*3.2. Inversion Period*

Table 2 shows the average absolute noise (×<sup>10</sup>4) of the BELMANIP sites' time series obtained when the full dataset and 3-year intervals are used to derive the correction parameters for the red, NIR and NDVI. Inversion using full and 3-year parameters is shown in green and brown font respectively. The percentage under every noise value indicates the improvement with respect to the raw data. For almost every band and dataset, the average noise of the time series using full parameters is lower than using 3-year parameters. In MODIS data, the difference between the full and 3-year parameters is of ~2.5%, 8% and 5% in the red, NIR and NDVI time series noise. This difference is lower for AVHRR-pre (~0%, 7% and 6%) and AVHRR (~2%, 5% and 1%) data.

**Table 2.** Average noise (×<sup>10</sup>4) of the Benchmark Land Multisite Analysis and Intercomparison Products (BELMANIP2) sites' time series obtained before (raw) and after directional effects for the red and Near Infrared bands and the NDVI, using MODIS, the Advanced Very High Resolution Radiometer (AVHRR) pre-MODIS data and AVHRR data from top to bottom. Full and 3-year columns describe the noise after computing the Bidirectional Reflectance Distribution Function (BRDF) parameters using the whole time series, or 3-year intervals, respectively. The percentage under every noise value indicates the improvement with respect to the raw data.


Effectively, these results show that the effect of the gradual surface change in coarse spatial resolution pixels has little impact on the noise of the time series, as compared with the effect of the number of observations used to compute the parameters. Using a larger number of years retrieves information about the surface which is used in the model to retrieve more accurate parameters and reduce the noise of the normalized time series by a significant amount, especially for the NIR and NDVI.
