Analysis of the Persistence of NDVI and Climatic Factors in the Shandong Peninsular
山东地区NDVI与气象因子持续性分析

Global change has been receiving much attention from the geosciences community in that the core of land cover change, vegetation, is one of the most sensitive factors in global change. Variations in vegetation can reflect the influence of meteorological factors as well as human activities in global change at short-term scales. It is an important issue for Land Use and Land Cover Change (LUCC) studies to explore fluctuation mechanisms and therefore to make reasonable estimation of the persistence of vegetation. To explore long-term correlation of vegetation with meteorological factors, i.e., temperature, precipitation, and sunshine duration, Normalized Difference Vegetation Index (NDVI) time series data were employed in this work to represent meteorological and vegetation factors respectively. Rescaled range analysis (R/S) was employed to analyze the persistence of the NDVI time series and three corresponding meteorological factors. The most obvious advantage of this method is that it is unnecessary to presume a specific distribution characteristic when analyzing time series data, i.e., the stability of the R/S-based results can reflect the time series data following other types of distributions. The point-based time series data are frequently unavailable in studying NDVI because of the discontinuity of remote sensing data. The SPOT/VEGETATION NDVI time series used in this study can, however, circumvent the problem. Meteorological data used were from the China Meteorological Administration. This study was conducted over Shandong Province with a long shoreline affected by Land Ocean Interactions in the Coastal Zone (LOICZ). In addition to rapid economic and societal development recent years, urbanization and industrialization have been speeded up during the past decade. Six meteorological stations generally representing the meteorological and climatic characteristics of the Shandong area were selected for this analysis. The R/S was used to unravel the long-term correlation on the basis of NDVI time series and corresponding meteorological observations from 1998 to 2008, particularly analyzing the Hurst index of the meteorological factors and NDVI time series for all the selected stations. Results showed that the temperature, precipitation and NDVI data essentially display the same distribution characteristics. The long-term correlation exists for all factors. Each of the long-term correlation can be divided into two phases, showing a stronger long-term correlation in the first phase, and a stronger short-term correlation in the second phase. The persistence of those factors was found to be very close. Except for the sunshine duration, positions of the turning points of the Hurst index were close in all factors, indicating roughly 550 d or 560 d. This demonstrates that the long-term correlation of NDVI is primarily affected by the persistence of temperature and precipitation, but less by sunshine duration. The time series of NDVI was of fractal chara