Fractional cover - MODIS, CSIRO algorithm

Vegetation Fractional Cover represents the exposed proportion of Photosynthetic Vegetation (PV), Non-Photosynthetic Vegetation (NPV) and Bare Soil (BS) within each pixel. The sum of the three fractions is 100% (+/- 3%) and shown in Red/Green/Blue colors. In forested canopies the photosynthetic or non-photosynthetic portions of trees may obscure those of the grass layer and/or bare soil. This product is derived from the MODIS Nadir BRDF-Adjusted Reflectance product (MCD43A4) collection 6 and has 500 meters spatial resolution. \n\nA suite of derivative products are also produced including monthly fractional cover, total vegetation cover (PV+NPV), and anomaly of total cover against the time series. \nMonthly: The monthly product is aggregated from the 8-day composites using the medoid method.\nAnomaly: represents the difference between total vegetation cover (PV+NPV) in a given month and the mean total vegetation cover for that month in all years available, expressed in units of cover. For example, if the mean vegetation cover in January (2001-current year) was 40% and the vegetation cover for the pixel in January 2018 was 30%, the anomaly for the pixel in Jan 2018 would be -10%. \nDecile: represents the ranking (in ten value intervals) for the total vegetation cover in a given month in relation to the vegetation cover in that month for all years in the time-series.\n\nMODIS fractional cover has been validated for Australia. \nLineage: Version 3.0: Fractional cover was derived using a linear unmixing methodology (Guerschman et al. 2015). The method uses all 7 MODIS bands and adds log transforms and band interaction terms to account for non-linearities in the spectral mixing. A cross-validation step was also included to select the optimal number of singular values to avoid over-fitting. The calibration and validation steps used 1171 field observations across Australia. Overall, the model fitted and applied to MCD43A4 fractional cover has a root mean square error (RMSE) of 12.9%, 18.1% and 16.6% for the PV, NPV and BS fractions respectively (percentage cover)\n\nVersion 3.1: ("v310") Same as Version 3.0 with the following modifications: 1- input data changed to MODIS Collection 6.0 (MCD43A4.006) (see source). 2- calibration dataset expanded to include ~3022 field measurement sites accross Australia. 3- overall accuracy improved to an RMSE of 11.3%, 16.1% and 14.7% for the PV, NPV and BS fractions respectively.\n\nMonthly vegetation cover is calculated from the 8-day composites using a medoid method as described in Gill et al. \n\nMonthly anomalies show the difference between total vegetation cover (PV+NPV) in a given month and the mean total vegetation cover for that month in all years available\\nMonthly deciles show the decile (ranking) for the total vegetation cover in a given month in relation to the vegetation cover in that month for all years in the time series.\n