Tropical-Forest Biomass Dynamics from Comparison and Combination ofInterferomeric SAR and MODIS Observations

Tropical forests account for about 50% of the world's forested biomass, and play a critical role in the control of atmospheric carbon dioxide. Large-scale (1000's of km) changes in forest structure and biomass bear on global carbon source-sink dynamics, while small-scale (< 100 m) changes bear on deforestation and degradation monitoring for programs like REDD+. We are developing an approach to carbon monitoring which capitalizes on X-band interferometric SAR's (InSAR) sensitivity to change of forest vegetation structure as measured by TanDEM-X. Changes in the interferometric phase height, a radar-power-profile average height, are related to aboveground biomass (AGB) rate of change via conversion factors which themselves depend on AGB. The approach in this paper relies equally on the sensitivity to spectroscopic and structural change from Moderate Resolution Imaging Spectroradiometer (MODIS) observations. AGB measurement from MODIS is based on the spectral reflectance of 7 bands from the visible to short wave infrared, and auxiliary metrics describing the variance in reflectance. Conversion of MODIS reflectance to AGB is enabled by training a machine learning algorithm with lidar-derived AGB data, which are in turn trained by field measurements for small areas.

Two modes of data fusion are considered to improve on the AGB rates of either observation technique: 1) Fill the spatio-temporal gaps of either technique with the other; 2) Use the AGB measurements from MODIS to derive the phase-rate-to-AGB-rate conversion factors mentioned above. While TanDEM-X has wall-to-wall coverage, and it made forest structure measurements in our Tapajós National Forest supersite in Brazil at monthly intervals between 2011 and 2014, many other tropical forests, including the rest of the Amazon basin, were not covered as frequently. And TanDEM-X has already outlived its planned lifespan of 5 years (2016). In contrast, MODIS has been running since 2002, and delivers cloud-free observations of any given area in the tropics 2-3 times per year. The first step in the 1^stmode of data fusion will be to compare coincident TanDEM-X phase height rate to MODIS AGB rate over Tapajós. For the 2^ndfusion mode, we will apply AGBs from MODIS to derive phase-height-AGB rate conversion factors for TanDEM-X. We will compare AGB rates from the two sensors.