Département Analyse des Systèmes, Evaluation Intégrée et Modélisation

Study regionThe Urumqi Glacier No.1 (UGN1) and Dongkemadi Glacier (DG) catchments in China, both highly glacierized alpine catchments, provide critical water resources for downstream rivers and are sensitive to climate change impacts.Study focusAccurate simulation of glacier runoff processes remains challenging due to limited observational data. Here, we introduce FLEXG-iso, an isotope-aided glacier hydrological model that couples stable water isotopes with glacier mass balance (GMB) to simulate runoff generation and partition runoff components in these catchments. We also assess the transferability of model structures and parameters between UGN1 and DG catchments.New hydrological insights for the regionResults demonstrate that incorporating stable water isotopes and GMB data enhances the identification of parameters controlling snow and ice accumulation and melt. FLEXG-iso provides more robust partitioning of runoff components, distinguishing throughflow and groundwater, even with limited tracer availability, and improves runoff simulation accuracy (KGE > 0.8) when transferring optimal parameter sets. These findings highlight the potential of isotope-aided modeling to improve both calibration and transferability of glacier hydrological models, offering a robust framework for understanding runoff dynamics in high mountain regions under climate change.

Land-use policies founded on the expertise and interests of key stakeholders are likely the most implementable and politically long-lasting, assuming stakeholders share a broader, similar perception of the future. We investigated whether there is a difference in how eleven indicator species of conservation concern may be affected given policy and forestry scenarios of four key forest stakeholders for a 100 000 ha Swedish landscape over the coming 100 years. We used colonization-extinction models and species distribution models. Most species had stable or increasing metapopulation sizes or occurrence probabilities after 100 years under all except the most production-oriented scenario by private landowners. For six wood-decaying fungi, forests protected or managed with continuous cover forestry (CCF) drove the positive developments. By contrast, these species essentially disappeared from stands with even-aged clearcutting forestry. Regarding the 14.5% strict protection applied by the state-owned forest company increased the area occupied by these species. Protecting an even larger area and application of CCF in the scenarios of the (non-)governmental conservation organizations increased fungal metapopulation sizes even more. Four bird species showed stable or positive developments in all scenarios. The same held true for the epiphytic lichen Lobaria pulmonaria, because the management applied projected increasing host tree numbers and densities. Thus, protecting up to 20% of the productive forest, applying more alternative management regimes and less even-aged clearcutting rotation forestry has the potential to greatly increase the (meta)population size and improve the red-list status of all the focal species, and presumably also of other species of conservation concern.

For large-sample hydrological studies over large spatial domains, large-scale meteorological forcing data are often desired. For Europe, the EStreams dataset and catalogue satisfies this demand. In EStreams, the meteorological time series are obtained from the Ensemble Observation (E-OBS) product which is available for all of Europe. Due to the large spatial extent of this dataset, limitations and regional variations of data quality have to be expected when the dataset is compared to smaller-scale datasets, e.g., at national level. In this study, we compare the meteorological time series included for 2682 catchments in EStreams to eight smaller datasets (mostly CAMELS datasets). We assess how the different meteorological data impact the performance of a bucket-type hydrological model. For most catchments, the precipitation amounts derived from E-OBS are lower than the ones from the CAMELS data, while the temperature and the potential evapotranspiration values are higher. Model performances tend to be lower when the E-OBS data are used than when the CAMELS datasets are used for calibration. Exceptions arise when the station density in the E-OBS data is high. This study provides the first assessment of the EOBS data at a continental scale for hydrological applications and shows that, despite some limitations, the dataset offers a reasonable basis for large-sample hydrological modelling across Europe.

La creciente frecuencia de eventos de inundación en entornos urbanos, asociada al cambio climático y a la urbanización, ha incrementado la demanda de herramientas de predicción que sean tanto precisas como computacionalmente eficientes. Los modelos físicos basados en las ecuaciones de aguas someras bidimensionales (2D-SWE) ofrecen resultados fiables, pero requieren elevados tiempos de simulación, lo que limita su uso en contextos operativos.
Para superar esta barrera, han surgido enfoques basados en modelos subrogados con técnicas de aprendizaje profundo (DL), que reducen drásticamente los tiempos de cálculo (Guo et al., 2021; Löwe et al., 2021). Sin embargo, la mayoría de estos modelos se entrenan utilizando lluvia bruta como única entrada, ignorando procesos hidrológicos como la infiltración o las condiciones de humedad antecedente (AMC), que influyen directamente en la generación de escorrentía (Cea y Fraga, 2018).
Esta omisión puede afectar la capacidad de generalización del modelo subrogado y requiere, además, grandes volúmenes de datos para cubrir una mayor variabilidad, lo que implica costes computacionales elevados (Pouyanfar et al., 2018). En este trabajo se propone un enfoque híbrido que incorpora el cálculo de la lluvia neta en el flujo de trabajo del modelo subrogado basado en DL. Este marco se estructura en dos objetivos:
1. Incluir de forma explícita el cálculo de la lluvia neta durante el entrenamiento y la predicción.
2. Evaluar la capacidad de generalización del modelo ante eventos con diferente extensión de inundación y condiciones de AMC.
El estudio se aplica a la localidad de Sada (noroeste de España) y compara los resultados del modelo propuesto con un modelo subrogado convencional, evaluando precisión y eficiencia computacional.