Search

Accurate spring flow forecasts, primarily driven by snow accumulation and melt processes, are essential for decision-makers aiming to optimize hydro-electricity production and mitigate potential flood damages in snow-dominated regions. By integrating snowpack data from diverse sources (in-situ, remote sensing, and reanalysis) with modeled snow-related state variables through Data Assimilation (DA), there is the potential to leverage both modeling and observations for more accurate estimation of the resulting spring flow. However, challenges such as the lack of sufficient snow station networks and issues with optical and microwave sensors to provide Snow Water Equivalent (SWE) can hinder progress, particularly in regions with heterogeneous snowpack characteristics. Overcoming these challenges is crucial to realizing the full potential of snow DA in advancing spring flow forecasts. This thesis aims to optimize spring flow forecasting effectiveness using snow DA, where snow-related observations are incorporated into hydrological models. It investigates the best hydrological model structure (a lumped model: HSAMI, and a distributed model: HYDROTEL) for leveraging distributed SWE data provided by SNODAS (SNOw Data Assimilation System) dataset, which serves as the observation. This DA framework is applied on a large, heterogeneous northern Québec watershed (Outardes 4). It does so by updating SWE model states of HYDROTEL and HSAMI using the Ensemble Kalman Filter (EnKF) DA method. The simulated spring flow are then compared to observed spring flow data. Another scope of the thesis is to compare two reanalysis products with varying spatial resolutions and SWE representations (SNODAS and ERA5-Land) for improving 1-day spring flow forecasts in HYDROTEL over a smaller southern Québec watershed (Au Saumon). Finally, the thesis assesses the forecasting skill of the optimal model-observation combination over an extended 30-day forecast horizon using various probabilist, Des prévisions fiable du débit de printemps, principalement influencées par les processus d’accumulation et de fonte de la neige, sont essentielles pour les décideurs visant à optimiser la production hydroélectrique et à atténuer les dommages dus aux inondations dans les régions dominées par la neige. En intégrant les données sur le manteau neigeux provenant de diverses sources (in situ, télédétection et réanalyse) avec les variables d’état modélisées liées à la neige par le biais de l’assimilation de données (AD), il est possible de tirer parti à la fois de la modélisation et des observations pour une estimation plus précise du débit printanier.Cependant, le manque de réseaux de stations et les problèmes avec les capteurs optiques et micro-ondes pour fournir l’equivalent en eau de neige (EEN) peuvent freiner les progrès, notamment dans les régions avec un manteau neigeux hétérogène. Surmonter ces défis est crucial pour exploiter pleinement le potentiel de l’assimilation de données sur la neige dans les prévisions de débits printaniers. Cette thèse vise à optimiser les prévisions de débits printaniers en utilisant l’AD sur la neige, en intégrant les observations de neige dans les modèles hydrologiques. Elle explore la meilleure structure de modèle hydrologique (global : HSAMI, distribué : HYDROTEL) pour exploiter les données distribuées EEN fournies par l’ensemble de données SNODAS (système d’assimilation de données sur la neige), qui sert d’observation. Ce cadre d’AD est appliqué sur un grand bassin versant hétérogène du nord du Québec (Outardes 4). Elle le fait en mettant à jour les états d’EEN du modèle de HYDROTEL et HSAMI en utilisant la méthode d’AD de filtre de Kalman d’ensemble (EnKF). Les débits printaniers simulés sont ensuite comparés aux données de débits printaniers observés. Une autre perspective de la thèse est de comparer deuxproduitsderéanalyse avecdesrésolutions spatiales et des représentations d’EENdifférentes(SNODASetERA5-Land) pour améliorer les

This study aims to investigate the potential of using SAR images from Sentinel-1 to predict snow depth in sub-arctic alpine regions by employing Random Forest regression. In addition to predicting snow depth, the correlation between radar backscatter and point snow depth measurements is explored. Random Forest regression is a powerful algorithm for classification and regression tasks. It is well-suited for applications in remote sensing, with strengths that include capturing non-linear relationships and effectively handling outliers and noisy data. This study investigates the following questions: Can snow depth be estimated using SAR? Are there other variables that can improve the accuracy of the models? Can medium-resolution ERA5-Land data be used with point measurements? For the study, field measurements of snow depth in the Vindelfjällen region in Sweden were used as the reference data for the Random Forest (RF) models. Additional data used to enhance the predictive performance of the algorithm includes terrain data and climate data from the ERA5-Land reanalysis dataset. Nine models with different combinations of Sentinel-1 resolutions and additional parameters were created to assess their ability to predict snow depth. The results show that SAR correlates poorly with the in situ snow depth measurements. The performance of the RF model to predict snow depth using SAR alone is also poor, with R² values of 0.03-0.05. With the addition of terrain data, the R² increases to 0.22-0.3. Finally, models incorporating SAR, terrain, and ERA5-Land data achieved predictive performance with R² values of 0.35-0.4. The Sentinel-1 images were processed at 100, 500, and 1000-meter resolutions, with no significant difference observed between the different resolutions. These findings indicate that while SAR data alone is insufficient for accurate snow depth prediction, integrating terrain and climate data significantly enhances model performance. The results do not show a significan

The permafrost region covers a vast area of land surface on the northern hemisphere,storing large amounts of carbon. Unfortunately, climate warming leads to permafrostthaw altering the hydrothermal state of permafrost soils. Due to the remoteness of thepermafrost region, access to field measurements is restricted. Therefore, remotesensing is an asset to study the permafrost region. Since permafrost is a sub-surfacephenomenon it cannot be directly observed from space. However, by using differenttypes of satellites the soil properties of the top soil layer, down to 10 cm depth, can beaccessed. To establish soil properties for the deeper soil layers modelling is required.The ERA5-Land (ERA5L) soil moisture is modelled based on climate reanalysis. Inthis study in-situ soil moisture data from the International Soil Moisture Network(ISMN) is used to evaluate the performance of the ERA5L soil moisture data withinthe permafrost region. The performance of the ERA5L soil moisture is found toperform best in soil layer 1 (0-7 cm depth) and worst in soil layer 3 (28-100 cm depth).For both soil layer 1 and 2 (0-7 and 7-28 cm depth) a moderate correlation(0.309 < R < 0.335) was found between ERA5L and in-situ soil moisture data, in Julyand August. The performance of the ERA5L soil moisture is best in Europe and worstin North-America. Compared to other evaluations of ERA5L soil moisture, within thepermafrost region, this study found a relatively low correlation. Therefore, this studyconcludes that on a global scale the ERA5L soil moisture is not ideal for directlyinforming permafrost research and decision making. However, integrating multisourcedatasets, resampled to a finer spatial resolution, could improve the performance ofERA5L soil moisture model on a global level. Moreover, on a local scale theapplication of a bias correction could also improve the performance of the ERA5L soilmoisture model.

In the high altitude Hindukush Karakoram Himalaya (HKH) mountains, the complex weather system, inaccessible terrain and sparse measurements make the elevation-distributed precipitation and temperature among the most significant unknowns. The elevation-distributed snow and glacier dynamics in the HKH region are also little known, leading to serious concerns about the current and future water availability and management. The Hunza Basin in the HKH region is a scarcely monitored, and snow- and glacier-dominated part of the Upper Indus Basin (UIB). The current study investigates the elevation-distributed hydrological regime in the Hunza Basin. The Distance Distribution Dynamics (DDD) model, with a degree day and an energy balance approach for simulating glacial melt, is forced with precipitation derived from two global datasets (ERA5-Land and JRA-55). The mean annual precipitation for 1997–2010 is estimated as 947 and 1,322 mm by ERA5-Land and JRA-55, respectively. The elevation-distributed precipitation estimates showed that the basin receives more precipitation at lower elevations. The daily river flow is well simulated, with KGE ranging between 0.84 and 0.88 and NSE between 0.80 and 0.82. The flow regime in the basin is dominated by glacier melt (45%–48%), followed by snowmelt (30%–34%) and rainfall (21%–23%). The simulated snow cover area (SCA) is in good agreement with the MODIS satellite-derived SCA. The elevation-distributed glacier melt simulation suggested that the glacial melt is highest at the lower elevations, with a maximum in the elevation 3,218–3,755 masl (14%–21% of total melt). The findings improve the understanding of the local hydrology by providing helpful information about the elevation-distributed meltwater contributions, water balance and hydro-climatic regimes. The simulation showed that the DDD model reproduces the hydrological processes satisfactorily for such a data-scarce basin., Water Resources

The permafrost region covers a vast area of land surface on the northern hemisphere,storing large amounts of carbon. Unfortunately, climate warming leads to permafrostthaw altering the hydrothermal state of permafrost soils. Due to the remoteness of thepermafrost region, access to field measurements is restricted. Therefore, remotesensing is an asset to study the permafrost region. Since permafrost is a sub-surfacephenomenon it cannot be directly observed from space. However, by using differenttypes of satellites the soil properties of the top soil layer, down to 10 cm depth, can beaccessed. To establish soil properties for the deeper soil layers modelling is required.The ERA5-Land (ERA5L) soil moisture is modelled based on climate reanalysis. Inthis study in-situ soil moisture data from the International Soil Moisture Network(ISMN) is used to evaluate the performance of the ERA5L soil moisture data withinthe permafrost region. The performance of the ERA5L soil moisture is found toperform best in soil layer 1 (0-7 cm depth) and worst in soil layer 3 (28-100 cm depth).For both soil layer 1 and 2 (0-7 and 7-28 cm depth) a moderate correlation(0.309 < R < 0.335) was found between ERA5L and in-situ soil moisture data, in Julyand August. The performance of the ERA5L soil moisture is best in Europe and worstin North-America. Compared to other evaluations of ERA5L soil moisture, within thepermafrost region, this study found a relatively low correlation. Therefore, this studyconcludes that on a global scale the ERA5L soil moisture is not ideal for directlyinforming permafrost research and decision making. However, integrating multisourcedatasets, resampled to a finer spatial resolution, could improve the performance ofERA5L soil moisture model on a global level. Moreover, on a local scale theapplication of a bias correction could also improve the performance of the ERA5L soilmoisture model.

In many high altitude river basins, the hydro-climatic regimes and the spatial and temporal distribution of precipitation are little known, complicating efforts to quantify current and future water availability. Scarce, or non-existent, gauged observations at high altitudes coupled with complex weather systems and orographic effects further prevent a realistic and comprehensive assessment of precipitation. Quantifying the contribution from seasonal snow and glacier melt to the river runoff for a high altitude, melt dependent region is especially difficult. Global scale precipitation products, in combination with precipitation-runoff modelling may provide insights to the hydro-climatic regimes for such data scarce regions. In this study two global precipitation products; the high resolution (0.1° × 0.1°), newly developed ERA5-Land, and a coarser resolution (0.55° × 0.55°) JRA-55, are used to simulate snow/glacier melts and runoff for the Gilgit Basin, a sub-basin of the Indus. A hydrological precipitation-runoff model, the Distance Distribution Dynamics (DDD), requires minimum input data and was developed for snow dominated catchments. The mean of total annual precipitation from 1995 to 2010 data was estimated at 888 mm and 951 mm by ERA5-Land and JRA-55, respectively. The daily runoff simulation obtained a Kling Gupta efficiency (KGE) of 0.78 and 0.72 with ERA5-Land and JRA-55 based simulations, respectively. The simulated snow cover area (SCA) was validated using MODIS SCA and the results are quite promising on daily, monthly and annual scales. Our result showed an overall contribution to the river flow as about 26% from rainfall, 37–38% from snow melt, 31% from glacier melt and 5% from soil moisture. These melt simulations are in good agreement with the overall hydro-climatic regimes and seasonality of the area. The proxy energy balance approach in the DDD model, used to estimate snow melt and evapotranspiration, showed robust behaviour and potential for being, Water Resources

In many high altitude river basins, the hydro-climatic regimes and the spatial and temporal distribution of precipitation are little known, complicating efforts to quantify current and future water availability. Scarce, or non-existent, gauged observations at high altitudes coupled with complex weather systems and orographic effects further prevent a realistic and comprehensive assessment of precipitation. Quantifying the contribution from seasonal snow and glacier melt to the river runoff for a high altitude, melt dependent region is especially difficult. Global scale precipitation products, in combination with precipitation-runoff modelling may provide insights to the hydro-climatic regimes for such data scarce regions. In this study two global precipitation products; the high resolution (0.1° × 0.1°), newly developed ERA5-Land, and a coarser resolution (0.55° × 0.55°) JRA-55, are used to simulate snow/glacier melts and runoff for the Gilgit Basin, a sub-basin of the Indus. A hydrological precipitation-runoff model, the Distance Distribution Dynamics (DDD), requires minimum input data and was developed for snow dominated catchments. The mean of total annual precipitation from 1995 to 2010 data was estimated at 888 mm and 951 mm by ERA5-Land and JRA-55, respectively. The daily runoff simulation obtained a Kling Gupta efficiency (KGE) of 0.78 and 0.72 with ERA5-Land and JRA-55 based simulations, respectively. The simulated snow cover area (SCA) was validated using MODIS SCA and the results are quite promising on daily, monthly and annual scales. Our result showed an overall contribution to the river flow as about 26% from rainfall, 37–38% from snow melt, 31% from glacier melt and 5% from soil moisture. These melt simulations are in good agreement with the overall hydro-climatic regimes and seasonality of the area. The proxy energy balance approach in the DDD model, used to estimate snow melt and evapotranspiration, showed robust behaviour and potential for being, Water Resources

Precipitation data availability plays a crucial role in many climatic, hydrological and agricultural-related applications. In this study, the use of alternative data sources (i.e. interpolation methods and ERA5-Land reanalysis data) was combined for improving the spatially distributed precipitation estimates at the Simeto river basin, located in Eastern Sicily (Italy). A total of 51 rain gauges were used to generate a monthly precipitation dataset for the reference period 2002–2019. Among the 6 tested interpolation methods, Natural Neighbour was the method that predicted precipitation the best at monthly level with a Distance between Indices of Simulation and Observation (DISO) of 0.51. Radial Basis Functions and Inverse Distance Weighting provided the highest precipitation accuracies, respectively, for winter and autumn (with DISO values of 0.44 and 0.50, respectively), and for spring and summer seasons (with DISO values of 0.50 and 0.63, respectively). Underestimations on the ERA5-Land precipitation estimates were observed when compared to the most accurate interpolation methods both at monthly (25%) and seasonal temporal scales (21% in winter and summer, 36% in autumn), with the exception for spring. The performance was significantly improved when the interpolation estimates were corrected with local observations (with RMSD values ranging from 35.29 mm to 26.46 mm at monthly scale, and from 23.33–55.34 mm to 23.15–37.88 mm at seasonal level). The spatial distribution of the estimation errors associated to precipitation obtained from ERA5-Land reanalysis revealed a significant positive correlation (p value <0.05) with the altitude variation in each ERA5-Land cell within the basin under study. These results confirm the good performance on the combined use of alternative precipitation data sources, while adjustments are required to reduce site-specific uncertainties due to local microclimatic conditions occurring at the basin scale.

The novel Correlated Triple Collocation (CTC) analysis allows to assess three different data sources of similar spatial resolutions, but with two of them being correlated. In this study, the CTC was applied to estimate the unbiased random errors of the global soil moisture (SM) data provided by two L-band satellite missions -the Soil Moisture and Ocean Salinity (SMOS) and the Soil Moisture Active Passive (SMAP)- and one numerical model -the ERA5-Land. The three existing SMOS SM products distributed by different research institutions were also analyzed. Preliminary results revealed that errors of SMOS and SMAP SM are correlated, with correlations of ~0.5-0.6. Thus, only ERA5-Land can be considered as independent. The lowest error was obtained for SMAP (0.025 m3m-3), followed by ERA5-Land (0.036 m3m-3). Among the SMOS SM, SMOS-IC had the lowest error (0.046 m3m-3), SMOS-BEC showed an intermediate value (0.048 m3m-3), and SMOS-CATDS had the highest error (0.055 m3m-3). © 2021 IEEE., This work has been supported by the Spanish Ministry of Science and Innovation through the projects ESP2017-89463-C3-1R and ESP2017-89463-C3-2R, the ICM-CSIC Severo Ochoa Excellence Award CEX2019-000928-S, the CommSensLab-UPC María de Maeztu Excellence Award MDM-2016-0600, and the CSIC Interdisciplinary Thematic Platform TELEDETECT., Peer Reviewed, Postprint (author's final draft)

Long-term and high-quality surface soil moisture (SSM) and root-zone soil moisture (RZSM) data is crucial for understanding the land-atmosphere interactions of the Qinghai-Tibet Plateau (QTP). More than 40% of QTP is covered by permafrost, yet few studies have evaluated the accuracy of SSM and RZSM products derived from microwave satellite, land surface models (LSMs) and reanalysis over that region. This study tries to address this gap by evaluating a range of satellite and reanalysis estimates of SSM and RZSM in the thawed soil overlaying permafrost in the QTP, using in-situ measurements from sixteen stations. Here, seven SSM products were evaluated: Soil Moisture Active Passive L3 (SMAP-L3) and L4 (SMAP-L4), Soil Moisture and Ocean Salinity in version IC (SMOS-IC), Land Parameter Retrieval Model (LPRM) Advanced Microwave Scanning Radiometer 2 (AMSR2), European Space Agency Climate Change Initiative (ESA CCI), Advanced Scatterometer (ASCAT), and the fifth generation of the land component of the European Centre for Medium-Range Weather Forecasts atmospheric reanalysis (ERA5-Land). We also evaluated three RZSM products from SMAP-L4, ERA5-Land, and the Noah land surface model driven by Global Land Data Assimilation System (GLDAS-Noah). The assessment was conducted using five statistical metrics, i.e. Pearson correlation coefficient (R), bias, slope, Root Mean Square Error (RMSE), and unbiased RMSE (ubRMSE) between SSM or RZSM products and in-situ measurements. Our results showed that the ESA CCI, SMAP-L4 and SMOS-IC SSM products outperformed the other SSM products, indicated by higher correlation coefficients (R) (with a median R value of 0.63, 0.44 and 0.57, respectively) and lower ubRMSE (with a median ubRMSE value of 0.05, 0.04 and 0.07 m3/m3, respectively). Yet, SSM overestimation was found for all SSM products. This could be partly attributed to ancillary data used in the retrieval (e.g. overestimation of land surface temperature for SMAP-L3) and to the fact that

Evaporation is a crucial driver of Congo Basin climate, but the dynamics controlling the seasonality of basin evaporation are not well understood. This study aims to discover why evaporation on the basin-wide average is lower at the November rainfall peak than the March rainfall peak, despite similar rainfall. Using 16-year mean LandFlux-EVAL data, we find that evaporation is lower in November than March in the rainforest and the eastern savannah. The ERA5-Land reanalysis, which effectively reproduces this pattern, shows that transpiration is the main component responsible for lower evaporation in these regions. Using ERA5-Land, we find the following contrasting controls on transpiration, and therefore evaporation, at the two rainfall peaks: (a) In the northern rainforest, there is lower leaf area index (LAI) in November, driven by lower surface downward shortwave radiation (DSR), and lower vapour pressure deficit (VPD) in November, driven by lower sensible heat flux that results from lower net radiation. The combination of lower LAI and VPD explains lower transpiration, and therefore lower evaporation, in November. (b) In the southern rainforest, and in the north-eastern savannah, there is lower LAI in November, driven by lower surface DSR, and this explains lower transpiration, and therefore lower evaporation, in November. (c) In the south-eastern savannah, there is lower LAI in November, driven by lower volumetric water content (VWC), and this explains lower transpiration, and therefore lower evaporation, in November. Collectively, these contrasting controls at the two rainfall peaks explain why the basin-wide average evaporation is lower in November than March.