RESUMEN
Floods are one of the most common natural disasters, with a disproportionate impact in developing countries that often lack dense streamflow gauge networks1. Accurate and timely warnings are critical for mitigating flood risks2, but hydrological simulation models typically must be calibrated to long data records in each watershed. Here we show that artificial intelligence-based forecasting achieves reliability in predicting extreme riverine events in ungauged watersheds at up to a five-day lead time that is similar to or better than the reliability of nowcasts (zero-day lead time) from a current state-of-the-art global modelling system (the Copernicus Emergency Management Service Global Flood Awareness System). In addition, we achieve accuracies over five-year return period events that are similar to or better than current accuracies over one-year return period events. This means that artificial intelligence can provide flood warnings earlier and over larger and more impactful events in ungauged basins. The model developed here was incorporated into an operational early warning system that produces publicly available (free and open) forecasts in real time in over 80 countries. This work highlights a need for increasing the availability of hydrological data to continue to improve global access to reliable flood warnings.
Asunto(s)
Inteligencia Artificial , Simulación por Computador , Inundaciones , Predicción , Predicción/métodos , Reproducibilidad de los Resultados , Ríos , Hidrología , Calibración , Factores de Tiempo , Planificación en Desastres/métodosRESUMEN
Seasonal precipitation forecasting is highly challenging for the northwest fringes of Europe due to complex dynamical drivers. Hybrid dynamical-statistical approaches offer potential to improve forecast skill. Here, hindcasts of mean sea level pressure (MSLP) from two dynamical systems (GloSea5 and SEAS5) are used to derive two distinct sets of indices for forecasting winter (DJF) and summer (JJA) precipitation over lead-times of 1-4 months. These indices provide predictors of seasonal precipitation via a multiple linear regression model (MLR) and an artificial neural network (ANN) applied to four Irish rainfall regions and the Island of Ireland. Forecast skill for each model, lead time, and region was evaluated using the correlation coefficient (r) and mean absolute error (MAE), benchmarked against (a) climatology, (b) bias corrected precipitation hindcasts from both GloSea5 and SEAS5, and (c) a zero-order forecast based on rainfall persistence. The MLR and ANN models produced skilful precipitation forecasts with leads of up to 4 months. In all tests, our hybrid method based on MSLP indices outperformed the three benchmarks (i.e., climatology, bias corrected, and persistence). With correlation coefficients ranging between 0.38 and 0.81 in winter, and between 0.24 and 0.78 in summer, the ANN model outperformed MLR in both seasons in most regions and lead-times. Forecast skill for summer was comparable to that in winter and for some regions/lead times even superior. Our results also show that climatology and persistence performed better than direct use of bias corrected dynamical outputs in most regions and lead-times in terms of MAE. We conclude that the hybrid dynamical-statistical approach developed here-by leveraging useful information about MSLP from dynamical systems-enables more skilful seasonal precipitation forecasts for Ireland, and possibly other locations in western Europe, in both winter and summer.
RESUMEN
Globally, few precipitation records extend to the 18th century. The England Wales Precipitation (EWP) series is a notable exception with continuous monthly records from 1766. EWP has found widespread use across diverse fields of research including trend detection, evaluation of climate model simulations, as a proxy for mid-latitude atmospheric circulation, a predictor in long-term European gridded precipitation data sets, the assessment of drought and extremes, tree-ring reconstructions and as a benchmark for other regional series. A key finding from EWP has been the multi-centennial trends towards wetter winters and drier summers. We statistically reconstruct seasonal EWP using independent, quality-assured temperature, pressure and circulation indices. Using a sleet and snow series for the UK derived by Profs. Gordon Manley and Elizabeth Shaw to examine winter reconstructions, we show that precipitation totals for pre-1870 winters are likely biased low due to gauge under-catch of snowfall and a higher incidence of snowfall during this period. When these factors are accounted for in our reconstructions, the observed trend to wetter winters in EWP is no longer evident. For summer, we find that pre-1820 precipitation totals are too high, likely due to decreasing network density and less certain data at key stations. A significant trend to drier summers is not robustly present in our reconstructions of the EWP series. While our findings are more certain for winter than summer, we highlight (a) that extreme caution should be exercised when using EWP to make inferences about multi-centennial trends, and; (b) that assessments of 18th and 19th Century winter precipitation should be aware of potential snow biases in early records. Our findings underline the importance of continual re-appraisal of established long-term climate data sets as new evidence becomes available. It is also likely that the identified biases in winter EWP have distorted many other long-term European precipitation series.
RESUMEN
Global and continental scale hydrological reanalysis datasets receive growing attention due to their increasing number of applications, ranging from water resources management, climate change studies, water related hazards and policy support. Until recently, their use was mostly limited to qualitative assessments, due to their coarse spatial and temporal resolution, large uncertainty and bias in the model output, and limited extent of the dataset in space and time. This research reports on the setup of a gridded hydrological model with quasi-global coverage, able to reproduce a seamless 39-year streamflow simulation in all world's medium to large river basins. The model was calibrated at 1226 river sections with a total drainage area of 51 million km2 within 66 countries, using ECMWF's latest atmospheric reanalysis ERA5. A performance assessment revealed large improvements in reproducing past discharge observations, in comparison to the calibration used in the current operational setup of the hydrological model as part of the Copernicus - Global Flood Awareness System (GloFAS, www.globalfloods.eu), with median scores of Kling-Gupta Efficiency KGE = 0.67 and correlation r = 0.8. The simulation bias was also dramatically reduced and narrowed around zero, with more than 60% of stations showing percent bias within ±20%. Pronounced regional differences in the simulation results remain, pointing out the need for detailed investigation of the hydrological processes in specific regions, including parts of Africa and South Asia. In addition, observed discharges with high data quality is key to achieving skillful model output. The new calibrated model will become part of the operational runs of GloFAS in the next system release foreseen for Spring 2020, together with a near real time extension of the streamflow reanalysis.
RESUMEN
Early warning systems (EWS) for river flooding are strategic tools for effective disaster risk management in many world regions. When driven by ensemble Numerical Weather Predictions (NWP), flood EWS can provide skillful streamflow forecasts beyond the monthly time scale in large river basins. Yet, effective flood detection is challenged by accurate estimation of warning thresholds that identify specific hazard levels along the entire river network and forecast horizon. This research describes a novel approach to estimate warning thresholds which retain statistical consistency with the operational forecasts at all lead times. The procedure is developed in the context of the Global Flood Awareness System (GloFAS). A 21-year forecast-consistent dataset is used to derive thresholds with global coverage and forecast range up to six weeks. These are compared with thresholds derived from ERA5, a state of the art atmospheric reanalysis used to run the baseline simulation for the years 1986-2017 and to give a best guess of the present hydrological states. Findings show that the use of constant thresholds for 30-day flood forecasting, as in the current operational GloFAS setup, is consistent throughout the entire forecast range in only 30% to 40% of the river network, depending on the flood return period. Findings show that range-dependent thresholds, of weekly duration, are a more suitable alternative to time-invariant thresholds, as they improve the model consistency as well as the skills in flood monitoring and early warning, particularly over longer forecasting range.
RESUMEN
Climate change has led to concerns about increasing river floods resulting from the greater water-holding capacity of a warmer atmosphere1. These concerns are reinforced by evidence of increasing economic losses associated with flooding in many parts of the world, including Europe2. Any changes in river floods would have lasting implications for the design of flood protection measures and flood risk zoning. However, existing studies have been unable to identify a consistent continental-scale climatic-change signal in flood discharge observations in Europe3, because of the limited spatial coverage and number of hydrometric stations. Here we demonstrate clear regional patterns of both increases and decreases in observed river flood discharges in the past five decades in Europe, which are manifestations of a changing climate. Our results-arising from the most complete database of European flooding so far-suggest that: increasing autumn and winter rainfall has resulted in increasing floods in northwestern Europe; decreasing precipitation and increasing evaporation have led to decreasing floods in medium and large catchments in southern Europe; and decreasing snow cover and snowmelt, resulting from warmer temperatures, have led to decreasing floods in eastern Europe. Regional flood discharge trends in Europe range from an increase of about 11 per cent per decade to a decrease of 23 per cent. Notwithstanding the spatial and temporal heterogeneity of the observational record, the flood changes identified here are broadly consistent with climate model projections for the next century4,5, suggesting that climate-driven changes are already happening and supporting calls for the consideration of climate change in flood risk management.
Asunto(s)
Cambio Climático/estadística & datos numéricos , Inundaciones/estadística & datos numéricos , Ríos , Cambio Climático/historia , Europa (Continente) , Inundaciones/historia , Inundaciones/prevención & control , Mapeo Geográfico , Historia del Siglo XX , Historia del Siglo XXI , Lluvia , Estaciones del Año , Factores de TiempoRESUMEN
A warming climate is expected to have an impact on the magnitude and timing of river floods; however, no consistent large-scale climate change signal in observed flood magnitudes has been identified so far. We analyzed the timing of river floods in Europe over the past five decades, using a pan-European database from 4262 observational hydrometric stations, and found clear patterns of change in flood timing. Warmer temperatures have led to earlier spring snowmelt floods throughout northeastern Europe; delayed winter storms associated with polar warming have led to later winter floods around the North Sea and some sectors of the Mediterranean coast; and earlier soil moisture maxima have led to earlier winter floods in western Europe. Our results highlight the existence of a clear climate signal in flood observations at the continental scale.