Detection of Ecosystem Functioning Using Object-based Time-series Analysis

Accurate detection of tree growth and ecosystem functioning in naturally vegetated areas is of interest to many applications, such as ecological and land degradation modelling, wildfire risk monitoring, carbon budgeting, and climate science. However, the changes are often small when compared to normal spatial and seasonal variability, or occur gradually in time. We use object-based image analysis (OBIA) for time-series analysis of vegetation functioning because objects coincide better with ecological units in the field than pixels. For this study, we combine 12 ASTER images recorded between 2002 and 2008 with information from 75 field sites that were visited multiple times to provide field reference. A single segmentation of multi-temporal data is used to align all available data to a single object framework. The images are recorded before, during, and after the dry summer season, and thus show the effect of summer drought on the vegetation in our study area. The object-based method of creating time series from imagery allows for a reliable detection of small changes in NDVI and TIR. Using this approach, we show a differentiation in vegetation response to drought that can be related to the underlying lithological substrate and its water holding capacity. The presented method is especially valuable in fragmented landscapes which are common to areas in the Mediterranean basin.