MODELLING NDVI TIME SERIES TO FILL GAPS OF METEOROLOGICAL DATA

GONZÁLEZ LOYARTE , MARGARITA; MENENTI MASSIMO; DIBLASI, ANGELA M.

Proceedings EARSeL Special Interest Group Temporal Analysisi of Satellite Images

Proceedings EARSeL - www.isn-oldenburg.de

Lugar: Oldenburg; Año: 2012 p. 44 - 51

0257-0521

ABSTRACT We applied a measure of foliar phenology to interpolate climate statistics and produce a bioclimate classification for a vast plain in Argentina, with sparse weather observations. As a measure of foliar phenology, we used parameters obtained by modelling NDVI time series with a Fast Fourier Transform (FFT) applied to a 9-year time series of NOAA-AVHRR NDVI GAC images. FFT decomposes the series into an average signal and two sinusoidal components. Selected FFT parameters were mean NDVI, amplitude and phase for a 1-year period. Climate data were annual rainfall (P) and mean temperature (T) expressed as Potential Evapotranspiration (PET) estimated by an empirical equation (PET= 68.64 T). P/PET ratio was related to FFT parameters by fitting a multiple linear regression model with P/PET as predicted variable and FFT parameters as predictive variables. The regression model, that explained 92% of the P/PET variation, was then applied to the entire selected images (parameters) to obtain a map of the P/PET ratio. The P/PET map was compared with existing climate maps and ancillary data to derive a consistent bioclimate map. Mean annual phenological rhythm was graphed for each bioclimate by reconstructing the yearly NDVI curve. This shows that aridity reduces the contrast between minimum and maximum NDVI and that time of maximal vegetation cover varies from January (semiarid) to April (subdesert). The proposed method is an adequate tool to extend meteorological data into regions where climate data have an uneven coverage or poor spatial resolution. This finding shows that each selected FFT parameter was necessary because none was significant by itself.