Trends in 15-year MODIS NDVI time series for Mexico

The number of studies exploring long data records have increased in recent years and trend analysis has become a frequently-used analysis approach. This study tested parametric and non-parametric regression methods for trend estimation in MODIS NDVI data. The full record of 15 years of MODIS data from the Terra satellite was filtered to exclude clouds, shadow or other data of inferior quality from this analysis. Yearly averages and dry and wet season means were analyzed using linear least-squares and Theil-Sen regression methods. Corresponding statistical tests (F-test and Mann-Kendall test) indicated the significance of each regression model. Consistent spatial patterns were found in MODIS NDVI trend analysis which match with local knowledge and studies exploring socioeconomic, environmental and demographic factors of vegetation and land cover change in Mexico. Significant (p<0.05) positive trends in NDVI were found in the states of Chihuahua, northern Durango and Nuevo Leon due to increased woody coverage or Oaxaca due to vegetation densification. Negative trends occur in Sonora and Coahuila due to climate variability and around larger cities. Seasonal trend analysis helped interpreting and discerning anthropogenic from natural drivers.