Evaluation of vegetation greenness metrics for characterizing tropical forest cover types–a case study using NOAA AVHRR datasets

Abstract

We used vegetation greenness metrics in conjunction with Fourier analysis, a signal detection algorithm to characterize different forest types from NOAA AVHRR datasets. We selected three widely different tropical forest patches in the Indian region, i.e. tropical evergreen forests of the Western Ghats, tropical moist mixed deciduous forests of the Eastern Ghats and tropical wet evergreen forests located in North East India, representing different climatic, physiographic and topographic gradients. In addition to Fourier analysis, we used different vegetation greenness metrics derived from NDVI. These included annual sum of NDVI, maximum, minimum, amplitude, mean, coefficient of variation in NDVI and time integrated NDVI to discriminate the forest types from space. Results suggested relatively high amplitude in vegetation greenness for mixed deciduous forests compared to other forest types. The phase i.e. timing of the peak of vegetation greenness has been found to be quite distinct for different forest types. For mixed deciduous forests (Eastern Ghats), the phase was high during the fourth week of December (phase angle of 322°), compared to the third week of December (phase angle of 312°) for wet evergreen forests of northeast India and early January (phase angle of 6°) for Evergreen forests of Western Ghats. Although, the phase angle for mixed deciduous forests and wet evergreen forests of northeast India were close, differences in amplitude were quite distinct. Evaluation of NDVI metrics for different forest types suggested that wet evergreen forests (northeast India) showed distinct higher values for sum of NDVI, max-NDVI, min-NDVI, integrated NDVI and mean NDVI. Results clearly suggested the potentiality of Fourier analysis and NDVI metrics for characterizing tropical forest types in the Indian region.

Keywords:

• Tropical forests
• Forest type discrimination
• Fourier analysis
• NDVI metrics
• NOAA, AVHRR data

Acknowledgements

The authors are grateful to Professor Tateishi, CERES, Chiba University, Japan for providing the datasets. We would also like to thank the Director of the Agroecosystem Management Program, Ohio Agricultural Research and Development Center, The Ohio State University, USA for encouragement and facilities.