Assessment of ALOS PALSAR L-band SAR for estimation of above ground biomass in tropical forests

Tropical forest biomass is one of the key parameters in addressing issues relating to climate change as it known to store large amount of carbon. Retrieving biomass over large forest areas has been challenging due to the limited data resource, accessibility, and various technical issues. Remote sensing data has been used actively for forest biomass estimation since the last three decades and it is proven to be effective. Although there are issues and arguments raised in the estimation accuracy, research is continuously being carried out. Synthetic aperture radar (SAR) system has its own potential in retrieving biomass but several issues are remaining unaddressed. One of the biggest issues in the system is signal saturation at high level of biomass and this has been identified as the most critical challenge. Studies concentrating on the use of SAR system for biomass retrieval in Malaysia are very limited and literatures on this subject matter are scarce. This study therefore was conducted to address this issue and attempted to produce estimations of aboveground biomass (AGB) on several types of forests in Peninsular Malaysia by using Phased Array L-band SAR (Palsar) on board the Japanese Advanced Land Observing Satellite (Alos). The study was conducted on three forest types, which are lowland dipterocarp, hill dipterocarp, and mangroves. A number of sampling plots have been established in respective forest types. By using these sample plots information, factors contributing to the variation in L-band SAR signal has been identified. Empirical models for the estimation of AGB in these forests have been developed. The level of saturation of AGB has been also determined for each forest types. The study found that the allometric equation used for AGB calculation did affects the estimation and pixel degradation from 25 m to 50 m was found most appropriate for AGB estimation in lowland and dipterocarp forests. While there is no significance difference of backscatter response towards natural and logged forests, the study found that biomass in canopies of trees with dbh ≥ 30 cm has better interaction with the backscatter. The inclusion of trees with dbh >10 cm also improved the estimation. The AGB in lowland and hill dipterocarp forests in the whole Peninsular Malaysia, which has an extent of 4,709,806.34 ha, has been estimated at 1,650,819,055 Mg with and root means square error (RMSE) of ±19.32 Mg ha-1 (at AGB < 200 Mg ha-1) and ±79.58 Mg ha-1 (at AGB > 200 Mg ha-1). The saturation level for these forests was at 200 Mg ha-1 at backscatter of -12 dB in HV polarization. The study also established empirical models for mangrove forest. It was estimated that the AGB in the whole mangroves of Peninsular Malaysia, which has an extent of 115,108 ha was at 11,441,795 Mg with RMSE at ±33.90 Mg ha-1. The saturation level for this forest type occurred at around 100 -150 Mg ha-1 at backscatter of -15 dB in HV polarization. The study revealed that the incorporation of texture measures in the images has improved the estimation accuracy as compared to images without texture measure. Overall, the study found that the HV polarization is the best for AGB estimations. The capability of Palsar data in estimating AGB on the forest types is evaluated. The potential, issues and challenges of Palsar data in the retrieval of AGB on these forests are elaborated in details and explained critically in this thesis.

Preview Text fh 2015 15 ir.pdf Download (1MB) | Preview

Actions (login required)

View Item