Historical relationship between climate and fire regime in Asagı Köprüçay Basin (Antalya, Turkey)

In this work, we analyzed fire regime, climatic trend and relationships between fire and climate in the Asağı Köprüçay basin (Antalya, Turkey), an area of about 205,000 hectares.
The study area is intensively affected by forest fires. Since the historical data on fires were consistent in detail after 1979, we analyzed the historical period 1979-2009. Regarding the climate analysis, data of four meteorological stations representing the different climatic conditions of the study area were taken into consideration. Daily records of maximum, mean and minimum temperatures, precipitation, wind speed and direction, relative humidity were used for the analysis. In the study period, mean and minimum temperatures showed a statistically significant increasing trend with time. The trends of fire number,
burned area and meteorological data were analyzed by using correlation and linear regression techniques. Regarding the historical trends in terms of fire number, no statistically significant trends were observed, because of the high inter-annual variability of the data. The burnt areas showed a general increasing trend that is not statistically significant. No statistically significant correlation between fire number and burnt area was observed. The relationships between weather parameters and the main indicators of fire activity: fire days (FD, at least 1 fire per day), large fire days (LFD, at least 20 hectares per day), multiple fire days (MFD, more than 1 fire per day), were investigated by the application of the logistic regression. The historical relationship between weather variables and the main indicators of the fire activity (FD, LFD, MFD) were analyzed by a set of logistic regression models. In particular, 4 models provided the best combined response in predicting the different fire activity indices on both annual and seasonal data. The different models were characterised by low estimation accuracies for FD, while the best results were obtained for LFD. Temperature and relative humidity are the weather variables mostly correlated with the fire activity probability as predicted by the logistic models. An increase in the accuracy was generally obtained where the 3 and 7 days minimum and maximum average values were used instead of the daily mean values of the weather variables.