Analysis and Modeling of Rainfed Crops Dynamics Based on NDVI Time Series in Central Spain

Abstract

Rainfed crops occupy 76% of the cultivated area of Spain being distributed throughout the whole country. The yield of these crops depends on the great interannual variability of meteorological factors. The monitoring and prediction of crop dynamics is a key factor for their sustainable management from an environmental and socioeconomic point of view. Long time series of remote sensing data, such as spectral indices, allow monitoring vegetation dynamics at different spatial and temporal scales and provide valuable information to predict these dynamics through time series analysis. The objectives of this study are as follows: (1) To assess the dynamics of rainfed crops in a typical dryland area of Spain and (2) to build dynamic models to explain and predict the evolution of these crops. The NDVI time series of a rainfed cereal crop area of central Spain have been analyzed using statistical time series methods and their values were predicted using the Box-Jenkins approach. At the model identification stage, the evaluation of their autocorrelation functions, periodogram, and stationarity tests has revealed that most of these series are stationary and that their dynamics are dominated by annual seasonality. The selected preliminary dynamic model presents a good degree of adjustment for a 30% of the studied pixels.