Long Range Correlation in Vegetation Over West Africa From 1982 to 2011

Satellite-derived Normalized Difference Vegetation index (NDVI) data records offer important sources for long term correlation modelling over West Africa. In this study, we assessed long range correlations in half monthly NDVI records over West Africa from 1982 to 2011 using GIMMS NDVI. In our analysis, we assessed (a) the annual and seasonal trends obtained using Ordinary Linear Regression, (b) the detrended lag-1-autocorrelation C(1), (c) the Detrended Fluctuation Analysis (DFA) scaling Hurst exponent h and (d) the Multifractal (MF) characteristics of NDVI. Results show that there exist some patterns or trends in the records that persist over time. The value of C(1) for NDVI was obtained as 0.989 is significant at 95% confidence interval. Consequently, the scaling h values of the Hurst DFA showed that about 37.4, 20.5, 41.7 and 0.5% of the vegetated areas are anti-correlated (h <; 0.5), un-correlated (h = 0.5), correlated (0.5 <; h <; 1) and uncorrelated random walk (h = 1), respectively. The trend analysis from Ordinary Least square Regression (OLR) shows that about 54.3, 0.1 and 45.6% of the vegetated areas are positively, uncorrelated and negatively correlated, respectively. Our findings revealed that the DFA method performed better than OLR and the findings could be useful in identifying areas with improved and degraded vegetation, which cannot be properly captured by the OLR method. Accordingly, the comparison of the MF-DFA results of original data to those of shuffled and surrogate series indicated that the multifractal nature of considered time-series is both from PDF and long-range correlations but arguably, MF due to long range correlation dominates over West Africa. The research is therefore helpful in the formulating crop and environmental management policies that may be used to improve ecosystem management using a long term plan (inter-annual) or short term (inter-seasonal) planning.