Assessing vegetation response to climate variability at different time scales

To evaluate human impacts on forests and other carbon-storing ecosystems, temporal patterns of vegetation cover status must be explained in terms of their causal factors. More specifically, the impacts of climate variability on vegetation dynamics must be identified, quantified, and separated from the impacts of human disturbances such as land use changes. A generic approach is to combine remote sensing-derived information about the vegetation cover status with meteorological data for a corresponding period and study area. To date, the most common approach consists of correlation analyses between time series of Normalized Difference Vegetation Index (NDVI) and precipitation depths. Various authors have examined the strength and the lag of the correlation between monthly, seasonal or annual vegetation indices on the one hand, and rainfall depths on the other hand. Thereafter, the findings were spatially related to various environmental conditions, such as soil and vegetation types. We propose a number of methodological improvements related to the choice of variables, and the handling of periodic fluctuations at different time scales. Previous studies have utilized standardized NDVI anomalies and anomalies of precipitation accumulated over variable time spans. However, we found that off-site and lag effects are better accounted for by using anomalies of soil moisture content as an indicator of climate variability. Since the soil pore space can be regarded as a buffering reservoir for precipitation, we observed a more instantaneous and site specific correlation of soil moisture content with vegetation growth over a study area spanning East and Central Africa. Furthermore, the fraction of absorbed photosynthetically active radiation (fAPAR) is, as opposed to NDVI, a biophysical variable and is considered here to replace NDVI for further temporal analysis. Moreover, it is derived from multispectral satellite imagery, and its sensor-independent characteristics allow the compilation of a 30-year long time series. The methodology to quantify the response of vegetation to variable climate conditions must be capable of detecting the time scales at which fluctuations occur. Time series of soil moisture content and vegetation indices display a clear annual periodicity, arising from the annual cycles of precipitation and the growing seasons of dominant vegation types. As the objective is to understand climatic forcing beyond this seasonal response, all possible time scales must be accounted for in the analysis of the temporal signals of the explanatory variables. After the detection of trends in the time domain, spectral analysis of the temporal signals may reveal significant underlying periodic components, other than the annual cycle. As it is expected that the studied phenomena show changing behavior over time, the applied spectral decomposition methods must allow for flexibility in the basic periodic function. Therefore, time adaptive expansions of regular harmonic analysis are considered, and applied on univariate time series. Ultimately, bivariate analysis on a meteorological variable and a vegetation index will indicate which climatic fluctuations affect vegetation growth most.status: publishe