| 英文摘要: | Huntingford et al. reply
Our Perspective1 on the potential factors in the UK floods of December, January and February 2013/14 (DJF1314) discusses potential links between remote drivers in the climate system, Atlantic atmospheric circulation and storminess and UK precipitation. Based on a correlational analysis, van Oldenborgh et al.2 are critical of these links, arguing for instance that winter rainfall amounts for parts of the UK have only a correlation of 0.23 when compared with the North Atlantic Oscillation (NAO). The disagreement is essentially one of spatial scale: their arguments are based on a particularly localized measure of rainfall for southern England, rather than the Atlantic atmospheric circulation or more UK-wide precipitation and river flows that we discussed.
We agree with van Oldenborgh et al.2 that the scale-dependence of the response to remote drivers has important practical implications: if the response to a particular driver is only evident on very large scales, and not at the scale of a river catchment, then its utility may be quite limited in terms of analysis and prediction of specific flood events. Improved understanding of the role of remote drivers in the overall synoptic situation in DJF1314 may, nevertheless, provide improved warnings for flood-prone areas. This is valid, even if it does not translate into a substantial improvement in a simple correlation skill for local rainfall.
van Oldenborgh et al. question our use of the NAO index when discussing DJF1314, arguing that other pressure patterns are better related to rainfall in the UK. While we agree that the observed sea-level pressure pattern from last winter is not simply characterized by the NAO, the sea-level pressure NAO index for winter 2013/14 was ~12hPa (1.5 standard deviations) above normal. This is an atmospheric pattern that is known to emerge in response to a multitude of external climate drivers, including those discussed in our Perspective and the references therein.
Interactions between drivers can also result in relationships being obscured in correlation analyses. As an example (albeit not specifically relevant to DJF1314), El Niño is well established as a particularly strong driver of the global climate state3 with recently verified influence on northern European winter climate4, 5yet it occurs only episodically, every five years or so. In the years when El Niño is inactive, the atmospheric circulation will continue to vary due to internal fluctuations and external drivers. A correlation across all years can therefore easily mask the influence of El Niño in the years when it is active. For many of the drivers we highlighted, multiple modelling and observational studies show statistically significant influences on Atlantic–European surface climate when they are active.
Our Perspective reviews the enormous literature that relates particular phases and drivers of the Earth system to Atlantic circulation and hence to the risk of extreme rainfall in the UK. Many of these drivers appear to have been contributing to a large-scale synoptic situation conducive to flooding in the UK in DJF1314. We remain confident that improved modelling of such drivers will improve our ability to interpret and predict both long-term and year-to-year variations in flood risk. However, we are particularly careful in our Perspective article not to attribute DJF1314 rainfall events to any specific driver. Instead, the purpose of our study was to highlight that it is important to correctly model known teleconnections to Atlantic circulation if we are to understand and predict changing flood risks. That said, as van Oldenborgh et al.2 correctly note, an ability to predict flood risk should not be confused with capability to predict individual flood events: the enormous importance of chance should always be acknowledged in any discussion of our chaotic weather. |