资源环境科技发展态势分析平台

There are concerns that recent climate change is altering the frequency and magnitude of river floods in an unprecedented way(1). Historical studies have identified flood-rich periods in the past half millennium in various regions of Europe(2). However, because of the low temporal resolution of existing datasets and the relatively low number of series, it has remained unclear whether Europe is currently in a flood-rich period from a long-term perspective. Here we analyse how recent decades compare with the flood history of Europe, using a new database composed of more than 100 high-resolution (sub-annual) historical flood series based on documentary evidence covering all major regions of Europe. We show that the past three decades were among the most flood-rich periods in Europe in the past 500 years, and that this period differs from other flood-rich periods in terms of its extent, air temperatures and flood seasonality. We identified nine flood-rich periods and associated regions. Among the periods richest in floods are 1560-1580 (western and central Europe), 1760-1800 (most of Europe), 1840-1870 (western and southern Europe) and 1990-2016 (western and central Europe). In most parts of Europe, previous flood-rich periods occurred during cooler-than-usual phases, but the current flood-rich period has been much warmer. Flood seasonality is also more pronounced in the recent period. For example, during previous flood and interflood periods, 41 per cent and 42 per cent of central European floods occurred in summer, respectively, compared with 55 per cent of floods in the recent period. The exceptional nature of the present-day flood-rich period calls for process-based tools for flood-risk assessment that capture the physical mechanisms involved, and management strategies that can incorporate the recent changes in risk.

Analysis of thousands of historical documents recording floods in Europe shows that flooding characteristics in recent decades are unlike those of previous centuries.

Since its introduction, the reward prediction error theory of dopamine has explained a wealth of empirical phenomena, providing a unifying framework for understanding the representation of reward and value in the brain(1-3). According to the now canonical theory, reward predictions are represented as a single scalar quantity, which supports learning about the expectation, or mean, of stochastic outcomes. Here we propose an account of dopamine-based reinforcement learning inspired by recent artificial intelligence research on distributional reinforcement learning(4-6). We hypothesized that the brain represents possible future rewards not as a single mean, but instead as a probability distribution, effectively representing multiple future outcomes simultaneously and in parallel. This idea implies a set of empirical predictions, which we tested using single-unit recordings from mouse ventral tegmental area. Our findings provide strong evidence for a neural realization of distributional reinforcement learning.

Analyses of single-cell recordings from mouse ventral tegmental area are consistent with a model of reinforcement learning in which the brain represents possible future rewards not as a single mean of stochastic outcomes, as in the canonical model, but instead as a probability distribution.