Inter-annual variability of rainfalls in the Amazon basin and its vicinity

An analysis of the rainfall series (12-month running means) of the 5° × 5° gridded data in the Amazon river basin and its vicinity (15° N – 20° S, 30° - 80° W) indicated that the rainfalls were highly variable both from year to year and from region to region. Correlations with even nearby regions hardly exceeded 0.50, though correlations were better (up to 0.70) in the regions near the eastern coast of Brazil. Moderate relationship with ENSO indices was obtained for the Amazon river basin and the regions to its north, and for NE Brazil, while moderate relationship with South Atlantic SST was obtained for NE Brazil and the region immediately to its west. All other relationships (with 30 hPa wind, North Atlantic Oscillation Index, etc.) were obscure.

Applying machine learning for drought prediction using a large ensemble of climate simulations

<p>There is strong scientific and social interest to understand the factors leading to extreme events in order to improve the management of risks associated with hazards like droughts. Recent events like the summer 2018 drought in Germany already had severe und unexpected impacts, e.g. forest fires and crop failures; in order to increase preparedness robust prediction tools are  urgently required. In this study, machine learning methods are applied to predict the occurrence of a drought with lead times of one to three months. The approach takes into account a list of thirty atmospheric and soil variables<strong> </strong>as predictor input parameters from a single regional climate model initial condition large ensemble (CRCM5-LE). The data was produced the context of the ClimEx project by Ouranos with the Canadian Regional Climate Model (CRCM5) driven by 50 members of the Canadian Earth System Model (CanESM2) for the Bavarian and Quebec domains.</p><p>Drought occurrence was defined using the Standardized Precipitation Index. The training and test datasets were chosen from the current climatology (1955-2005) for the Munich and Lisbon subdomain within the CRCM5-LE. The best performing machine learning algorithms managed to obtain a correct classification of drought or no drought for a lead time of one month for around 60 % of the events of each class for the both domains. Explainable AI methods like feature importance and shapley values were applied to gain a better understanding of the trained algorithms. Physical variables like the North Atlantic Oscillation Index and air pressure one month before the event proved to be of high importance for the prediction. The study showed that better accuracies can be obtained for the Lisbon domain, due to the stronger influence of the North Atlantic Oscillation Index on Portugal’s climate.</p>

Significant Extremal Dependence of a Daily North Atlantic Oscillation Index (NAOI) and Weighted Regionalised Rainfall in a Small Island Using the Extremogram

Extremal dependence or independence may occur among the components of univariate or bivariate random vectors. Assessing which asymptotic regime occurs and also its extent are crucial tasks when such vectors are used as statistical models for risk assessment in the field of Climatology under climate change conditions. Motivated by the poor resolution of current global climate models in North Atlantic Small Islands, the extremal dependence between a North Atlantic Oscillation index (NAOI) and rainfall was considered at multi-year dominance of negative and positive NAOI, i.e., −NAOI and +NAOI dominance subperiods, respectively. The datasets used (from 1948–2017) were daily NAOI, and three daily weighted regionalised rainfall series computed based on factor analysis and the Voronoi polygons method from 40 rain gauges in the small island of Madeira (∼740 km2), Portugal. The extremogram technique was applied for measuring the extremal dependence within the NAOI univariate series. The cross-extremogram determined the dependence between the upper tail of the weighted regionalised rainfalls, and the upper and lower tails of daily NAOI. Throughout the 70-year period, the results suggest systematic evidence of statistical dependence over Madeira between exceptionally −NAOI records and extreme rainfalls, which is stronger in the −NAOI dominance subperiods. The extremal dependence for +NAOI records is only significant in recent years, however, with a still unclear +NAOI dominance.

An integration and assessment of multiple covariates of nonstationary storm surge statistical behavior by Bayesian model averaging

Projections of coastal storm surge hazard are a basic requirement for effective management of coastal risks. A common approach for estimating hazards posed by extreme sea levels is to use a statistical model, which may use a time series of a climate variable as a covariate to modulate the statistical model and account for potentially nonstationary storm surge behavior (e.g., North Atlantic Oscillation index). Previous works using nonstationary statistical approaches to assess coastal flood hazard have demonstrated the importance of accounting for many key modeling uncertainties. However, many assessments have typically relied on a single climate covariate, which may leave out important processes and lead to potential biases in the projected flood hazards. Here, I employ a recently developed approach to integrate stationary and nonstationary statistical models, and characterize the effects of choice of covariate time series on projected flood hazard. Furthermore, I expand upon this approach by developing a nonstationary storm surge statistical model that makes use of multiple covariate time series, namely, global mean temperature, sea level, the North Atlantic Oscillation index and time. Using Norfolk, Virginia, as a case study, I show that a storm surge model that accounts for additional processes raises the projected 100-year storm surge return level by up to 23 cm relative to a stationary model or one that employs a single covariate time series. I find that the total model posterior probability associated with each candidate covariate, as well as a stationary model, is about 20 %. These results shed light on how including a wider range of physical process information and considering nonstationary behavior can better enable modeling efforts to inform coastal risk management.