Quantifying aboveground vegetation water storage combining Landsat 8 OLI and Sentinel-1 imageries

Abstract

Although optical remote sensing has been widely used to monitor vegetation characteristics, its use on aboveground vegetation water storage (AVWS) is quite scare. Therefore, we combined the Landsat 8 OLI and Sentinel-1 imageries to quantify AVWS using generalized linear regression (GLM), artificial neural network (ANN) and random forest (RF) with the linkage of field observations in Mao Country, Southwest China. Field observations showed that the AVWS varied significantly among different ecosystems (p < 0.001). In terms of model efficiency and root mean square error, ANN (0.66, 56 Mg ha⁻¹) performed best compared to RF (0.52, 66 Mg ha⁻¹) and GLM (0.48, 69 Mg ha⁻¹). Total AVWS was 3.8 × 10⁷ Mg for the whole study area with 76% contributions from coniferous forests. Strong spatial patterns of AVWS were observed among different ecosystems, which were highly consistent with the spatial distributions of vegetation types. This research highlights a potential way to estimate AVWS through the combination of multispectral remote sensing and SAR data by linking machine learning algorithms, particular in mountainous areas.

Keywords:

• Aboveground vegetation water storage
• Landsat 8 OLI
• Sentinel-1
• ANN
• RF

Acknowledgements

The authors thank Yan Wen, Qin Zhao and Rui Liu for their aids in the field work. We are also thanks to the Geospatial Data Cloud and European Space Agency for the assistance in providing data.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [grant no. 41671432], foundation for University Key Teacher of Chengdu University of Technology [grant no. 10912-2019JX-06910] and the Key Laboratory of Geoscience Spatial Information Technology of Ministry of Land and Resources (Chengdu University of Technology).