The management of irrigation water storages involves a consumption-storage problem: water can either be released from storage and consumed in the current period or held in storage for use some time in the future. While this type of consumption-storage problem is common in economics, the water storage problem does have a number of unique features.

One of the unique features of the irrigation water storage problem is the magnitude of uncertainty surrounding the future supply of water. The primary source of this uncertainty is weather variability, with inflows into storages being the product of highly variable rainfall and resulting catchment run-off. Another relatively unique feature of the water storage problem is the presence of centralised storage. While private on farm storage of water is possible, it is often costly and inefficient compared with collective water storage in major dams.

One main role of large irrigation water storages is to help mitigate this variability, particularly supply variability. By retaining a proportion of available water in storage, dam managers can accumulate a buffer stock which effectively smooths variability in water supply. Determining how much water to retain in storage is a complex problem, given uncertain future inflows and water needs.

The water storage problem

For the purposes of this discussion, the storage problem facing a simple irrigation scheme is considered, as shown in figure a. While in practice irrigation schemes may involve a range of additional complexities not included in this example, the essence of the storage problem is relatively universal. In this simple irrigation scheme there is a single storage (dam) which receives variable inflows, upstream of an irrigation area comprising a number of irrigated farms.

Irrigation water is released from the storage and transported to farms via natural water courses (for example rivers) and irrigation channels. Losses occur both in storage and in the delivery of water through evaporation and seepage. It is assumed there exists no on farm storage and no in stream or tributary flows downstream of the storage.

In each time period there is a certain amount of water available, equal to the start of period storage level (water carried over from the previous period), plus within period inflows, less any storage losses. Each period a proportion of this available water is released for irrigation and a proportion remains in storage for use in future periods. These time periods can be thought of as water years (financial years) although theoretically the unit of time could equally be months or days.

Increasing irrigation water use generates benefits to irrigators by increasing the yield of crops, which in turn increases revenues and profits. Marginal increases in crop yields tend to decline as total water use increases, and at some point additional water may cease to generate increases in yield (a point referred to as maximum yield). Conversely as total water use declines, there may exist a threshold below which crops die because of water stress; at this point the marginal benefit of additional water use will be high. Crop death is particularly costly for irrigators with perennial plantings.

There may be a marginal cost associated with using irrigation water (the cost of transporting water from dam to farm and applying it to crops). For most levels of water use, the marginal cost is likely to be well below the marginal benefit. Therefore, for the purposes of this discussion it is assumed that the marginal cost is zero.

In the hypothetical case where inflows into storage are certain; where each period the dam is filled and enough water is available to provide all irrigators with their maximum level of water use, the marginal value of water would be equal in each time period and there would be no incentive to store any water for future periods. In contrast, in the case where inflows are variable; and there exists a chance of a drought occurring, water availability in a drought year would be low and the marginal value of water use high. Therefore, it may be optimal to forgo some water use during periods of high availability (low marginal value) in order to store more water for low availability (high value) periods.

The essence of the water storage problem is to compare the marginal value of consuming water with the marginal value of storing water, where the marginal value of water in storage is equal to the expected marginal value of future water use, discounted into today’s dollars and adjusted for storage losses. The total benefits of water use will be maximised in the long run where the marginal value of water use is equal to the marginal value of water in storage, subject to constraints on water storage.

Centralised storage management

In Australia, the storage management decision tends to be centrally controlled. This occurs via an announced allocation system, where each season the dam manager (generally a state government body or a state government owned entity) announces a percentage allocation – the percentage of nominal water entitlement volume available for use or trade by the entitlement holder within that season. For a centralised storage management policy to achieve an efficient allocation of water across time and space, a number of conditions must be met. In this section those conditions are outlined, along with reasons that they may not be met in practice.

For the purposes of this discussion a simplified example of an irrigation water system, as outlined above, can be used. In this example it is assumed that the water storage decision is managed by a central agency, that there exists one class of entitlement and that irrigators have no access to carryover rights. Further, it is assumed that there are multiple irrigators and there exists unrestricted intra-regional trade in water allocations. A number of potential complications to this simple example, such as high and low reliability entitlement systems, intra and inter-seasonal allocation and the potential for unused allocations, are considered later.

The first condition required for centralised management to achieve an optimal allocation of water is that the dam manager has perfect information. Specifically, the dam manager requires perfect information on the aggregate demand curve for water use and the expected aggregate demand curve for future water use (the marginal value of water use and marginal value of water storage). Effectively the dam manager needs to know the marginal value of water for each point in time and each set of circumstances. Given this information, the dam manager would be in the position to develop an optimal release rule, which would for each point in time and for each state (situation) specify the optimal aggregate amount of water to be released from the storage (the optimal allocation percentage).

The second condition required for centralised management to be optimal, is that water trade is efficient and costless, that is there are zero transaction costs. Under these conditions, the optimal aggregate amount of water would be released each period, with this water being efficiently allocated across irrigators via trade in water allocations. Under these conditions the allocation of water across time and space would be efficient. The next section considers two reasons why these conditions may not be met: the existence of asymmetric information between the storage manager and irrigators; and the presence of transaction costs or restrictions on water trade.

Asymmetric information

Dam managers may obtain approximations of aggregate water demand (current and expected) through observing traded prices on water markets and through regular discussions with representative irrigators. However, it is unlikely that dam managers will obtain full information on aggregate water demand without knowing all of the individual water demands. In practice, the costs of acquiring this information from individual irrigators are likely to be prohibitive. Hence the term asymmetric information; individual irrigators have better information on their water needs than is available at a reasonable cost to central managers.

Obtaining information on water needs from irrigators is difficult for a number of reasons. Firstly, irrigators are likely to display highly diverse preferences for water. For example, different irrigators may be engaged in different activities, each with specific water requirements. A common example is the significant differences in water requirements between annual and perennial plantings. Other potential sources of diversity include spatial variation (differences in soil type and local climate) and variation in risk preferences across irrigators.

An additional complication is that irrigators’ water preferences are likely to change significantly over time. For example, they may change in response to changes in relative prices of commodities, which could alter the mix of irrigated activities, or they could change if irrigators’ attitudes to risk change. The more diverse irrigators’ preferences are and the more they change over time, the more difficult and costly it will be for a central agency to obtain full information.

With asymmetric information, a central manager may implement a sub-optimal release (allocation) policy, which will lead to a sub-optimal distribution of water over time and ultimately lower average returns to irrigators in the long run. A central manager with incomplete information may choose to adopt a simple aggressive release policy, for example release all available irrigation water in each period (a detailed discussion of prevailing centralised storage management polices in Australia is located at the end of this chapter). Such an aggressive allocation policy may not result in an ideal level of water reliability, from an irrigators’ perspective. In practice, asymmetric information may affect both inter-seasonal water allocation (between years) as well as intra-seasonal allocation (within water years). This distinction is discussed in detail below.

Intra and inter-seasonal allocation

In practice water allocation occurs on an annual water year or irrigation season cycle, approximately coinciding with the financial year. In each water year, an allocation applies specifying the volume of water available for use within the year. Given uncertainty over seasonal inflows, allocations are announced progressively, often at monthly intervals. Typically, the initial allocation made at the commencement of the season is relatively conservative and the allocation is increased as additional inflows arrive later in the season. Water managers are understandably averse to decreasing allocations once they have been announced, although recent drought conditions did result in small allocation reductions in some regions.

In the absence of any carryover rights, the central manager maintains control over inter-seasonal water allocation. However, irrigators do have a degree of flexibility with regard to intra-seasonal water use, since irrigators can use announced allocations at any time within the current season. The market price of water reflects the option value (the option to delay water use) attached to early season allocations; there is generally a premium placed on early season allocations which gradually reduces as the market price of allocations converges to a spot price at the end of the season (Brennan 2006).

While the announced allocation system does provide irrigators with the right to delay the use of allocations until later in the water year, it does not allow them to bring forward later year allocations. This can be a problem where the intra-seasonal allocation of water is overly conservative – where early season allocations are too low and there is water available in storage that remains unallocated.

Dam managers may adopt a conservative approach as a result of uncertainty over expected storage and delivery losses. Under the current entitlement system, dam managers must ensure that enough water is available to deliver all announced allocation volumes and cover all of the associated losses. Where the dam manager has imperfect information on water demands, specifically where there is uncertainty over when allocations will be used, this will increase uncertainty over expected losses. For example, the longer irrigators delay using allocations, the greater the associated storage losses. In order to avoid the possibility of reductions in announced allocations, dam managers may hold excess water in storage to insure against the risk of higher than anticipated losses.

In the short run, the intertemporal allocation of water may be adversely affected by lag times in allocation announcements. For example where allocation announcements occur at monthly intervals, it may be up to a month before inflows received in the dam become allocated and available for use.

Transaction costs

Transaction costs refer to costs incurred when making an economic exchange. These can include search costs, information costs, bargaining costs and costs incurred in administering and enforcing a transaction. There is evidence to suggest that irrigators face significant transaction costs when trading water allocations in the Murray-Darling Basin (see Allen Consulting 2006). Transaction costs in water allocation trade can include both direct financial costs, such as fees paid to water brokers and exchanges and application fees paid to governments, and non-financial costs, such as time costs incurred by irrigators.

box 1

Unused allocations

The discussion above has implicitly assumed that all allocated water is used within the period it is allocated, yet in practice this may not be the case. While the proportion of allocations not used by irrigators has declined in recent times as irrigation water has become increasingly scarce, unused allocations remain a common occurrence. Under an announced allocation system, with no carryover rights, any unused allocations at the end of the season are returned to the common pool and shared among all water users.

Unused allocations may occur if there are constraints in the delivery or trade of water, or where the marginal benefit of water use (or the market value of water) is less than marginal cost of water use. Unused allocations are more likely to arise in wet years when the marginal value of water is low and where there are restrictions on intra or inter-regional water trade and/or restrictions on intertemporal water management.

For example, situations may arise where large allocation increases occur late in the water year, potentially as a result of overly conservative early season allocations (for reasons discussed in the previous section). At the end of a season, the marginal benefit of applying water to a crop may be very low. Moreover, there may be restrictions on carrying this water over to the next season and/or transaction costs and constraints on water trade. Irrigators may attempt to store any unused allocations on farm, either by investing in on farm storage or by applying the water to the farm in order to increase the soil moisture level. However, on farm storage will tend to be less efficient and more costly than central storage.

Unused allocations result in an increase in storage levels and an improvement in the reliability level of water entitlements. It has been noted by Brennan (2007) that the removal of institutional constraints on trade may result in an increase in the utilisation of allocations. This may inadvertently have a detrimental effect on the reliability of water entitlements within an irrigation system. This is not to say that constraints on trade should be retained for this purpose. Ideally, appropriate reliability should be achieved explicitly with some form of storage property rights, as discussed in the following chapter.

The size of government application fees and the amount of paper work involved vary significantly across jurisdictions (see Allen Consulting 2006). The time taken by governments to approve trades imposes costs on irrigators – again this varies by jurisdiction (in the range of one to seven days according to Allen Consulting (2006)). In addition to direct transaction costs, irrigators may face significant search costs (the costs associated with finding willing sellers or buyers) particularly in regions where the market for water is ‘thin’ (where there is little trade volume) or where there exists a lack of well developed exchanges or brokers (or a lack of irrigator knowledge about exchanges and or brokers).

As noted by Freebairn and Quiggin (2006), while transaction costs may be expected to decline as water markets mature and improvements are made to property rights systems and associated institutions, the fundamental complexity of water rights (reflective of the complexity of water as a commodity) would suggest that transaction costs are likely to remain significant for the foreseeable future. In addition to transaction costs, there also exist a range institutional constraints on water trade, although these less commonly apply to intra-regional temporary trade (trade in allocations within a region).

Under a simple announced allocation system, with a single class of entitlement and no carryover rights, substantial temporary trade in water allocations may be required to achieve an efficient allocation of available water across irrigators in each season or period. This can be illustrated with a simple example as shown in figure e.

box 2

Constraints on water trade

Currently there exist a range of constraints on permanent water trade at an inter-regional level, especially on trade out of an irrigation region (see Goesch et al. 2006; Peterson et al. 2004; Goesch et al. 2008). For example, trade in permanent entitlements out of irrigation areas is currently capped at 4 per cent of total water entitlements in any one year. While there has been a commitment under the National Water Initiative (NWI) to remove this cap, this will not occur until at least 2014 in the northern catchments of the basin (Goesch et al. 2008).

In this report the focus is predominantly on the efficiency of temporary trade (trade in allocations) rather than permanent trade (trade in entitlements). However, there are a number of constraints placed on inter-regional temporary water trade, although these constraints are often less explicit. For instance, restrictions on inter-regional trade may take the form of lengthy delays in the processing of trades. In some cases these delays can be more than a month in duration (Goesch et al. 2008).

Intra-regional temporary trade tends to be subject to fewer instructional constraints (Peterson et al. 2004). Often, constraints imposed on temporary intra-regional water trade reflect real world hydrological constraints or environmental concerns (Peterson et al. 2004; Brennan 2006). Hydrological constraints may involve delivery capacity constraints, for example, a river choke point or simply the inability to transfer water upstream, which can place a practical limit on the volume of water than can be traded upstream of confluences in multiple river systems (Peterson et al. 2004; Brennan 2006).

Demand elasticity refers to the responsiveness of demand to changes in price (defined as the percentage change in the quantity demanded in response to a given percentage change in price). In this example there are two irrigators, irrigator A has an elastic demand curve for water (responsive to price) while irrigator B has an inelastic demand curve for water (less responsive to price). Irrigator A is representative of more flexible farming activities (for example, broadacre) and irrigator B is representative of activities which require consistent levels of water use (for example, horticulture). P_W represents the market price of water in a wet (high allocation) year while P_D is the higher market price of water during a dry (low allocation) year. For simplicity, it is assumed that the demand curves are the same in both states (wet and dry).

From the diagram it can be seen that irrigator B demands a similar amount of water in each state of nature (that is, wet and dry), while the demand for water by irrigator A varies significantly between the states of nature. Under an announced allocation system, with a single class of water entitlement, substantial temporary trade will need to occur to generate an efficient allocation of water. For example, if both irrigators own identical water entitlements (A and B receive equal allocations in both states), irrigator B will buy water from irrigator A in a dry year, while in a wet year irrigator A will buy water from irrigator B.

High and low reliability entitlements

High and low reliability entitlement systems (referred to as general and high security entitlements in New South Wales) are relatively common in the Murray-Darling Basin. High and low reliability entitlement systems have the potential to reduce temporary water trade requirements, and reduce irrigators’ exposure to transaction costs by providing water rights which closely match the reliability preferences of individual irrigators (Freebairn and Quiggin 2006). However, a two reliability level approach does have a number of limitations.

In practice there exists a spectrum of reliability preferences. Under a two level reliability system, irrigators will need to hold a mix of the two entitlement classes to achieve a specific reliability level. This may involve some additional cost for irrigators, particularly where there are transaction costs associated with permanent trade. A two level reliability system also places an artificial upper and lower bound on available reliability levels. A high and low reliability entitlement system may also involve additional administrative effort, with governments needing to make two separate allocation announcements. Moreover, governments will need to define and communicate reliability levels clearly. This may be difficult given uncertainty over the future climate.

Under a high and low reliability system there is a need to ensure that the mix of high and low reliability entitlements in the system at any point in time is appropriate. Freebairn and Quiggin (2006) consider a system where the water authority takes an active role in the market for water entitlements, to ensure the optimal mix of high and general security entitlements is achieved. For example, where there is a market preference for high security over general security entitlements, a water authority could purchase general security entitlements and sell high security entitlements (at a ratio determined by hydrological constraints).

Such a system would use market preferences for high and general security entitlements, to reveal information about the regions aggregate reliability preference. The system would be one way of addressing the information problems of a standard announced allocation system. However, there are obvious costs associated with a water authority taking such an active role, including the transaction costs of engaging in the market and any additional administrative effort and regulatory requirements. Further, determining the appropriate conversion ratio between high and low security at any point in time may be problematic given significant uncertainty over future climatic conditions. An embargo was recently placed on the conversion of entitlements between general and high security in the Murrumbidgee region because of concern over the accuracy of conversion rates (New South Wales Department of Water and Energy 2008b).

Implications for investment

In the previous discussion it has been assumed that the irrigation capital stock is fixed. Consistent with this approach, in the model presented later, it is assumed that irrigators cannot make any additional investments to increase the area set up for irrigation, improve water use efficiency or change irrigation activities. For example, in the applied example of the model, it is assumed there is a fixed maximum demand for water (fixed nominal volume of entitlements) and fixed proportions of less flexible activities (for example, horticulture) and more flexible activities (for example, broadacre).

In practice, it is likely there will be significant interdependence between storage management polices (the yield-reliability of water entitlements) and irrigator investment decisions. This interdependence has been demonstrated in a number of other models. For example, Dudley (1988) develops a model where total area irrigated and storage policies are jointly determined by a single decision-maker. Brennan (2006) presents a model where the proportion of available irrigation land devoted to three broad activities (horticulture, dairy and broadacre) is a function of the water availability probability distribution (the yield and reliability of water entitlements) and the relative costs and returns.

To this point the focus has been on how a centralised storage management policy may not adequately match existing water demands. However in the long run, a fixed centralised storage policy may also act as a constraint on irrigator investment, for example preventing an optimal distribution of low and high flexibility irrigation activities. Where a fixed aggressive storage policy is adopted, this may constrain investment in more intensive forms of agriculture which require more reliable water supply. A potential example of this would be the significantly greater proportion of horticultural activity in Victorian irrigation systems relative to New South Wales systems, where the storage policy is significantly more aggressive.

Further, the higher water supply and price variability associated with aggressive storage polices may act as a general constraint on investments in water use efficiency. For example Hafi et al. (2006) and McClintock (2009) demonstrated how water price variability is likely to significantly slow the adoption of water efficiency technologies.

Centralised storage management in practice

As discussed, the state governments have responsibility for managing the major water storages (announcing water allocations). As such, storage management policies (allocation rules) can differ significantly across jurisdictions. For example, the storage management polices employed in Victoria tend to be significantly more conservative (for example lower yield and greater reliability) than those employed in New South Wales (Murray-Darling Basin Commission 1999) and Queensland. While there are a number of differences between states, there are a number of elements common to most storage management polices. These are outlined below.

Many irrigation systems in the Basin operate a two reliability level entitlement system. The allocation of water to these entitlement classes generally occurs on a priority basis: water is allocated to high reliability entitlements first, before remaining water is allocated to low reliability entitlements. Reserves may be held in storage to maintain the reliability of high reliability entitlements. Other than reserves held for high reliability entitlements, water may remain in storage between seasons as a result of: carryover rights, unused allocations, or where total available water exceeds the maximum limit on water entitlements (generally 100 per cent of the nominal value of entitlements). Clearly the maximum level of water use in a system has implications for system reliability. For a given supply system, a higher maximum level of water use will result in a reduction in mean storage levels and reliability. This issue is not considered in detail in this paper.

Dam managers must also provide for minimum environmental water requirements (as specified in water sharing plans), town water, and stock and domestic water. These water requirements generally take priority over irrigation water needs. The dam manager must also hold enough water to cover any losses associated with storing and delivering water. Generally, once all of the above water requirements have been met, remaining water is allocated to low reliability entitlement holders (up to the maximum limit). Below is a simplified representative allocation rule for low reliability entitlement allocations.

Victoria

Water entitlements in Victoria have undergone a number of reforms (introduced 1 July 2007) including the unbundling of water entitlements into separate water, use and delivery rights. These reforms also involved the definition of separate high reliability and low reliability entitlements, with the newly created low reliability entitlements replacing the previous system of sales water. In Victoria, water allocations are determined by state government water corporations, such as Goulburn-Murray Water. As discussed, the storage management polices implemented in Victoria tend to be relatively conservative. This conservative approach is evidenced by the relatively high proportion of high reliability water entitlements in most Victorian systems (table 1).

The Victorian approach to water allocation involves allocating water to high reliability entitlements first, then providing storage reserves to ensure water is available for next season’s high reliability entitlements (under drought inflows), and only then allocating any remaining water to low reliability entitlements (Goulburn-Murray Water 2008).

New South Wales

In New South Wales, water allocations are determined by the Department of Water and Energy. High reliability entitlements are referred to as high security and low reliability entitlements general security. Relative to Victoria, New South Wales irrigation systems have a relatively small proportion of high reliability entitlements.

The water allocation rules used in determining water allocations are listed in the relevant water sharing plans (New South Wales Department of Water and Energy 2004). The Murrumbidgee plan states that the water supply system must be managed such that a minimum of 95 per cent of high reliability entitlements can be provided under a repeat of the worst period of low inflows on record. The plan states that sufficient volumes of water must be set aside in storage to meet the 95 per cent requirement. According to the water sharing plan, general security allocations are only to be made where the high security allocation is at least 95 per cent.

Given the relatively small volume of high reliability entitlements and the relatively reliable historical inflows into the Murrumbidgee region, it is likely that minor reserves (relative to those in Victoria) would be required to achieve the 95 per cent requirement.

Queensland

In Queensland, water allocations are announced by water corporation SunWater, in accordance with rules specified in the resource operation plans (water sharing plans) for each catchment (QLD Department of Natural Resources 2007, 2008). High reliability entitlements are represented as high priority entitlements and low reliability represented as medium priority. In general, only a relatively small proportion of water entitlements are defined as high priority. In the St George, MacIntyre Brook and Border rivers irrigation systems, all irrigation entitlements are defined as medium priority. Individual systems with substantial proportions of high priority entitlements include Bundaberg and Emerald.

The resource operation plans specify that reserves equal to one year’s supply should be held in storage for high priority entitlements and that water will not be allocated to medium priority entitlements unless high priority entitlements have received a 100 per cent allocation. Allocations to medium priory allocations are calculated as total available water, less high priority allocations and reserves, and less storage and delivery loss provisions. Given the small proportion of high reliability entitlements in most Queensland systems, the aggregate storage management policy (allocation rule) can be adequately approximated (as for New South Wales) as shown in figure g.

Decentralised storage management

An alternative to centralised storage management is to decentralise the storage decision – allow irrigators to make their own storage decisions. In this report, two decentralised approaches to storage management are considered: carryover rights and capacity sharing. These property rights structures have the potential to address some of the problems of centralised storage management, including asymmetric information and transaction costs in water trade.

A decentralised approach to storage management involves irrigators making their own storage decisions, taking into account their private information on water needs. In the presence of asymmetric information (where the central manager has less information on water demands than irrigators), a decentralised storage management policy may result in releases from storage more closely aligning with the preferences of irrigators, which could potentially increase returns to irrigators in the long run.

By making their own storage decisions, irrigators can influence the reliability level of their entitlement such that it better matches their individual preferences. For example, irrigators who value high reliability can leave more water in storage and use less now, effectively increasing the reliability of their entitlement. More closely aligning individual water entitlements with individual reliability preferences will tend to reduce the volume of temporary water trade required and thus reduce irrigators’ exposure to transaction costs associated with trade.