Publications

Yin, GF; Li, AN; Wu, CY; Wang, JY; Xie, QY; Zhang, ZJ; Nan, X; Jin, HA; Bian, JH; Lei, GB (2018). Seamless Upscaling of the Field-Measured Grassland Aboveground Biomass Based on Gaussian Process Regression and Gap-Filled Landsat 8 OLI Reflectance. ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION, 7(7), 242.

Abstract
The spatially explicit aboveground biomass (AGB) generated through upscaling field measurements is critical for carbon cycle simulation and optimized management of grasslands. However, the spatial gaps that exist in the optical remote sensing data, underutilization of the multispectral data cube and unavailability of uncertainty information hinder the generation of seamless and accurate AGB maps. This study proposes a novel framework to address the above challenges. The proposed framework filled the spatial gaps in the remote sensing data via the consistent adjustment of the climatology to actual observations (CACAO) method. Gaussian process regression (GPR) was used to fully exploit the multispectral data cube and generated the pixelwise uncertainty concurrent with the AGB estimation. A case study in a 100 km x 100 km area located in the Zoige Plateau, China was used to evaluate this framework. The results show that the CACAO method can fill almost all of the gaps, accounting for 93.1% of the study area, with satisfactory accuracy. The generated AGB map from the GPR was characterized by a relatively high accuracy (R-2 = 0.64, RMSE = 48.13 g/m(2)) compared to vegetation index-derived ones, and was accompanied by a corresponding uncertainty map that provides a new source of information on the credibility of each pixel. This study demonstrates the potential of the joint use of gap-filling and machine-learning methods to generate spatially explicit AGB.

DOI:
10.3390/ijgi7070242

ISSN:
2220-9964