Background+on+Trend+Map

= = =**Background and Trend Analysis**=

Changes in the character of the vegetated land surface can be expressed in terms of temporal trends in the Normalized Difference Vegetation Index (NDVI) retrieved from spaceborne sensors. The NDVI exploits a spectral contrast between red and near infrared reflectance to indicate presence of green vegetation ( //3// ). Trends in NDVI as a function of time have typically been evaluated by the slope of a linear regression model, but many factors can degrade the reliability of the parameter estimates. Instead, we apply the Seasonal Kendall (SK) trend test, which is routinely used in analyses of climatological and hydrological time series ( //4, 5// ). We did not calculate trends in areas lacking either sufficient data or seasonality (SI). These thresholds filtered out deserts, inland waters, and cloudy and/or hazy areas, omitting <26% of the terrestrial surface (Table 1).

=** Materials and methods **= We selected a NASA MODIS product (Terra+Aqua Nadir BRDF-Adjusted Reflectance data; MCD43C4) at a 0.05° spatial resolution and a 16-day temporal resolution to calculate the NDVI. Reflectance data are adjusted using models of bidirectional reflectance distribution functions to simulate reflectance from a nadir view. To attenuate noise, we resampled the data to 0.10°. We divided the terrestrial surface into three regions: Northern Hemisphere (NH; >20°N), the Equator (Eq; 20°N - 20°S) and the Southern Hemisphere (SH; >20°S). We selected for analysis all observations in NH from March 2000 through October 2008 (16 obs/yr), for EQ all observations from February 2000 through December 2007 (23 obs/yr), and all observations in SH from September 2000 through April 2008 (15 obs/yr). Missing composites were replaced by their composite period means, which does not affect the trend analysis. Pixel time series that either (1) lacked more than 20% of the data or (2) exhibited low NDVI seasonality, //viz//., average NDVI <0.10 and a coefficient of variation of seasonal NDVI <5%, were omitted from the analysis. The Seasonal Kendall (SK) trend test, corrected for autocorrelation, is a robust non-parametric alternative for standard linear regression that relies on fewer statistical assumptions; thus, it is well-equipped to pick up linear or non-linear trends or step changes that could result from changes in aboveground biomass or land cover or changes in the length of the growing season. The SK trend test provides indication of the direction of the trend and its significance, but does not estimate a rate of change.

.