Aboveground Biomass Changes in a Tropical Landscape Revealed by TanDEM-X InSAR Height Model Differences

Abstract

Tropical landscapes are relevant in their contribution to global climate regulation as potential carbon sink or source. The sequestered and emitted carbon is commonly approximated with the aboveground biomass (AGB). However, the estimation of AGB is to date mostly focussed on the estimation of one single point in time, where the accuracy is limited in particular on large spatial scales. The interferometric information (i.e. coherence and height) of high-frequency synthetic aperture radar (SAR) systems are considered particularly useful to estimate vegetation height and AGB. It is frequently assumed that the interfometric height of X-band systems like TanDEM-X represent the canopy surface height. Consequently, these interferometric SAR (InSAR) heights can be combined with terrain information to estimate vegetation canopy height and subsequently AGB. No spaceborne system exists to date to estimate the terrain height consistently on a global scale and thus the combination of TanDEM-X InSAR height and terrain information is normally limited to small spatial coverages. The potential to estimate AGB and in particular its change with such an approach is limited. In contrast, TanDEM-X InSAR heights can be directly compared over time. The differences can be assumed as differences in the canopy height assuming that the TanDEM-X InSAR heights represent the canopy surface height at a single point in time. The canopy height differences estimated by calculating the difference between bi- or multi-temporal TanDEM-X InSAR heights can be related to AGB differences. However, it was frequently found that the X-band signal penetrates into the canopy. This results in the fact that the X-band InSAR height is an approximation of the canopy surface height is not true. More importantly, the penetration depth can differ between different acquisitions depending on the acquisition properties and properties on the ground (e.g. moisture), which would result in pseudo-changes in the difference calculation of InSAR height models at different point in times. In our study, we used two TanDEM-X acquisitions from 2012 and 2019 covering a dynamic tropical area in Sumatra, Indonesia. We derived the InSAR heights for each acquisition date individually and calculated their difference. In addition, we assessed the penetration depth of the individual InSAR heights and modelled the penetration to compensate potential pseudo-changes. The absolute accuracy of the individual TanDEM-X heights was assessed with a LiDAR height model used as a reference. The TanDEM-X height differences were further related to ground-based AGB estimations from 2012 and 2019. This resulted in a significant linear relationship between height model and AGB differences, where the penetration compensated height models had a higher coefficient of determination and accuracy compared to the original InSAR heights. However, the accuracy was generally high in both cases with relative root mean square errors below 15%. This suggests that X-band height from TanDEM-X can be used to estimate canopy height differences and subsequently AGB changes on large spatial scale. However, the differences in penetration depth should not be neglected in order to avoid pseudo-changes and to estimate also small changes like degradation or growth.