Estimating the fractional cover of photosynthetic vegetation, non-photosynthetic vegetation and bare soil from MODIS data: Assessing the applicability of the NDVI-DFI model in the typical Xilingol grasslands

• Published in: International journal of applied earth observation and geoinformation Vol. 76; pp. 154 - 166
• Main Authors: Wang, Guangzhen, Wang, Jingpu, Zou, Xueyong, Chai, Guoqi, Wu, Mengquan, Wang, Zhoulong
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2019
• Subjects: Bare soil; carbon sequestration; desertification; DFI; Fractional cover; fuels (fire ecology); grasslands; linear models; MODIS; Non-photosynthetic vegetation; normalized difference vegetation index; phenology; photosynthesis; soil; spatial data; time series analysis; wind erosion; Xilingol grassland; Bare soil; Fractional cover; Xilingol grassland; Non-photosynthetic vegetation; DFI; MODIS
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2018.11.006

[Display omitted] •Relationships between non-photosynthetic vegetation (NPV) coverage and 6 NPVIs were explored.•Develop an NDVI-optimal NPVI model and quantitatively assess the accuracy of the model retrieval of fPV, fNPV and fBS.•Produce a MODIS-based time series of fPV, fNPV and fBS at 500 m resolution for the typical steppe.•Discussed several difficult factors for estimating fPV, fNPV and fBS using remote sensing. Quantitative estimations of the fractional cover of photosynthetic vegetation (fPV), non-photosynthetic vegetation (fNPV) and bare soil (fBS) are critical for soil wind erosion, desertification, grassland grazing, grassland fire, and grassland carbon storage studies. At present, regional and large-scale fPV, fNPV and fBS estimations have been carried out in many areas. However, few studies have used moderate resolution imaging spectroradiometer (MODIS) data to perform large-scale, long-term fPV, fNPV and fBS estimations in the Xilingol grassland of China. The objective of this study was to quantitatively estimate the time series of fPV, fNPV and fBS in the typical grassland region of Xilingol from MODIS image data. Field measurement spectral and coverage data from May and September 2017 were combined with the 8-day composite product (MOD09A1) acquired during 2017. We established an empirical linear model of different non-photosynthetic vegetation indices (NPVIs) and fNPV based on the sample scale. The linear correlation between the dead fuel index (DFI) and fNPV was best (R2 = 0.60, RMSE = 0.15). A normalized difference vegetation index (NDVI)-DFI model based on MODIS data was proposed to accurately estimate the fPV, fNPV and fBS (estimation accuracies of 44%, 71%, and 74%, respectively) in the typical grasslands of Xilingol in China. The fPV, fNPV and fBS values for the typical grassland time series estimated by the NDVI-DFI model were consistent with the phenological characteristics of the grassland vegetation. The results show that the application of the NDVI-DFI model to the Xilingol grassland is reasonable and appropriate, and it is of great significance to the monitoring of soil wind erosion and fires in grasslands.