Regulated rivers often have associated wetlands with declining ecological health due to reduced inundation frequency. One innovative option to improve the ecological condition of such wetlands is to use them as temporary off-river water storages, where the water used to inundate them is subsequently allocated to consumptive use. The hydrologic feasibility of this option has yet to be demonstrated. We investigated three physical aspects of a floodplain wetlands system that must be considered, relative to irrigation demand, to determine feasibility: (a) historical inundation frequencies and the effect of regulation and climate change, (b) natural storage volumes and enhanced volumes using retaining walls, and (c) estimated loss rates. We found that inundation frequencies are reduced under regulation and that this reduction is even greater for projected climate change scenarios. Natural volumes were found to be 5% of annual irrigation demand, increasing to 20% with retaining walls; a small proportion at the system scale but significant at farm scale, especially with opportunities for multiple fillings per season. Losses are estimated at 36%–63% of the initial volume, depending on timing of wetland inundation and drawdown. Careful consideration must be applied to issues of frequency and timing of inundation, drawdown rates, and connectivity when considering the ecological benefits of using wetlands as storages. Environmental benefits will be a trade-off with capital, operational, and maintenance costs and water pricing.
 River regulation often leads to a reduction in the frequency and duration of wetland inundation and has contributed to their decline and loss throughout the developed world [Vorosmarty et al., 2010]. It has been estimated that more than 50% of floodplain wetlands in Australian rivers may no longer be subject to inundation resulting in adverse ecological impacts for a range of freshwater organisms, such as fish, invertebrates, and plants [Kingsford, 2000]. With climate change scenarios predicting a warmer, drier future, such as that identified for the Murray-Darling Basin (MDB) in Australia [CSIRO, 2008], further reductions in natural wetland inundation and less available water for irrigation increases the need to optimize the efficiency of water use.
 This paper explores the innovative option that wetlands can be used as a distributed network of off-river storages, where the consumptive water allocation might first be employed for wetland inundation before subsequent extraction. Although this idea promotes the exchange of one “unnatural” hydrological regime (reduced inundation due to regulation) for another (regular inundation aligned to agricultural demands), the latter may provide a better environmental outcome while simultaneously providing an agricultural benefit. This pragmatic approach adopts the goals of “ecological engineering”: the restoration of human-disturbed ecosystems and the development of new sustainable ecosystems that benefit both humanity and the environment [Mitsch, 1993]. Wetlands are typically naturally ephemeral and have the potential to be artificially drawn down without excessive impacts on biota [Ning et al., 2012; Watkins et al., 2013]. The final determination of the feasibility of using wetlands as off-river storages requires ecological and economic risk analyses; however, the hydrologic viability (the subject of this study) must be demonstrated as a necessary first step.
 We present an approach to assessing the physical suitability of wetlands as storages by the spatial and temporal assessment of inflows, outflows, and storage under natural, regulated, and climate change scenarios. We used the Broken River, a subcatchment of the MDB, Australia, as our case study (Figure 1). We developed a repeatable method for determining wetland storage volume using light detection and ranging (lidar) to determine storage (natural and “enhanced” volumes, i.e., the latter incorporating bund walls) and investigated water loss rates from wetlands and the implications for water availability over the irrigation period. We used the results to address the following questions: (1) What is the natural timing and frequency of inundation for wetlands and what are the predicted effects of regulation and climate change? (2) Are wetland storage volumes of adequate size relative to irrigation demands? (3) What are the losses in wetlands due to leakage and evapotranspiration (ET) if filled in the peak flow season (winter-spring) and drained in the summer irrigation period? We also highlight the principal ecological, economic, and institutional considerations that need to be addressed to determine the ultimate feasibility of this option.
 The lower Broken River flows across alluvial plains where the channel is sinuous and contains numerous anabranches and paleo-channels creating a network of wetlands in the lower reaches. The Broken River is a medium-sized river relative to other comparable rivers in the MDB, and although rivers of increasing discharge and catchment area would be expected to have larger and more numerous wetlands that could be considered as storages, the level of consumptive use also generally increases with discharge (Table 1). The Broken River study provides both a method and an estimate of the volumes that can be available to supplement irrigation demand for a river of comparable size. The methods presented here are applicable to other lowland systems of similar anabranching morphology, and in theory, the results of this case study should be scalable to alluvial river systems with morphologically similar floodplain wetlands.
Table 1. Catchments of the Murray Darling Basin of Comparable Geomorphology to the Broken River, in Order of Increasing Unregulated Discharge (Data Sourced From CSIRO Sustainable Yields 2008)
Mean Annual Flow (Regulated) (×106 m3/yr)
Mean Annual Flow (Unregulated) (×106 m3/yr)
Percentage of Flow Diverted
Catchment Area (km2)
Mean Annual Rainfall (mm)
 The Broken River catchment has an average annual rainfall ranging from 550 mm in the lowland region to 1450 mm in the uplands, with a mean annual areal potential evapotranspiration (APET) rate between 950 and 1150 mm [Wang et al., 2001]. Peak flow occurs in late winter/spring with low flows through summer/autumn. Unregulated mean annual flow is 212 × 106 m3 and has been reduced to 175 × 106 m3 under a regulated system through extraction and increased evaporation. The main catchment storage (Lake Nillahcootie) has a capacity of 40 × 106 m3. Regulation has primarily increased summer flows while reducing some late autumn/winter flows [Murray-Darling Basin Authority, 2010; Vietz et al., 2013]. The central area of the catchment is used for grazing and mixed cereals with a large area of intensive irrigation and cropping. The catchment has a current surface water diversion limit of 57 × 106 m3/yr with the proposed new sustainable diversion limit being reduced to 51.4 × 106 m3/yr. Actual diversions as of 2010 totaled 17 × 106 m3/yr [Murray-Darling Basin Authority, 2010].
 Figure 2 outlines the method that we have employed in the hydrologic assessment of the Broken river wetlands, described in detail in the following section.
 The Broken River channel and floodplain was surveyed using lidar in late summer 2007 during long-term (12 year) drought conditions [Ummenhofer et al., 2009], yielding a digital elevation model with 1 m pixel resolution and elevation accuracy of ±0.15 m. The discharge at Casey's Weir on the day of the lidar survey was 14.01 × 103 m3/d—just below the 10th percentile of flow for the Casey-gauged record (for regulated flow). These conditions of low flow and minimal standing water provided the ideal opportunity to survey both the channel, wetland and floodplain morphology, since lidar in the infrared band cannot penetrate the water surface. 148 wetlands along 80 km of river were surveyed between Benalla and Shepparton (Figure 1). The digital elevation model was analyzed using ESRI ArcMap (version 9.3) in three stages to: (1) identify suitable wetlands for storage, (2) calculate natural storage volumes (Figure 3a) based on commence-to-flow (CTF) elevations, and (3) calculate enhanced storage volumes through the application of bund walls (Figures 3b–3d). We have analyzed wetlands showing an enhanced (i.e., with bund walls) volume of greater than 3 × 103 m3: a threshold based on the mean farm dam volume for the MDB [Nathan and Lowe, 2010]. Further details of the analysis procedure are contained in the supporting information accompanying this article.
 Field measurements of changes in the depths of three wetlands located on the Broken River were assessed between September 2010 and March 2011 using Odyssey pressure water level recorders (Odyssey Data Recording Systems, Christchurch, NZ). This period corresponds to the time where wetlands would typically be filled (during late spring rainfall) until use in the major irrigation demand period of late summer. The logger data were used to produce a relation curve of storage volume versus water surface elevation (Australian Height Datum, AHD) for each wetland. A Leica VIVA GNSS Rover differential GPS was used to accurately (±3 cm) determine elevations at reference points adjacent to each wetland. These points were used to determine water surface elevation of each wetland at a known time using a Sokkia 30R Series Total Station. This elevation was used to convert water level measurements to AHD and, thus, produce a time series of wetland storage volumes for the study period from the relation curve.
 Wetland CTF elevations were fitted to river chainage (i.e., the distance along the centre of the main river channel measured from the furthest downstream point) in a linear regression model. The interquartile range of model residuals was used to determine the range of discharges that would inundate wetlands at the field reference sites (Figure 1) using HEC-RAS (US Army Corps of Engineers, 2011) derived rating curves. This range was assumed to be representative of all wetlands along the reach with the lower bounds applying to wetlands with a lower CTF level that will accordingly have a higher frequency of inundation.
 A 100 year daily flow time series was modeled for the Broken River for the natural (predevelopment) hydrology and for two levels of regulation by George et al. . The natural series corresponds to unregulated flow (but with current land use); light regulation includes a 40 × 106 m3 upstream storage with a 30 × 106 m3/yr diversion; Heavy regulation includes a 240 × 106 m3 upstream storage with a 76 × 106 m3/yr diversion. The two regulated series were also modeled using the 2030 dry climate scenario used in the CSIRO Murray Darling Basin Sustainable Yields study [CSIRO, 2008], which predicts lower rainfall and higher temperatures for this region. A spells analysis of each time series was then conducted to determine inundation frequencies using the range of CTF discharges determined from the study sites. Spells were considered independent if they occurred four or more days apart.
 The time series of water level data from the monitored wetlands were used to evaluate their loss rate characteristics. Periods of continual stage recession below the CTF level (i.e., where wetlands were isolated from the river) were extracted and used to define mean daily loss rates for each month of the data record. The partitioning of the loss rates between seepage and evaporation was investigated by comparing the observed loss rates to long-term mean APET rates [Australian Bureau of Meteorology, 2002] and to pan evaporation rates from a nearby climate station (Tatura, Station 081049, 42 km west of instrumented wetlands). The APET was considered as a minimum bound to evaporative losses while the pan evaporation was considered as a maximum [Costelloe et al., 2007]. Where the observed loss rates exceeded the pan evaporation rate, leakage to groundwater is inferred. Mean monthly loss rates from the three representative wetlands for the monitoring period were used to estimate losses from a wetland with an estimated enhanced storage volume using bund walls.
 Complete data for irrigation demand along the Casey's Weir-Orrvale Weir reach of the river were available for 2001–2009 (from the Goulburn-Murray Water Corporation) with delivery volumes for 68 specific off-take locations (each assumed to supply individual farms). Because of drought restrictions on entitlements for the years 2007–2009, only the first 6 years of data were considered to determine demands more representative of long-term conditions.
3.1. Wetland Inundation Characteristics
 CTF elevation as a function of chainage was described by a linear relationship (R2 = 0.99) (Figure 4a) with an interquartile range of model residuals of ±0.7 m. Both field sites were predicted to inundate at a discharge of 6 × 106 m3/d with upper and lower bounds of 4 to 9 × 106 m3/d. This range of discharges was used in a spells analysis to model the effect of regulation and climate change.
 The frequency of wetland inundation is reduced under regulation and further reduced by climate change. Under natural conditions, wetland filling occurs most frequently in August (late winter) with a mean number of events per month of 0.64, 0.41, and 0.08 for 4, 6, and 9 × 106 m3/d flows, respectively (Figure 5a). Natural annual mean inundation frequency is 2.2, 1.3, and 0.4 events per year, respectively. Regulated conditions result in the reduction of the number of 6 × 106 m3/d high flow events that are likely to cause wetland inundation (Figure 5b). Only a small change is observed for the light regulation scenario (40 × 106 m3 storage/30 × 106 m3 demand), with only a 5% reduction in annual mean inundation frequency (1.2 per year). The heavy regulation scenario (240 × 106 m3 storage/76 × 106 m3 demand) has a more pronounced effect, with a 47% reduction (0.7 per year). Incorporating climate change in regulated scenarios reveals a large reduction in the number of 6 × 106 m3/d high flow events (Figure 5c); including the light regulation scenario (85% reduction in annual mean inundation frequency, 0.2 per year) and heavy regulation (96% reduction, 0.06 per year).
3.2. Wetland Volumes and Distribution
 The application of bund walls to the wetlands greatly enhanced the potential storage volumes available for irrigation purposes. Over the 80 km reach assessed, 148 wetlands were identified that demonstrated enhanced (i.e., with bund walls) storage volumes of greater than 3 × 103 m3. The distribution of natural storage volumes is highly skewed with a median of 1.5 × 103 m3 and a total storage of 497 × 103 m3 (Figure 6a). Enhanced volumes have a median volume of 10.3 × 103 m3 and total storage of 2065 × 103 m3 (Figure 6b). Estimated errors for volume calculations were ±20% on average, based on systematic lidar elevation errors of ±0.15 m.
 The median monthly demand per off-take point during peak irrigation season (December to April) was 19.5 × 103 m3/month (11.7 × 103 m3/month averaged over the year). Monthly total irrigation orders for the study reach show peak demands between December and April with a peak of almost 2 × 106 m3/month in March (Figure 7a). Figure 7b shows the skewed distribution of individual off-take demands for the same period, indicating that monthly orders are predominantly below 50 × 103 m3 but with some larger orders up to 300 × 103 m3. Natural storage volumes equate to 5% of the mean annual demand; enhanced volumes are approximately 20%.
 For natural storages, seven were greater than the median monthly demand per off-take point during peak irrigation season (19.5 × 103 m3) and five greater than the median monthly demand per off-take point considering the whole year (11.7 × 103 m3): for enhanced storages, this increased to 65 and 30, respectively (Figures 7 and 8). On average, wetlands are spaced at 0.5 km intervals (Figure 8).
3.3. Loss Rates From Wetlands
 Typical wetland losses are between 1.2 and 2.3 m per year with an average of 1.6 m. Observed mean daily loss rates were calculated from water levels of four instrumented wetlands and a floodplain dam (Figure 9). Loss rates from the wetlands were consistently greater than the APET rate, ranging from 10% to 107% higher across the monitoring period. The months with the lowest loss rates were those with only short periods of receding water levels and were likely to be affected by rainfall. Loss rates above APET are consistent with evapotranspiration, and seepage arising from losing hydraulic gradients between the wetlands and the unconfined groundwater in the floodplain (head differences of 0.9 to 2.0 m observed near one of the instrumented wetlands in September and December 2010). Across all four wetland sites and across the months, loss rates exceeded the APET rate by an average of 50%. The variation in loss rates between wetlands likely reflects the effects of variations in substrate and hydraulic gradients on seepage rates and degree of shading, wetland orientation, and presence of aquatic macrophytes on evapotranspiration rates. Mean loss rates were used to estimate losses from the enhanced volume of the largest of the field sites.
 The largest of the instrumented wetlands (enhanced storage of 20 × 103 m3), if providing irrigation water in March, would lose between 36% and 63% of the initial volume, depending on the time of filling (Figure 10). Mean observed loss rates from the monitored wetlands increase as summer progresses but with the cumulative loss increasing with earlier filling (Figure 10). For extraction in March, losses are greatest for August fillings (12.6 × 103 m3, 63% of initial volume) and least for December fillings (7.2 × 103 m3, 36%).
 Given their relatively low volumes and high loss rates, the Broken River wetlands are unlikely to be able to provide a large proportion of the reach scale irrigation demand; however, they provide an appreciable volume at a farm scale. But with inundation frequencies predicted to fall dramatically under regulation in a drier climate, their use as an intermediate storage may go some way to solving the inevitable problem of environmental watering while still delivering an agricultural benefit. This study demonstrates hydrologic feasibility only. A comprehensive feasibility study would further require a rigorous assessment of the complex ecological and economic considerations (Figure 11). Figure 2 outlines the method that we have employed in the hydrologic assessment that could be adopted as a template for other river systems.
4.1. Wetland Inundation and the Consequences of Regulation and Climate Change
 A reduced frequency of wetland inundation under regulation and climate change is likely to lead to a progressive transition of wetlands from aquatic to terrestrial ecosystems [Kingsford, 2000]. Under such conditions, the dual use of environmental water to both conserve wetland health and provide for agriculture may be an efficient option for water allocation. Using wetlands for storage would likely improve their ecological condition [Ning et al., 2012; Watkins et al., 2013] that would otherwise continue to decline without some form of artificial watering.
 Using wetlands as interim storages for irrigation use implies filling them on an annual or biannual basis to match the agricultural period (Figure 7a). The timing of natural annual inundation (winter-spring, Figure 5) only partially overlaps this agricultural demand (summer-autumn peak, Figure 7a). Consideration should be given to the option of multiple wetland fillings in a single season, which would have a positive economic outcome, but an uncertain ecological one. Natural occurrences of multiple annual fillings are only likely in wetlands that are filled at lower discharges (Figure 5a). There may also be a trade-off between the environmental penalties incurred through imposing a uniform inundation frequency across all wetlands (to maximize agricultural benefit) with promotion of wetland plant biodiversity at a landscape level through a diversity of flood regimes (Figure 11).
4.2. Wetland Storage Volumes Relative to Irrigation Demands
 At the farm scale, the use of wetlands as storages could provide a significant volume for agricultural demand, but the reach scale demand far exceeds the available volume. The natural volume of wetlands of the lower Broken River is considerably less than the total annual demand volume, approximately 5%; however, the use of bund walls can increase this to 20% (Figures 6 and 7). Assuming that consumer off-take points are evenly distributed, then each property has 2 to 4 wetlands available for use, equating to a farm scale available storage of 3–6 × 103 m3 (natural) and 20 to 40 × 103 m3 (enhanced), which is comparable to the median summer monthly off-take demand (Figure 8). There are a smaller number of very large off-takes (>50 × 103 m3/month) where the contribution of local wetlands would likely fall far short of the volume required (Figure 7b). Multiple wetland fillings in a single season could further increase the available volume but would alter the natural hydroperiod of some wetlands.
 Catchments with larger mean annual discharge and similar meandering channel morphology (see examples in the MDB in Table 1) are expected to have larger volume of wetlands per river length. As channel width and depth increases with dominant bankfull discharge [Huang and Nanson, 2000], we also expect wetlands, being ancient channels themselves, to increase in size (and therefore volume). This trend would be expected to extend to meandering rivers globally. However, relationships between river channel and wetland size and abundance have not been demonstrated so this assumption is uncertain. Nevertheless, larger catchments are also likely to be subject to greater consumptive demands (Table 1). Given that the Broken River is a lightly regulated system, it is likely that the volume of floodplain wetlands of other catchments would provide less than the 20% of annual demand estimated here.
 Our estimates of enhanced wetland volumes have simply considered the maximum achievable storage volume (Figure 3), whereas optimum bund wall size would include water prices, projected losses, and infrastructure costs (Figure 11). Direct pumping from the river presents an additional cost that would become increasingly necessary under climate change: the driest scenario resulting in rare occurrences of any natural filling for all wetlands (Figure 5c). Operable sluice gates would be required to allow natural filling by river flows and subsequent retention in enhanced volumes, resulting in greater capital and ongoing costs. Economic benefits include greater flexibility in the delivery system for the irrigator with rapid access to irrigation water without the delay of transfer through the system and a decrease in “rainfall rejection” orders: water delivered to the irrigator that remains unused due to rainfall occurring between order lodgment and the delivery time [Chong and Ladson, 2003] (Figure 11).
 In addition to the environmental benefit of wetland watering, a distributed network of off-river wetland storages could also contribute to a reduction in unseasonally high flows—“seasonal reversals”—and their negative implications (Figure 11). Environmental costs may include, among others, the results of reduced connectivity (due to bund walls) and altered hydroperiods (due to unseasonal and/or multiple fillings and drawdown) and increased risk of invasive species (Figure 11).
4.3. Wetland Losses and the Implications for Timing of Use
 Wetlands in southeastern Australia are most likely to fill in the winter-spring period (Figure 5a) and naturally drain through evaporation and leakage as the drier summer progresses. To align the use of wetlands as storages with ecological requirements, the timing of natural filling followed by a period of prolonged inundation is required [Ning et al., 2012]. Wetlands in the Broken River would be filled August to October, for later use during peak irrigation times. The relatively high losses observed in the field study wetlands (Figure 9) indicate that the timing of natural filling has important implications in the choice of suitable wetlands. The natural variability of wetland morphology will have a great effect on evaporative losses based on empirical evidence from the depth/volume relation curve (Figure 10) and losses to subsurface leakage will depend on wetland substrate and the hydraulic gradient with local groundwater. This variability is shown in our data set; however, it is difficult to form conclusions about the pattern of losses in individual wetlands because of the short period of monitoring.
 The timing and extent of irrigation water extraction has implications for biota. Ning et al.  and Watkins et al.  found that partial drawdown of wetlands after a 2 month inundation period did not result in a reduction in plant or zooplankton species richness or diversity, relative to natural drawdown rates; however, a complete drawdown resulted in reduced diversity and richness. Nevertheless, infiltration losses from wetlands are also recognized to provide an environmental benefit of groundwater recharge and can support the provision of river base flow (Figure 11). Evaporative losses are an indirect environmental demand due to the requirement for wetlands to remain inundated for a sufficient period to allow biota to complete lifecycles.
4.4. Management Considerations
 There are three main stakeholders that would require involvement in the implementation of the use of floodplain wetlands as distributed irrigation stores: irrigators, water supply authorities, and environmental managers. Decisions on water allocation involving wetlands as irrigation storages may be best handled by the environmental water manager, where goals for ecosystem health are prioritized. Some form of discounting on the price for the remainder of this “dual use” water would then be determined to ensure that irrigator costs are matched or outcompete that of the conventional water delivery. There is an impediment to this in Australia where the Commonwealth environmental water holder is not entitled to sell water [Australian Federal Government, 2007], and adjustments would be required to legislation in order for the environmental manager to manage this conjunctive use of water. A rigorous cost/benefit analysis of this option would be required to balance environmental outcomes with the associated capital, operational and maintenance costs.
 We gratefully acknowledge Ali Aien, Nathan Ning, and Susanne Watkins for their assistance with the fieldwork for this study and Steve Wealands and Ali Aien for developing the GIS methods. The lidar and irrigation demand data were provided courtesy of the Goulburn Broken Catchment Management Authority (GBCMA). This project was part of a larger Farms, Rivers and Markets Project, which is an initiative of Uniwater and funded by The National Water Commission, The Victorian Water Trust, The Dookie Farms 2000 Trust (Tallis Trust), and The University of Melbourne. The Farms, Rivers and Markets Project is supported by the Departments of Sustainability and Environment and Primary Industry, the Goulburn Broken Catchment Management Authority, and Goulburn Murray Water.