Regional estimation of litterfall in a subtropical forest

Litterfall represents the major losses of aboveground carbon and nutrients to forest floors, which are pivotal components in terrestrial biogeochemical cycles. It is also an indicator to assess the impacts of perturbations on forest ecosystems such as tropical cyclones and insect infestation. The objective of study is to integrate field and satellite observations to estimate litterfall in a subtropical forest in Taiwan (23°54'N, 121°54'E). Eighteen litterfall traps across a wide range of topo-edaphic gradients were installed to capture litterfall during May 2008 to April 2009. A probabilistic spectral mixture analysis model was used to estimate time-series fractional photosynthetically active vegetation cover (PV) from seven sets of Landsat Enhanced Thematic Mapper (ETM+) images acquired during the field sampling period. Based on preliminary field investigation, litterfall pattern might be related to vegetation phenology. This would permit the estimation of litterfall abundance (g m-2) using the difference of PV (ΔPV) between two consecutive images through time. Results showed that the mean (standard deviation, minimum-maximum) litterfall and ΔPV of the observation period were 149.6 (206.9, 7.5-663.8) g m-2 and 1.0 (8.0, -18.8-13.2)%, respectively; there was a strong correlation (r = 0.63, p-value = 0.012) between these two measures. This relationship allowed us to map the temporal variation of litterfall over a vast region. High values of ΔPV and litterfall during the growing season (April-September) may be due to leaf expansion and tropical cyclones, respectively, while low values of ΔPV and litterfall during the non-growing season (October-March) may reflect the normal phenological pattern. This study demonstrates the feasibility of using remote sensing to assess the temporal dynamics of litterfall, which makes it possible for the regional monitoring of metabolism of subtropical forest ecosystems.