Structural change points of NDVI in Mexico driven by climate oscillatio

Abstract

Based on the climatology of air temperature, precipitation, and the normalized vegetation index (NDVI), a regionalization of Mexico for the rainy season is presented through a non-parametric clustering algorithm known as DBSCAN. Thirty years of data, spanning from 1984 to 2013, are used to detect structural change points with the Mann-Kendall and Pettitt non-parametric tests applied on the NDVI, mean daily precipitation, 99th percentile precipitation, and mean daily air temperature. The relative predictive importance of the parameters examined was estimated using a Machine-Learning Random Forest algorithm that allows establishing a connection between changes in the NDVI and changes in air temperature, average precipitation, and extreme precipitation for some regions. Modulation by large-scale climate phenomena, such as the Atlantic Multidecadal Oscillation (AMO) and the Pacific Decadal Oscillation (PDO), as well as interannual modulation by El Niño/Southern Oscillation (ENSO) are explored. Structural change points in the series appear to be modulated mainly by the phase shift of the AMO and those of the ENSO and PDO in 1997.