In cold regions, ice jams frequently result in severe flooding due to a rapid rise in water levels upstream of the jam. Sudden floods resulting from ice jams threaten human safety and cause damage to properties and infrastructure. Hence, ice-jam prediction tools can give an early warning to increase response time and minimize the possible damages. However, ice-jam prediction has always been a challenge as there is no analytical method available for this purpose. Nonetheless, ice jams form when some hydro-meteorological conditions happen, a few hours to a few days before the event. Ice-jam prediction can be addressed as a binary multivariate time-series classification. Deep learning techniques have been widely used for time-series classification in many fields such as finance, engineering, weather forecasting, and medicine. In this research, we successfully applied convolutional neural networks (CNN), long short-term memory (LSTM), and combined convolutional–long short-term memory (CNN-LSTM) networks to predict the formation of ice jams in 150 rivers in the province of Quebec (Canada). We also employed machine learning methods including support vector machine (SVM), k-nearest neighbors classifier (KNN), decision tree, and multilayer perceptron (MLP) for this purpose. The hydro-meteorological variables (e.g., temperature, precipitation, and snow depth) along with the corresponding jam or no-jam events are used as model inputs. Ten percent of the data were excluded from the model and set aside for testing, and 100 reshuffling and splitting iterations were applied to 80 % of the remaining data for training and 20 % for validation. The developed deep learning models achieved improvements in performance in comparison to the developed machine learning models. The results show that the CNN-LSTM model yields the best results in the validation and testing with F1 scores of 0.82 and 0.92, respectively. This demonstrates that CNN and LSTM models are complementary, and a combination of both further improves classification.

Understanding the hydrological dynamics of boreal wetland ecosystems (peatlands) is essential in order to better manage hydropower inter-annual productivity at the La Grande basin (Northern Quebec, QC, Canada). Given the remoteness and the huge dimension of the La Grande basin, it is imperative to develop remote sensing monitoring techniques to retrieve hydrological parameters. The main objective of this study is to find out if multi-date and multi-polarization Radar Satellite 2 (RADARSAT-2) (C-band) image analysis could detect seasonal variations of surface soil moisture conditions of the acrotelm. A change detection approach through the use of multi temporal indexes was chosen based on the assumption that the temporal variability of surface roughness and natural vegetation biomass is generally at a much longer time scale than that of surface soil moisture (Δ-Index is based on a reference image that represents dry soil, in order to maximize the sensitivity of σ° to changes in soil moisture with respect to the same location when soil is wet). The Δ-Index approach was tested with each polarization: σ° for fully polarimetric mode (HH, HV, VV) and the cross-polarization coefficient (HV/HH). Results show that the best regression adjustment with regard to surface soil moisture content in boreal wetlands was obtained with the cross-polarization coefficient. The cross-polarization multi-temporal index enables precise volumetric surface soil moisture estimation and monitoring on boreal wetlands, regardless of the influence of vegetation cover and surface roughness conditions (bias was under 1%, standard deviation and RMSE were under 10% for almost all estimation errors). Surface soil moisture estimation was more precise over permanently flooded areas than seasonally flooded ones (standard deviation is systematically greater for the seasonally flooded areas, at all analyzed scales), although the overall quality of the estimation is still precise. Cross-polarization ratio image analysis appears to be a useful mean to exploit radar data spatially, as we were able to relate changes in wetland eco-hydrological dynamics to variations in the intensity of the ratio.

Floods resulting from river ice jams pose a great risk to many riverside municipalities in Canada. The location of an ice jam is mainly influenced by channel morphology. The goal of this work was therefore to develop a simplified geospatial model to estimate the predisposition of a river channel to ice jams. Rather than predicting the timing of river ice breakup, the main question here was to predict where the broken ice is susceptible to jam based on the river's geomorphological characteristics. Thus, six parameters referred to potential causes for ice jams in the literature were initially selected: presence of an island, narrowing of the channel, high sinuosity, presence of a bridge, confluence of rivers, and slope break. A GIS-based tool was used to generate the aforementioned factors over regular-spaced segments along the entire channel using available geospatial data. An ice jam predisposition index (IJPI) was calculated by combining the weighted optimal factors. Three Canadian rivers (province of Québec) were chosen as test sites. The resulting maps were assessed from historical observations and local knowledge. Results show that 77 % of the observed ice jam sites on record occurred in river sections that the model considered as having high or medium predisposition. This leaves 23 % of false negative errors (missed occurrence). Between 7 and 11 % of the highly predisposed river sections did not have an ice jam on record (false-positive cases). Results, limitations, and potential improvements are discussed.

In Canada, climate change is expected to increase the extreme precipitation events by magnitude and frequency, leading to more intense and frequent river flooding. In this study, we attempt to map the flood hazard and damage under projected climate scenarios (2050 and 2080). The study was performed in the two most populated municipalities of the Petite Nation River Watershed, located in southern Quebec (Canada). The methodology follows a modelling approach, in which climate projections are derived from the Hydroclimatic Atlas of Southern Quebec following two representative concentration pathways (RCPs) scenarios, i.e., RCP 4.5 and RCP 8.5. These projections are used to predict future river flows. A frequency analysis was carried out with historical data of the peak flow (period 1969–2018) to derive different return periods (2, 20, and 100 years), which were then fed into the GARI tool (Gestion et Analyse du Risque d’Inondation). This tool is used to simulate flood hazard maps and to quantify future flood risk changes. Projected flood hazard (extent and depth) and damage maps were produced for the two municipalities under current and for future scenarios. The results indicate that the flood frequencies are expected to show a minor decrease in peak flows in the basin at the time horizons, 2050 and 2080. In addition, the depth and inundation areas will not significantly change for two time horizons, but instead show a minor decrease. Similarly, the projected flood damage changes in monetary losses are projected to decrease in the future. The results of this study allow one to identify present and future flood hazards and vulnerabilities, and should help decision-makers and the public to better understand the significance of climate change on flood risk in the Petite Nation River watershed.

This study has been conducted within the "Fonds pour la formation de chercheurs et l'aide à la recherche (FCAR) - Action concertée RADARSAT" program. The study shows the potential of RADARSAT-1 standard mode data (S1 and S7 beams) for mapping natural and exploited wetlands in southern Quebec. The best period for the acquisition of S1 beam data is during the growing season. However, an S7 beam mode image acquired in February (winter) can help to discriminate different vegetation densities within wetlands. With a maximum likelihood classification method, the data set giving the best results is the winter S7 image and two summer S1 images (May 7 or June 11 and July 28 or August 3). The large wetlands can be easily classified amongst other areas (i.e., agricultural, open water, forest, etc.) by this method and data set, but the different categories of vegetation communities within wetlands cannot be well discriminated. However, the use of a texture parameter (mean, 7 × 7 windows) can significantly improve the... Cette étude, réalisée dans le cadre du programme "Fonds pour la formation de chercheurs et l'aide à la recherche (FCAR) - Action concertée RADARSAT", montre le potentiel des images en mode S1 et S7 de RADARSAT-1 pour la cartographie des tourbières naturelles et exploitées du sud du Québec. Les périodes pertinentes pour l'acquisition des images S1 demeurent à l'intérieur de la saison végétative. Toutefois, les images hivernales en mode S7, sensibles à la densité du couvert, permettent de distinguer les forêts de certaines tourbières arbustives. Pour la classification par maximum de vraisemblance, la meilleure combinaison de trois images demeure l'image en mode S7 de début février et deux images en mode S1 (7 mai ou 11 juin et 28 juillet ou 3 août) et ce pour les deux sites expérimentaux. Bien que les tourbières puissent être bien séparées des autres utilisations du sol, les différentes catégories de tourbières n'ont toutefois pu être distinguées. Néanmoins, en utilisant un paramètre de texture (" moyenne "...

A new method for sensitivity analysis of water depths is presented based on a two-dimensional hydraulic model as a convenient and cost-effective alternative to Monte Carlo simulations. The method involves perturbation of the probability distribution of input variables. A relative sensitivity index is calculated for each variable, using the Gauss quadrature sampling, thus limiting the number of runs of the hydraulic model. The variable-related highest variation of the expected water depths is considered to be the most influential. The proposed method proved particularly efficient, requiring less information to describe model inputs and fewer model executions to calculate the sensitivity index. It was tested over a 45 km long reach of the Richelieu River, Canada. A 2D hydraulic model was used to solve the shallow water equations (SWE). Three input variables were considered: Flow rate, Manning’s coefficient, and topography of a shoal within the considered reach. Four flow scenarios were simulated with discharge rates of 759, 824, 936, and 1113 ${\mathrm{m}}^3/\mathrm{s}$ . The results show that the predicted water depths were most sensitive to the topography of the shoal, whereas the sensitivity indices of Manning’s coefficient and the flow rate were comparatively lower. These results are important for making better hydraulic models, taking into account the sensitivity analysis.