Shukla, Yashvini
(2024)
Assessing mangrove canopy heights in Myanmar using GEDI & Sentinel-2 for effective monitoring.
MRes thesis, University of Nottingham.
Abstract
Mangrove forests are crucial ecosystems that store over three times as much carbon per hectare as terrestrial tropical forests (Donato et al., 2011) and host an essential role in regulating global and local climate systems (Estoque et al.2018). Alarmingly, the global mangrove area shrank by 50% between 1997 and 2016, with the most significant losses occurring in Southeast Asia (Estoque et al., 2018; Romañach et al., 2018). Projections suggest that mangroves could vanish entirely within the next century (Polidoro et al., 2010), highlighting the urgent need for accurate mapping of their structural and spatial characteristics to aid conservation and restoration efforts. While readily available multispectral satellite data like Sentinel-2 provide insights into mangrove coverage, they offer limited information on three-dimensional (3D) structural characteristics such as Canopy Height (CH). Traditional methods for 3D mapping, such as Airborne Lidar Surveying (ALS) and Synthetic Aperture Radar (SAR), are expensive and geographically unscalable. Spaceborne Lidar missions, and specifically the launch of the Global Ecosystem Dynamics Investigation (GEDI), offer a new opportunity to obtain 3D mangrove canopy data. As GEDI samples only 4% of the Earth, it is often fused with contiguous imagery like Landsat and Sentinel-2 to produce Global Canopy Height Maps (CHMs), albeit with limitations at local scales and for non-standard forest structures like mangroves (Potapov, 2021; Lang et al., 2022). This study leverages a Random Forest (RF) algorithm to combine Sentinel-2 and GEDI data for producing a contiguous Mangrove CHM for a restored local region in Myanmar for the year 2019. Field heights were obtained from
Worldview International, the project facilitators in 32 sample field plots (Vanniarachchy and Jayakody, 2020). Three models, trained on data sets from 2019 and 2020, were tested against a GEDI validation set and the field heights. The Relative Height (Rh) at the 60th Percentile Waveform Energy Return (Rh60) from GEDI's Level 2A product (that provides elevation and height metrics) was identified as the best predictor of field heights out of other Rh metrics, yielding
an R2 value of 0.24, a Mean Error (ME) of 0.28 m, and a Root Mean Squared Error (RMSE) of 0.37m. These results were compared with 3 baseline Global CHMs by Potapov et al. (2021), Lang et al. (2022), and Simard et al. (2019). These comparisons revealed that the CH predictions by Lang et al.
(2022) had the highest ME and RMSE, followed by that of Simard et al. (2019) and lastly Potapov et al. (2021). Further, vertical structural analysis using GEDI's L2B (that provides biophysical metrics) product indicated that the mangroves studied are vertically uniform, unlike typical forests.
This could explain why the mean Rh metrics provide a better approximation of true CH as opposed to the commonly used Rh90+ metrics in the baseline Global C estimates are overestimated by over 50%, and as a consequence Above Ground Biomass (AGB) estimates could be grossly inaccurate in short stature (<3m) mangroves if current Global CHMs are used, emphasizing the wider need for their local calibration. These findings have a direct impact on estimating National Carbon Stocks, contribute to the accreditation of community-based forest conservation and afforestation projects, and aid wider efforts in understanding and mitigating climate change.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
