Results and discussion

Storm generated surface runoff resulted in 18 ponding events at the Hauraki site and 19 ponding events at the Awahou site during this 12-month study. Ponding events occurred most often during the winter months compared to the other seasons (Fig. 2) . Water samples were analysed for 13 of these ponding events at the Hauraki site, and 14 events at the Awahou site, since not all ponding events generated flow rates high enough to trigger auto-samplers. Discharge samples were collected during 10 events at the Hauraki site, and 13 events at the Awahou site, since not all events generated discharge flows to be sampled due to leakage and soil infiltration.

Concentrations

The annual SS inflow concentration was 17 g m-3 at the Hauraki site, and 96 g m-3 at the Awahou site. Concentrations peaked in the winter at the Hauraki site, while there was no clear temporal trend for inflow concentrations at the Awahou site (Fig. 2) . Event inflow concentrations did not tend to correspond to event runoff magnitudes and varied widely between events.
The annual MFP SS discharge concentration was 28% lower than inflows at the Hauraki site, and 29% lower at the Awahou site. These results suggest that DBs effectively facilitated sedimentation during ponding, and attenuated deposited sediments in the ponding area. Discharge concentrations were lower than inflows during 7 of the 10 events analysed at the Hauraki site, and 10 of the 13 events analysed at the Awahou site (Fig. 2 ). Inconsistencies in concentration treatment efficiencies were observed between and within event types at both sites (Table 2 ).
The wide range of concentration treatment efficiencies observed in this study were influenced by multiple factors. Treading damage, deposited animal excreta and previously deposited sediments in the ponding area could have contributed to SS discharged from the DB that were not accurately accounted for by the pro rata correction of the contributing catchment area and affected the concentration treatment efficiency results.
Variations in particle sizes delivered to the DBs, which were not measured in this study, could have also been a contributing factor to the varying concentration treatment efficiencies observed between events and the sites in this present study. Large particles that have greater densities settle more readily than smaller sized particles with lower densities, which would be less likely to settle and more likely to be remobilised and discharged from the DBs (R. W. McDowell et al., 2003). A greater number of larger particles, which require more energy to mobilise and transport, could have been delivered to the DBs during higher magnitude runoff events, particularly Overflow Events, and could be partially responsible for SS concentrations decreasing during all Overflow events in this study, while this was not the case for all Non-Overflow Events (Table 2 ).
During Overflow Events at both sites, the SS concentrations between the portions of inflow contributing to ponded water going over the top of the upstand riser and emergency spillway (i.e. overflow discharge) and the overflow discharge, did not decrease to the same extent as the concentration decreased between the between overflow discharge and the following release discharges (Table 3) . These results are somewhat surprising since we would expect the decanting of the uppermost layer of water performed by the upstand riser (Fig. 1 ) and emergency spillway would be highly effective at preventing SS discharge. The data suggests however, that longer pond residence times experienced by the release discharge compared to the overflow discharge (an average of 14 hours between overflow discharge and the following release discharge at both sites) allowed for greater sedimentation to occur. Longer retention times have been found to increase sediment removal efficiencies in a study of sedimentation ponds (Brown et al., 1981).
The data suggests ponding runoff for longer than the currently suggested 3 days could achieve greater concentration treatment efficiencies, however, this could risk damaging pasture productivity. Removing the upstand riser/outlet valve/discharge pipe installation (Fig. 1 ), and allowing all ponded water to infiltrate the soil, would avoid discharging the bottommost portion of ponded water where SS are likely to concentrate and/or be stirred up by turbulence when unplugging the outlet valve to drain the pond. Also, placing the outlet valve 10 cm above ground level would enable a small portion of the ponded water left after draining the pond to infiltrate the soil. This later revision would prevent the discharge of a lower portion of ponded runoff, and would decrease the area potentially affected by prolonged inundation, compared to avoiding the release procedure entirely. Lastly, approaches to achieve greater SS concentration treatment efficiencies could include the use of flocculants.

Yields and loads

The key finding of this study was that impeding stormflow with DBs effectively attenuated 789 kg and 1280 kg of SS at the Hauraki and Awahou sites, respectively. The 51% and 60% decrease in annual SS loads discharged from the DB catchments was a result of the DBs’ ability to facilitate sedimentation, which often decreased event SS concentrations, and soil infiltration, which decreased the volume of runoff discharged from the DB. These results suggest DBs should be effective at reducing P losses from pastures in the Lake Rotorua catchment, due to the high proportion of sediment bound P delivered to the lake (Hamill, 2018).
Annual SS inflow yields were 28 kg ha-1 at the Hauraki site, and 109 kg ha-1 at the Awahou site, although runoff inflow yields were greater at the Hauraki site than the Awahou site. The annual SS inflow yields at both sites in this study were much lower than the estimated annual SS yields entering streams in the same area of the Lake Rotorua catchment from May 2010 to May 2012 (479-741 kg ha-1 y-1) (Abell, Hamilton, & Rutherford, 2013). Factors affecting the catchments’ hydrological responses to precipitation, including antecedent soil conditions and localised differences in storm rainfall intensity and duration, and differences between the catchment sizes, geomorphologies, and land use and management factors, affected runoff generation and erosion (Dougherty, Fleming, Cox, & Chittleborough, 2004), and likely accounted for the SS inflow yield differences between study sites in this present study and the results reported by Abell et al. (2013).
At both sites during this present study, runoff and SS inflow yields were lowest in the spring and increased during each subsequent season, peaking during the winter period (Fig. 3 ). This was not surprising, as the contributing catchment is grazed by dairy cattle soil treading damage and erosion is likely to increase when soils are wet (R. W. McDowell et al., 2003) . Additionally, greater SS yields tended to correspond with greater runoff yields, particularly during the high runoff magnitude Overflow Events (Fig. 4 ). These results are consistent with other studies that found greater runoff magnitudes tend to mobilise and transport greater quantities of sediments and nutrients from pastures in New Zealand (Cooke, 1988; Smith & Monaghan, 2003) and the Lake Rotorua catchment, specifically (Abell et al., 2013; Dare, 2018).
The results of this study suggest the DBs at both sites were able to consistently decrease SS loads discharged from the DB catchments, even during rare, high runoff magnitude events. The greater inflow magnitudes during Overflow Events at the Hauraki site contributed to a greater portion of runoff undergoing overflow discharge, and consequently, the difference in the portion of inflow undergoing soil infiltration and SS yield treatment efficiencies between the sites during these high magnitude events (Fig. 5) .
The results from the high magnitude Overflow Events emphasize the importance of being effective during these high magnitude events since they were responsible for 61% and 66% of the annual SS inflow loads at the Hauraki and Awahou sites, respectively, and 39% and 59% of annual SS yields attenuated. This finding is important to note since large storm events have been found to be responsible for the majority of SS loading to streams in the Lake Rotorua catchment (Abell et al., 2013).
Although DBs effectively attenuated SS loads during Overflow Events, these large magnitude events still generated 84% of the annual SS discharge yields at the Hauraki site, and 77% at the Awahou site. These results are likely related to the majority of annual runoff discharges also occurring during Overflow Events at both sites (Table 4) . These results suggest that reducing runoff discharges from DBs by facilitating soil infiltration played a key role in effectively decreasing SS loads, which highlights the importance of optimising DB designs to maximise soil infiltration of ponded runoff and avoiding excess overflow discharge.
The ability of DBs to consistently decrease SS loads, particularly during high magnitude runoff events, is also significant because some land management strategies may be overwhelmed by extreme hydrologic conditions (Kleinman et al., 2006; R. W. McDowell & Sharpley, 2002; McKergow et al., 2007). Sediments deposited as a blanket across the relatively wide DB ponding area was observed during this present study, and likely contributed to the consistency in DB performance during this study (Table 1 ) (McKergow et al., 2007). Importantly, it is likely sediments deposited in the DB ponding area will be attenuated for longer periods of time than other mitigation strategies, such as buffer strips and treatment wetlands, that have more concentrated sediment deposition areas and are susceptible to flushing during high magnitude events (McKergow et al., 2007). The ability of the DB to impede the stormflow of each runoff event reduced the kinetic energy of water, which enables the transfer and/or remobilisation sediments, and particularly the ‘first-flush’ of the initial runoff, could have had a major influence on the DBs’ ability to decrease SS loads transported in surface runoff during each event in this study (Bieroza et al., 2019).
The role of soil infiltration in annual SS yield treatment efficiencies is important to note. The concurrent study by (Levine et al., In review) found lower infiltration rates in the ponding area than outside the ponding area (Table 1 ). The decreased soil infiltration rates in the ponding area could be due to a combination of factors including treading damage in the lower lying and wetter areas (R. W. McDowell et al., 2003), the large volumes of water moving through the soil causing stress and deterioration in soil structure, and sediments deposited in the ponding area clogging soil pores and/or forming a less permeable surface soil layer (Hendrickson, 1934; Reddi, Ming, Hajra, & Lee, 2000; Rice, 1974). Therefore, the data suggests infiltration rates, and consequently SS yield treatment efficiencies, may be highest in newly constructed DBs, and may decrease over time. Also, soil infiltration rates, and consequently SS yield treatment efficiencies, could decline more rapidly if DBs are located in areas where erosion is more intense and a greater amount of sediments are deposited in the ponding area compared to this present study, if deposited sediments are responsible for infiltration rate declines. Causes of declining soil infiltration rates in the ponding areas, as well as methods of mitigating soil infiltration rate declines, such as aerating the pond area soils or employing subsoil amendments, should be investigated.
During this study, event discharge concentrations were lower than inflows more often than higher, and SS yield treatment efficiencies were greater than runoff yield treatment efficiencies (Fig. 5 ). These results indicate that sedimentation facilitated by impeding stormflows with DBs caused lower SS discharge concentrations. Therefore, DBs would likely be able to decrease SS discharge yields in areas where soil infiltration rates and pond storage to catchment area ratios are lower than those in this present study. Other factors influencing the proportion of runoff infiltrating the soil and sediment sizes delivered to the DBs would affect yield treatment efficiencies.
Lastly, revising the DB design to remove the upstand riser/outlet valve/discharge pipe installation would prevent SS leak and release discharges, which could be remobilised and delivered to downstream surface waters in subsequent runoff events. We calculated that applying this revised DB design would have prevented an additional 147 kg of SS from being discharged from the Hauraki site, and an addition 216 kg at the Awahou site, increases of 16% and 14% of the annual load attenuated at each site, respectively. The costs and benefits of revising the DB design should be considered since the increased inundation period could damage pasture productivity.