14 May 2020 in Hydrology and Earth System Sciences
The European Centre for Medium-Range Weather Forecasts (ECMWF) recently released its most advanced reanalysis product, the ERA5 dataset. It was designed and generated with methods giving it multiple advantages over the previous release, the ERA-Interim reanalysis product. Notably, it has a finer spatial resolution, is archived at the hourly time step, uses a more advanced assimilation system and includes more sources of data. This paper aims to evaluate the ERA5 reanalysis as a potential reference dataset for hydrological modelling by considering the ERA5 precipitation and temperatures as proxies for observations in the hydrological modelling process, using two lumped hydrological models over 3138 North American catchments. This study shows that ERA5-based hydrological modelling performance is equivalent to using observations over most of North America, with the exception of the eastern half of the US, where observations lead to consistently better performance. ERA5 temperature and precipitation biases are consistently reduced compared to ERA-Interim and systematically more accurate for hydrological modelling. Differences between ERA5, ERA-Interim and observation datasets are mostly linked to precipitation, as temperature only marginally influences the hydrological simulation outcomes.
9 January 2023 in Hydrology and Earth System Sciences
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.
12 March 2009 in Water Resources Management
16 June 2021 in Hydrology and Earth System Sciences
Climate change impact studies require a reference climatological dataset providing a baseline period to assess future changes and post-process climate model biases. High-resolution gridded precipitation and temperature datasets interpolated from weather stations are available in regions of high-density networks of weather stations, as is the case in most parts of Europe and the United States. In many of the world's regions, however, the low density of observational networks renders gauge-based datasets highly uncertain. Satellite, reanalysis and merged product datasets have been used to overcome this deficiency. However, it is not known how much uncertainty the choice of a reference dataset may bring to impact studies. To tackle this issue, this study compares nine precipitation and two temperature datasets over 1145 African catchments to evaluate the dataset uncertainty contribution to the results of climate change studies. These deterministic datasets all cover a common 30-year period needed to define the reference period climate. The precipitation datasets include two gauge-only products (GPCC and CPC Unified), two satellite products (CHIRPS and PERSIANN-CDR) corrected using ground-based observations, four reanalysis products (JRA55, NCEP-CFSR, ERA-I and ERA5) and one merged gauged, satellite and reanalysis product (MSWEP). The temperature datasets include one gauged-only (CPC Unified) product and one reanalysis (ERA5) product. All combinations of these precipitation and temperature datasets were used to assess changes in future streamflows. To assess dataset uncertainty against that of other sources of uncertainty, the climate change impact study used a top-down hydroclimatic modeling chain using 10 CMIP5 (fifth Coupled Model Intercomparison Project) general circulation models (GCMs) under RCP8.5 and two lumped hydrological models (HMETS and GR4J) to generate future streamflows over the 2071–2100 period. Variance decomposition was performed to compare how much the different uncertainty sources contribute to actual uncertainty. Results show that all precipitation and temperature datasets provide good streamflow simulations over the reference period, but four precipitation datasets outperformed the others for most catchments. They are, in order, MSWEP, CHIRPS, PERSIANN and ERA5. For the present study, the two-member ensemble of temperature datasets provided negligible levels of uncertainty. However, the ensemble of nine precipitation datasets provided uncertainty that was equal to or larger than that related to GCMs for most of the streamflow metrics and over most of the catchments. A selection of the four best-performing reference datasets (credibility ensemble) significantly reduced the uncertainty attributed to precipitation for most metrics but still remained the main source of uncertainty for some streamflow metrics. The choice of a reference dataset can therefore be critical to climate change impact studies as apparently small differences between datasets over a common reference period can propagate to generate large amounts of uncertainty in future climate streamflows.
1 May 2010 in Journal of Water Resources Planning and Management
The impacts of climate change on medium-term reservoir operations for the Peribonka water resource system (Quebec, Canada) were evaluated with annual and seasonal hydropower production indicators and flood control criteria. According to simulations under the current operating rules in a climate change context, the tendency is for a reduction in mean annual hydropower production and an increase in spills, despite an increase in the annual average inflow to the reservoirs. The main results indicate that annual mean hydropower would change by −12 to +2% , and spills by −49 to +152% . A broad range of climate projections—a combination of five general circulation models with two greenhouse gas scenarios each—were used in order to evaluate the uncertainty of these future potential climates on floods and hydroelectric production. Climate projections were downscaled with the change factor method (also called the Delta method) at a horizon centered in 2050. To represent natural variability, a stochastic weather ge...
1 April 2017 in Journal of Hydrologic Engineering
It is well known that multimodel averaging can considerably improve hydrological model simulation skill. However, the need to set up and run different models can be a time-consuming task. This work expands on the classic multimodel averaging approach by feeding models with different climate datasets and treating each version as a unique model in the ensemble. Three hydrological models and four climate datasets were combined to produce multimodel/multi-input, multimodel/monoinput and monomodel/multi-input combined flows using a weighting scheme that minimizes the root mean square error (RMSE) between the combined and observed hydrographs. The results show that model averaging improves performance significantly and that the proposed multi-input averaging provides higher skill than multimodel averaging. A combination of all models run with all datasets (12 members in total) produced the best results with the averaged hydrograph being more accurate than any single member on 70% of the catchments. The ...