Page - 71 - in Loss and Damage from Climate Change - Concepts, Methods and Policy Options

3 ObservedandProjected Impacts fromExtremeWeatherEvents… 71 In sum,while increasing trends are found for losses frompast extremeweather events, increasing exposure has been the main driver, and climate change (both anthropogenic climate change aswell as natural climate variability) could have an additional role, but this role was not substantiated. The confidence of the role of anthropogenicclimatechangeasdriver in theupward trend indisaster loses ishow- ever low. Results from these previous reviews, as well as more recent studies on disaster loss databases, are displayed inTable 3.2. In total 34 studies are included. Mostof thesestudieshaveanalysedmonetised losses fromextremeweatherevents, although in some cases the losses concern quantified impacts, such as volume of damaged timber wood. Andmost studies account for increasing exposure, using eitherdataonexposedcapital assets, population,wealth indicators, andan inflation correction. While this overview is perhaps neither exhaustive nor complete, it provides a comprehensiveoverviewofscientificstudiesonimpactsfrommajorextremeweather types,suchastropicalandextra-tropicalcyclones,rainfallflooding,hailstorms,wild- firesandconvectiveweathertypes.Whilecoastalfloodingisoftenincludedintropical cyclone losses, drought events areunderrepresented in these studies.A fewstudies detect trendsat the regionalornational level, theoverall conclusion is thatvery few upward trendsare found, afternormalising for changes inexposure. Thereareseveral issuesrelatedtothenormalisationapproach,aswellastheinter- pretationofnormalised losses.Firstofall, thegeneralassumption is that thechange in themajor driver of losses, that is increasing exposure of assets, has a propor- tional(linear)relationwiththelosses(e.g.PielkeandLandsea1998;Bouwer2011a; Handmeret al. 2012).Butalternativeapproaches suchas fromEstradaet al. (2015) show that alternative formulations of statisticalmodelswith explanatory variables may lead to different trends in losses, such as forUShurricanes. Such approaches arehowevernotyetconclusive,andneedfurtherconfirmationinconsecutivestudies (Hallegatte2015). Inaddition, theinterpretationofthenormalisedrecordisalsonotstraightforward. As Visser et al. (2014) and Visser and Petersen (2012) show, different statistical methodsfortrenddetectionmayleadtodifferentinterpretationofupward,downward or no trends found in the normalised loss records of extremeweather events. And howfluctuations in the normalised loss-record are interpreted, possibly related to natural climatevariability, is anothermatterofdiscussion. What is clear from the normalisation studies listed here (Table 3.2) is thatmost donotfindan increasing trend in losses, after the recordshavebeennormalised for increasingexposure.This implies that themaindriver of theobserved losses likely has been an increasing number of population and assets, and not a change in the hazardfrequencyorseverity.Afewstudieshoweverdofindincreases in losses,also afternormalisation.Theseincludemostnotablyconvectiveweatherevents,including thunderstorms and hailstorms, where three studies find increasing trends for over several decades.With increasing temperatures, there is a possibility that extremes related to convectiveweather couldbecomemore frequent.However, IPCC(2012)