Votre recherche
Résultats 5 ressources
-
Abstract. This study investigates the ability of long short-term memory (LSTM) neural networks to perform streamflow prediction at ungauged basins. A set of state-of-the-art, hydrological model-dependent regionalization methods are applied to 148 catchments in northeast North America and compared to an LSTM model that uses the exact same available data as the hydrological models. While conceptual model-based methods attempt to derive parameterizations at ungauged sites from other similar or nearby catchments, the LSTM model uses all available data in the region to maximize the information content and increase its robustness. Furthermore, by design, the LSTM does not require explicit definition of hydrological processes and derives its own structure from the provided data. The LSTM networks were able to clearly outperform the hydrological models in a leave-one-out cross-validation regionalization setting on most catchments in the study area, with the LSTM model outperforming the hydrological models in 93 % to 97 % of catchments depending on the hydrological model. Furthermore, for up to 78 % of the catchments, the LSTM model was able to predict streamflow more accurately on pseudo-ungauged catchments than hydrological models calibrated on the target data, showing that the LSTM model's structure was better suited to convert the meteorological data and geophysical descriptors into streamflow than the hydrological models even calibrated to those sites in these cases. Furthermore, the LSTM model robustness was tested by varying its hyperparameters, and still outperformed hydrological models in regionalization in almost all cases. Overall, LSTM networks have the potential to change the regionalization research landscape by providing clear improvement pathways over traditional methods in the field of streamflow prediction in ungauged catchments.
-
Natural disasters have been demonstrated to cause devastating effects on economic and social development in China, but little is known about the relationship between natural disasters and income at the household level. This study explores the impact of natural disasters on household income, expenditure, and inequality in China as the first study of this nature for the country. The empirical analysis is conducted based on a unique panel dataset that contains six waves of the Chinese Household Income Project (CHIP) survey data over the 1988–2018 period, data on natural disasters, and other social and economic status of households. By employing the fixed effects models, we find that disasters increase contemporaneous levels of income inequality, and disasters that occurred in the previous year significantly increase expenditure inequality. Natural disasters increase operating income inequality but decrease transfer income inequality. Poor households are found to be more vulnerable to disasters and suffer significant income losses. However, there is no evidence to suggest that natural disasters significantly reduce the income of upper- and middle-income groups. These findings have important implications for policies aimed at poverty alleviation and revenue recycling, as they can help improve economic justice and enhance resilience to natural disasters.
-
Agricultural activities can result in the contamination of surface runoff with pathogens, pesticides, and nutrients. These pollutants can enter surface water bodies in two ways: by direct discharge into surface waters or by infiltration and recharge into groundwater, followed by release to surface waters. Lack of financial resources makes risk assessment through analysis of drinking water pollutants challenging for drinking water suppliers. Inability to identify agricultural lands with a high-risk level and implement action measures might lead to public health issues. As a result, it is essential to identify hazards and conduct risk assessments even with limited data. This study proposes a risk assessment model for agricultural activities based on available data and integrating various types of knowledge, including expert and literature knowledge, to estimate the levels of hazard and risk that different agricultural activities could pose to the quality of withdrawal waters. To accomplish this, we built a Bayesian network with continuous and discrete inputs capturing raw water quality and land use upstream of drinking water intakes (DWIs). This probabilistic model integrates the DWI vulnerability, threat exposure, and threats from agricultural activities, including animal and crop production inventoried in drainage basins. The probabilistic dependencies between model nodes are established through a novel adaptation of a mixed aggregation method. The mixed aggregation method, a traditional approach used in ecological assessments following a deterministic framework, involves using fixed assumptions and parameters to estimate ecological outcomes in a specific case without considering inherent randomness and uncertainty within the system. After validation, this probabilistic model was used for four water intakes in a heavily urbanized watershed with agricultural activities in the south of Quebec, Canada. The findings imply that this methodology can assist stakeholders direct their efforts and investments on at-risk locations by identifying agricultural areas that can potentially pose a risk to DWIs.