Modified vegetation indices for estimating crop fraction of absorbed photosynthetically active radiation

Abstract

The fraction of absorbed photosynthetically active radiation (FPAR) is an important biophysical parameter of vegetation. It is often estimated using vegetation indices (VIs) derived from remote-sensing data, such as the normalized difference VI (NDVI). Ideally a linear relationship is used for the estimation; however, most conventional VIs are affected by canopy background reflectance and their sensitivity to FPAR declines at high biomass. In this study, a multiplier, the ratio of the green to the red reflectance, was introduced to improve the linear relationship between VIs and crop FPAR. Three widely used VIs -NDVI, the green normalized difference VI (GNDVI), and the renormalized difference VI (RDVI) -were modified this way and were called modified NDVI (MNDVI), modified GNDVI (MGNDVI), and modified RDVI (MRDVI), respectively. A sensitivity study was applied to analyse the correlation between the three modified indices and the leaf area index (LAI) using the reflectance data simulated by the combined PROSPECT leaf optical properties model and SAIL canopy bidirectional reflectance model (PROSAIL model). The results revealed that these new indices reduced the saturation trend at high LAI and achieved better linearity with crop LAI at low-to-medium biomass when compared with their corresponding original versions. This has also been validated using in situ FPAR measurements over wheat and maize crops. In particular, estimation using MNDVI achieved a coefficient of determination (R 2 ) of 0.97 for wheat and 0.86 for maize compared to 0.90 and 0.82 for NDVI, respectively, while MGNDVI achieved 0.97 for wheat and 0.88 for maize, compared to 0.90 and 0.81 for GNDVI, respectively. Algorithms based on the VIs when applied to both wheat and maize showed that MNDVI and MGNDVI achieved a better linearity relationship with FPAR (R 2 = 0.92), in comparison with NDVI (R 2 = 0.85) and GNDVI (R 2 = 0.82). The study demonstrated that applying the green to red reflectance ratio can improve the accuracy of FPAR estimation.

Remote sensing for agricultural applications: A meta-review

“…Purely empirical approaches, also named as "regressions", consist of calibrating a numerical relationship between one or several measured biophysical variables and the remote sensing signal or a numerical transformation of this signal (using vegetation indices for instance, for an exhaustive list of such indices see Henrich et al (2009)). The simplest examples are linear or nonlinear relationships established between: (i) reflectances and Leaf Area Index (Viña et al, 2011), fraction of Absorbed Photosynthetically Active Radiation (Dong et al, 2015), chlorophyll (Gitelson et al, 2005), or water content (Chen et al, 2005); (ii) backscattering coefficient and LAI, crop water content, or crop height (McNairn and Shang, 2016); and (iii) LIDAR and chlorophyll content (Eitel et al, 2014). More advanced techniques include machine learning regression such as support vector machine (Mountrakis et al, 2011), random forest (Zhu and Liu, 2015), or Gaussian processes and neural networks (Camacho et al, 2017;Yuan et al, 2017).…”

Hyperspectral remote sensing for tobacco quality estimation, yield prediction, and stress detection: A review of applications and methods

“…(2020) . The empirical models mainly rely on data collection and statistics, also known as “regressions”, such as partial least squares regression (PLSR) ( Dong et al., 2015 ), support vector machine (SVM) ( Mountrakis et al., 2011 ), random forest (RF) ( Johansen et al., 2020 ), neural networks ( Yuan et al., 2017 ), etc. Its main task is to fit the numerical relationship between the measured agronomy traits in practical and spectral features.…”

Assessing the Impact of Climate Variability on Cropland Productivity in the Canadian Prairies Using Time Series MODIS FAPAR

“…FAPAR is a biophysical variable indicative of vegetation photosynthetic status [30][31][32][33]. It is directly related to primary productivity and is widely used in vegetation productivity modeling based on the light use efficiency theory [30,31].…”