Effectiveness of the BFAST algorithm for detecting vegetation response patterns in a semi-arid region

Abstract

The effective use of satellite image time series for examining vegetation response patterns across regional extents requires a method which accounts for variation at the seasonal scale while simultaneously detecting abrupt changes in any long term trends. The Breaks for Additive Seasonal and Trend (BFAST) algorithm has been developed to do this. However, its effectiveness in semi-arid regions, where vegetation response is typically aseasonal, has yet to be assessed.In this research the BFAST algorithm was assessed for a semi-arid study area in the Paroo catchment of far north-western New South Wales, Australia. Moderate Resolution Imaging Spectroradiometer (MODIS) EVI time series were decomposed using BFAST for 270 sample pixels to assess the algorithm's ability to detect abrupt changes in vegetation response caused by known fires and floods. The algorithm was also applied across the study area to examine regional patterns in the timing and magnitude of abrupt changes and the direction and magnitude of the long term trends.The timing of breaks detected by BFAST corresponded with the timing of known floods in the study region for between 68% and 79% of breaks detected across the sample pixels, depending on the parameters used in the decomposition. BFAST was not, however, able to accurately detect fires in the Paroo region, with agreement between the timing of breaks and fires occurring in only 3% of breaks detected. This most likely reflects the low EVI values present before a fire event, which would be typical of semi-arid zones. Spatial patterns in the timing of abrupt changes and greening and browning trends across the study area were a function of land cover and vegetation type. These results indicate that BFAST is able to detect abrupt changes in vegetation greening caused by known floods in semi-arid regions. The presence of spatial patterns in the results also indicates that the algorithm is sensitive to vegetation cover type. BFAST is therefore able to detect abrupt trend changes in regions where vegetation response is not expected to show strong seasonal patterns and could be used in further applications such as classification or regional vegetation modelling in semi-arid environments.