Coupled Spatiotemporal Characterization of Monsoon Cloud Cover and Vegetation Phenology

Abstract: Many biogeophysical processes continuously evolve in both space and time. Some of these processes change gradually enough across both space and time to be oversampled by the satellite image archive. In these cases, image time series of Earth observations can support robust spatiotemporal analysis. In this analysis, we leverage the spectral differences of green vegetation, soil, and clouds to perform an EOF analysis which simultaneously characterizes monsoon cloud cover and vegetation phenology in a temperate cloud forest in Dhofar, southern Oman. The Normalized Difference Vegetation Index of daily MODIS reflectance clearly distinguishes clouds and bare soil from both grassland and cloud forest vegetation. Using the temporal feature space of the low order principal components of the NDVI time series, we identify distinct annual cycles of cloud cover and vegetation phenology corresponding to geographically contiguous areas of coupled monsoon-phenological activity (phenoregions). The EOF analysis characterizes the temporal dynamics of these phenoregions. The wadi cloud forests of the Jabal Al Qara and Jabal Al Qamar can each be described with 2 distinct spatiotemporal patterns of monsoon phenology, corresponding to the eastern and western portions of each range. In addition, the rangetop grasslands of the Jabal Al Qara show 3 distinct spatiotemporal patterns, corresponding to eastern, central, and western portions of the range. While each phenoregion has a clearly repeating annual cycle, each also demonstrates substantial interannual variability in both cloud cover and phenology. This interannual variability is of a sufficient magnitude to mask any trends in cloud cover or phenology in any of the regions in the Dhofar Mountains. Finally, using the degree of asymmetry of the cumulative NDVI time series, we infer a causal link between duration of cloud cover and amplitude of phenology.