Retrieval of Grassland Aboveground Biomass through Inversion of the PROSAIL Model with MODIS Imagery

• Published in: Remote sensing (Basel, Switzerland) Vol. 11; no. 13; p. 1597
• Main Authors: He, Li, Li, Ainong, Yin, Gaofei, Nan, Xi, Bian, Jinhu
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.07.2019
• Subjects: aboveground biomass; aboveground biomass (AGB); algorithms; Biomass; Canopies; carrying capacity; China; climate change; Dry matter; empirical research; grassland; Grasslands; Growth models; Imagery; Landsat; Landsat satellites; Leaf area; Leaf area index; leaf dry matter content; Leaves; livestock; Mathematical models; MCD43A4; Methods; moderate resolution imaging spectroradiometer; MODIS; monitoring; Neural networks; Parameters; PROSAIL; Radiative transfer; Remote sensing; Retrieval; Scale models; sustainable development; time series analysis; Variables; Vegetation; China
• DOI: 10.3390/rs11131597
• ISSN: 2072-4292

The estimation of aboveground biomass (AGB), an important indicator of grassland production, is crucial for evaluating livestock carrying capacity, understanding the response and feedback to climate change, and achieving sustainable development. Most existing grassland AGB estimation studies were based on empirical methods, in which field measurements are indispensable, hindering their operational use. This study proposed a novel physically-based grassland AGB retrieval method through the inversion of PROSAIL model against MCD43A4 imagery. This method relies on the basic understanding that grassland is herbaceous, and therefore AGB can be represented as the product of leaf dry matter content (Cm) and leaf area index (LAI), i.e., AGB = Cm × LAI. First, the PROSAIL model was parameterized according to the literature regarding grassland parameters retrieval, then Cm and LAI were retrieved using a lookup table (LUT) algorithm, finally, the retrieved Cm and LAI were multiplied to obtain the AGB. The method was assessed in Zoige Plateau, China. Results show that it could reproduce the reference AGB map, which is generated by upscaling the field measurements, in terms of magnitude (with RMSE and R-RMSE of 60.06 g·m−2 and 18.1%, respectively) and spatial distribution. The estimated AGB time series also agreed reasonably well with the expected temporal dynamic trends of the grassland in our study area. The greatest advantage of our method is its fully physical nature, i.e., no field measurement is needed. Our method has the potential for operational monitoring of grassland AGB at regional and even larger scales.