GIS and Remote Sensing Unit at FEM-CRI

MODIS NDVI/EVI analysis

From MODIS (flown on Terra and Aqua satellites), Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI) maps are available. They are aggregated to 16 days to minimize cloud contamination. We prefer to use EVI in our studies as it tends to perform better than NDVI. EVI is less prone to saturation as well as less sensitive to haze due to the inclusion of the blue channel (Huete et al., 2002). This is of interest in mountainous regions where valleys are often relatively hazy.

Ecological indicators from MODIS Vegetation indices

EVI permits to detect inter-annual seasonal vegetation differences (pixel-wise map subtraction), spring/autumn detection (EVI thresholding) and the calculation of the growing season length (period with EVI over certain value). In Trentino, we observe for example that even over short distances the seasons are slightly shifted due to effect of valley orientation and exposition.

(click to enlarge)

Landuse/Landcover and EVI

The detailed identification of zones relevant to a particular ecological problem can be extracted from the combined analysis of LULC (e.g., the European CORINE maps) and MODIS EVI. The screenshot shows CORINE polygon boundaries over MODIS EVI from June 2003 (Trento, Italy, with Molveno and Caldonazzo lakes in grey).

(click to enlarge)

Comparison of NDVI and EVI performance

Both NDVI and EVI maps are colored with identical color table (MODIS/Terra scene MOD13, composite of 21 March - 5 April 2000, Calabria, Southern Italy). EVI is less prone to atmospheric distortion. Both maps have been processed according to the list shown below.

Calabria, Southern Italy: NDVI and EVI (composite of 21 Mar-5 Apr 2000)

GIS processing of MODIS NDVI/EVI (MOD13/MYD13)

The following preprocessing steps are required to obtain usable NDVI/EVI maps from MOD13:

• reprojection from SIN to a commonly used projection (e.g., UTM)
• application of the quality maps (requires pixelwise bitpattern analysis)
  
• division by 10000.0 to get back the range from -1.0 .. +1.0
  
• elimination of pixel with value outside of that range
  
• application of color table

Reference

M. Neteler, 2008: Free GIS Software meets zoonotic diseases: From raw data to ecological indicators, Proc. FOSS4G 2008, 29 Sept-3 Oct 2008, Cape Town, South Africa [ Abstract ]