Albedo, fAPAR, Land Surface Temperature, Light Use Efficiency (LUE), NDVI , Precipitation, Solar radiation, Statics and Weather data

NDVI

The Normalized Difference Vegetation Index (NDVI) correlates well with the amount of photosynthetically active vegetation present and is therefore a measure of the greenness of the earth’s surface. Since it only requires a red and near infrared (NIR) band, the NDVI is a commonly used vegetation index that can easily be derived using most multispectral sensors. NDVI values range between -1 and 1. Vegetated areas have positive values closer to 1, bare soil/artificial surfaces have values of around 0, and water has negative NDVI values.

WaPOR application

Dekadal NDVI composites are produced and used internally as input for the computation of various data components, such as fAPAR, E and T. Furthermore it is used to characterize the temporal dynamics of vegetation conditions and hence to define the start and end of a growing season. NDVI can be derived from any sensor/platform that provides red and near infrared reflectances. For WaPOR version 3 NDVI is primarily based on VIIRS, Sentinel-2 and Landsat.

Table 10: Overview of NDVI intermediate data component (see also the chapter on quality layers)

Methodology

NDVI is based on the spectral reflectance of the red and near-infrared wavelengths. It is calculated as follows:

Level 1 preprocessing: NDVI is based on VIIRS reflectances. The NDVI time series is smoothened and interpolated by applying a pixel based temporal fill algorithm: the Whittaker smoother (Eilers, 2003, Eilers et al., 2017).

Level 2 preprocessing: NDVI is based on Sentinel-2 L2a reflectances which are resampled to 100m prior to calculating NDVI. The NDVI time series is smoothened and interpolated by applying the Whittaker smoother (Eilers, 2003, Eilers et al., 2017) for every 100x100m pixel.

Level 3 preprocessing: NDVI at Level 3 is based on both Sentinel-2 and Landsat bottom of atmosphere reflectance data. Once the Landsat and Sentinel-2 input data is pre-processed, the NDVI is calculated using the Red and NIR bands. The missing data in the daily NDVI time series (resulting from cloud masking) is then modelled using the gap-filling approach developed by Weiss, Daniel J., et al. (2014). This approach uses both spatial and temporal information within the NDVI time-series to fill the missing pixels. The noise in the filled NDVI time series is later smoothed out by applying the Whittaker smoother (Eilers, 2003, Eilers et al., 2017), a fast, flexible smoother technique.

Challenges

One of the main challenges when producing NDVI time series is the high cloud cover that occurs over certain areas. NDVI composites are produced to fill gaps and missing data that occur in the input satellite imagery. When an insufficient number of data observations are available within a composite period, the results of smoothing and gap filling are less accurate. Data layers that indicate the quality of each of the dekadal NDVI data composites are produced and should help the end user in correctly interpreting the data. The larger the gaps due to cloud coverage, the lower will be the accuracy of the prediction as a date as close as possible from the missing data should be used for reconstruction to ensure similar spectral characteristics.