Both NDVI time-series and LAI curves contain some missing or bad quality data over the seasonal trajectory due to various reasons. Based on the assumption that no rapid changes in variable values should occur between adjacent time periods, the missing values could be interpolated from adjacent available data. In order to investigate the effect of different interpolation methods, we have examined linear, cubic spline and Fourier transform interpolation methods for
NDVI and LAI. Due to total amount of available data and data quality in the VEGETATION and MODIS original data, this investigation was only carried out for AVHRR NDVI time series data. Downloaded weekly AVHRR NDVI was first smoothed using the best index slope extraction (BISE) to reduce the noise (Viovy et al., 1992). Table 2
shows the linear regression models based on the data interpolated from each method. The results confirmed the
relationships described in the results section. For each pair of interpolation methods, NDVI showed strong linear relationships in leaf production and leaf senescence periods but quite weak linear relationships during the leaf constant period. Interannual variations in the relationship were noted for each interpolation method and each NDVI data source.