Global Monthly Vegetation Cover Caveat About Using The NDVI Global Monthly Time Series

Introduction Page, Preface

---

The NDVI dataset is unique in that it is global, multi seasonal, multi annual, and multi spectral. These features are useful for environmental studies and numerical modeling. Despite its potential, these data have certain limitations (Gutman and Ignatov 1995) that are described below and should be heeded. There are known trends and biases in the data. The trends result from residual errors in corrections for sensor degradation of AVHRR solar channels (sensitivity changes, see Rao and Chen 1996) and satellite orbit degradation (equator crossing time changes, see Price 1991), whereas biases are caused by the compositing method used for data compression and primary cloud screening. Additional biases may be introduced in the secondary (post-composite) cloud screening by using climatological temperature thresholds for different years and in the areas of persistent clouds (see Gutman et al. 1994). Large-scale atmospheric perturbations, such as the Mt. Pinatubo eruption, contribute to the uncertainty in the NDVI data (Stowe et al. 1992). The user of the NDVI data should pay attention to the following comments.

Calibration

Figures 1 and 2 from Rao and Chen 1996 demonstrate the sensitivity drift of solar channels. The difference of degradation of the two AVHRR channels (1 and 2), NDVI causes trends and discontinuities as shown in Figure 2. However, residual errors in the corrected calibration, albeit small, are seen in these figures and will propagate in the analysis of NDVI time series.

Satellite orbit drift

Data from the afternoon polar orbiters is preferred for producing the NDVI because of the high sun elevation angle (low solar zenith angle). However, the equator crossing time drifts to a later hour as the satellites age (Price 1991). Satellite orbit drift results in a systematic change of illumination conditions which is one of the main sources of non-uniformity in multi annual NDVI time series. Figure 3 (personal communication with Garik Gutman) shows the general degradation over time of the orbits for NOAA-9, NOAA-11 and NOAA-14. Note that when NOAA-11 was first launched in 1988, it had a 1330 local time equator crossing. By 1994, its orbit had degraded to 1700 local time.

Residual cloud contamination

The issues of cloud contamination and calibration have been addressed in the new, enhanced Third Generation NDVI product. Although weekly composites contain fewer cloudy pixels than the original daily images, cloud contamination persists in global composite imagery (Gutman et al. 1994). A "post-composite" cloud screening was applied to the Second Generation NDVI composites (Gutman et al. 1994). Some residual cloud contamination, however, may be expected in the areas of persistent clouds. The screening technique relies on climatological temperature thresholds. This may result in biases due to the satellite orbit drift during the satellite's lifetime.

Mt. Pinatubo contamination

Large-scale atmospheric perturbations from volcanic eruptions, like Mt. Pinatubo, contribute to the nonuniformity of time series because stratospheric aerosols produced by these eruptions have a strong and persistent impact on observed radiances, mostly in the solar channels (Stowe et al. 1992).

Mt. Pinatubo erupted in June 1991, affecting satellite observation through December 1992. The impact can be seen in the NDVI data, with a more pronounced effect in the tropical and sub-tropical areas of the world.

The following months are considered problematic:

There is a gap between October 1994 - January 1995, when NOAA-11 AVHRR failed and before NOAA-14 was launched. Only two weeks in September 1994 were used.

The NDVI data for NOAA-14 from February 1995 - December 1997 are suspected of having a residual calibration error and will probably have to be regenerated.