Estimating fractional vegetation cover and the vegetation index of bare soil and highly dense vegetation with a physically based method

Highlights

• Method to estimate NDVI of pure components and fractional vegetation cover (FVC).

• Proposed method quantitatively derives NDVI_v and NDVI_s and needs no statistics.

• Proposed method performs well with simulated data and real satellite data.

• In contrast, traditional methods estimate NDVI_v, NDVI_s and FVC with higher errors.

• Flexible to generate NDVI_v, NDVI_s and FVC at different spatial scales.

Abstract

Normalized difference vegetation index (NDVI) of highly dense vegetation (NDVI_v) and bare soil (NDVI_s), identified as the key parameters for Fractional Vegetation Cover (FVC) estimation, are usually obtained with empirical statistical methods However, it is often difficult to obtain reasonable values of NDVI_v and NDVI_s at a coarse resolution (e.g., 1 km), or in arid, semiarid, and evergreen areas. The uncertainty of estimated NDVI_s and NDVI_v can cause substantial errors in FVC estimations when a simple linear mixture model is used. To address this problem, this paper proposes a physically based method. The leaf area index (LAI) and directional NDVI are introduced in a gap fraction model and a linear mixture model for FVC estimation to calculate NDVI_v and NDVI_s. The model incorporates the Moderate Resolution Imaging Spectroradiometer (MODIS) Bidirectional Reflectance Distribution Function (BRDF) model parameters product (MCD43B1) and LAI product, which are convenient to acquire. Two types of evaluation experiments are designed 1) with data simulated by a canopy radiative transfer model and 2) with satellite observations. The root-mean-square deviation (RMSD) for simulated data is less than 0.117, depending on the type of noise added on the data. In the real data experiment, the RMSD for cropland is 0.127, for grassland is 0.075, and for forest is 0.107. The experimental areas respectively lack fully vegetated and non-vegetated pixels at 1 km resolution. Consequently, a relatively large uncertainty is found while using the statistical methods and the RMSD ranges from 0.110 to 0.363 based on the real data. The proposed method is convenient to produce NDVI_v and NDVI_s maps for FVC estimation on regional and global scales.

Introduction

Fractional vegetation cover (FVC) is the ratio of vertically projected area of vegetation to the total surface extent, generally expressed in relation to a unit area. FVC plays an important role in climatic, hydrologic, and geochemical cycles (Jiang et al., 2006, Jiménez-Muñoz et al., 2009) and is widely used to describe vegetation quality and ecosystem change. It is also a controlling factor in transpiration, photosynthesis, global climate changes and other terrestrial processes and climate models (Gutman and Ignatov, 1998, Barlage and Zeng, 2004, Jiapaer et al., 2011, Arneth, 2015).

Remote sensing provides the most efficient way to derive FVC at regional and global scales. Among all the estimation methods, the linear mixture model has been widely applied (Gutman and Ignatov, 1998, Van Der Meer, 1999, Zeng et al., 2000, Scanlon et al., 2002, Montandon and Small, 2008, Jiapaer et al., 2011, Delamater et al., 2012, Yang et al., 2013, Zhang et al., 2013, Iordache et al., 2014, Guo et al., 2015, Waldner et al., 2015). The linear mixture model requires a definition of the normalized difference vegetation index (NDVI) values of highly dense vegetation (NDVI_v) and bare soil (NDVI_s) (Montandon and Small, 2008, Jiapaer et al., 2011). The values of NDVI_v and NDVI_s reflect vegetation species, soil types and other factors (Baumgardner et al., 1986, Jensen, 2000). The model describes the vegetation proportion by scaling NDVI according to NDVI_v and NDVI_s (Eq. (1)).

$$FVC=\frac{NDVI-NDV{I}_s}{NDV{I}_v-NDV{I}_s}$$

However, the NDVI_v and NDVI_s are often difficult to obtain, particularly in sparsely or highly vegetated areas. The traditional methods for NDVI_v and NDVI_s estimation can be summarized as: 1) the maximum and minimum NDVI in a study area (Gutman and Ignatov, 1998); 2) the accumulative maximum and minimum over a long-term series dataset (Zeng et al., 2000); 3) NDVI_v and NDVI_s based on field measurements or high-resolution remotely sensed data (Jiapaer et al., 2011, Zhang et al., 2013). Of these approaches, the statistical methods (methods 1 and 2) are easy to implement but usually have large uncertainty due to the different empirical thresholds that have been used (Zeng et al., 2000). The statistical methods also do not work well in extreme cases. For example, in arid and semiarid areas where vegetation is usually sparse all year, it is hard to obtain NDVI_v based on empirical statistics. NDVI_s in evergreen areas is also hard to obtain where vegetation is always dense. In cases where higher resolution images are needed for some methods the data might not be available at all. All of these issues can lead to considerable errors in FVC estimation. It has been reported that the underestimation of NDVI_s in sparse vegetation areas may cause the overestimation of FVC as high as 0.2 (Montandon and Small, 2008, Ding et al., 2016).

This study provides a method to estimate NDVI for highly dense vegetation and bare soil in a linear mixture model using multi-angle observations and leaf area index (LAI) products. A specific angle, i.e., 57°, is used to construct equations, where the mean projection coefficient of leaves keeps invariable in a variety of leaf inclination angle distributions (Chen and Black, 1991, Zou et al., 2014). After combining the angular information and LAI, NDVI_v and NDVI_s are quantitatively obtained without other empirical knowledge or higher resolution data. The proposed method has been validated in two experimental settings: experiments using a vegetation canopy radiative transfer model, and experiments using the operational Moderate Resolution Imaging Spectroradiometer (MODIS) Bidirectional Reflectance Distribution Function (BRDF) products.