Use of virtual and augmented reality technologies to inventory Miombo woodland carbon stocks in support of REDD

Background/Question/Methods: How do traditional field measures compare to measurements in virtual (VR) and augmented (AR) reality? From 2005 – 2017, traditional forest inventory methods were used to collect data on Miombo woodland structure: particularly tree and crown diameters, height, biomass, and vegetation carbon stock density (CSD), from fifty 30-m diameter circular sample stands in the 42,000 km² Niassa National Reserve’s (NNR) in northern Mozambique. The 30-m stand diameter corresponds to a Landsat 5, 7, and 8’s pixel size, which allowed the development of statistical models between pixel reflectance values and stand metrics. These models resulted in the spatial scaling of woodland stand attributes, particularly CSD, to the NNR landscape scale. In 2015, we digitized 35 of the 50 stands using a 1550-nm wavelength terrestrial laser scanner (TLS) to produce a 3-D VR environment of each stand. Each stand was scanned at a sampling rate of ~ 1 million points sec^-1. Five overlapping GPS positioned scans were collected per plot and registered to each other within ~ 4-mm RMSE into a single 3-D cloud of ~23 million points at a mean spacing of 1-cm. Commercial and open source software were used to both manually and automatically measure the same traditionally-measured stand metrics within each virtual stand. Manual and automatic measures were made within each stand either on a point cloud, a rendered mesh (3-D object model) or interpolated to a grid (a digital surface model). We then input a subset of a 3-D object model of stand 4 into a Microsoft HoloLens where the wearer can simultaneously see reality and the overlaid object model.

Results/Conclusions: The 3-D Miombo stand was overlaid within our lab and I was able to walk around the trees in the 30-m diameter stand and measure tree height, diameter-at-breast height, crown area and tree density. Field measures were comparable to measures in VR and AR (r > 0.85). Future research will concentrate on 1) improved 3-D rendering of stand objects and 2) develop a toolkit of AR forest mensuration tools.