detection of extreme events / anomalies
from Julian Rogger research plan Master: For the analysis of extreme events, an approach introduced by Zscheischler et al., 2013 will be followed. In a rst step, the data of the CO2- uxes and the driving environmental variables needs processing. This includes the exploration and removal of linear (or maybe non-linear) trends as well as seasonal cycles for those variables exhibiting a such. This procedure makes the measurements comparable in time. A possible method of detrending is to use parametric modelling, where a cyclic function of the following character is tted to the data: Xt = 0 + 1 t + 2 sin(2t) + 3 cos(2t) After detrending, a percentile threshold is used for identifying \extreme" periods. For each extreme event in CO2- uxes it is examined how anomalous or extreme the environmental variables behaved during, before and after the respective period. To depict the magnitude of a specic extreme event, the \perfectde cit approach" as proposed by Yi et al., 2012 can be used. In this approach, the maximum potential of GPP at the respective site on a specic day of the year is dened as the maximum GPP measured on the respective day of the year throughout the entire study period. Hereby, smoothed GPP measurements of 30 min resolution are used. The deviation of GPP from this potential GPP during a specic extreme period marks the eect size of the respective extreme period.