Tackling some statistical challenges of wildfire seasonality identification

Vortrag von Dr. Jose Ameijeiras-Alonso

Sprecher eingeladen von: Prof. Dr. Reinhard Furrer

Datum: 23.11.17  Zeit: 09.30 - 10.30  Raum: Y27G28

Forest fires may be caused by different factors, both climatological and anthropogenic. If fires are just considered as a natural hazard, then the number of fire seasons along a year will be just driven by climate conditions, but if human activity is not taken into consideration, this may lead to an underestimation of the number of fire seasons. In this setting, such periods can be identified by analyzing the pattern of fire occurrences along a year, which can be viewed as a circular variable. The statistical problem of determining the number of modes in the underlying density will provide a way of identifying significant fire seasons. A new method is provided presenting a good calibration and power properties in simulated scenarios, and it has been applied to analyze fire seasons on different regions of the Earth.