Investigating event-specific drought attribution using self-organizing maps

Journal article

Previous studies evaluating anthropogenic influences on the meteorological drivers of drought have found mixed results owing to (1) the complex physical mechanisms which lead to the onset of drought, (2) differences in the characteristics and time scales of drought for different regions of the world, and (3) different approaches to the question of attribution. For a midlatitude, temperate climate like New Zealand, strongly modulated by oceanic influences, summer droughts last on the order of 3 months, and are less strongly linked to persistent temperature anomalies than continental climates. Here we demonstrate the utility of a novel approach for characterizing the meteorological conditions conducive to extreme drought over the North Island of New Zealand, using the January–March 2013 event as a case study. Specifically, we consider the use of self‐organizing map techniques in a multimember coupled climate model ensemble to capture changes in daily circulation, between two 41 year periods (1861–1901 and 1993–2033). Comparisons are made with seasonal pressure and precipitation indices. Our results demonstrate robust (>99% confidence) increases in the likelihood of observing circulation patterns like those of the 2013 drought in the recent‐climate simulations when compared with the early‐climate simulations. Best guess estimates of the fraction of attributable risk range from 0.2 to 0.4, depending on the metric used and threshold considered. Contributions to uncertainty in these attribution statements are discussed.

Authors: Harrington, LJ; Gibson, PB; Dean, SM; Mitchell, D; Rosier, SM

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: American Geophysical Union
• Journal: Journal of Geophysical Research: Atmospheres
• Volume: 121
• Issue: 21
• Pages: 12,766-12,780
• Publication date: 2016-11-11
• Acceptance date: 2016-10-15
• DOI: 10.1002/2016JD025602
• EISSN: 2169-8996
• ISSN: 2169-897X
• Keywords: self‐organizing maps; climate modeling; drought; event attribution