Arctic and Subarctic environments are among the most vulnerable regions to climate change. Increases in liquid precipitation and changes in snowmelt onset are cited as the main drivers of change in ...streamflow and water temperature patterns in some of the largest rivers of the Canadian Arctic. However, in spite of this evidence, there is still a lack of research on water temperature, particularly in the eastern Canadian Arctic. In this paper, we use the CEQUEAU hydrological‐water temperature model to derive consistent long‐term daily flow and stream temperature time series in Aux Mélèzes River, a non‐regulated basin (41 297 km2) in the eastern Canadian subarctic. The model was forced using reanalysis data from the fifth‐generation ECMWF atmospheric reanalyses (ERA5) from 1979 to 2020. We used water temperature derived from thermal infrared (TIR) images as reference data to calibrate CEQUEAU's water temperature model, with calibration performed using single‐site, multi‐site, and upscaling factors approaches. Our results indicate that the CEQUEAU model can simulate streamflow patterns in the river and shows excellent spatiotemporal performance with Kling‐Gupta Efficiency (KGE) metric >0.8. Using the best‐performing flow simulation as one of the inputs allowed us to produce synthetic daily water temperature time series throughout the basin, with the multi‐site calibration approach being the most accurate with root mean square errors (RMSE) <2.0°C. The validation of the water temperature simulations with a three‐year in situ data logger dataset yielded an RMSE = 1.38°C for the summer temperatures, highlighting the robustness of the calibrated parameters and the chosen calibration strategy. This research demonstrates the reliability of TIR imagery and ERA5 as sources of model calibration data in data‐sparse environments and underlines the CEQUEAU model as an assessment tool, opening the door to its use to assess climate change impact on the arctic regions of Canada.
RMSE of water temperature model for different locations. The RMSE has units of °C.
The development of nonstationary frequency analysis models is gaining popularity in the field of hydro‐climatology. Such models account for nonstationarities related to climate change and climate ...variability but at the price of added complexity. It has been debated if such models are worth developing considering the increase in uncertainty inherent to more complex models. However, the uncertainty associated to nonstationary models is rarely studied. The objective of this article is to compare the uncertainties in stationary and nonstationary models based on objective criteria. The study is based on observed rainfall data in the United Arab Emirates (UAE) where strong nonstationarities were observed. In this study, a nonstationary frequency analysis introducing covariates into the distribution parameters was carried out for total and maximum annual rainfalls observed in the UAE. The generalized extreme value (GEV) distribution was used to model annual maximum rainfalls and the gamma (G) distribution was used to model total annual rainfalls. A number of nonstationary models, using time and climate indices as covariates, were developed and compared to classical stationary frequency analysis models. Two climate oscillation patterns having strong impacts on precipitation in the UAE were selected: the Oceanic Niño Index and the Northern Oscillation Index. Results indicate that the inclusion of a climate oscillation index generally improves the fit of the models to the observed data and the inclusion of two covariates generally provides the overall best fits. Uncertainties of estimated quantiles were assessed with confidence intervals (CIs) computed with the parametric bootstrap method. Results show that for the small sample sizes in this study, the width of the CIs can be very large for extreme nonexceedance probabilities and for the most extreme values of the climate index covariates. The weaknesses of nonstationary models revealed by the bootstrap uncertainties are discussed and words of caution are formulated.
We compare here the uncertainties obtained for predictions of rainfall quantiles between stationary and nonstationary models. The uncertainties of quantile estimates are assessed with confidence intervals (CIs) computed with the parametric bootstrap method. The figure shows a comparison of extreme rainfall quantile estimates with CIs for a stationary model and a nonstationary model including the Northern Oscillation Index as covariate.
We present a potential growth thermal index (PGTI) and assess its correlation with juvenile Atlantic salmon Salmo salar fork length data collected near the end of the growth season in a range of ...latitudinal locations and geographic scales (watershed, regional, continental) across the American north‐east. The PGTI is based on two components: a water temperature‐dependent growth curve and a water temperature time series continuously describing the thermal environment preceding fish sampling. Testing various shapes and characteristics of the temperature–growth curve against fish length data revealed strong positive correlations for all combinations. PGTI warming, calculated only from the beginning of the growth season until maximum summer temperature is reached, consistently performed well in explaining fish size‐at‐age across the latitudinal gradient and the three geographic scales that were considered. Varying thermal contrasts created by repeat subsampling of the dataset showed that fish length is better explained by the level of thermal contrast within the dataset than the geographical scale of analysis. A simple generalized linear model using a log link function with PGTI warming, fish density and water discharge as predictors explained 87% of the variance of size‐at‐age of 0+ and 1+ juvenile Atlantic salmon.
Evaporative flux is a key component of hydrological budgets. Water loss through evapotranspiration reduces volumes available for run‐off. The transition from liquid to water vapour on open water ...surfaces requires heat. Consequently, evaporation act as a cooling mechanism during summer. Both river discharge and water temperature simulations are thus influenced by the methods used to model evaporation. In this paper, the impact of evapotranspiration estimation methods on simulated discharge is assessed using a semidistributed model on two Canadian watersheds. The impact of evaporation estimation methods on water temperature simulations is also evaluated. Finally, the validity of using the same formulation to simulate both of these processes is verified. Five well‐known evapotranspiration models and five evaporation models with different wind functions were tested. Results show a large disparity (18–22% of mean annual total evapotranspiration) among the evapotranspiration methods, leading to important differences in simulated discharge (3–25% of observed discharge). Larger differences result from evaporation estimation methods with mean annual divergences of 34–48%. This translates into a difference in mean summer water temperature of 1–15%. Results also show that the choice of model parameter has less influence than the choice of evapotranspiration method in discharge simulations. However, the parameter values influence thermal simulations in the same order of magnitude as the choice of evaporation estimation method. Overall, the results of this study suggest that evapotranspiration and open water evaporation should be represented separately in a hydrological modelling framework, especially when water temperature simulations are required.
The impact of evaporation/evapotranspiration estimation methods on flow and water temperature simulations is evaluated using the CEQUEAU model implemented on two Canadian watersheds. Five well‐known evapotranspiration models and five evaporation models with different wind functions were tested, yielding important differences in simulated flows and temperatures. The selection of the evapotranspiration model is more important than the process of calibration of model parameters, whereas appropriate calibration is as important as the choice of an evaporation algorithm for the temperature model.
ABSTRACT
Quantile estimates are generally interpreted in association with the return period concept in practical engineering. To do so with the peaks‐over‐threshold (POT) approach, combined ...Poisson‐generalized Pareto distributions (referred to as PD‐GPD model) must be considered. In this article, we evaluate the incorporation of non‐stationarity in the generalized Pareto distribution (GPD) and the Poisson distribution (PD) using, respectively, the smoothing‐based B‐spline functions and the logarithmic link function. Two models are proposed, a stationary PD combined to a non‐stationary GPD (referred to as PD0‐GPD1) and a combined non‐stationary PD and GPD (referred to as PD1‐GPD1). The teleconnections between hydro‐climatological variables and a number of large‐scale climate patterns allow using these climate indices as covariates in the development of non‐stationary extreme value models. The case study is made with daily precipitation amount time series from southeastern Canada and two climatic covariates, the Arctic Oscillation (AO) and the Pacific North American (PNA) indices. A comparison of PD0‐GPD1 and PD1‐GPD1 models showed that the incorporation of non‐stationarity in both POT models instead of solely in the GPD has an effect on the estimated quantiles. The use of the B‐spline function as link function between the GPD parameters and the considered climatic covariates provided flexible non‐stationary PD‐GPD models. Indeed, linear and nonlinear conditional quantiles are observed at various stations in the case study, opening an interesting perspective for further research on the physical mechanism behind these simple and complex interactions.
Using statistical tools like the cross‐wavelet analysis illustrated in the figure, common features of variability are found between precipitation extreme events and the Artic Oscillation index at the Upper Stewiacke station located in Nova Scotia (Canada). Using this index as covariate, we developed non‐stationary Poisson‐generalized Pareto models, which allow observing conditional quantiles with concave form. The proposed models are more flexible than classical extreme value non‐stationary models which often used prior assumption of linear dependence.
Compounding the joint impact of extreme river temperature and low flow characteristics can harm the aquatic habitat of certain organisms (e.g., ecototherm fish) and freshwater ecosystems. Considering ...only river temperature or low flow via univariate frequency distribution as a stress indicator would be incomplete. Maximum water temperature and low flow series are strongly negatively correlated; thus, their joint probability distribution can be helpful to assess better the risks associated with joint extreme events. This study incorporated the 2-D parametric copulas in the bivariate joint modelling of annual maximum river water temperature and corresponding low flow. This proposed bivariate framework is applied to 5 independent and identically distributed stations in Switzerland. Parametric 1-D probability density functions are employed in modelling the univariate marginal distribution of both variables separately. The efficacy of eighteen different parametric class negatively dependent 2-D copulas is tested. The best-fitted copulas and selected marginals are used to estimate joint return periods for quantiles corresponding to multiple return periods. The joint return periods of annual maximum temperatures conditional to low flows or vice versa are also estimated. Investigation reveals that the occurrence of bivariate events simultaneously is less frequent in the AND-joint case than in the OR-joint event case for all stations. Also, OR-return periods are less (nearly half) the value of univariate return periods. Secondly, higher conditional return periods are observed in annual maximum temperature (or low flow) when increasing the percentile value of the conditioning variable, i.e., low flow (or maximum temperature). Also, when the low flow (or water temperature) conditioning variable is fixed, higher bivariate event return periods are observed at a higher water temperature (or low flow) value. In conclusion, these estimated bivariate statistics can help provide a more complete picture for an adequate assessment of the risks associated with cold-water species.
Display omitted
•In Swiss Rivers, a parametric copula model was used to estimate the joint exceedance probability of negatively dependent extreme river temperature and low flow.•Simultaneous occurrences of bivariate events are less frequent in the AND-joint case than in the OR-joint event.•Higher bivariate return periods are observed in river temperature (or low flow) when increasing the percentile value of the conditioning variable, low flow (or river temperature).•Also, higher bivariate event return periods occur at higher river temperatures (or low flow) values when fixing conditioning variables (river temperature or low flow).•These bivariate statistics can better describe the cold-water species real risk during extreme events and help in their management.