A comparison of desktop hydrologic methods for determining environmental flows

Abstract

Determining environmental flows can be a daunting task because fluvial systems are physically and biologically complex, and there can be difficult trade-offs between instream and out-of-stream water uses. Desktop methods are primarily office-based exercises that often use more readily available hydrologic information to assess a proposed diversion of stream flows. Two-tiered water allocation decision processes, in which desktop methods represent the first tier or filter, and detailed methods the second tier, are common in many jurisdictions. A primary objective of desktop and detailed methods is to protect environmental resources, but the required information, time and cost may be substantially different. This paper reviews six desktop hydrologic methods commonly used in Canada. Seventeen performance measures (PMs), for ecosystem health (hydrology, habitat, geomorphology, connectivity and water quality) and out-of-stream water use are used to compare and contrast the predicted outcome of the desktop methods for three mid-size rivers with different hydrologic regimes. Results of the PM calculations demonstrate that absolute and relative performance of the six desktop methods is not consistent across the three streams. As a group, the methods clearly underscored the trade-off between water use and ecosystem health; however, the methods had widely divergent performance for water-use PMs, and variable performance on the different ecosystem health PMs. All desktop methods performed well on the fish habitat PMs, but performance on the other ecosystem health PMs was variable and not consistent among methods. The results emphasize the need for clearly articulated objectives as an integral component of any desktop method, and the benefit of transparent trade-off considerations during any water allocation decision.

Abstract

La détermination et gestion des débits environnementaux pour les cours d’eaux peuvent être intimidantes, étant donné les complexités physiques et biologiques des systèmes fluviaux. En plus, la gestion de l’eau doit contempler différents usages de cette ressource et les conflits d’intérêts entre les valeurs écologiques et les demandes anthropiques. Les approches méthodologiques utilisées pour évaluer les débits environnementaux sont principalement basées sur des techniques de bases qui prennent des données hydrologiques historiques pour évaluer différentes débits réservés. Pour plusieurs juridictions, le processus de décisions d’allocation d’eaux est arrivé en deux étapes avec l’évaluation générale représentant la première étape ou filtre, et la méthode détaillée comme la deuxième étape. L’objectif principal des deux étapes est la protection de cette ressource, mais l’information et les données obligatoires, le coût du processus, et le temps nécessaires peut-être considérablement différents pour chaque méthode. Cet article examine six méthodes d’évaluation des débits environnementaux connues. Dix-sept mesures de performances (MPs), évaluant la santé de l’écosystème aquatique (hydrologie, habitat, géomorphologie, connectivité et qualité d’eau) et l’utilisation et consommation d’eau hors-fleuve, sont utilisées pour comparer les résultats prédits des six méthodes pour trois rivières de taille moyenne ayant des régimes hydrologiques différentes. Les résultats des MPs démontrent que les performances relatives ainsi qu’absolus des six méthodes ne sont pas consistantes pour les trois rivières. En groupe, les six méthodes soulignent clairement le compromis entre l’usage d’eau et les valeurs écologiques. Cependant, les méthodes ont donné des MPs de l’usage d’eau divergentes et contradictoire, en plus les MPs représentant les différentes santés d’écosystèmes étaient variables. Toutes les méthodes ont bien performées pour les MPs d’habitats, mais leurs performances sur les autres MPs de la santé d’écosystèmes étaient variables ainsi qu’inconsistantes parmi les méthodes. Ces résultats démontrent la nécessité de bien et clairement spécifier les objectifs de l’étude comme éléments intégraux pour chaque méthode, donnant le bénéfice de bien établir les considérations et les compromis pour toutes décisions d’allocation d’eau.

Introduction

Environmental flows are defined as “the quantity, quality and timing of water flows required to sustain freshwater and estuarine ecosystems and the human livelihoods and well-being that depend on these ecosystems.” (Global Environmental Flows Network 2011). This definition acknowledges both environmental and socioeconomic benefits that are provided by intact ecosystems, and is referred to as the “Brisbane Declaration,” which was agreed to at the 10th International River Symposium and Environmental Flows Conference, held in Brisbane, Australia, on 3–6 September 2007. The term “environmental flows” is used in this paper in preference to “instream flows,” because it better reflects environmental considerations that extend beyond the wetted area of the river, such as the adjacent riparian community, and more directly recognizes potential trade-offs with out-of-stream uses such as diversion for power, agricultural and industrial use, or drinking water, that are part of economic or social considerations.

There have been several excellent reviews of environmental flow assessment (EFA) methods and their various strengths and weaknesses (e.g. Jowett 1997; Electric Power Research Institute [EPRI] 2000; Tharme 2003; Annear et al. 2004). Other techniques or adjustments to existing methods are continually being developed (Jowett 1997). In one of the more recent reviews, Tharme (2003) lists more than 200 EFA methods that have been used around the world. The sheer number of available assessment methods, and the fact that the list continues to grow, is testament to both the urgency of the need and the frustration with the present set of tools. The list of techniques continues to grow due to the improving ecological and physical science of rivers, including advances in technology; the shifting nature of social and economic values placed on riverine resources; and increasing demand for water associated with human population and economic growth.

Available EFA techniques can be categorized in different ways (e.g. Jowett 1997; Tharme 2003). This paper follows Tharme’s classification of hydrologic (sometimes called historical flow methods), hydraulic, habitat-rating and holistic methods. In practice, some methods span classifications, and often multiple methods are used during an assessment. Desktop methods (sometimes called standard-setting methods) are primarily office-based scoping exercises that make use of existing information to assess a proposed diversion of stream flows. These may be explicitly precautionary (i.e. environmental protection is the primary objective) to account for uncertainty in predicted effects (e.g. Locke and Paul 2011; Richter et al. 2012). Desktop methods are typically based on hydrology information.

Many jurisdictions defer to more detailed studies than desktop methods when assessing intensive water uses. These methods (sometimes called incremental methods) are often data-intensive, and require collection and analyses of biological and physical data, which are used to determine a schedule of flow requirements, often in a negotiation context. Two-tiered processes, in which desktop methods represent the first tier or filter, and detailed methods the second tier, are common in many jurisdictions (Kulik 1990; Dunbar et al. 1998). A primary objective of desktop and detailed methods is to protect environmental resources, but the amount of information and the time and cost needed may be substantially different.

This paper reviews several commonly used desktop hydrologic methods to demonstrate the inherent trade-offs between ecological, economic and social values. Several simple performance measures (PMs) for hydrology, habitat, geomorphology, connectivity and water quality (termed the ecosystem health PMs) and out-of-stream uses (economic or social PMs) are used to compare and contrast the predicted outcome of different desktop methods for three mid-size rivers with different hydrologic regimes. The primary purpose is to emphasize the need for clearly articulated objectives as an integral component of any desktop method, and the benefit of transparent trade-off considerations during any water allocation decision.

Six desktop hydrologic methods

Historical flow methods, as the name implies, rely entirely or mostly on a long-term time series of recorded or estimated flows in a target stream; the empirical or estimated flows should be naturalized to account for existing alterations to the flow regime. A fixed percentage of flow, or some other derived flow index, is selected as a flow recommendation to maintain an ecosystem feature at a predetermined level or standard. The recommended flows may be set at an annual to instantaneous time step, depending on the desktop method and logistics of the water-use activity. The indices are generally derived using expert opinion or structured observations on the health of a subset of rivers studied in more detail combined with statistical analysis.

The Tennant method is the best-known method in this category, but there are other methods that rely on this general approach to produce an environmental flow guideline. King et al. (1999) noted that there are at least 15 frequently referenced, hydrology-based desktop methods, but many are region-specific or context-specific in their application. Jowett (1997), Dunbar et al. (1998), King et al. (1999) and Annear et al. (2004) review a variety of these methods. Historical flow methods are most appropriate at a reconnaissance level, and in cases where negotiation among competing interests is not a substantial part of the decision-making process.

Tennant method

The Tennant method (Tennant 1976), also known as the Montana method, was one of the first techniques for determining flow needs for fish. It has been especially influential and is still widely used throughout the world (Reiser et al. 1989; Jowett 1997; Tharme 2003). According to Tennant (1976), the method is based on 17 years of experience on hundreds of coldwater and warmwater streams, and tested with field studies on 11 streams in Nebraska, Wyoming and Montana. The test used empirical hydraulic data from cross-channel transects combined with subjective assessments of habitat quality. From these measurements Tennant defined relationships between flow and aquatic habitat quality, and found that they were similar for each of his study streams. He developed stream flow recommendations based on percentages of mean annual discharge (MAD) (Table 1).

Table 1. Instream flow regimens for fish, wildlife, recreation and related environmental resources, as described in Tennant (1976). Flows are expressed as percentages of mean annual discharge.

	October–March (%)	April–September (%)
Flushing or maximum	200	200
Optimum range	60–100	60–100
Outstanding	40	60
Excellent	30	50
Good	20	40
Fair or degrading	10	30
Poor or minimum	10	10
Severe degradation	0–10	0–10

Modified Tennant (Tessmann) method

It was recognized that the original Tennant method may not apply to geographic locations outside the region for which it was originally devised (e.g. Mann 2006). Various modifications have made the technique more applicable to other biogeoclimatic regions. For example, the Tessmann (1979) modification incorporates consideration of natural variations in flow on a monthly basis (Table 2). This type of modification is common, and has led to modifications that make the original Tennant method more applicable to regions with different hydrological and biological cycles (e.g. see Estes 1995 for modifications appropriate for Alaska, and Locke 1999 for modifications appropriate for Alberta). Tessmann also specified an annual 14-day high flow of 200% MAD in the month of highest runoff as a flushing flow to maintain substrate quality and flood riparian habitat (Tessmann 1979). However, in practical experience, this latter recommendation is often omitted when applying the method.

Table 2. Tessmann's (1979) rules for calculating minimum flow thresholds for each calendar month.

Natural flow	Recommended minimum flow
Mean monthly discharge < 40% of mean annual discharge	Mean monthly discharge
Mean monthly discharge > 40% of mean annual discharge and 40% of mean monthly discharge < 40% of mean annual discharge	40% of mean annual discharge
40% of mean monthly discharge > 40% of mean annual discharge	40% mean monthly discharge

Another form of modification to the original Tennant method incorporates local biological information and provides Tennant-like streamflow criteria based on the presence of different life stages for fish. In this type of modification, the timing window for each flow threshold is adjusted depending on the fish life histories and ecological information for the target stream. This approach is described by Estes and Orsborn (1986) and Ptolemy and Lewis (2002).

Presumptive standard

Richter et al. (2012) reviewed several case studies that developed and applied risk thresholds for water management, and made a case for a “presumptive standard” to be used when site-specific standards are absent. They argue that the standard should be based on cumulative departures from the natural flow regime (i.e. total effect of all water uses) and should be graded as low risk (± < 10%), medium risk (±10–20%) and high risk (± > 20%). Although it was not explicitly designed as a desktop method for determining environmental flows, the presumptive standard can be used to design a “low-risk” flow alternative that is equivalent to a fixed proportion (≤ 10%) of natural flows. Diversion of a fixed percentage of flow may have different ecological cost depending on whether the flow is removed during a period of low or high flows, but this is not incorporated into the method.

Alberta Desktop method

The Alberta Desktop method (Locke and Paul 2011) is a method for setting environmental flows in the Province of Alberta. The method provides withdrawal rules, based on natural historical streamflows, which are expected to result in an ecologically based flow regime that incorporates the spatial and temporal flow conditions necessary to ensure long-term protection of aquatic environments. The stated objective of the method is full protection of the riverine environment, in the absence of site-specific studies, where full protection is defined as:

“No measurable environmental decline over the long term due to human changes in the flow regime away from natural conditions” (Locke and Paul 2011, iv). For example, fish population structure and function are similar to communities in the natural flow regime” (Locke and Paul 2011, iv).

The qualifier, “in the absence of site-specific studies” implies that more detailed studies may provide support for additional flow abstractions or protection. In essence, this implies the two-tiered process of using the Alberta Desktop method as a screening tool to determine whether more detailed studies may be necessary. Much of the supporting analysis and discussion of the method (Clipperton et al. 2002; Clipperton et al. 2003; Locke and Paul 2011) relate to instream conditions for fish, but the method attempts to retain aspects of the hydrograph that are important for the maintenance of both instream and riparian habitats, water quality and channel morphology.

The Alberta Desktop method provides flow abstraction rules throughout the year. Using a weekly time-step, there is a “cut-off” below which no water abstractions can occur, and the maximum diversion is capped at 15% of instantaneous natural flows provided natural flow is in excess of the cut-off. The cut-off flow for each week is specified as the 80% exceedance in that period (i.e. 20th percentile), but where insufficient data are available, the time step may be expanded to monthly or seasonal time steps. The method relies on observed or modeled daily natural flows over a long period of record; however, a minimum record length is not specified. In essence, the method implies that water is not available for abstraction in 1 out of 5 years within each time step. Likewise, a full 15% of available flow may not be accessible if stream flow only slightly exceeds the threshold. The cut-off thresholds lead to periods where no water is available for withdrawal (and these may be extended periods).

BC Instream flow thresholds for fish and fish habitat

The BC Instream flow thresholds for fish and fish habitat (BCIFN) were designed to support a two-tiered review process for proposed water uses on BC streams; the guidelines were developed for run-of-river small hydropower and therefore apply primarily to small, steep streams (Hatfield et al. 2003). The first tier of the review is a scoping-level process that provides thresholds for alterations to natural stream flows that are expected to result in low risk to fish, fish habitat and productive capacity. These thresholds are meant to act as a “coarse filter” during the review of proposed water uses on BC streams when there is little or no biological or physical data available. Projects that propose to exceed these flow thresholds must collect additional data for a more detailed project review (the second tier).

The thresholds are calculated from historical flow data, with two sets of flow thresholds, one for non-fish-bearing streams and another for fish-bearing streams. This creates two specific data requirements. The first is an adequate assessment of fish presence (or absence); the second is an adequate time series of mean daily flows. A minimum 20-year synthetic or empirical flow record is recommended to reflect natural flow variation.

The flow thresholds for fish-bearing streams are monthly-adjusted percentiles of natural mean daily flows. These percentiles vary through the year on a sliding scale from 20% (during the month of highest median flow) to 90% (during the month of lowest median flow). The sliding scale means that more water is available for diversion during high-flow months than during low-flow months. Although it was designed for application to small hydropower projects, the conservative thresholds have meant that most projects have chosen to undertake detailed, site-specific studies in support of additional diversions. The method was explicitly designed for non-consumptive uses, and adjustments to the method would likely be required (e.g. the maximum diversion rate) for consumptive uses.

DFO framework for ecological flow requirements

The Department of Fisheries and Oceans Canada (DFO) released its Framework for assessing the ecological flow requirements to support fisheries in Canada in 2013 (DFO 2013). It is a desktop method for determining environmental flows that can be applied to a natural (or “naturalized”) flow series. The stated objective of the method is “to guide the assessment of ecological flows required to sustain a fishery (commercial, recreational, or Aboriginal), including potential future fisheries” (DFO 2013, 4). The method has two main components that are similar to other methods described above; it has a cut-off flow of 30% MAD, and a diversion restriction of 10% of instantaneous natural flow above 30% MAD. These components are intended to interact to provide a flow regime that has a “low probability of detectable impacts to ecosystems that support commercial, recreational or Aboriginal fisheries” (DFO 2013, 10), relative to a natural flow regime. The method is meant to apply throughout Canada.

Advantages and disadvantages

Not surprisingly, the key advantage of any desktop method is its potential ease of implementation, since by design they may require little or no field work. The Tennant and Modified Tennant methods are among the simplest, since they are based on one or a few hydrologic statistics such as MAD or mean monthly discharge (MMD) that are easy to obtain from empirical time series or simple runoff models. Despite this simplicity, decisions based on the Tennant method are common in the United States and Canada (Reiser et al. 1989). The modified methods are intended to provide a better biological or hydrologic “fit” to different time periods or geographic regions.

Other hydrologic methods reviewed here are also inexpensive, rapid, desktop exercises, but rely on empirical or modelled hydrologic time series rather than a single statistic. These methods typically result in diversions that vary temporally based on water availability because the flow abstraction rules respond to natural flow patterns rather than a fixed seasonal schedule that applies to all streams. The exceptions are desktop methods that allocate a single, fixed percentage of natural flow to out-of-stream use, as no historical data are required. All methods require corrections to account for existing water uses.

Critiques of desktop methods tend to focus on two aspects: the high degree of professional judgement embedded in the method, and the lack of biological validation of assumptions and outcomes. These criticisms are, in fact, valid for most EFA methods, including detailed methods – all techniques require subjective judgements during the collection of data and in the final recommendation of an environmental flow schedule, and in practice exceedingly few decisions are rigorously assessed after they are implemented (Moyle et al. 2011). A potential disadvantage of the Tennant method is that it was developed from three states east of the Rockies (Tennant 1976). The streams in that area have similar hydrologic regimes, and may therefore be more consistently characterized using Tennant’s %MAD statistics than streamflow patterns in other regions. However, various modifications can compensate for this weakness. An additional weakness is that many desktop methods do not scale with stream size or type, despite evidence of differences in sensitivity to withdrawal that are related to stream size (Hatfield and Bruce 2000; Rosenfeld et al. 2007).

In summary, strengths and weaknesses of hydrologic desktop methods depend in part on the method (Table 3). In general, there is risk that the criteria developed by historical flow methods will be applied across different geographic regions and river types, without sufficient understanding of their ecological implications. Likewise, different methods may use different thresholds and thereby assume different levels of ecological or socioeconomic risk.

Table 3. Summary of data requirements and outputs for six evaluated desktop hydrologic methods.

Method	Source	Data requirements	Withdrawal rule	Cut-off flow	Environmental criteria addressed
Tennant	Tennant (1976, 6)	Mean annual discharge	Different withdrawal rules for each of two annual periods of 6 months	Yes	“protect the aquatic resources in both warmwater and coldwater streams”
Modified Tennant (Tessmann)	Tessmann (1979, 1)	Mean annual discharge, mean monthly discharge	Variable rules based on hydrology, applied to each of 12 annual periods of 1 month	Yes	“minimum stream flows [are] needed to sustain a healthy semblance of the ecosystem in rivers and major tributaries…”
Presumptive Standard	Richter et al. (2012, 7)	Real time discharge estimate	Same rule throughout year	No	“a high level of ecological protection will be provided when daily flow alterations are no greater than 10%; a high level of protection means that the natural structure and function of the riverine ecosystem will be maintained with minimal changes.”
Alberta Desktop	Locke and Paul (2011, 81)	Historical flow series	Same rule applied separately to each week of the year	Yes	“No measurable environmental decline over the long term due to human changes in the flow regime away from natural conditions. For example, fish population structure and function are similar to communities in the natural flow regime.”
BCIFN	Hatfield et al. (2003, 5)	Historical flow series	Variable rules based on hydrology, applied separately to each of 12 annual periods of 1 month	Yes	“ensure protection of fish and fish habitat, where the level of protection is consistent with current legislation and regulations...alteration to flows based on the thresholds should be low risk to fish, fish habitat, and the productive capacity of a stream.”
DFO Framework for Ecological Flow Requirements	DFO (2013, 2)	Historical flow series	Same rule applied separately to each month of the year	Yes	“Cumulative flow alterations of less than +/– 10% of the magnitude of actual (instantaneous) flow in the river relative to a ‘natural flow regime’ have a low probability of detectable negative impacts to ecosystems including those that support CRA [commercial, recreational and Aboriginal] fisheries. In addition, cumulative flow alterations that result in instantaneous flows less than 30% of the Mean Annual Discharge (MAD) have a heightened risk of impacts to ecosystems that support fisheries.”

Notes:

BCIFN, BC instream flow thresholds for fish and fish habitat; DFO, Department of Fisheries and Oceans Canada.

Application of desktop methods

Stream examples

Three mid-size rivers from different biogeoclimatic zones at approximately the same latitude (Table 4) were selected to demonstrate the potential outcomes of the different desktop hydrologic methods. Twenty-one years of daily flow data (1990–2010) were compiled from Water Survey of Canada gauges for the rivers. To produce complete time series for calculations, missing data were linearly interpolated from the end points of the missing period; in all cases, missing data were less than 3% of the record.

Table 4. Three example streams from similar latitudes, but with different hydrology.

Stream	Gauge	Latitude	Longitude	Watershed area (km^2)	Hydrology	MAD(m^3 s^−1)	Median annual peak flow (m^3 s^−1)
Chemainus River	08HA001	48°52'45" N	123°42'7" W	355	Coastal dry	18.1	265
Coquihalla River	08MF068	49°22'10" N	121°23'20" W	720	Coastal wet	30.9	175
Similkameen River	08NL007	49°27'35" N	120°30'7" W	1810	Interior dry	22.6	163

Note: MAD, mean annual discharge.

The three streams are representative of a range of hydrologic patterns in western North America: coastal dry regions with U-shaped hydrographs, interior dry regions with snowmelt-driven hydrographs, and coastal wet regions with a mixture of the two (Figure 1); respectively, these correspond to type 1 and type 5 from Monk et al.’s (2011) classification, and a mixture of these two types. The observed hydrology time series for the three streams is natural or close to natural; there are no mainstem dams and water use is minor upstream of the gauge locations. The examples were selected from a single jurisdiction (British Columbia), at similar latitude (49° N), in relatively close proximity but with diverse hydrology to emphasise that the performance of a desktop method is not a simple geographic exercise (i.e. desktop methods may not be associated with a particular region). The approach described in this paper can be applied to any stream for which there are sufficient hydrologic data.

Figure 1. Hydrographs for (a) Chemainus River, (b) Coquihalla and (c) Similkameen River, using 21 years of hydrology data (1990–2010) from the Water Survey of Canada. Light grey lines indicate daily flows. The heavy black line indicates median flows, the thinner black lines indicate the 10th and 90th percentile and the horizontal dashed line indicates mean annual discharge (MAD).

Water withdrawal alternatives

Water withdrawals and resulting environmental flows (i.e. flows remaining in the stream) were modelled for six desktop methods assuming maximum withdrawal rates for all time periods, as specified by each method (Table 5). For the Tennant method, withdrawal rates were modelled resulting in environmental flows that meet the “good” criteria of the method (Table 1), which are 20% MAD in winter and 40% MAD in summer. For the presumptive standard, withdrawal rates of 10% were modelled, which corresponds to the upper threshold for low risk. For the other desktop methods, maximum withdrawal rates specified by the method were modelled.

Table 5. Summary of rules used to model environmental flow and out-of-stream use for the six desktop hydrologic methods examined.

Method	Adjustment of rules	Withdrawal	Cut-off flow
Tennant	Two seasons	All flow above cut-off	20% MAD (October–March)40% MAD (April–September)
Modified Tennant (Tessmann)	12 months	All flow above cut-off	Mean monthly flow, 40% of mean monthly flow or 40% of mean annual flow, depending on natural flow condition (see rules in Table 2)
Presumptive Standard	None	10% of instantaneous	None
Alberta Desktop	52 weeks	15% of instantaneous	80% exceedence flow for each week in the year
BC Instream Flow Thresholds	12 months	20% exceedence flow from flow record	10% exceedence flow for lowest flow month, 80% exceedence flow for highest flow month, sliding scale for other months
DFO Framework for Ecological Flow Requirements	None	10% of instantaneous	30% MAD

Note: MAD, mean annual discharge; DFO, Department of Fisheries and Oceans.

Performance measures (PMs)

Several PMs were developed to compare estimated water withdrawals and environmental flows for the six desktop hydrologic methods. In general, PMs are used to compare different environmental decisions by consistently exposing trade-offs, generating productive discussion about better methods, prioritizing information needs and communicating the rationale for decisions (Gregory et al. 2012). A total of 17 PMs were developed within two categories: (1) PMs that address out-of-stream water use; and (2) PMs that address ecosystem health using the five riverine components described by Annear et al. (2004), namely hydrology, biology, geomorphology, connectivity and water quality.

The five water use PMs were developed under the assumption that water use will be sensitive to the volume of water available and continuity of access. The water use PMs are:

•	Water Use PM1: mean annual diversion rate (m3 s−1) – the diversion rate for each year of record is calculated, and the average taken for all years;
•	Water Use PM2: mean annual diversion rate (m3 s−1) in February – the diversion rate for all days in February is calculated for each year, and the average taken for all years. February is used to indicate winter conditions.
•	Water Use PM3: mean annual diversion rate (m3 s−1) in August – the diversion rate for all days in August is calculated for each year, and the average taken for all years. August is used to indicate summer conditions;
•	Water Use PM4: mean days per year with no diversion – the total number of days in each year in which no diversion occurs, and the average taken for all years;
•	Water Use PM5: maximum duration (days) with no diversion – the maximum number of consecutive days in which no diversion occurs for the entire period of record. Maximum duration can span across years.

The 12 ecosystem health PMs were further subdivided into their respective riverine components. Five hydrology PMs were used based on indicators of hydrologic alteration and environmental flow components (Richter et al. 1996; Mathews and Richter 2007):

•	Hydrology PM1: extreme low flow (m3 s−1) – the 10th percentile of annual 3-day minimum flows. This provides a measure of low flows across all years on record and is representative of low flows during drought conditions;
•	Hydrology PM2: low flow (m3 s−1) – the median of annual 30-day minimum flows. This provides a measure of typical low flows across the record;
•	Hydrology PM3: high flow pulses (number) – the median number of annual flow reversals. This provides a measure of typical flood pulses each year during rainstorms or brief periods of snowmelt; the magnitude of these events would not overtop the channel banks;
•	Hydrology PM4: small floods (m3 s−1) – 75th percentile of annual 3-day maximum flows. Small floods occur every 2–10 years and overtop the main channel banks, providing sediment load movement and localized connection to floodplains;
•	Hydrology PM5; large floods (m3 s−1) – 90th percentile of 1-day maximum flows. This provides a measure of large floods across all years on record. Large floods occur rarely but play a critical role in determining channel geomorphology and broad connection to floodplains.

Four biology PMs were developed as indicators of instream ecological values for fish, using a simple habitat-rating curve based on 20% MAD, which corresponds to the widely used Tennant value for “good” habitat (Tennant 1976). Support for a ~20% MAD threshold comes from Rosenfeld and Ptolemy (2012), who showed that food availability for rearing Coho juveniles declined substantially at flows below this level. Also, 20% MAD was a value that represented maximum or near-maximum weighted useable area for rearing juveniles of several species of salmonids in western North America streams (Hatfield and Bruce 2000).

For the biology PMs, all flows ≥ 20% MAD were assigned a habitat score of 1 (i.e. maximum suitability), zero flow was assigned a score of 0 (i.e. no suitability). All flows < 20% MAD were assigned a habitat score based on a linear relation between 0 and 1. This simple rating curve means that a flow of just under 20% MAD will score close to the maximum, whereas a substantially lower flow will score proportionally less. The rating curve is a useful tool for comparing withdrawal alternatives, but is a rather blunt tool and should not be applied universally. For example, the curve suggests that flows above 20% MAD are of no additional value and that there is no upper flow beyond which suitability to fish or other biota declines. The four biology PMs are:

•	Biology PM1: mean days per year ≥ 20% MAD – the total number of days in each year in which flows exceed 20% MAD, and the average taken for all years;
•	Biology PM2: low-flow annual habitat score – the “habitat score” is calculated for each day on record using the habitat rating curve, the average score is calculated for each year, and then the average taken for the bottom quartile of years (i.e. the average of the lowest 5 years);
•	Biology PM3: low-flow habitat score for February – the “habitat score” is calculated for each day in February, and then the average taken for the bottom quartile of years;
•	Biology PM4: low-flow habitat score for August – the “habitat score” is calculated for each day in August, and then the average taken for the bottom quartile of years.

Two simple geomorphology PMs were used to highlight differences in high flows under each alternative. High flows are important for maintaining gravel quality, alluvial sediment dynamics, connectivity with off-channel habitats, and riparian communities (Annear et al. 2004). The first PM is based on bankfull discharge, which is considered to be the effective channel-forming discharge (Leopold 1994). Leopold (1994) recommends a recurrence interval of 1.5 years in annual peak flows to estimate bankfull discharge. A 2-year recurrence interval was used for the PM since only daily discharge data were available, which underestimates peak flow; furthermore, the PM was used to assess lateral connectivity, so estimates somewhat higher than bankfull would be appropriate. The second geomorphology PM used a threshold value of 200% MAD, based on the widely used Tennant value for “flushing flows” (see Table 1). The geomorphology and connectivity PMs are:

•	Geomorphology and Connectivity PM1: count of years with flows exceeding the median annual peak flow;
•	Geomorphology PM2: mean days per year > 200% MAD – the total number of days in each year in which flows meet or exceed 200% MAD, and the average taken for all years.

A single water quality PM was used based on the 7Q10 (lowest 7-day average flow that occurs on average once every 10 years) value, the 7-day, consecutive low flow with a 10-year return frequency. The 7Q10 value can be used as a check to ensure water quality standards are not violated (Annear et al. 2004). The water quality PM is:

•	Water Quality PM1: the 7Q10 value estimated as the 10th percentile of a log-Pearson Type III probability distribution fit to the 7-day moving average of daily flows.

PMs for each desktop method were calculated for the three example streams. All calculations were completed with R (R Core Team 2012) code applied to time series of daily flows. Indicators of hydrologic alteration were calculated within R using the package “IHA” (Law 2013).

Results

Results from the PM calculations are presented in Table 6 and indicate that no single desktop method performs best across all PMs. Furthermore, relative performance of each desktop method is not consistent among streams. There is a general trend for methods that perform well on ecosystem health to score poorly on water use.

Table 6. Consequence table comparing performance measures (PMs) for the three example streams. PM results are bolded to indicate best performance and underlined to indicate worst performance among the alternatives for each class of PM.

Chemainus	PM description	Units	Natural	Tennant	Tessmann	Pres. Std.	ABD	BCIFN	DFO
Water use PM1	Mean annual diversion	m^3 s^−1	0	14.73	11.33	1.81	2.60	5.18	1.74
Water use PM2	Mean annual diversion in February	m^3 s^−1	0	22.81	16.42	2.64	3.96	7.78	2.63
Water use PM3	Mean annual diversion in August	m^3 s^−1	0	0.27	0.39	0.11	0.15	0.17	0.03
Water use PM4	Mean days per year with no diversion	days yr^−1	365.0	148.0	175.0	0.0	59.4	215.2	143.9
Water use PM5	Maximum duration with no diversion	days	7665	166	163	0	49	188	162
Geomorph and connectivity PM1	Years with flows exceeding the median annual peak flow	years	12	0	0	7	2	9	7
Geomorph PM2	Mean days per year > 200% MAD	days yr^−1	43.4	0.0	0.0	37.0	34.0	21.0	37.0
Hydrology PM1	Extreme low flow	m^3 s^−1	0.14	0.14	0.14	0.13	0.14	0.14	0.14
Hydrology PM2	Low flow	m^3 s^−1	0.44	0.44	0.44	0.39	0.38	0.44	0.44
Hydrology PM3	High flow pulses	count	97.5	68.5	70.5	97.5	96.0	80.5	97.5
Hydrology PM4	Small floods	m^3 s^−1	192.83	7.24	17.32	173.55	163.91	168.13	173.55
Hydrology PM5	Large floods	m^3 s^−1	308.80	7.24	17.32	277.92	262.48	284.10	277.92
Biology PM1	Mean days per year ≥ 20% MAD	days yr^−1	238.5	238.5	232.0	234.1	235.0	232.8	238.5
Biology PM2	Low-flow annual habitat score	value yr^−1	0.71	0.71	0.71	0.70	0.70	0.71	0.71
Biology PM3	Low-flow habitat score for February	value yr^−1	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Biology PM4	Low-flow habitat score for August	value yr^−1	0.09	0.09	0.09	0.08	0.09	0.09	0.09
Water quality PM1	7Q10 value estimated from the 10^th percentile of the 7-day moving average of daily flows	m^3 s^−1	0.17	0.17	0.17	0.15	0.17	0.17	0.17
Coquihalla	PM description	Units	Natural	Tennant	Tessmann	Pres. Std.	ABD	BCIFN	DFO
Water use PM1	Mean annual diversion	m^3 s^−1	0	22.40	17.07	3.09	4.12	7.21	2.96
Water use PM2	Mean annual diversion in February	m^3 s^−1	0	15.14	9.59	2.13	3.00	3.17	2.05
Water use PM3	Mean annual diversion in August	m^3 s^−1	0	1.22	1.62	0.98	1.21	0.54	0.54
Water use PM4	Mean days per year with no diversion	days yr^−1	365.0	72.6	106.0	0.0	69.1	240.4	67.1
Water use PM5	Maximum duration with no diversion	days	7665	105	117	0	72	153	91
Geomorph and connectivity PM1	Years with flows exceeding the median annual peak flow	years	12	0	0	9	8	6	9
Geomorph PM2	Mean days per year > 200% MAD	days yr^−1	43.9	0.0	0.0	34.7	30.3	8.8	34.7
Hydrology PM1	Extreme low flow	m^3 s^−1	3.42	3.42	3.42	3.08	3.42	3.42	3.42
Hydrology PM2	Low flow	m^3 s^−1	6.24	6.09	6.24	5.62	5.51	6.24	6.24
Hydrology PM3	High flow pulses	count	102.0	58.0	50.5	102.0	96.0	82.5	102.0
Hydrology PM4	Small floods	m^3 s^−1	154.17	12.36	27.28	138.75	131.04	107.35	138.75
Hydrology PM5	Large floods	m^3 s^−1	277.00	12.36	27.28	249.30	235.45	230.18	249.30
Biology PM1	Mean days per year ≥ 20% MAD	days yr^−1	335.8	335.8	335.8	327.8	328.5	335.8	335.8
Biology PM2	Low-flow annual habitat score	value yr^−1	0.97	0.97	0.97	0.95	0.96	0.97	0.97
Biology PM3	Low-flow habitat score for February	value yr^−1	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Biology PM4	Low-flow habitat score for August	value yr^−1	0.91	0.91	0.91	0.85	0.90	0.91	0.91
Water quality PM1	7Q10 value estimated from the 10^th percentile of the 7-day moving average of daily flows	m^3 s^−1	3.7	3.7	3.7	3.3	3.7	3.7	3.7
Similkameen	PM description	Units	Natural	Tennant	Tessmann	Pres. Std.	ABD	BCIFN	DFO
Water use PM1	Mean annual diversion	m^3 s^−1	0	16.80	12.37	2.26	3.07	4.75	2.04
Water use PM2	Mean annual diversion in February	m^3 s^−1	0	2.19	1.48	0.62	0.77	0.57	0.30
Water use PM3	Mean annual diversion in August	m^3 s^−1	0	1.80	1.78	0.80	1.07	0.48	0.52
Water use PM4	Mean days per year with no diversion	days yr^−1	365.0	142.3	194.1	0.0	78.2	277.3	172.2
Water use PM5	Maximum duration with no diversion	days	7665	214	273	0	122	327	254
Geomorph and connectivity PM1	Years with flows exceeding the median annual peak flow	years	11	0	0	9	9	8	9
Geomorph PM2	Mean days per year > 200% MAD	days yr^−1	55.3	0.0	0.0	50.0	48.1	29.3	50.0
Hydrology PM1	Extreme low flow	m^3 s^−1	1.48	1.48	1.48	1.33	1.48	1.48	1.48
Hydrology PM2	Low flow	m^3 s^−1	3.11	3.11	3.11	2.80	2.89	3.11	3.11
Hydrology PM3	High flow pulses	count	93.0	56.5	62.0	93.0	92.0	86.0	92.0
Hydrology PM4	Small floods	m^3 s^−1	132.33	9.04	31.65	119.10	112.48	99.43	119.10
Hydrology PM5	Large floods	m^3 s^−1	265.80	9.04	31.65	239.22	225.93	232.90	239.22
Biology PM1	Mean days per year ≥ 20% MAD	days yr^−1	272.0	272.0	272.0	251.0	241.4	272.0	272.0
Biology PM2	Low-flow annual habitat score	value yr^−1	0.84	0.84	0.84	0.81	0.83	0.84	0.84
Biology PM3	Low-flow habitat score for February	value yr^−1	0.71	0.71	0.71	0.64	0.69	0.71	0.71
Biology PM4	Low-flow habitat score for August	value yr^−1	0.77	0.77	0.77	0.71	0.76	0.77	0.77
Water quality PM1	7Q10 value estimated from the 10^th percentile of the 7-day moving average of daily flows	m^3 s^−1	1.6	1.6	1.6	1.4	1.6	1.6	1.6

Note: ABD, Alberta Desktop Method, BCIFN, BC Instream flow thresholds for fish and fish habitat; DFO, DFO Framework for Ecological Flow Requirements; MAD, mean annual discharge; Pres. Std., Presumptive Standard.

Water use

There were substantial differences in availability of out-of-stream water use among the desktop methods (Table 6). Annual diversion rates spanned more than an order of magnitude among methods. The Tennant method consistently allowed the greatest annual rate and the DFO method consistently allowed the least. Performance on a seasonal basis was less consistent, as indicated by different relative performance among desktop methods in February and August diversion rates. The presumptive standard, the only method with no cut-off flow (Table 3), was the only method that provided continuous availability for water use. The BCIFN had, by far, the worst performance for continuous water use availability. Other methods varied considerably for continuous availability, but for most real-life considerations, performance on this PM would be rated as low for all methods but the presumptive standard. For example, even the best of the other methods (Alberta Desktop) indicates an average of 59 days across the year with no diversion in the Chemainus, and climbs to an average of 69 and 78 days for the Coquihalla and Similkameen, respectively. These averages are made up of substantial continuous durations with no access to water (Water Use PM5): a maximum of 49 consecutive days with no diversion in the Chemainus and 122 days for the Similkameen.

Ecosystem health

Geomorphology and connectivity

The Tennant and Tessmann methods eliminated geomorphic and connectivity flows in all three streams (Table 6). For other methods, performance varied considerably among streams. The BCIFN method was variable and performed poorly for the geomorphology PMs. For example, the BCIFN method reduced the mean annual number of days with flushing flows from natural by 80% on the Coquihalla and 53% on the Similkameen. For channel-forming flows, the BCIFN method was the best performer of all methods on the Chemainus, but the worst performer (outside of the Tennant or Tessmann methods) on the Coquihalla and Similkameen rivers. The Alberta Desktop method was less variable and performed better for the Geomorphology PMs than the BCIFN method, but with one notable exception. The Alberta Desktop method applied to the Chemainus River reduced occurrence of median peak flows by 83%. The presumptive standard and DFO methods performed best across all streams. The PM values for these methods varied from 58% to 90% of natural flows, with the greatest decline occurring in channel-forming flows on the Chemainus.

Hydrology

The five environmental flow components for hydrology showed substantially different results among the desktop methods (Table 6). Extreme low flows were only affected by the presumptive standard. Low flows were only affected by the presumptive standard and the Alberta Desktop methods; the presumptive standard reduced this PM by 10%, whereas the Alberta Desktop reduced the PM by 7–13%. High flow pulses (i.e. number of flow reversals) were marginally affected by the Alberta Desktop, but were more impacted by the Tennant, Tessmann and BCIFN methods. The Tennant and Tessmann methods both performed poorly, decreasing this PM by 30–50%. The BCIFN performed better but still altered flow reversals by 8–19%. All methods reduced small and large flood magnitudes. The Alberta Desktop, DFO and presumptive standard decreased small and large flood flows by the percent-of-flow reduction factors used by the method (i.e. 10% or 15%). The BCIFN method reduced small and large flood flows by 8–30%; the Tennant and Tessmann methods reduced flood flows by 76–98%.

Biology

Results of the biology PMs indicate that all of the desktop methods performed well both in absolute terms and relative to each other (Table 6). There were only subtle differences in biology PM results among methods and streams. Because the DFO and Tennant methods allow no withdrawals below 20% MAD (i.e. the threshold in the habitat rating curve used for the PM), by definition, the biology PM results for these two desktop methods were identical to natural flows. The BCIFN and Tessmann methods had near-identical results to natural flows. Results for the Alberta Desktop and presumptive standard were within 95% of natural in most cases, with the worst relative comparison being 89% for average rearing days in the Similkameen using the Alberta Desktop. Although close to natural, the presumptive standard performed worst out of the six methods in all streams for the habitat scores.

Water quality

All methods, except the presumptive standard, maintained the 7Q10 flow at natural in all three streams (Table 6). As would be expected by the approach prescribed by the presumptive standard, the 7Q10 flow was reduced by 10% from natural in the three streams.

Hydrograph summary

As a final summary, median hydrographs for each of the six desktop methods applied to the three rivers can be compared (Figure 2). The median hydrographs show general patterns previously captured by the various water use and ecosystem health PMs. For example, loss of peak flows from the Tennant and Tessmann methods are evident in the hydrographs, and periods of no water withdrawals for some of the methods are also apparent. The hydrographs provide a useful visual representation of how different desktop methods apply cut-off thresholds, percent-of-natural flow reductions or a combination of the two rules. Three of the desktop methods (Tennant, Tessmann and BCIFN) allow median flows diverted for water use to exceed median flows in the river, although the BCIFN method only allows this to occur in the Chemainus River. It should be emphasized that these graphs demonstrate differences in medians and therefore patterns in central tendency for natural, residual and withdrawals across a year. Patterns on any one day may differ markedly. For example, cut-off periods in a year may differ from these patterns when streamflows are substantially lower than median flow.

Figure 2. Median hydrographs for the Chemainus, Coquihalla and Similkameen rivers using the six desktop method. The upper bold line shows median natural flows for the three rivers; the grey region represents median flows after applying the given desktop method (i.e. median flows left in the river); and grey lines are median flows allowed for water use by the method. Periods where grey lines are absent indicate median water use is zero for that time of year.

Discussion

This work demonstrates that absolute and relative performance of water use and ecosystem health resulting from six desktop methods are not consistent across hydrologically diverse streams. As a group, the desktop methods clearly underscored the trade-off between out-of-stream water use and ecosystem health; however, the methods had divergent performance for water use, and variable performance on different aspects of ecosystem health. The trade-off between out-of-stream water use and ecosystem health is not surprising in itself; however, the desktop methods assessed here purport to provide a high degree of environmental protection, so the trade-off may not be apparent among these methods.

The six desktop methods performed well for the biology PMs, which were surrogate measures for fish and aquatic habitat in general. This is not overly surprising because derivation of the six desktop methods shared a common desire for the protection of instream habitat. In fact, the biology PMs could be argued to be circular in their application to some of the methods. Three of the desktop methods explored (Tennant, Tessmann and DFO) are directly related to Tennant’s (1976) work indicating that 20–40% of MAD is required to maintain good aquatic habitat. However, the approach is considered to be reasonable because other studies have shown both aquatic habitat and fish productivity can peak around 20% MAD for some species (Hatfield and Bruce 2000; Rosenfeld and Ptolemy 2012).

Performance of the six desktop methods on ecosystem health PMs for hydrology, geomorphology, connectivity and water quality was much more variable. For geomorphology and connectivity, the Tennant and Tessmann methods as applied here completely failed, a result that was also supported by the hydrology PMs for small and large floods. This implies the two methods would result in significant changes to channel morphology, and lateral connectivity to floodplains and riparian areas would be substantially reduced or eliminated. Such large-scale physical changes would be expected to cascade into other components of the riverine environment (Annear et al. 2004). For example, morphologic changes and accumulation of fine sediments would be expected to impact fish habitat. However, it is important to remind readers that both Don Tennant and Stephen Tessmann recommended periods of higher flows (200% MAD) for flushing the stream’s silt load (Tennant 1976; Tessmann 1979) and flooding streamside habitat (Tessmann 1979), recommendations that are often omitted in applications of their methods. Interestingly, a 200% MAD rule would not have provided channel-forming flows (using our PM) in the Chemainus, Coquihalla or Similkameen rivers (Table 4).

Variability or poor performance of the different desktop methods to the ecosystem health PMs other than fish habitat may reflect that other riverine components were not considered, not incorporated into the method, assumed to have been covered by protecting fish habitat or thought to be of less importance. Consider the following three examples. First, the Alberta Desktop method (Locke and Paul 2011) states that the method does not specifically account for lateral connectivity to floodplains. On the Coquihalla and Similkameen rivers, the Alberta Desktop performed well, similar to the presumptive standard and DFO methods, for protecting channel morphology and lateral connectivity, whereas on the Chemainus River, the additional 5% withdrawal by the Alberta Desktop resulted in near loss of bankfull flows and, hence, expected changes to channel morphology and loss in lateral connectivity. Second, the fundamental objective of the BCIFN method was the protection of fish and fish habitat (Hatfield et al. 2003). However, the method performed poorly at maintaining flushing flows and high-flow pulses relative to some of the other desktop methods. Third, three of the four case studies reviewed by Richter et al. (2012) when developing their presumptive standard had cut-off flows for protection of a variety of ecosystem health components including water quality; however, their final recommendation did not include a cut-off flow. It was not surprising this was the only method to affect the extreme low-flow and water-quality PMs.

The intent here is not to critique, rank or endorse individual methods; however, there were consistent trends that are worth emphasizing, particularly since some of these desktop methods are in widespread use. The Tennant method, as applied here, consistently provided the greatest volume for out-of-stream water use on an annual basis, but has frequent periods within every year that would deny access to water (Figure 2). As noted previously, the method did not protect channel geomorphology or allow lateral connectivity, which could be argued is a prerequisite for other riverine components. The Tessmann method performed similarly to the Tennant method across most PMs, but was superior to the Tennant method on a few of the hydrology PMs because the rules are based on temporal hydrology patterns rather than calendar dates.

The presumptive standard was the only method that allowed continuous out-of-stream water use (Figure 2). As a result, it was also the only method that had an impact on extreme low flows and water quality. The presumptive standard was the worst of the methods examined for protecting instream aquatic habitat as measured by the biology PMs, but relative to natural its performance was still high. For protection of channel morphology and lateral connectivity, the presumptive standard tied the DFO method for best performance (both allow a maximum of 10% withdrawals at higher flows).

The Alberta Desktop, BCIFN and DFO methods are similar in combining a withdrawal amount (either percent-of-flow or fixed amount) with a cut-off threshold (Figure 2). Of the three, BCIFN allowed the greatest mean annual diversion, but the method is the most stringent in protecting periods of seasonal low flows. As a result, BCIFN consistently had the worst performance for continuous access to water for out-of-stream uses, and also performed worst for absolute diversion rate in some months. Of these three methods, BCIFN and DFO methods performed marginally better than the Alberta Desktop method on the biology PMs, but the differences from natural were minimal for all three methods. Performance of the three methods was variable for geomorphology and connectivity, both among methods and among streams. BCIFN had the worst performance of the three on the hydrology high-flow pulse PM; since this PM is the median number of annual flow reversals, a decrease in this metric can be interpreted as a flattening of the hydrograph via clipping of variability associated with modest rainfall and snowmelt. It is widely accepted that rivers and their biota require seasonal and annual variation (Poff et al. 1997). Flattening of the hydrograph is therefore assumed to have negative consequences; however, the magnitude of effect related to this PM is uncertain. The DFO method performed worst of all six methods on annual diversion rates, but was not the worst on continuous access. The Alberta Desktop method tended to allow better average annual diversion rates and shorter periods of no diversion than either the DFO or BCIFN methods. The incorporation of a cut-off threshold in these three methods, and also the Tennant and Tessmann methods, results in frequent and potentially long periods in which access to water would be denied. For example, the maximum duration in which water would be unavailable for use in the Similkameen River exceeds 100 days for all methods except the presumptive standard.

The use of natural, actual and water-use hydrographs provides an important visual tool for managers and stakeholders to understand how the rules of different desktop methods are applied (e.g. Figure 2). The median hydrographs for the Chemainus, Coquihalla and Similkameen for the six desktop methods shows the Tennant method consists of two seasonal cut-off thresholds, the Tessmann method consists of several seasonal cut-off thresholds, the presumptive standard is a percent-of-natural flow reduction and the remaining methods are a combination of cut-off thresholds and percent-of-natural. However, hydrographs themselves should be avoided as performance measures as they either must be simplified using statistics (e.g. Figure 2) or include too much information to be interpretable (e.g. Figure 1). The statistical simplification of hydrographs can lead to erroneous overgeneralizations. For example, a visual comparison of the presumptive standard and Alberta Desktop method using Figure 2 might conclude the two methods provide relatively similar ecosystem health and water-use performance (differing only by 5% natural flow reduction). However, assessment of the PMs (Table 6) shows substantial differences: the Alberta Desktop has on average 60–80 days each year with no diversion compared to zero days for the presumptive standard, and the presumptive standard resulted in declines to the extreme low-flow and water-quality ecosystem health PMs whereas the Alberta Desktop was equal to natural.

Analyses presented in this paper that compare hydrology-based desktop methods bear resemblance to other approaches, in particular Richter et al.’s (1997) range of variability and Caissie et al.’s (2014) hydrology analyses. Like their metrics, the PMs presented here are also hydrology-based metrics, but there is an important distinction in that the focus here is on multiple criteria that relate directly to economic, social or environmental performance rather than hydrology or a single assessment criterion alone. The strength of hydrology-based methods lies in their simplicity and focus on protecting the hydrologic character of rivers as a means of protecting ecological integrity (Poff et al. 1997; Richter et al. 1997). However, the PMs used here (with the exception of the hydrology PMs) have explicit criteria that transparently relate economic, social or ecological components to hydrology. Improvements to the PMs may be possible through empirically testing the underlying assumptions.

These results depend on the structure, and in some cases knife-edge threshold values, of the PMs, and a different set of PMs with different thresholds may lead to different conclusions. However, the PMs rely on neutral assumptions from a wide range of environmental studies or basic out-of-stream water use. Furthermore, the use of threshold values is not unreasonable, as some riverine components may be expected to respond in a strongly non-linear fashion (e.g. lateral floodplain connectivity). One could design more sensitive PMs for any of the different PM categories, but these would likely require less neutral assumptions and might only accentuate the trade-offs among ecosystem health components, and trade-offs between ecosystem health and water use. The PMs used here are not necessarily universally applicable, or even the most useful for a specific water use decision, but they offer insight into a range of relevant factors that should be considered in allocation decisions. Overall, the general approach of using a diversity of PMs is certainly broadly applicable to water-use decision-making.

Desktop methods are born from the desire for quick, simple answers to the question “How much water does a river need?” (sensu Richter et al. 1997). The conclusion from the presented results is no single “right” answer exists to this question because different evaluators, and therefore different desktop methods, emphasize different values when making trade-offs. For example, some methods emphasize protection of low flows over protection of geomorphic flows, or ecosystem values over out-of-stream water use. Certainly, the science of environmental flows has evolved, and differences in scientific methods may explain some of the differences among methods. Approaches to dealing with uncertainty may also explain differences among methods. However, it is important to understand that the evaluators’ approaches to balancing risks among different ecosystem health components or between ecosystem health and out-of-stream water use likely had a large influence on the design of each method. The designers of each of the six desktop methods undoubtedly understood that ecological risks increase with increasing withdrawal, but came to different conclusions on the acceptability of ecological or social consequences of water allocation. Each of the desktop methods evaluated indicates some impact to ecological values; however, each method provides a different judgement on what is an acceptable ecological risk in the face of demand for out-of-stream water use. Just as there is no single answer to how much water a river needs, individual desktop methods should not be viewed as providing a single environmental standard across rivers. This is an essential point. The methods assessed here are all meant to provide a “high” level of environmental protection, yet the methods can be altered to provide lesser levels of environmental protection if policy allows greater water use (e.g. see approach described in King and Louw 1998). It is important for stakeholders to understand explicitly the performance and trade-offs from any desktop method. It is also essential that any desktop method employed for water allocation decisions be rigorously tested through monitoring.

Demand for water to support human population and economic growth is certain to continue for the foreseeable future, which will increase pressure on a limited resource and the ecological services that streams provide. At the same time, society’s environmental values and limited governance capacity will inevitably lead to development and use of rapid desktop approaches for establishing environmental flow requirements. The cost, time and expertise to undertake detailed environmental flow assessments tend to be prohibitive for all but the most intensive water-use decisions, sustaining the need for rapid desktop methods to assess applications for less intensive uses. Desktop methods that are highly risk averse for ecosystem health may be less useful for water allocation decisions as the methods could be frequently ignored. Such a desktop method will merely indicate that there will be some level of environmental risk, which in itself is unlikely to be very useful unless it is accompanied by a legal or policy objective for meeting a specific risk level.

In most contexts, there is a fundamental trade-off between water use and riverine ecosystem health. Most involved in water allocation decisions understand this intuitively and try to achieve a balance among competing objectives. What has been less clear, yet underscored by the continuing development of new desktop methods, is there is no “silver bullet” to solving this trade-off; some as-yet-undefined desktop method will not preclude the need to address this trade-off. Therefore, the best way forward is to provide, as an integral component of any desktop method, clearly articulated objectives that describe how the trade-off has been made, and to include in the assessment a wide array of performance measures and how they relate to these objectives. This point has been made by others (King and Louw 1998; Richter et al. 2012). These objectives should be clear and quantitative, rather than qualitative objectives communicated as part of the desktop methods reviewed here (Table 3); obtaining such a level of clarity in policy is likely to be a real challenge, but will force risk-balancing to be done more transparently.

Performance measures and their objectives may be stated as quantitative criteria or policy objectives. For example, it could be stated that a decision based on a desktop method should achieve 90% relative to natural conditions on pre-defined performance measures. The objective may come from the stated policy of the regulatory agency, an opinion of the individual regulator or from stakeholder input, but the statement of these objectives is at least as important in a desktop method as the quantitative thresholds and calculation methods. At best, these types of objectives are embedded within desktop methods and not stated explicitly; at worst, the objectives were never evaluated. Transparency in objectives will communicate to stakeholders what the regulator is trying to achieve, allow a better understanding for selection of one desktop method over another and direct monitoring to determine whether objectives are achieved. Another benefit is to allow those involved in the science of assessing and monitoring effects of water management to focus on science and defer discussion of the objectives to a separate, more appropriate policy arena.

The answer to “what is the best desktop method?” is a method that consistently achieves the stated objective for balance between instream and out-of-stream water uses while supporting an efficient allocation process. Results presented here show that it is achievable to develop PMs to assess whether a desktop method meets multiple instream and out-of-stream performance criteria.

Acknowledgements

We thank J. Rosenfeld for information to support a generic habitat-rating curve, and D. Lacroix and W. Vibert for French translation. Comments from T. Linnansari and an anonymous reviewer substantially improved the presentation.