankton concentrations sampled at the NIWA monitoring site in the 15-year, three-monthly time series showed DIN increased by 5% per year (p < 0.01) from 1998 to 2013 (seasonal Kendall trend tests (Jowett 2014)). DON increased by 2% per year (p < 0.02) and micro-phytoplankton cell numbers increased by 7% per year (p < 0.00). DIP, on the other hand, had no trend (not shown).
Zeldis et al. (2015) examined whether this enrichment arose from changed offshore oceanographic circulation patterns. Trend analysis of salinity and temperatures recorded during the three-monthly surveys showed no evidence of higher salinity and cooler water (with respect to freshened and warmer coastal water), which would indicate increased upwelled supply of N from offshore, nor did satellite sea-surface temperature for the adjacent shelf exhibit a long-term cooling trend. Southern Oscillation Index (SOI) state is related to wind directions and upwelling frequency in the NE North Island (MacDiarmid et al. 2013). SOI varied between negative and positive states over frequencies considerably higher than the long-term increasing trend observed in the DIN time series, including negative (upwelling favourable) phases when DIN concentrations were low and positive phases when they were high. It therefore did not predict the nutrient trends.
Freshwater flows from Firth rivers also showed no time trends, indicating that rate of estuarine circulation would not have changed. This was confirmed with current meter data from the NIWA monitoring site showing no trends in rates of exchange of outer Firth waters with the seaward Hauraki Gulf. Firth river water quality changes also were not a strong driver of the trends in enrichment. Published water quality trends between 1993 and 2012 (Vant 2013) showed, for TN, improvements in the Piako River and deteriorations in the Waihou River (Fig. 1). The Waihou River (which contributed approximately 60% of the TN loadings to the Firth in the mass balances) showed increasing TN slopes of 0.5, 1.0 and 1.7% per year at its gauging sites. These rates of change were considerably less than the trends in DIN in the marine record (Fig. 10). Thus, there was little evidence of either ocean- or catchment-side change that would explain the decadal enrichment observed in the Firth.
The key role of denitrification in loss of N from the Firth system described above means that decreases in denitrification efficiency could have important effects on water quality. Studies in the USA and Australian coastal systems have shown that denitrification efficiency drops with degree of organic enrichment in sediments (Cook et al. 2004; Eyre and Ferguson 2009; Hale et al. 2016; Harris et al. 1996; Kemp et al. 2005; Kemp et al. 1990). In the Firth, the mass balance analyses showed reduced denitrification rates (by 58%) between 2000–2001 and 2012–2013, consistent with the observed Firth water-column enrichment over the 15-year period (Fig. 10).
Shipboard studies of Firth benthic nutrient, oxygen and carbon fluxes showed sediment oxygen consumption (SOC) in 2003 (Giles et al. 2007) and 2012 (Zeldis et al. 2015) increased by 1.5–1.7-fold between the dates (although this comparison was compromised by differences in sampling seasons (summer vs autumn)). Published relationships between denitrification efficiency and CO2 efflux rates indicate that efflux rates > 1000 μmol CO2 m−2 h−1 correspond with lowered denitrification efficiency (Cook et al. 2004; Eyre and Ferguson 2009). In 2012, measured SOC (assumed equal and opposite to CO2 efflux) in the inner Firth averaged 1800 μmol O2 m−2 h−1, where efficiency was predicted to drop to around 50% (ibid.).
These findings were consistent with the hypothesis that water quality in the Firth has declined over the last decade because of reduced denitrification efficiency. The accelerated availability of nutrients to primary producers would be a positive feedback on eutrophication and organic enrichment, as indicated by the coincident increase in phytoplankton abundance (Fig. 10c). Although the trends of increasing nutrient and phytoplankton concentrations do not appear to be a consequence of contemporaneous increases in inflowing nutrients, we hypothesise that they may be a consequence of cumulative effects of its historically high nutrient loading.
Reliability of Mass Balance Budget Approach
It is important to examine reliability of our mass balance approach with respect to assumptions implicit in the mass balance methodology and with respect to the size of effects in NEM and denitrification across the systems.
Potential errors in the LOICZ mass balance approach were described by Webster et al. (2000). They identified errors introduced by neglecting seasonality in modelling the system. We included seasonality, so minimising this error. They identified the importance of setting the locations of marine sampling down-estuary of major nutrient and freshwater inputs, which we also achieved. Finally, they signified the importance of vertical structure in modelling of highly stratified estuaries, noting that two-box representations (in the vertical) would be most appropriate in highly stratified systems (in their case, with salinity varying 3-fold between upper and lower strata). They concluded that single-layer representations were most appropriate in weakly stratified systems, such as the systems we modelled.
Phosphorus demand (or supply), related to P adsorption/desorption processes, can be significant in highly turbid or anoxic systems, potentially biasing estimates of its non-conservative behaviour (Swaney 2011). However, these conditions do not apply to the systems investigated here, which have relatively clear water columns (Gall and Zeldis 2011; MacKenzie and Gillespie 1986) and oxic conditions near the sea bed (Fig. 9) and within the surficial sediments (Giles et al. 2006), (C. Depree, NIWA pers. comm., 2015).
To further assess the reliability of the conclusions regarding NEM (p − r) and denitrification (nfix − denit) made from mass balance, in Supplementary S2, we considered the sensitivity of those estimates to variability in their components. Those results suggest that cumulative error involved in the (p − r) calculation for its most sensitive parameters (river flow, marine DIP concentrations, Redfield estimates) is on the order of 15%. This is considerably less than the observed contrasts in (p − r) between, for example, the Firth and Golden Bay (Fig. 7). The cumulative errors involved in the (nfix − denit) calculation (Supplementary S2) for its most sensitive parameters (river flow, marine DIN concentrations, Redfield estimates) suggest that (nfix − denit) error could be on the order of 25–30%. Given this scale of error, we consider that the Firth budgetary estimates made between the 2000–2001 and 2012–2013 showing a 58% decrease in denitrification support the hypothesis that ongoing, high nutrient enrichment has decreased denitrification efficiency. This is supported by the findings over the same period of increased water column DIN, DON and micro-plankton biomass (Fig. 10), and results of SCOC experiments described above.
Our conclusions could be further tested using independent approaches toward ‘whole-system’ description (field survey and experiment-based, spatially and temporally resolved dynamic modelling, mass balance) and checking whether their conclusions converge, similar to the cross-comparison of methods by Gazeau et al. (2005). Toward this, we are presently undertaking dynamic biogeochemical modelling of the Hauraki Gulf/Firth of Thames system. Also, our NEM results will be compared with field data on Hauraki Gulf/Firth carbonate system dynamics (pCO2, DIC, alkalinity, pH) that were sampled coincidentally with the 2012–2013 mass balance budgets (Zeldis et al. 2015).
Implications for Resource Management of the Systems
We believe our budget findings are useful for resource management of Golden and Tasman Bays, Hauraki Gulf and Firth of Thames. First, the nutrient budgets of the systems provide inventories and comparisons among their catchment nutrient loading terms (Table 1). For Golden and Tasman Bays, submarine groundwater discharges were small relative to surface water (river) inputs in terms of system loading (13–14% for NO3−), even though the aquifers have concentrated nutrients. Wastewater was a minor nutrient contributor relative to rivers (1% for NO3−) in Golden Bay but a larger contributor (10%) in the more urbanised Tasman Bay adjacent to Nelson city. There are many reasons to manage nutrient levels in groundwater and wastewater at local scales, but it is useful for managers to know that these sources are not important at bay-wide scales in Golden and Tasman Bays relative to river inputs. In contrast, for the Hauraki Gulf adjacent to NZ’s largest city, Auckland, wastewater inputs were comparable to inputs from its catchments (Table 1). Also, direct atmospheric inputs were found to exceed those from both wastewater and catchment sources. This was the first evaluation of atmospheric deposition for NZ coastal waters, and this finding, along with uncertainties surrounding its estimation (Supplementary S1), indicates that it should receive further research.
Irrespective of their catchment inputs, oceanic mixing usually dominated the nutrient budgets of Golden and Tasman Bays and the Hauraki Gulf. On occasion, however, mixing with the inshore waters of the Firth and by extension, its agricultural catchment, can affect the Gulf’s nutrient climate. The influences of agricultural nutrient inputs on the Firth and the wider Hauraki Gulf are a key concern for its regional managers (Kelly et al. 2017). The results therefore signify to managers that understanding both offshore and onshore circulation dynamics is critical for predicting the provisioning and regulating ecosystem services of the four systems (MacDiarmid et al. 2013).
We found that NEM of Golden and Tasman Bays and the Hauraki Gulf was approximately balanced between net autotrophy and net heterotrophy. This suggests that these bays may not exhibit significant levels of stressors associated with strong heterotrophy (high DIC, reduced O2 and pH: Wallace et al. 2014). To date, there has been no direct sampling of the carbonate systems of Golden and Tasman Bays (e.g. surveys of pCO2/DIC/alkalinity) and relatively little study of their O2 climates. These would further inform the implications of their relatively neutral NEM for DIC and O2-associated stressors. As noted, surveys of carbonate dynamics of the Hauraki Gulf/Firth system have been carried out and will be reported in a future publication (see also Law et al. 2017).
The mean percentages of denitrification-to-total N export were high across the four systems, showing that denitrification was an important term in their N budgets. Clearly, maintaining this regulating ecosystem service is critical in terms of preventing N enrichment. Results presented here, that evidence of declining denitrification efficiency is coincident with N enrichment of the Firth, is a concern of managers of Hauraki Gulf/Firth of Thames regional marine resources (Kelly et al. 2017). Negative feedbacks between eutrophication and denitrification show that from this standpoint alone, eutrophication should be avoided. This should include management of fine sediment loading (mud), because of the known association of increased sediment muddiness and enrichment by solid and aqueous phase nutrients (%OM, %N) (Engelsen et al. 2008). Morrison et al. (2009) showed that fine sediment accumulation rate (SAR) in NZ estuaries has accelerated often by an order of magnitude with agricultural and forestry intensification. This includes the Firth, which now sustains high SAR values (7 mm year−1 in the inner Firth: Pritchard et al. 2015).
Calculations using these mass balance budgets have been used to place anthropogenic N sources arising from fish farms and sinks arising from mussel farms in ecological context (Zeldis et al. 2011; Zeldis 2005; reviewed in Swaney et al. 2011). Example findings were that under scenarios of maximal mussel farm development in the Firth, phytoplankton depletion by the farms in terms of N and C were very small with respect to system-wide fluxes, while proposed fish farm N loadings to the system were significant proportions of bay-wide denitrification (~ 25%). Other reports (Kelly et al. 2017; Zeldis 2008) to central and regional NZ government have used the mass balance information to advise on risks associated with biogeochemical stressors.
The present day loadings received by Golden and Tasman Bays and Hauraki Gulf represent much less of an ‘anthropogenic legacy’ than those entering the Firth, in terms of effects of catchment nutrient loading on trophic state. The findings for land-ocean nutrient balance, NEM and denitrification signal that understanding ocean-coastal interactions of Golden and Tasman Bays and Hauraki Gulf is essential for predicting their ecosystem services, whereas managing water quality of catchment inputs will be most effective in sustaining Firth of Thames ecosystem health. Thus, mass balance budgeting has provided inventories of important state variables including catchment loading, the balance of land and ocean nutrient loading and important biogeochemical responses, relevant to coastal management.