Remote sensing data fusion as a tool for biomass prediction in extensive grasslands invaded by L. polyphyllus

Abstract

Remote sensing data fusion is a powerful tool to gain information of quantitative and qualitative vegetation properties on field level. The aim of this study was to develop prediction models from sensor data fusion for fresh and dry matter yield (FMY/DMY) in extensively managed grasslands with variable degree of invasion by L. polyphyllus. Therefore, a terrestrial 3d laser scanner (TLS) and a drone-based hyperspectral camera was used to collect high resolution 3d point clouds and hyperspectral aerial orthomosaics of four extremely heterogenous grasslands. From 3d point clouds multiple features (vegetation height, sum of voxel, point density and surface structure) were extracted and combined with hyperspectral data to develop an optimized biomass model from random forest regression algorithm to predict FMY and DMY (n train = 130, n test = 33). Models from hyperspectral data solitarily had the lowest prediction performance (FMY: R 2 = 0.61, nRMSE r = 17.14; DMY: R 2 = 0.59, nRMSE r = 19.37). Higher performance was gained by models derived from 3d laser data (FMY: R 2 = 0. 76, nRMSE r = 13.3; DMY: R 2 = 0. 74, nRMSE r = 15.1). A fusion of both sensor systems increased the FMY prediction performance up to R 2 = 0.8; nRMSE r = 12.02 and the DMY prediction performance to R 2 = 0.81 and nRMSE r = 12.06. The fusion of complementary sensor systems can increase the power to predict biomass yields of heterogenous and extensively managed grasslands. It is a novel alternative to labour-intensive, traditional biomass prediction methods and to remote sensing methods using only single sensor data.

Global application of an unoccupied aerial vehicle photogrammetry protocol for predicting aboveground biomass in non‐forest ecosystems

“…Non‐destructive estimates of AGB are conventionally obtained from in situ measurements of attributes such as plant cover, height and stem diameters, using functions fitted to harvested biomass observations (Paul et al., 2016 ; Rudgers et al., 2019 ). Canopy volume, the product of height and cover, is often the strongest predictor of AGB for low‐stature plants like shrubs and herbs (Alonzo et al., 2020 ; Bendig et al., 2014 ; Cunliffe et al., 2020a ; Grüner et al., 2019 , 2020 , 2021 ; Huenneke et al., 2001 ; Kröhnert et al., 2018 ; Schulze‐Brüninghoff et al., 2020 ; Wijesingha et al., 2019 ). Remote sensing approaches have been widely used to extend the coverage of biomass predictions.…”

Identifying plant species in kettle holes using UAV images and deep learning techniques

“…Results of this study utilizing machine learning methods to UAV images achieved F 1‐scores that are comparable to previous studies of plant segmentation (Elkind et al., 2019; Martins et al., 2021; Torres et al., 2020). Some studies implemented additional information such as digital surface models (DSMs) or multi‐spectral data to classify species to obtain similar or better F 1‐scores (Benjamin et al., 2021; Chabot et al., 2018; Durgan et al., 2020; Husson et al., 2016; Schulze‐Brüninghoff et al., 2021). However, retrieving DSMs might be difficult in kettle holes, for example, under windy conditions due to moving vegetation (Pätzig et al., 2020) Moreover, the application of multi‐spectral cameras makes the approach more expensive.…”

Comparison of Different Important Predictors and Models for Estimating Large-Scale Biomass of Rubber Plantations in Hainan Island, China

“…However, this approach is both timeconsuming and spatially constrained [5]. Owing to its high spatial, temporal, and spectral resolution, as well as its capacity for continuous, long-term scientific observations, remote sensing technology has become a widely employed tool for biomass estimation [10][11][12][13]. Broadly speaking, forest biomass estimation via remote sensing has evolved through three stages: (1) estimation solely based on optical, radar, or Light Detection and Ranging (LiDAR) data [14,15]; (2) effective incorporation of multiple types of satellite imagery, mitigating the limitations of solitary satellite imagery data [16]; (3) selection of optimal variables from multi-source satellite imagery and ancillary data [5], thereby enhancing model accuracy [12,17].…”