Estimating biomass of economically important palms in Peru using UAV and satellite remote sensing

Abstract

Palm swamps in the Amazon are regarded as important carbon-dense ecosystems, with palms regarded as important non-timber forest products for local communities. Knowing the abundance and biomass of palms in these areas allows for better sustainable forest planning, but data collecting in the field is difficult and expensive. However, commercial UAV’s present opportunities for mapping palm abundance and biomass in a cost-effective way.

The main objective of this study is to understand how effectively commercial RGB UAV can be used in dense tropical palm forests to detect palms and estimate the biomass. Palm biomass is estimated using allometric models that require palm height as input. The UAV imagery is used to create canopy height maps and is compared with other remote sensing derived height maps to determine which height dataset is best suitable for estimating palm biomass. The UAV derived palm locations from the Tagle Casapia et al. (2020) study are used to extract the palm crown height values from each height map. A total of six height maps were used to estimate palm biomass. The results showed that the detection rate of the UAV was in important factor when estimating palm biomass in plots. The palm heights mapped by the UAV has large errors and underestimated the palm heights. These UAV errors were mostly caused by the dense and complex canopies of tropical palm forests, where the ground is also not visible or covered by water, making palm crown identification and heights estimations difficult. The palm height maps by Potapov et al. (2021) and Asner (2021) had much lower errors, however all of the height maps underestimated the heights of palms taller than 34m. A linear model was also created to estimate the palm heights by using the UAV, Potapov and Asner maps as input. The Potapov and linear model height data had the lowest errors when estimating biomass.

UAV provides a cost-effective solution for mapping palms and their biomass, but has varying results based on the local forest structure. The RGB UAV palm detection method used for this study could however still be used for forest management and planning purposes, as the UAV is able to give an estimation of the number of palms and their biomass in an area. Remote sensing derived heights can also supplement field data collection, offering an alternative to labor intensive palm height measurements in the field.