Multiscale leaf area index assimilation for Moso bamboo forest based on Sentinel-2 and MODIS data

Abstract

Leaf area index (LAI) is an important driving factor in forest ecosystems carbon cycle, the acquisition of multiscale LAI can illustrate the carbon cycle at different scales and its dynamic response to environmental changes. Moso bamboo forest (MBF) has strong carbon sequestration capability and unique phenological characteristics of on-years and off-years. In this study, based on the MODIS LAI and Sentinel-2 reflectance products in 2018–2019, we coupled the Hierarchical Bayesian Network (HBN) algorithm, LAI dynamic model and PROSAIL model to acquire multiscale spatiotemporal assimilated LAI (500-m, 100-m, 20-m) of MBF. First, the results showed that band7 (B7) and band8a (B8a) of Sentinel-2 were highly sensitive to LAI; the simulated leaf and canopy reflectance by PROSAIL demonstrated higher accuracies and lower errors in these two bands. Second, after applying the Savitzky-Golay smoothing, compared with the observed LAI (Obs_eLAI), the accuracy of MODIS LAI was improved by 121%, and the error was reduced by 24%. Third, there was a significant correlation between the multiscale assimilated LAI (HBN_LAIs) and Obs_eLAI (R500m_HBN_LAI2=0.80,R100m_HBN_LAI2=0.82,R20m_HBN_LAI2=0.80). Finally, the assimilated results in 2018 (off-year) were better than those in 2019 (on-year); the variations in multiscale spatiotemporal assimilated LAI were consistent with the actual growth trends of MBF. This study provides a feasible way to accurately obtain multiscale high-resolution LAI for the carbon cycle simulation of bamboo forests.

Keywords

• Sentinel-2
• MODIS LAI
• Moso bamboo forest
• Hierarchical Bayesian network