By Marian Vittek
Abstract
Subtropical thicket in South Africa comprises a unique biome due its notable species composition and the significant carbon storage under arid conditions. However, thicket vegetation is threatened by several factors that result in different levels of degradation. Despite the human presence has been relatively short, the effects of human activities have had a major impact. In order to support the current restoration initiatives, suitable methodologies for spatial monitoring and mapping of degradation patterns are required.
Remote sensing approaches offer several techniques dealing with change detection and mapping of spatial patterns of vegetation. In this study, a methodology using time series analysis and phenology parameter extraction from hyper-temporal MODIS Enhanced Vegetation Index (EVI) data was implemented for assessment of thicket degradation. The Savitzky-Golay smoothing filter was applied to raw satellite data with the purpose of producing a more representative time series dataset.
In order to quantify differences between the time series curve of degraded and non-degraded thicket vegetation, phenology parameters were determined for each growing season. Subsequently, three parameters were selected based on their suitability for degradation mapping utilizes decision tree approach. Parameters showing lower suitability are related to temporal properties of the time series as length, start and end of the season and peak time. Parameters showing more suitability were those related to the magnitude of EVI values in a particular growing season (peak, base value and amplitude). Finally, parameters characterizing the vegetation productivity as the small and large integral also seemed to be more usable for degradation monitoring purpose.
Based on the individual parameters selected, three thicket degradation maps with three classes were created (severely, moderately degraded and pristine thicket). Outputs of this analysis were validated with the results from previous studies in thicket biome and field measurements. From the comparison between the geophysical parameters and EVI vegetation curve, it was concluded that vegetation growth and EVI values are highly correlated with rainfall. It suggests that the decrease of biomass in thicket biome during the nine years study period may be caused by the climatic conditions and not by degradation processes.
Results demonstrate the applicability of phenology parameters derived from MODIS EVI time series for degradation monitoring in the thicket biome in South Africa. The extraction of phenological parameters from a temporal pattern of time series uniquely assessed the vegetation conditions in multiple growing seasons. However, some improvements may be done in the methodology in order to obtain more information and to achieve more reliable results. These improvements are particularly related with the field sampling and the parameters selection for degradation level classification.
Keywords: subtropical thicket biome, time series, remote sensing phenology, degradation, EVI, MODIS