*4.2. Inversion Period*

Originally, the VJB method was performed using 5-years of data. The results showed an improvement of ~30% in the NDVI using MODIS data [18], which agree with our results. When including a higher number of years in the inversion, we are increasing the number of observations that are used in the model. This means increasing the range of observation geometries and of possible NDVI values that the pixel can have during the years. However, the accuracy gained by including these observations in the models might be counterbalanced by the change in vegetation characteristics of the target over the years. The larger the amount of years, the more likely it is that this change occurs, and the less likely it is for the assumptions behind the VJB method to hold. Nonetheless, these land cover changes are contemplated to some extent through the NDVI regression implicit in the VJB method. The analysis from this section on the different bands and sensors has shown that this change in vegetation is insignificant compared to the information gained by the increased number of observations. This might only be true, however, so long as the vegetation dynamics or human factors such as deforestation or agricultural practices don't change the surface value significantly. In these cases, the noise of the time series after BRDF normalization is likely to be high, and a shorter inversion period is recommended that correctly quantifies these land cover changes. When using high spatial resolution, these land cover changes are more discernible than at moderate or high spatial resolution. In this case, a further analysis is required to determine what the adequate inversion period to retrieve the V and R parameters would be.
