Monitoring Vegetation Phenology Using Daily Nadir BRDF-Adjusted VIs from VIIRS

1. Introduction

Vegetation phenology is the timing of the biological events in plants such as flowering, leafing, fruiting and leaf-out (Lieth, 1974; Reed, 1994). Longtime accurate and consistent estimation of phenology at global scale can help understanding of inter-annual variability of vegetation and how climate changes affect vegetation. Vegetation Index (VI) from AVHRR (Advanced Very High Resolution Radiometer) and MODIS (Moderate Resolution Imaging Spectroradiometer) have been used to monitor timing of vegetation greenup, maturity, senescence and dormancy at regional to global scales in the past decades (Reed, 1994; Moulin, 1997; Zhang, 2003).The Visible Infrared Imaging Radiometer Suite (VIIRS) on the Suomi-NPP satellite (launched on October 28, 2011) can continue providing global observations to monitoring vegetation phenology. By fitting VIIRS Nadir BRDF- adjusted VIs of 2012 to a series of piecewise logistic functions, intra-annual vegetation dynamics (vegetation phenology) at SURFRAD (Surface Radiation) sites can be estimated.

2. Methodology

Three major steps were performed. First, Nadir BRDF- adjusted Reflectance(NBAR) is derived by utilizing a similar approach that used for the Collection V006 daily MODIS Bidirectional Reflectance Distribution Function (BRDF)/Albedo product. Second, NBAR are used to calculate VIs using Eq. (2) and Eq. (3). Then the VIs are fitted to estimate the timing of greenup, maturity, senescence and dormancy.

Multidate, multispectral, cloud-cleared, atmospherically-corrected VIIRS surface reflectances are used to fit the Ross-Thick/Li-Sparse-Reciprocal (RTLSR) semi-empirical BRDF model to generate BRDF model parameters.

        (1)

Where and ϕ are solar zenith, view zenith and relative azimuth angles; iso, vol and geo mean isotropic, volumetric and geometric. K_vol is the volumetric kernel derived from Ross-Thick volume scattering radiative model, and K_geo is the geometric kernel derived from Li-Sparse geometric shadow casting theory. f_iso, f_vol and f_geo are the weights given to the BRDF model parameters (Schaaf, 2002).

Clear sky observations are weighted based on their quality (shadowed or affected by aerosol), observation footprint, and proximity to the production date of interest (Wang, 2012). A Least-Squares Error function (LSE) is used to establish the analytical solutions for the model parameters f_k (Lucht, 2000). When a high quality full inversion is not possible due to insufficient of reflectance observations, a lower quality magnitude inversion is produced by using the latest full inversion retrieved as a priori. The model parameters f_k are then utilized to calculate NBAR (Schaaf, 2002).

Two VIs are used, NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced vegetation index).

                                   (2)

              (3)

To monitor vegetation phenology, same approach used by Zhang (2003) is utilized to fit time series VIs of SURFRAD sites to identify phonological transition dates.

3. Results

Figure 1 is the annual VIs of Fort Peck. It clearly shows the annual cycle of vegetation phenology. More results at other sites will be provided, but this initial case demonstrates that VIs from VIIRS can detect the inter-annual variability of vegetation, and assure data continuity for land surface climate and biosphere models.

Figure 1 VIs of Fort Peck