Investigating dynamical trends in burned and unburned vegetation covers using SPOT-VGT NDVI data

Abstract

Fires induce dynamical trends in vegetation covers. In order to investigate the effects of fires in dynamical patterns of vegetation cover, normalized difference vegetation index data from the SPOT-VEGETATION sensor over the times series 1998–2003 were analysed for burned and unburned test sites located in the Italian Peninsula. The statistical analysis was carried out by means of three different methods: (i) power spectral density (PSD), which reveals scaling as well as periodic trends; (ii) the multiple segmenting method (MSM), which is well suited to analysing scaling behaviour for short time series; and (iii) detrended fluctuation analysis (DFA), a method that allows persistence in non-stationary signal fluctuations to be captured. Results from the statistical analyses showed that the scaling exponents α of the pixel time series for fire-affected sites range around mean values of ~1.38 (PSD), ~1.19 (MSM) and ~1.22 (DFA), while those for fire-unaffected sites vary around mean values of ~0.86 (PSD), ~0.63 (MSM) and ~0.65 (DFA). The two classes of vegetation (fire affected and fire unaffected) are significantly discriminated from each other (with the t-Student test, p < 0.0001) for all three methods adopted. The scaling exponents of both fire-affected and fire-unaffected sites show the persistent character of the vegetation dynamics though the fire-affected sites show larger exponents. Such a result shows that fires contribute by increasing the persistence of the time dynamics of vegetation and, therefore, drive unstable behavioural trends in vegetation dynamics of burned areas.