Trend detection addresses the question of whether a trend exists in the data. The trend of interest can be purely biophysical, as for example, in measuring sea-level rise or average mean global temperature. Trend detection can also address questions that involve both socio-economic and biophysical systems, as is the case with detecting trends in damages from tropical storms.

Impact attribution methods explain observed impacts by either using process-based models or building statistical models of the relationship between observed impacts and a number of explanatory variables. Most scholarly work in the field of CCVIA focuses on the explanation of impact through biophysical variables and goes under the label of impact attribution. Some work also tries to explain impacts through socio-economic variables. These approaches are sometimes also called vulnerability or adaptive capacity indicators in the literature (Yohe and Tol 2002; Hinkel 2011a). In order to avoid confusion, we only use the term vulnerability indicator for those approaches that do not use data of observed impacts. The attribution of impacts to social system variables, face two major challenges. First, in contrast to the case of natural systems, general simulation models of social system are not available and the development of statistical models is only promising when systems can be described by few variables and a lot of data is available. The systems considered in the context of climate change are, however, generally complex social-ecological systems, which means that in principle, they cannot be described using few variables and simple (or linear) statistical regression models (Barnett et al. 2008; Hinkel 2011a). Second, the time-series data records available for impacts caused by climate change are often not long enough for building reliable statistical models (e.g., Bouwer 2011)

Example cases from literature

Trend detection
Emanual (2005) develops an index of accumulated annual power-dissipation from tropical storms in 5 ocean basins. The index is based on measures of wind-speed and precipitation in the storms. Using a statistical methods an upward trend in the index is observed over the period since the 1970s.

Pielke et al. (2005) find no trend in the annual hurricane damage in the US normalised for inflation, population and wealth.

Impact attribution
Checkley et al., (2000), for example, explain changes in daily hospital admissions in Lima through the stimuli variables temperature, humidity and rainfall.

Singh et al., (2001) explain observed incidences of diarrhoea in Fiji based on variations in temperature and rainfall.

Tol and Yohe (2007) address the question whether national level socio-economic variables can explain observed impact data found in the EM-DAT database. An initial list of 34 variables was selected based on the IPCC 's eight determinants of adaptive capacity. Six alternative impact indicators such as number of people affected by natural disasters, infant mortality and life expectancy were selected for which data was available in the EM-DAT database. 24 of the 34 indicating variables were found to be statistically not significant. Amongst the statistical significant ones, different ones were found significant for different impact indicators. They conclude that there are no universal explanations; mechanisms that cause impact vary from case to case and hazard to hazard.