Climate and water availability indicators in Canada: Challenges and a way forward. Part I – Indicators

Abstract

Climate variability influences the availability of water resources throughout Canada, and projected climate change is anticipated to affect future water availability. This is the first paper of a three-part analysis of water availability indicators in Canada (Parts II and III, this issue). The concept of water availability has been described in different ways in the literature. In Part I, the various approaches for estimating water availability are reviewed and compared, with a focus on Canadian studies. Global examples are used when necessary. The approaches to estimate water availability are organized into three categories: (1) climate-based indicators, (2) hydrology-based indicators and (3) water demand/supply-based indicators. Climate-based indicators use variables such as precipitation, and potential or actual evapotranspiration to calculate water budgets. Widely used meteorological drought indices that calculate moisture surpluses and deficits are also examined. Hydrology-based indicators focus on variables such as observed or modeled streamflow, or runoff. Water demand/supply-based indicators tend to focus on comparing the volume of available water with the amount of water used. Some conclusions on the status of water availability estimates in Canada are provided.

La variabilité du climat a des répercussions sur la disponibilité des ressources en eau partout au Canada et on s’attend à ce que les changements climatiques projetés influent sur la disponibilité de l’eau future. Le présent rapport est le premier d’une analyse en trois parties (parties II et III, cette série) des indicateurs de la disponibilité de l’eau au Canada. Les ouvrages scientifiques décrivent de nombreuses façons le concept de la disponibilité de l’eau. Dans la partie I, les diverses méthodes pour estimer la disponibilité de l'eau sont examinées et comparées, en privilégiant les études canadiennes. Exemples internatinaux sont utilisés comme nécessaire. Les méthodes d’estimation de la disponibilité de l’eau sont réparties en trois catégories: (1) les indicateurs climatiques, (2) les indicateurs hydrologiques et (3) les indicateurs fondés sur l’approvisionnement en eau et la demande d’eau. Les indicateurs climatiques servent à calculer les bilans hydriques à partir de diverses variables comme la précipitation et l’évapotranspiration potentielle ou réelle. L’étude examine aussi les indices de sécheresse météorologique, grandement utilisés, qui servent à calculer les surplus et les déficits en eau. Les indicateurs hydrologiques portent sur des variables comme le ruissellement ou l’écoulement fluvial simulés ou observés. Les indicateurs fondés sur l’approvisionnement en eau et la demande d’eau ont tendance à mettre l’accent sur la comparaison du volume d’eau disponible à la quantité d’eau utilisée. L’étude présente des conclusions sur l’état des estimations de la disponibilité de l’eau au Canada.

Introduction

The sustainability of freshwater supplies is a growing concern. Urbanization, industrial expansion, agricultural intensification and climate change impacts affect water supplies and aquatic ecosystem health. Accurate, up-to-date water availability information is crucial for planning, development or operational purposes (e.g. agriculture, energy, municipal water use, fisheries, manufacturing and transportation), and for ecosystem health. Yet there appears to be insufficient understanding of how much water is presently available in Canada, where available water is located, the changes of water availability through space and time, and how severe Canada’s regional water availability issues are. To ensure continued freshwater sustainability in Canada, it is necessary to assess water availability. This is difficult as there is no common definition of water availability in Canada. To our knowledge, there is also no consistent method to measure historical, current and future water availability across Canada.

As a way of beginning to consider these knowledge gaps, as they relate to conditions in Canada, a synthesis of primarily Canadian literature has been undertaken. The objective is to improve our understanding of the state of knowledge on the climate–water availability relationship within Canada’s watersheds, so as to provide context for future research efforts. This synthesis includes comparisons of historical trend and scenario case studies from across Canada, highlighting similarities among cases within regions, but also differences arising due to investigators’ choices of methods, scenarios (where applicable) and underlying databases.

This synthesis is presented in three parts. The first paper reviews the literature on defining water availability: the relevant indicators, their purposes and the strengths and weaknesses of these metrics. This is a challenge as there is no clearly established definition of water availability. Definitions vary depending on the discipline or purpose of the application. The review focuses on indicators of surface water availability. It does not include groundwater-based indicators. However, given the importance of groundwater throughout Canada, and the emergence of some new literature (e.g. Tremblay et al. 2011), such a review is certainly warranted. The second paper in this analysis explores some methods and associated caveats for assessing water availability in a temporal or spatial context in order to determine trends. In addition, key trends for Canada are synthesized from these sources (Mortsch et al. 2015). The third paper in this analysis reviews some common approaches for developing and assessing future scenarios of water availability, and reports on emerging potential trends and impacts from a range of recently published case studies (Cohen et al. 2015). Each of these three papers concludes by outlining a way forward on addressing important research questions related to indicators, trends and future scenarios of water availability in Canada.

What is water availability?

Water availability can be defined as: (1) the amount of water present at the mouth of a particular drainage basin, or (2) a climatological index of humidity or aridity expressed as a single number for a region, or shown continuously over space. Units of measurement vary depending on whether the index is expressed as a quantity of liquid (not frozen) water, or as a ratio that describes a level of humidity/aridity, or an amount of water per unit of area or population.

Indicators are one of the most commonly used methods for estimating water availability, as they enable the synthesis of large amounts of complex information to be presented in a simplified format that can be easily understood. In Canada, different levels of government (federal, provincial and municipal) have developed both water quality and water quantity indicators. Tracked over time, an indicator can provide information on the condition of a phenomenon and have significance extending beyond that associated with the properties of particular statistics (Bond et al. 2005, in Dunn and Bakker 2009). Indicators enable the state of Canada’s water to be assessed.

However, to our knowledge, systematic monitoring of water availability is not done in Canada (in contrast to other countries). This knowledge gap impedes our ability to adequately manage water supply and demand. Indicators can be used to create baselines against which water-related variables can be measured over time. Indicators can also be used to compare conditions in different locations and help to build a comparative picture of water availability across Canada (Dunn and Bakker 2009).

Climate-based indicators

Climate-based indicators have been used to measure key aspects of the climate system. Climate station point data, gridded data (national or global) and/or paleo-data from pre-instrumental records can be used.

Three types of climate-based indicators are currently used to estimate water availability. Precipitation minus evapotranspiration, or P – E, can be used to calculate water availability. Meteorological drought indices such as the Palmer Drought Severity Index (PDSI) and the Standardized Precipitation Index (SPI) can be used to calculate historical and current moisture surpluses and/or deficits for a location (Heim 2002; Steinemann et al. 2005). Soil moisture is an important component in determining water availability, especially in the Prairie Provinces. Soil moisture levels in this region are being assessed using the Variable Infiltration Capacity (VIC) model (Wen et al. 2011).

P – E

The difference between precipitation (P) and evaporation (E) (or potential evapotranspiration [PET], or actual evapotranspiration [AET]) is referred to as a water balance (or budget) approach, which enables calculation of water availability directly from climate variables. In this case, runoff (RO) is expressed as RO = P – E, or RO = P – AET.

There was an early attempt to calculate P – AET but this was not used to predict water availability (Hydrologic Atlas of Canada 1978). The 1978 Atlas contained a map depicting water balance throughout Canada, but this map was not updated in subsequent versions of the Atlas. The methodology was based on applying streamflow data as a correction for P and AET. In other words, RO, as represented by streamflow, was used as the basis for adjustment of P and AET.

For the Atlas mapping, average annual P for 1941–1970 was obtained from the station network and gridded. Each of the 1290 grid squares had an area of 10,000 km², and was oriented along lines of latitude one degree apart. Average annual RO was derived from annual streamflow at gauging stations, for the same grid. AET was obtained by first calculating PET from class A evaporation pan data, and then applying published AET/PET ratios for characteristic land cover types (from Thornthwaite and Mather 1957) and assigning ratios to the same grid. For example, the ratios for open water and urban areas would be 1 and 0.1, respectively. A ratio between 1.0 and 0.1 was assigned to areas covered by forest, grassland or tundra.

As mentioned earlier, evapotranspiration may be estimated by calculating the water balance (or budget) for a drainage basin. ET is an important climate variable, especially for the Prairie Provinces and Great Lakes region. But there are few actual evaporation (or evapotranspiration) monitoring stations in Canada.

ET is the sum of evaporation and plant transpiration from the Earth’s land surface to the atmosphere. ET can be measured or estimated using several methods. Pan evaporation is a measurement that has been used for decades to measure the quantity of evaporation at a given location. The most commonly used evaporation pan is the “Class A” evaporation pan. Pan evaporation data can be used to estimate lake evaporation, but transpiration and evaporation of intercepted rain on vegetation are unknown (Bruce and Clark 1966; Barry and Chorley 1987).

Potential evapotranspiration (PET) is the amount of water that would be evaporated and transpired if there were sufficient water available. Actual evapotranspiration (AET) is said to equal potential evapotranspiration when there is no limit on water supply. Several researchers have used different approaches to estimate or model evaporation, actual and potential evapotranspiration values for Canada (Bruce and Clark 1966; Barry and Chorley 1987).

Burn and Hesch (2006) compared trends for pan and potential evaporation for 11 sites on the Prairie Provinces for the months of May to September for different periods from the 1960s to early 2000s. All time series were analyzed for trends using the Mann–Kendall trend test.

A more recently developed indicator of this type is the Climate Moisture Index (CMI), in which the ratio of P/PET or PET/P is used to create an aggregate measure of water availability that ranges from –1 for arid conditions to +1 for humid conditions (Vörösmarty et al. 2005):

•	Humid, where P > PET; CMI = 1 – (PET/P);
•	Arid, where P < PET; CMI = (P/PET) – 1.

The original PET function is found in Shuttleworth and Wallace (1985), which is based on measuring the energy fluxes among soil, vegetation canopy and the atmosphere immediately above the canopy. A comparison of this model with other evapotranspiration models is provided by Stannard (1993).

The CMI has been produced for the 1960–1995 period as part of the Global Water System Project (Vörösmarty et al. 2005). Data sets and maps are available online (Global Water System Project [GWSP] 2014).

For Canada, much of the Atlantic Provinces, Quebec, Ontario, Nunavut and coastal British Columbia (BC) are classified as humid (CMI = 0.25 to 1.0). Areas considered sub-humid (0.0 to 0.25) include lands adjacent to the semi-arid areas (–0.25 to 0) of Alberta, southern Saskatchewan, and the Okanagan and northeast regions of BC. Parts of southwest Yukon and the southern Mackenzie Valley within the Northwest Territories are also classified as semi-arid. There are no regions in Canada classified as arid (–1.0 to –0.25), but there are arid areas within North Dakota, Montana and eastern Washington, just south of the Canada–United States border.

Sauchyn (2010) calculated historical trends in the summer CMI (P – PET) for the Prairie Provinces. Barrow (2010) calculated PE and CMI (P – PE) using two different methodologies (Thornthwaite, and simplified Penman–Monteith method) for the water year (October–September) and for three summer months (May–June–July) for the reference period 1971–2000 for the Canadian Prairies. Gridded observed data (ANUSPLIN) and climate model output were used (Hutchinson 2004).

Palmer Drought Severity Index (PDSI)

The Palmer Drought Severity Index (PDSI) employs both precipitation and temperature in its calculation, as well as local available soil water capacity, thus allowing for the effects of evapotranspiration. The PDSI was the first comprehensive drought index developed in the United States in 1965 by W. Palmer, and is still one of the most widely used meteorological drought indices. It is particularly useful in providing drought comparisons over relatively large areas, and historical drought information over a long period of record (computations in the United States are available for more than 100 years; (Steinemann et al. 2005; Heim 2002). However, the empirical nature of the PDSI and the fact that it was specifically developed for American agricultural regions limit its application in other countries (Heim 2002). In particular, PDSI does not consider variability in precipitation and thus does not perform well in regions with extreme variability in rainfall or runoff (Steinemann et al. 2005).

PDSI has been used to assess historical drought trends for some regions of Canada, especially in the Prairie Provinces, and in global-scale assessments. A real-time national drought model was developed in Canada by Agriculture and Agri-Food Canada (AAFC)2012). An example is shown in Figure 1. This version of PDSI uses real-time drought indices to help create the Canadian component of the North American Drought Monitor (NADM). The PDSI calculation used in Canada has modifications which attempt to account for the Canadian climate. These modifications include adjustments to the soil moisture component by coupling the Versatile Soil Moisture Budget with the original PDSI (Akinremi et al. 1996; Baier et al. 2000). Instead of the original two-layer soil moisture model, the modified PDSI currently uses a six-layer soil moisture model, which extends to a depth of 120 cm. Through other modifications, below-freezing conditions are now considered, with a new regional climate correction factor, where snow accumulation and snow melt relationships are applied (Akinremi et al. 1996; Baier et al. 2000). Finally, a major modification to the original PDSI is in the relationship used for calculating evapotranspiration; the Priestley and Taylor equation is now used instead of the Thornthwaite method (AAFC 2012; Baier et al. 2000). Current and historical maps of the modified PDSI are available on the Canadian Drought Watch website (AAFC 2012). National PDSI maps are produced each month for the agricultural land extent.

Figure 1. Palmer Drought Index for July 2014. Prepared by Agriculture and Agri-Food Canada’s National Agroclimate Information Service (NAIS). Data are provided through partnership with Environment Canada. The original version of the NAIS Drought Model was supplied by Alberta Agriculture and Rural Development, which partners with NAIS to foster ongoing development. Created 5 August 2014. Map obtained 26 August 2014, from http://www.agr.gc.ca/drought (Agriculture and Agri-Food Canada 2014, with permission).

Sheffield et al. (2012) looked at global drought trends for the period 1948–2008 using both the original and self-calibrating PDSI models, and both the Thornthwaite and Penman–Monteith algorithms for PE. They also combined atmospheric reanalysis data with available remote sensing and ground observation data. Millett et al. (2009) calculated PDSI trends from 1906 to 2000 for the Prairie Pothole Region (PPR) of North America, which includes portions of southern Alberta, Saskatchewan and Manitoba.

Standardized Precipitation Index (SPI)

The Standardized Precipitation Index, or SPI, is a probability index that was developed to give a representation of abnormal wetness and dryness. The SPI is a universal index which is referenced to the local climate, and can be used to monitor conditions on a variety of time scales, such as 1-, 2-, 3-, 6-, 9- … 72-month periods. This temporal flexibility allows the SPI to be useful in both short-term agricultural and long-term hydrological applications.

The SPI, as originally developed by McKee et al. (1993), is based on four arbitrarily defined drought categories or thresholds: SPI values of 0 to –0.99 for mild drought, –1.0 to –1.49 for moderate drought, –1.5 to –1.99 for severe drought and less than –2.0 for extreme drought. The SPI is referenced to the long-term precipitation record for a particular location. The actual distribution is fitted to a theoretical probability distribution (in this case, log-Pearson III). The percentile of the fitted distribution is then associated with the same percentile on a Gaussian distribution, which gives a Z score for the standard distribution. The Z score is the value of the SPI. Thus, a particular SPI can be thought of as the number of standard deviations from the median, mean and the cumulative probability (or percentile) which can be derived from statistical tables.

Experts participating in the World Meteorological Organization (WMO) Inter-Regional Workshop on Indices and Early Warning Systems for Drought in 2009 had a consensus agreement (i.e. the Lincoln declaration) that the SPI should be used by all national meteorological and hydrological services around the world to characterize meteorological droughts (WMO 2009). However, since SPI relies solely on precipitation data as input for the calculations, this can be a problem in semi-arid or arid climates where evaporation (or evapotranspiration) is an important variable that is not included.

Richards and Burridge (2006) calculated SPI values for agricultural areas of Canada using gridded precipitation data from 1900 to 2004. The Canadian national drought model uses the same SPI scale as the NADM mentioned above, but there are some differences. One difference between the two SPI calculations lies in the type of distribution considered; the Canadian method uses a gamma distribution while the NADM uses a Pearson Type III distribution. However, the difference between the two distributions has been found to be insignificant (Richards and Burridge 2006). As well, the time series on which the SPI calculations are based differ; the Canadian method currently uses a base period from 1960 to the most current available data, while the NADM uses a 50-year period (1951 to 2000). The Canadian model uses a base period starting in 1960 in order to maximize the number of stations. Richards and Burridge (2006) found that using a differing base period for historical SPI calculations can create inconsistent results.

On an operational basis, the accuracy of a calculated value will increase as more data become available. As for the PDSI, current and historical Canada-wide SPI maps are available monthly from the Canadian Drought Watch website (AAFC 2012).

Bonsal et al. (2013) used monthly gridded temperature and precipitation data that were interpolated at a 10-km resolution using ANUSPLIN (Hutchinson 2004). Summer PDSI and SPI values were calculated at the original 10-km resolution and then spatially aggregated to a 50-km grid to reduce the amount of data and to correspond with the resolution of the future drought scenarios. Bonsal et al. (2013) reconstructed areally averaged summer PDSI and SPI values for the southern Prairies study area using a network of tree rings from sites in southwestern Alberta. The PDSI and SPI reconstruction uses a principal component analysis (PCA) procedure on the 23 site chronologies that are significantly correlated with areally averaged summer PDSI and SPI values for the instrumental record (1901–2005).

One way to address the weakness of SPI is to use the Standardized Precipitation-Evapotranspiration Index (SPEI) instead. The SPEI uses the monthly (or weekly) difference between precipitation and the Precipitation–Evapotranspiration Index (PEI), which is a simple water balance methodology based upon the Thornthwaite method. Complete data sets of monthly (or weekly) temperature and precipitation data are needed to calculate the SPEI. A new global gridded SPEI data set has been developed which covers time scales from 1 to 48 months at a spatial resolution of 0.5 degrees for the period 1901–2006. The global gridded SPEI data set is freely available under an Open Database license with the Spanish National Research Council (Consejo Superior de Investigaciones Cientificas [CSIC] 2012). Work has started to use this global gridded SPEI data set for Canada (V. Wittrock, pers. comm., 21 December 2012; B. Bonsal, pers. comm., 11 February 2013).

Other precipitation and drought indices

Mekis and Vincent (2008) calculated trends for six different precipitation indices for southern Canada and all of Canada for the periods 1900–2007 and 1950–2007, respectively. The six indices are: annual total precipitation, summer total precipitation, days with rain > trace, days with rain >10 mm, maximum number of consecutive dry days, and very wet days ≥ 95^th percentile.

Mekis and Vincent (2011) made improvements in the adjusted daily precipitation data set for Canada. The data set has been updated and now includes data for 464 stations for the period 1900–2009, from the National Climate Data Archive of Environment Canada (Environment Canada 2014a). Daily rainfall and snowfall amounts were adjusted for known measurement issues, such as wind undercatch, evaporation and wetting losses for each type of rain gauge, snow water equivalent from ruler measurements, trace observations and accumulated amounts from several days. Trend analyses are summarized in paper II (Mortsch et al. 2015).

Current and historical precipitation deficit maps (difference from normal, percentile and dry spell) are available from the Canadian Drought Watch website (AAFC, 2012). Regional (Pacific, Prairie, Ontario, Quebec and Atlantic) and national maps are produced for different time steps: 7-, 14-, 30-, 60-, 90-, 180-, 270- and 365-day rolling time scales, as well as for the agricultural year (September–August), growing season (April–October) and winter season (November–March).

Soil moisture

Soil moisture levels are important, especially in the Prairie Provinces, to assess and monitor for agricultural drought. However, there is no consistent long-term soil moisture monitoring network in Canada. Instead, historical, real-time soil moisture levels and soil moisture forecasts are being calculated using the University of Washington’s VIC (Variable Infiltration Capacity) model by Wen et al. (2011). VIC is a land surface macroscale hydrologic model that can model soil moisture, snow water equivalent (SWE) and runoff levels for model grid cells over large areas.

Wen et al. (2011) reconstructed 60 years (1950–2009) of daily soil moisture values for three soil layers (0–20 cm, 20–100 cm and 0–100 cm) for the Prairie Provinces. Historical (daily, monthly, seasonal) and real-time Soil Moisture Anomaly Percentage Index (SMAPI) maps for the Prairie Provinces are produced and archived at Wen’s McGill University web page (Wen 2012). Real-time VIC Prairie drought forecast maps were also produced by Wen (until November 2012) and are available on the same website.

Wen’s real-time SMAPI forecasts were used as input data to the AAFC Drought Monitor. However, it is not clear whether these data will be available, as part of this research was funded by the Canadian Foundation for Climatic and Atmospheric Sciences’ DRI (Drought Research Initiative) which ended in March 2012.

Hydrology-based indicators

Hydrologic indicators have been developed to describe key aspects of the hydrologic regime and to characterize the hydrologic behavior of a basin. These metrics simplify, in a robust, quantitative manner, the complexity of the hydrologic system, and facilitate analyses such as comparisons among basins or examination of temporal trends within a basin. This section briefly describes three indicators: (1) streamflow, (2) runoff (RO) and (3) water yield. Runoff and water yield, as well as precipitation and evapotranspiration, are spatially distributed quantities, while streamflow or discharge is measured at a point in the river system.

Streamflow

A traditional metric for hydrology is streamflow, or discharge, representing the volume of water flowing past a point on a river in a unit of time (reported in L/s or m³/s, or for longer time periods, cubic decameters [dam³], one of which is 1000 m³). Streamflow can include contributions from surface runoff, interflow, groundwater flow and natural or artificial regulation. Discharge is considered a good measure of integrated terrestrial runoff, but a highly instrumented system is required to represent the spatial distribution of runoff within a large basin (Fekete et al. 2002). Streamflow data are obtained either through measurements and analyses – observed flow or modelling – defining “naturalized” flow or flow in ungauged basins.

Observed flow or discharge in many cases is a derived value and not a direct measurement per se. Systematic measurements of stage (water level) or stage and discharge are obtained at gauging stations. These records are used to develop a stage–discharge (or “rating”) curve. This relationship is then used to generate a time series of daily streamflow data from water level data; the latter being easier to automatically record in many cases. The quality of these data is dependent on the accuracy of the stage–discharge relationship and the frequency of field measurements used to calibrate the curves. These relationships or calibrations may change with seasons and conditions; there can be a high degree of uncertainty under conditions associated with ice cover, spring break-up and fall freeze-up, the spring freshet, floods, extensive plant growth and extreme low flow. These hydrometric sites are frequently/dominantly located on the mainstem of rivers and represent large-scale integrated basin hydrology.

Human influences, such as reservoirs, dams, water-taking, diversions, and land cover and land use change, are imbedded in the streamflow record, affecting the rate, magnitude, timing, duration and frequency of flow. Hydrometric stations that have had streamflow timing or amount, for example, altered by water management operations related to dams, reservoirs and/or diversions are identified as having “regulated” flow. Stations without these influences are considered to have “natural” flow, although that does not mean they are devoid of anthropogenic perturbations to streamflow.

Daily and monthly means of streamflow and water levels from the extensive Canadian network are contained in the hydrometric database, HYDAT, of the Water Survey of Canada (WSC; Environment Canada 2014b). A subset of long-term observing stations have been organized in the Reference Hydrometric Basin Network (RHBN); these stations are assumed to represent relatively pristine or stable land use and hydrologic conditions with little influence of regulation or diversions, and high accuracy of observations (Burn and Hag Elnur 2002). Specifically the criteria used to select the stations included: breadth of spatial coverage, degree of development in the basin (pristine with less than 10% of surface area modified or more developed but stable over time); no significant diversions or regulation (“natural” versus regulated) in the river system, although basins with structures controlling less than 5% of the area of a basin were considered; more than 20 years of data; and operational judgment of the accuracy of the records, with a particular focus on the accuracy of the rating curve under open-water and ice conditions and the stability of the control (Brimley et al. 1999; Harvey et al. 1999). Over 240 streamflow (continuous and seasonal discharge) stations and six lake level stations were identified.

The RHBN is a complement to the Reference Climate Station (RCS) network of over 300 stations, intended for determining trends in temperature and precipitation data (Environment Canada 2008); these data sets are of the highest quality and can be used in climate change detection, monitoring and assessment studies. While the RHBN stations have been carefully selected, the streamflow time series have not undergone a similar assessment (and adjustment) for homogeneity as the daily temperature and precipitation data in Canada.

The RHBN has been used extensively to assess trends in Canada’s water availability related to a changing climate; the data set has also been used to develop the national Water Availability Indicator (WAI) as part of the Canadian Environmental Sustainability Indicators (CESI) program (Environment Canada 2012a) described further below.

Reviews have identified indicators for assessing climate change and hydrologic trends (Kite and Harvey 1992; Lawford 1992; Mitosek 1992). Analyses have focused on streamflow variables as indicators since they represent a spatially integrated hydrologic response (Burn and Hag Elnur 2002). Indicators are used to represent important components of the hydrologic regime such as average water availability, seasonal variation, timing and magnitude of key events such as spring freshet, and extremes such as high- and low- flows, although other important phenomena such as duration of ice cover, break-up and freeze-up have also been included since they are tied to temperature change (Table 1; Burn et al. 2012).

Table 1. Selected list of indicators used for hydro-climate trend detection in Canada.

Variable	References
Annual mean streamflow	McCabe and Wolock (1997); Adamowski and Bocci (2001); Zhang et al. (2001); Burn and Hag Elnur (2002); Burn et al. (2004); Swansburg et al. (2004); Déry and Wood (2005); Rood et al. (2005); Abdul Aziz and Burn (2006); McClelland et al. (2006); St. George (2007); Brabets and Walvoord (2009); Déry et al. (2009b); St. Jacques et al. (2010)
Total annual discharge	Déry and Wood (2005); Déry et al. (2005); Déry et al. (2009a, 2009b); Déry et al. (2009); McClelland et al. (2006)
Date of centroid of annual streamflow	Zhang et al. (2001); Déry et al. (2009b)
Date of 5th, 10th, 90th, 95th percentile of annual streamflow volume	Burn (2008)
Annual minimum daily mean streamflow	Adamowski and Bocci (2001); Zhang et al. (2001); Burn et al. (2004); Swansburg et al. (2004); Déry et al. (2005); Abdul Aziz and Burn (2006); Monk et al. (2012)
Date of annual minimum daily mean streamflow	Abdul Aziz and Burn (2006); Monk et al. (2012)
Annual maximum daily mean streamflow	Adamowski and Bocci (2001); Zhang et al. (2001); Burn and Hag Elnur (2002); Woo and Thorne (2003); Burn et al. (2004); Swansburg et al. (2004); Déry et al. (2005); Abdul Aziz and Burn (2006); Cunderlik and Ouarda (2009); Assani et al. (2010); Burn et al. (2010); Monk et al. (2012)
Date of annual maximum daily mean streamflow	Burn (1994); Zhang et al. (2001); Burn and Hag Elnur (2002); Burn et al. (2004); Abdul Aziz and Burn (2006); Déry et al. (2009b); Burn et al. (2010); Monk et al. (2012)
1-, 7-, 15- and 30-day annual low flows	Khaliq et al. (2008); Burn et al. (2010)
Monthly mean streamflow	Adamowski and Bocci (2001); Zhang et al. (2001); Burn and Hag Elnur (2002); Woo and Thorne (2003); Burn et al. (2004); Déry et al. (2005); Abdul Aziz and Burn (2006); St. George (2007); Rood et al. (2008); Brabets and Walvoord (2009)
Monthly maximum flow	Cunderlik and Burn (2004)
Average spring flow	Woo and Thorne (2003); Rood et al. (2008); Brabets and Walvoord (2009)
Spring maximum daily flow	Abdul Aziz and Burn (2006); Cunderlik and Ouarda (2009); Burn et al. (2008); Burn et al. (2010)
Date of spring maximum daily flow	Burn (1994); Zhang et al. (2001); Burn et al. (2004); Swansburg et al. (2004); Déry et al. (2005); Abdul Aziz and Burn (2006); Burn (2008); Burn et al. (2008); Cunderlik and Ouarda (2009); Burn et al. (2010); Monk et al. (2012)
Date of spring freshet	Zhang et al. (2001); Swansberg et al. (2004); Déry et al. (2005); Abdul Aziz and Burn (2006); Burn (2008); Monk et al. (2012)
Average summer flow	Abdul Aziz and Burn (2006); Schindler and Donahue (2006); Rood et al. (2008); Brabets and Walvoord (2009)
Largest summer rainfall event flow	Burn et al. (2008)
Date of largest summer rainfall runoff event	Burn et al. (2008)
Average fall flow	Abdul Aziz and Burn (2006); Rood et al. (2008); Brabets and Walvoord (2009)
Fall rainfall maximum flood flow	Cunderlik and Ouarda (2009)
Date of fall rainfall maximum flood flow	Cunderlik and Ouarda (2009)
Average winter flow	Woo and Thorne (2003); Abdul Aziz and Burn (2006); Rood et al. (2008); Brabets and Walvoord (2009)
1-, 7-, 15- and 30-day winter low flow	Khaliq et al. (2008); Ehsanzadeh and Adamowski (2010)
1-, 7-, 15- and 30-day summer low flow	Khaliq et al. (2008); Ehsanzadeh and Adamowski (2010); Assani et al. (2012)
Cumulative (1 March to 31 October) seasonal runoff volume	Burn et al. (2008)
Magnitude of flow during maximum rainfall event	Burn et al. (2010)
Number of rain high flow events	Burn et al. (2010)
Date of spring ice break-up	Zhang et al. (2001); Burn and Hag Elnur (2002); Swansberg et al. (2004); Brabets and Walvoord (2009); Monk et al. (2012)
Date of river freeze-up	Zhang et al. (2001); Burn and Hag Elnur (2002); Woo and Thorne (2003); Monk et al. (2012)
Duration of ice cover	Zhang et al. (2001); Burn and Hag Elnur (2002)
Flow at freeze-up, spring freshet and ice break-up	Monk et al. (2012)
Number of flow reversals	Monk et al. (2012)

Déry et al. (2009a) accounted for different runoff-generating mechanisms for rivers of western Canada and stratified the data representing pluvial, nival and glacial rivers to assess changes in the timing and quantity of runoff. Monk et al. (2011) identified 32 hydro-ecological variables to explore temporal and spatial trends in streamflow attributes (magnitude, duration, timing, frequency and rate of change) and examined implications for aquatic habitat suitability. Subsequently, they considered cold-region processes (snow and ice) and identified 13 indicators relevant to ecological flow decision making (Monk et al. 2012). These indicators represent a shift in approach – from choosing indicators to determine whether a change in the time series is occurring, to developing indicators that are relevant to decision making, in this case an eco-hydrological context.

Runoff

Runoff is defined as that part of precipitation that does not evaporate nor transpire, but eventually leaves the watershed as a surface streamflow, whatever the flow pathway that the water has followed on its way to the stream channel. This includes both surface and subsurface runoff pathways (Beven 2003). This means that runoff is dependent on two larger independent processes, P and AET, as well as on the vegetation, soils and other characteristics of the landscape within the watershed. Runoff is expressed as a volume or depth for a drainage area, as is the case with the water yield term discussed below. An example of a depth indicator is mm per unit time, such as day or week, to standardize the effects of the size of a drainage area (Monk et al. 2011).

The Water and Global Change Project (WATCH) includes a comparison of five global hydrology models (GHMs) and six land surface models (LSMs) that are constructed within the framework of global climate models. WATCH is described by Haddeland et al. (2011), including its Water Model Intercomparison Project (WaterMIP). Each of these 11 models employs a runoff scheme, defined either as saturation excess or infiltration excess of a soil column. These are tied to the other components of these models, including water balance, energy balance, PET and snow accumulation/melt. These models were first run with the 40-year European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) product, provided at half-degree resolution. Haddeland et al. (2011) describe how this data set was applied, including elevation adjustments for temperature and corrections for snowfall undercatch.

Runoff from both GHMs and LSMs appear to overestimate global terrestrial runoff, when compared with data available from the Global Runoff Data Centre (GRDC 2013). Haddeland et al. (2011) suggest that this is due in part to the snowfall undercatch correction, and in part to underestimation of evapotranspiration. Also, there is wide variation among models’ partitioning of precipitation into snow and rain in regions where winter temperatures are close to 0°C, so some models will have higher estimates of snow water equivalent (SWE) in winter and spring.

Runoff fraction tends to be lower for the LSMs than for the GHMs. This means that LSMs will tend to be drier than GHMs in most regions. In a comparison of model results for the Mackenzie Basin, the LSMs actually had slightly higher runoff ratios, and so were slightly wetter. The GHMs showed earlier higher seasonal snowmelt peaks in runoff, while the LSMs were wetter during the summer. There was no consistent overprediction or underprediction in the Arctic.

Another attempt to model global runoff is the land surface scheme coupled to a river routing model within the HadCM3 and HadGEM1 climate models. Falloon et al. (2011) describe the Met Office Surface Exchange Scheme (MOSES), which calculates runoff for each grid within the climate models, and Total Runoff Integrating Pathways (TRIP), which is a 1° gridded river routing network which interpolates runoff values from the latest version of MOSES, called MOSES2. It should be noted that global-scale estimates of streamflow for areas of complex topography (such as BC) may be underestimated. Whitfield and Spence (2011) showed that for BC, the use of three representative stations by WMO as the basis for estimates of flow to ocean for the GRDC likely underestimated flow in all months, when compared with application of a larger set of 74 stations. Much of this is due to variations in precipitation, snowmelt and glacier melt across the complex topography of this region, and areas of very high runoff were likely missed in the global study.

In Canada, the MESH modelling system (Modélisation Environnementale Communautaire – MEC; MESH = MEC Surface and Hydrology) is being developed to couple atmosphere, land and hydrology models (Pietroniro et al. 2007). It is still being tested (e.g. MacLean et al. 2010), and has not yet been applied to climate change scenarios.

Water yield

The term “water yield” was used in a study by Statistics Canada (2010). It has been expressed as a total volume (km³) for a drainage region, volume per unit area (m³ per m², or a column expressed as “m”), and volume per capita (m³ per person). Freshwater volume is derived from unregulated flow measurements and is an estimate of renewable water over a defined period of time. Average annual volume for the 1971–2004 period for Canada was estimated (Statistics Canada 2010).

Demand/supply-based indicators

Demand/supply indicators focus on comparison of the relative availability of water, beyond water volume alone. In Canada, the only indicator that was found in the literature is the Water Availability Indicator (WAI). The WAI was developed by a federal interdepartmental working group, led by Environment Canada, to describe water availability across the country (Environment Canada 2012b).

Water availability refers to the volume of water in rivers compared with the amount of water used. This indicator is derived by calculating the ratio of annual water demand to annual water supply for select years at the sub-drainage area (SDA) scale for 164 watersheds across Canada (Environment Canada 2011). To calculate the ratio, Environment Canada estimated annual water demand for each SDA as the sum of municipal, industrial and agricultural water withdrawals from all flowing water. Water demand data were taken from three surveys: Statistics Canada’s Industrial Water Use Survey 2007 and Estimation of Water Used in Canadian Agriculture 2001, as well as Environment Canada’s Municipal Water and Wastewater Survey 2006 (Environment Canada 2011). Water supply is calculated using streamflow data collected by the Water Survey of Canada’s hydrometric stations (HYDAT). Water supply is estimated by extracting annual water flow (m³/s) data for the hydrometric station located at the basin outlet. The WAI does not account for water supply in lakes or groundwater aquifers, thus water availability may be underestimated (Environment Canada 2011).

The WAI was presented in maps and graphs at a national scale, but is also intended to be regionally relevant. All SDAs are assigned one of the Organisation for Economic Co-operation and Development’s (OECD) water availability threat classifications (see Environment Canada 2011) based on the water availability ratio:

•	High (more than 40% of available water is used): severe water stress;
•	Medium (between 20% and 40% of available water is used): both water supply and water demand need to be managed; conflicts among competing uses will need to be resolved;
•	Moderate (between 10% and 20% of available water is used): water availability becomes a constraint on development; significant investment is needed to provide for adequate water supply;
•	Low (less than 10% of available water is used): low water stress (OECD 2009; Environment Canada 2011).

The first nationwide results of the WAI were produced for 2005, 2007 and 2009 survey years, as well as a historical comparison map based on a 30-year yearly average water supply, and were released in the 2010–2011 Canada Water Act annual report (Environment Canada 2012b). The WAI was also an indicator in the Canadian Environmental Sustainability Indicators (CESI) program which provides data and information to track Canada’s performance on key environmental sustainability issues (Environment Canada 2012a). In 2012, the WAI indicator was discontinued due to organizational changes at Environment Canada in 2012 (F. Savignac, pers. comm., 9 May 2012).

A way forward?

Atmospheric components are essential factors in water availability. For example, even slight changes in precipitation and temperature can affect water supply through shifts in water storage, runoff and evaporation. Ideally, decision-makers would have easy access to data on streamflow, precipitation, temperature, groundwater, water quality, water balance, soil moisture, water use and ecosystem requirements, as well as an understanding of potential future variations. This is not the case in Canada.

Climate-based indicators for water availability are typically used for measuring drought frequency, severity, duration and extent. Much of the past work has focused on the Prairie Provinces, yet other regions of Canada are vulnerable to drought. It is possible to calculate both historical and future water budget data and drought indices such as PDSI, SPI (and SPEI) for climate stations across Canada as long as there is a complete monthly (or weekly) data set of temperature and precipitation data.

A common and important data gap is the lack of actual ET data. Different methods of calculating PET have been used in its place, with varying results. Unfortunately, as was shown for the Great Lakes, uncertainty in our knowledge of evaporation and ET leads to uncertainty in projections of future water availability. There is an increased need for monitoring of evaporation.

Another area that warrants consideration is the choice of hydrological indicators. The most prevalent indicators found in the literature were monthly mean and annual mean streamflow, and date of spring maximum flow. Some studies have provided daily streamflow information, particularly for analysis of ecological (or in-stream) flow (e.g. Monk et al. 2011), and these may be user-specific and/or watershed-specific. To enable comparisons of hydrological trends, is there a standard set of indicators, including some key ecological flow indicators, that could be defined by the research community in consultation with user interests? Also, is there a minimum length of record that the community can adhere to for trend detection studies? There is a need to explore the uncertainty associated with and methodological challenges in statistical analysis in hydro-climate research, for example, related to autocorrelation, and assumptions about the data.

The majority of Canada’s weather and hydrometric stations are located in the southern half of the country, where the population and economic pressures are greatest. As a result, Canada’s ability to monitor and assess its climatological and hydrological characteristics, both spatially and temporally, decreases significantly in the north. This has caused significant limitations in our ability to estimate freshwater availability in different parts of the country (Government of Canada 2009). With these knowledge and monitoring gaps, we will continue to face challenges in being able to manage our water resources sustainably (Environment Canada 2004, Government of Canada 2009; Council of Canadian Academies 2013).

Acknowledgements

We would like to thank Elaine Wheaton, Sean Fleming, Wendy Monk, Blaise Gauvin St-Denis, René Roy, Yonas Dibike and Donald Burn for their comments on earlier drafts of an unpublished report that was presented at the 2013 Canadian Meteorological and Oceanographic Society–Canadian Water Resources Association–Canadian Geophysical Union (CMOS–CWRA–CGU) Congress in Saskatoon, which has led to this series of papers. We would also like to thank Jenna Disch for assisting with the table and reference list. Opinions expressed here are those of the authors, and not necessarily those of Environment Canada.