The research described in this thesis examines how the hydrologic cycle is affected by climate changes in the Mackenzie River Basin (MRB) in northern Canada. The study focuses on five hydro-meteorological variables; runoff, evapotranspiration, storage, temperature and precipitation. Two different climate input data sets were used: Environment Canada gridded observed data and the European Center for Medium range Weather Forecasting (ECMWF) Re-Analysis climate data (ERA-40). In both data sets, runoff and evapotranspiration were modelled using the WATFLOOD hydrological model for the period of 1961 to 2002 on a 20 by 20 km grid. Trends were assessed on a monthly and annual basis using the Mann-Kendall non-parametric trend test. The hydrologic cycle in the MRB appears to be strongly influenced by climate change. The results reveal a general pattern of warming temperatures, and increasing precipitation and evapotranspiration. Overall decreases in runoff and in storage were detected from the Environment Canada data set while increases in runoff and in storage were detected from the ECMWF data set. The trends in runoff and evapotranspiration reflected changes in both precipitation and temperature. The spatial pattern of changes in runoff followed the pattern of change in precipitation very closely in most of the months, with the exception of March and October. The effect of changes in temperature is much more noticeable than that of changes in precipitation in March and October. The change in spatial distribution of evapotranspiration, on the other hand, matched the pattern of changes in temperature better; yet its seasonal pattern follows more closely to that of precipitation. The sensitivity of annual runoff to changes in climate was also estimated using a nonparametric estimator. Among the most important findings are: 1) runoff was more sensitive to precipitation and less sensitive to temperature; 2) runoff was positively correlated with precipitation and evapotranspiration; 3) runoff was negatively correlated with temperature, implying any increase in melt runoff from glaciers caused by increases in temperature were offset by losses due to evapotranspiration within the basin; 4) soil moisture storage may play an important role in the runoff and evapotranspiration processes; and 5) the sensitivity of mean annual runoff to changes in precipitation and evapotranspiration is typically lower along the Rocky Mountain chain, higher in the central zone of the Interior Plain, and highly varied in the Canadian Shield region in the basin. Correlation analysis suggested that the agreement between the two data sets is very weak at the grid-cell level. However, there was broad degree of consistencies in the seasonal and spatial patterns of trends between the two data sets, suggesting that the data are more reliable for identifying hydrological changes on a regional scale than at grid-cell level.

The research described in this thesis examines how the hydrologic cycle is affected by climate changes in the Mackenzie River Basin (MRB) in northern Canada. The study focuses on five hydro-meteorological variables; runoff, evapotranspiration, storage, temperature and precipitation. Two different climate input data sets were used: Environment Canada gridded observed data and the European Center for Medium range Weather Forecasting (ECMWF) Re-Analysis climate data (ERA-40). In both data sets, runoff and evapotranspiration were modelled using the WATFLOOD hydrological model for the period of 1961 to 2002 on a 20 by 20 km grid. Trends were assessed on a monthly and annual basis using the Mann-Kendall non-parametric trend test. The hydrologic cycle in the MRB appears to be strongly influenced by climate change. The results reveal a general pattern of warming temperatures, and increasing precipitation and evapotranspiration. Overall decreases in runoff and in storage were detected from the Environment Canada data set while increases in runoff and in storage were detected from the ECMWF data set. The trends in runoff and evapotranspiration reflected changes in both precipitation and temperature. The spatial pattern of changes in runoff followed the pattern of change in precipitation very closely in most of the months, with the exception of March and October. The effect of changes in temperature is much more noticeable than that of changes in precipitation in March and October. The change in spatial distribution of evapotranspiration, on the other hand, matched the pattern of changes in temperature better; yet its seasonal pattern follows more closely to that of precipitation. The sensitivity of annual runoff to changes in climate was also estimated using a nonparametric estimator. Among the most important findings are: 1) runoff was more sensitive to precipitation and less sensitive to temperature; 2) runoff was positively correlated with precipitation and evapotranspiration; 3) runoff was negatively correlated with temperature, implying any increase in melt runoff from glaciers caused by increases in temperature were offset by losses due to evapotranspiration within the basin; 4) soil moisture storage may play an important role in the runoff and evapotranspiration processes; and 5) the sensitivity of mean annual runoff to changes in precipitation and evapotranspiration is typically lower along the Rocky Mountain chain, higher in the central zone of the Interior Plain, and highly varied in the Canadian Shield region in the basin. Correlation analysis suggested that the agreement between the two data sets is very weak at the grid-cell level. However, there was broad degree of consistencies in the seasonal and spatial patterns of trends between the two data sets, suggesting that the data are more reliable for identifying hydrological changes on a regional scale than at grid-cell level.

Mercury (Hg) in rivers and streams of the Mackenzie River Basin (MRB) presents a risk to fish, aquatic mammals, and humans. This thesis makes use of newly-released water quality data, including total mercury (THg) and dissolved mercury (DHg), gathered through the NWT-Wide Community-based Water Quality Monitoring (CBM) program. Landscape metrics extracted from geospatial datasets (e.g. vegetation type cover and ground ice presence), along with the water quality parameters were transformed, as needed, and then assessed for relationships with Hg concentration in rivers using principle component analysis, correlation and linear regression. Transformed turbidity, total phosphate, total sulfate, and the 1st principle component representing total metals were positively correlated with log10 THg and log10 particulate Hg (PHg) (p < 0.05) in major tributaries on the MRB. Major tributaries had a greater proportion of THg as PHg. Dissolved organic carbon and total organic carbon were positivity correlated to log10 DHg in minor tributaries (p < 0.05) of the MRB. Logit-transformed ground ice presence was positively correlated with median log10 PHg and log10 THg (PHg; p < 0.05, THg; p < 0.05). Median log10 DHg was positively correlated with logit-transformed landcover metric mixed forest (R2 = 0.67), and negativity correlated with logit-transformed landcover metric sub-polar taiga needleleaf forest (R2 = 0.64) (p < 0.05). This suggests suspended particles, derived potentially from suspended mineral matter, kerogens, and/ or weathering of sulfides could be a control on PHg in major tributaries of the MRB, while thicker, organic soils, potentially in forested areas, releasing TOC and DOC, could control export of DHg to minor tributaries in the MRB.

Soil moisture plays a major role in the hydrologic water balance and is the basis for most hydrological models. It influences the partitioning of energy and moisture inputs at the land surface. Because of its importance, it has been used as a key variable for many hydrological studies such as flood forecasting, drought studies and the determination of groundwater recharge. Therefore, spatially distributed soil moisture with reasonable temporal resolution is considered a valuable source of information for hydrological model parameterization and validation. Unfortunately, soil moisture is difficult to measure and remains essentially unmeasured over spatial and temporal scales needed for a number of hydrological model applications. In 2002, the Gravity Recovery And Climate Experiment (GRACE) satellite platform was launched to measure, among other things, the gravitational field of the earth. Over its life span, these orbiting satellites have produced time series of mass changes of the earth-atmosphere system. The subsequent outcome of this, after integration over a number of years, is a time series of highly refined images of the earth's mass distribution. In addition to quantifying the static distribution of mass, the month-to-month variation in the earth's gravitational field are indicative of the integrated value of the subsurface total water storage for specific catchments. Utilization of these natural changes in the earth's gravitational field entails the transformation of the derived GRACE geopotential spherical harmonic coefficients into spatially varying time series estimates of total water storage. These remotely sensed basin total water storage estimates can be routinely validated against independent estimates of total water storage from an atmospheric-based water balance approach or from well calibrated macroscale hydrologic models. The hydrological relevance and implications of remotely estimated GRACE total water storage over poorly gauged, wetland-dominated watershed as well as over a deltaic region underlain by a thick sand aquifer in Western Canada are the focus of this thesis. The domain of the first case study was the Mackenzie River Basin wherein the GRACE total water storage estimates were successfully inter-compared and validated with the atmospheric based water balance. These were then used to assess the WATCLASS hydrological model estimates of total water storage. The outcome of this inter-comparison revealed the potential application of the GRACE-based approach for the closure of the hydrological water balance of the Mackenzie River Basin as well as a dependable source of data for the calibration of traditional hydrological models. The Mackenzie River Basin result led to a second case study where the GRACE-based total water storage was validated using storage estimated from the atmospheric-based water balance P-E computations in conjunction with the measured streamflow records for the Saskatchewan River Basin at its Grand Rapids outlet in Manitoba. The fallout from this comparison was then applied to the characterization of the Prairie-wide 2002/2003 drought enabling the development of a new drought index now known as the Total Storage Deficit Index (TSDI). This study demonstrated the potential application of the GRACE-based technique as a tool for drought characterization in the Canadian Prairies. Finally, the hydroinformatic approach based on the artificial neural network (ANN) enabled the downscaling of the groundwater component from the total water storage estimate from the remote sensing satellite, GRACE. This was subsequently explored as an alternate source of calibration and validation for a hydrological modeling application over the Assiniboine Delta Aquifer in Manitoba. Interestingly, a high correlation exists between the simulated groundwater storage from the coupled hydrological model, CLM-PF and the downscaled groundwater time series storage from the remote sensing satellite GRACE over this 4,000 km2 deltaic basin in Canada.

Soil moisture plays a major role in the hydrologic water balance and is the basis for most hydrological models. It influences the partitioning of energy and moisture inputs at the land surface. Because of its importance, it has been used as a key variable for many hydrological studies such as flood forecasting, drought studies and the determination of groundwater recharge. Therefore, spatially distributed soil moisture with reasonable temporal resolution is considered a valuable source of information for hydrological model parameterization and validation. Unfortunately, soil moisture is difficult to measure and remains essentially unmeasured over spatial and temporal scales needed for a number of hydrological model applications. In 2002, the Gravity Recovery And Climate Experiment (GRACE) satellite platform was launched to measure, among other things, the gravitational field of the earth. Over its life span, these orbiting satellites have produced time series of mass changes of the earth-atmosphere system. The subsequent outcome of this, after integration over a number of years, is a time series of highly refined images of the earth's mass distribution. In addition to quantifying the static distribution of mass, the month-to-month variation in the earth's gravitational field are indicative of the integrated value of the subsurface total water storage for specific catchments. Utilization of these natural changes in the earth's gravitational field entails the transformation of the derived GRACE geopotential spherical harmonic coefficients into spatially varying time series estimates of total water storage. These remotely sensed basin total water storage estimates can be routinely validated against independent estimates of total water storage from an atmospheric-based water balance approach or from well calibrated macroscale hydrologic models. The hydrological relevance and implications of remotely estimated GRACE total water storage over poorly gauged, wetland-dominated watershed as well as over a deltaic region underlain by a thick sand aquifer in Western Canada are the focus of this thesis. The domain of the first case study was the Mackenzie River Basin wherein the GRACE total water storage estimates were successfully inter-compared and validated with the atmospheric based water balance. These were then used to assess the WATCLASS hydrological model estimates of total water storage. The outcome of this inter-comparison revealed the potential application of the GRACE-based approach for the closure of the hydrological water balance of the Mackenzie River Basin as well as a dependable source of data for the calibration of traditional hydrological models. The Mackenzie River Basin result led to a second case study where the GRACE-based total water storage was validated using storage estimated from the atmospheric-based water balance P-E computations in conjunction with the measured streamflow records for the Saskatchewan River Basin at its Grand Rapids outlet in Manitoba. The fallout from this comparison was then applied to the characterization of the Prairie-wide 2002/2003 drought enabling the development of a new drought index now known as the Total Storage Deficit Index (TSDI). This study demonstrated the potential application of the GRACE-based technique as a tool for drought characterization in the Canadian Prairies. Finally, the hydroinformatic approach based on the artificial neural network (ANN) enabled the downscaling of the groundwater component from the total water storage estimate from the remote sensing satellite, GRACE. This was subsequently explored as an alternate source of calibration and validation for a hydrological modeling application over the Assiniboine Delta Aquifer in Manitoba. Interestingly, a high correlation exists between the simulated groundwater storage from the coupled hydrological model, CLM-PF and the downscaled groundwater time series storage from the remote sensing satellite GRACE over this 4,000 km2 deltaic basin in Canada.

Hydrological data extracted directly from Water Survey of Canada archives covering the 1913-2002 time period is used to assess river ice break-up in the Mackenzie River basin. A return-period analysis indicates that 13 (14) of 28 sites in the basin are dominated by peak water-levels occurring during the spring break-up (open-water) period. One location has a mixed signal. A map of flooding regimes is discussed in terms of physical, hydrological and climatic controls. Annual break-up is found to progress from south to north, over a period representing ~¼ of the year. Average annual duration is ~8 weeks. The at site break-up period, recognized as the most dynamic time of the year on cold-regions river systems is found to last from 4 days to 4 weeks. Break-up timing (1966-1995) is found to be occurring earlier in the western portions of the basin (~3 days/decade), concurrent with late 20th century warming.