Retrieving multi-scale climatic variations from high dimensional time-series MODIS green vegetation cover in a tropical/subtropical mountainous island

Abstract

There are knowledge gaps in our understanding of vegetation responses to multi-scale climate-related variables in tropical/subtropical mountainous islands in the Asia-Pacific region. Therefore, this study investigated inter-annual vegetation dynamics and regular/irregular climate patterns in Taiwan. We applied principal component analysis (PCA) on 11 years (2001∼2011) of high-dimensional monthly photosynthetically active vegetation cover (PV) derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) and investigated the relationships between spatiotemporal patterns of the eigenvectors and loadings of each component through time and multi-scale climaterelated variations. Results showed that the first five components contributed to 96.4% of the total variance. The first component (PC1, explaining 94.5% of variance) loadings, as expected, were significantly correlated with the temporal dynamics of the PV (r = 0.94), which was mainly governed by regional climate. The temporal loadings of PC2 and PC3 (0.8% and 0.6% of variance, respectively) were significantly correlated with the temporal dynamics of the PV of forests (r = 0.72) and the farmlands (r = 0.80), respectively. The low-order components (PC4 and PC5, 0.3% and 0.2% of variance, respectively) were closely related to the occurrence of drought (r = 0.49) and to irregular ENSO associated climate anomalies (r = −0.54), respectively. Pronounced correlations were also observed between PC5 and the Southern Oscillation Index (SOI) with one to three months of time lags (r = −0.35 ∼ −0.43, respectively), revealing biophysical memory effects on the time-series pattern of the vegetation through ENSO-related rainfall patterns. Our findings reveal that the sensitivity of the ecosystems in this tropical/subtropical mountainous island may not only be regulated by regional climate and human activities but also be susceptible to large-scale climate anomalies which are crucial and comparable to previous large scale analyses. This study demonstrates that PCA can be an effective tool for analyzing seasonal and inter-annual variability of vegetation dynamics across this tropical/subtropical mountainous islandin the Pacific Ocean, which provides an opportunity to forecast the responses and feedbacks of terrestrial environments to future climate scenarios.