Close Window

MODIS vegetation indices for detecting the 2005 Amazon drought

Liana O. Anderson, Oxford University Centre for the Environment,, lander@ouce.ox.ac.uk (Presenting)
Yadvinder Malhi, Oxford University Centre for the Environment,, ymalhi@ouce.ox.ac.uk
Luiz E.O.C. Aragao, Oxford University Centre for the Environment,, laragao@ouce.ox.ac.uk
Sassan Saatchi, Jet Propulsion Lab National Aeronautics and Space Administration, saatchi@congo.jpl.nasa.gov

In the last decades, the detection of drought occurrences and assessment of its severity using satellite data are becoming popular in disaster, desertification, crop production, phenology, land cover change and climate change studies. To detect the drought effects on different vegetation types, many methodologies have been developed, mostly relying on the use of vegetation indices. This communication reports the first attempt to assess the capability of MODIS NDVI, Enhanced Vegetation Index (EVI) and Normalized Difference Water Index (NDWI) time-series to detect the spatial pattern of the 2005 drought in Amazonia. To reach this objective, monthly composites of the MOD13A2 product were generated for the 2000 to 2006, based on maximum NDVI pixel value, for the entire basin. A South American land cover map updated with deforestation until 2005 for the Brazilian Amazon associated with a rainfall anomaly derived from TRMM data were used as basis for the sampling scheme. To identify intensity and duration of the canopy change / stress due to the drought across Amazonia, we calculated vegetation indices anomalies for 2005 and 2006 (NDVIanomaly, EVIanomaly, NDWIanomaly) as the departure from the 2000 - 2006 mean (VI2000 - 2006), normalized by the standard deviation (σ) in a pixel-by-pixel basis, based on 5 samples in 3 distinct areas affected by the rainfall anomaly in 2005. The spatial distribution analyses were based on re-sampling data to 0.25 degrees to diminish cloud coverage and noise in the dataset. Then, vegetation indices anomalies were calculated. To support the data interpretation, literfall data for 2 sites (Southern Colombia and Eastern Brazil) from 2004 to 2006 were used. Our preliminary results showed that despite the high variability in the vegetation indices response in the temporal series, they detected a persistence of an anomalous signal during 2005/2006. The spatial analysis showed NDWI anomaly in Jun/Jul 2005 in a region that is not used to water deficit, suggesting that this areas can be more sensitive to drought events and climate change. Finally, for the first site evaluated (Colombia), vegetation indices seems to not reflect literfall variability, suggesting that shade and other factors might be affecting vegetation indices response.

Science Theme:  LC (Land Use and Land Cover Change)

Session:  1A: Remote Sensing and the Carbon Cycle

Presentation Type:  Oral (view presentation (26978 KB))

Abstract ID: 55

Close Window