During this decade, the Amazon region has suffered two severe droughts in the short span of five years – 2005 and 2010. Studies on the 2005 drought present a complex, and sometimes contradictory, picture of how these forests have responded to the drought. Now, on the heels of the 2005 drought, comes an even stronger drought in 2010, as indicated by record low river levels in the 109 years of bookkeeping. How has the vegetation in this region responded to this record-breaking drought? Here we report widespread, severe and persistent declines in vegetation greenness, a proxy for photosynthetic carbon fixation, in the Amazon region during the 2010 drought based on analysis of satellite measurements. The 2010 drought, as measured by rainfall deficit, affected an area 1.65 times larger than the 2005 drought – nearly 5 million km2 of vegetated area in Amazonia. The decline in greenness during the 2010 drought spanned an area that was four times greater (2.4 million km2) and more severe than in 2005. Notably, 51% of all drought-stricken forests showed greenness declines in 2010 (1.68 million km2) compared to only 14% in 2005 (0.32 million km2). These declines in 2010 persisted following the end of the dry season drought and return of rainfall to normal levels, unlike in 2005. Overall, the widespread loss of photosynthetic capacity of Amazonian vegetation due to the 2010 drought may represent a significant perturbation to the global carbon cycle.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 On the heels of the once-in-a-century [Marengo et al., 2008] drought in 2005, comes an even more severe drought in the Amazon region [Lewis et al., 2011]. The causes of the 2010 drought still need to be investigated and are presently unknown, but like the 2005 drought it was intense and coincided with the dry season. The Rio Negro water level at the Manaus harbor is one of the most useful drought characterization indexes in Amazonia because it integrates rainfall totals over the entire western Amazon basin and is the longest available time series record in the region (since 1902). This index was at its lowest level (13.63 m above the local reference level, not 13.63 m lower in October than the long-term average for that month, as stated by Lewis et al. ) since 1902 on October 23, 2010 (Figure 1). The lowest level in 2005 was 14.75 m, or eighth lowest in the 109-year Rio Negro Manaus time series (Table S1). The main Amazon channel, Rio Solimões, also reached record low levels between October 14 and October 23, 2010 at various stations on its course (Tabatinga, Itapéua, Careiro, and Parintins). The river levels began to ascend with the arrival of rains in mid- to late-October 2010. As of November 25th 2010, the Rio Negro level is tracking the minimum-ever recorded river stage recovery (Figure 1). Year 2010 is now the driest year on record according to these river stage data.
 There is presently only a single report on the impact of the 2010 drought on Amazon vegetation, namely Lewis et al. , who evaluated changes in above ground biomass as a function of maximum climatological water deficit using satellite-based precipitation data. However, direct evidence of the impact of the 2010 drought on Amazonian vegetation is currently lacking, and this is presented in this article using satellite-based estimates of vegetation greenness, which is a proxy for photosynthetic carbon fixation. In addition, a comparative analysis of the impacts of the 2005 and 2010 droughts with regards to their areal extents, severity and post-drought effects is presented in this article to assess the two droughts that occurred in a relatively short time span.
2. Data and Methods
 We used the latest versions of satellite-based datasets of precipitation and vegetation greenness in this study. The greenness data consisted of Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra satellite. These indices, when evaluated from space-based reflectance measurements free of atmospheric corruption, that is, contamination of vegetation-reflected radiation with cloud- and/or aerosol-reflected radiation, represent direct observations of the physiologically functioning greenness level of vegetation canopies [Myneni et al., 1995; Huete et al., 2006]. Although the satellite data were cloud-filtered and corrected for aerosol-corruption effects, the data were further processed to remove residual atmosphere-corruption to produce the best possible signals from the vegetation. The precipitation data consisted of monthly precipitation rate (millimeters/hour, mm/hr) available from the Tropical Rainfall Measuring Mission (TRMM). The details of each dataset are provided in the auxiliary material. The analyses presented in this paper mainly consist of evaluating standardized anomalies of satellite-based data, expressed as
Where, a is the standardized anomaly of a given quantity (e.g., precipitation, vegetation greenness etc.) in a specific year (2005 or 2010) calculated using its value, x, in that year and long term mean, m, and standard deviation, s, over a reference period. Full details of data processing and anomaly calculation are provided in the auxiliary material.
 The Rio Negro water levels started to recede fast in August 2010, setting the start of 2010 drought, whereas in 2005, the drought started in July (Figure 1). Plant-available soil moisture data across the wide Amazon basin are not available to assess the spatial extent and impacts of droughts here. Therefore, we resorted to characterizing the drought through surrogates such as precipitation anomalies, as has been done before [Aragao et al., 2007; Saleska et al., 2007; Anderson et al., 2010]. Rain gauge network is also too sparse across the basin and the available gauge data merged with rainfall amounts inferred from satellite scatterometer (Tropical Rainfall Measuring Mission, TRMM) observations [Huffman et al., 1995] offer the best characterization of precipitation deficit, and possibly drought [Aragao et al., 2007], in the Amazon region. Therefore, we analyzed merged precipitation data from TRMM satellite and other sources for the period January 1998 to December 2010 to assess the spatial extent and severity of droughts in the Amazon, in spite of well-known limitations of this merged precipitation data set [Adler et al., 2000] (cf. auxiliary material).
3. Results and Discussion
 We characterize the 2010 dry season drought as July to September (JAS) precipitation anomalies less than −1 to be consistent with previous studies [Aragao et al., 2007; Saleska et al., 2007; Samanta et al., 2010b]. The 2010 drought impacted nearly the entire tropical region of South America south of the Equator unlike the 2005 drought, which affected mostly the southwestern Amazon (Figure S1). These patterns of precipitation deficit are approximately consistent with river stage data – the Rio Negro and the main stem Amazon river are hydrologically connected in Manaus, so receding levels on the main stem, which drains the most affected southwestern parts of the basin, can be measured at the Manaus harbor. About 41% of the vegetated area between 10°N–20°S and 80°W–45°W experienced JAS precipitation standardized anomalies less than −1 in 2010 (4.94 million km2) compared to 25% in 2005 (3 million km2). Notably, 50% of all forests within this vegetated region were subject to third quarter precipitation anomalies less than −1 std. in 2010 (3.3 million km2) compared to 34% in 2005 (2.3 million km2). The 2010 drought thus impacted a larger area and more rainforests than the 2005 drought, consistent with the analysis presented by Lewis et al. .
 To assess the impact on vegetation from these two droughts, we analyzed two different satellite-derived vegetation index data (cf. section 2). The Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) data, which are proxies for photosynthetic carbon fixation [Myneni et al., 1995; Huete et al., 2006; Yang et al., 2007; Brando et al., 2010], show wide spread declines, especially south of the Equator, during the 2010 drought, in contrast to the 2005 drought (Figure 2). About 49.1% of the vegetated area that was subject to drought shows greenness index declines (July to September NDVI standardized anomalies less than −1) in 2010 (2.43 million km2) compared to 19.9% in 2005 (0.60 million km2). Notably, 51.4% of all forests subjected to drought show similar declines in 2010 (1.68 million km2) compared to 14.3% in 2005 (0.32 million km2). The areas of greenness declines in 2005 and 2010 (Figure S3) generally coincide with the drought epicenters identified by Lewis et al. . In addition, there are large areas of vegetation greenness declines in 2010 (Figure S3) that were not identified in the Lewis et al.  study. Overall, these declines represent a significant loss of photosynthetic capacity of Amazonian vegetation [Myneni et al., 1995; Huete et al., 2006; Yang et al., 2007; Brando et al., 2010] and thus may represent a significant perturbation to the global carbon cycle, as the Amazon rainforests contribute a disproportionately large fraction to global annual net primary production (about 15%) relative to their area [Nemani et al., 2003]. The scale of this perturbation, though, is still to be quantified.
 Undisturbed Amazon rainforests were reported to have greened-up during the 2005 drought based on analysis of a previous version of the Enhanced Vegetation Index (EVI) data [Saleska et al., 2007]. This has now been shown not to be the case with current versions of both EVI [Samanta et al., 2010a, 2010b] and NDVI data (Figure S3). With respect to the 2010 drought, both EVI and NDVI data show widespread declines in vegetation greenness (Figure 2). The two droughts coincided with the dry season (July to September), and in both cases, rainfall returned to its normal level in the following months (Figures S4b and S4d). However, the greenness declines observed during the dry season persisted into the following three months, October through December, in 2010, but not in 2005 (Figures S4a and S4c), clearly indicating the severity and potential damage to the vegetation in this region.
 A comparison of NDVI and EVI anomalies from vegetated areas affected by both droughts also reveals the varied impacts of these two droughts (Figure 3). The spatial extent of greenness declines increased nearly five-fold in 2010 compared to 2005, which is consistent with TRMM precipitation analysis (Figure S1) that shows a much larger area under precipitation deficit in 2010 compared to 2005. The intensification of these declines in 2010 is also evident in the distributions of NDVI and EVI anomalies (Figure 3). The NDVI anomalies in 2010 display a strong positive skew, i.e. characterized by a majority of negative anomalies, with a peak value between −1 and −1.5 std., which is statistically different than that observed in 2005 (p < 0.001 from a two-sided t-test). A similar positive skew is also observed in the distribution of EVI anomalies in 2010, with a peak value at about −1 std., which is also statistically different (p < 0.001) than the 2005 EVI anomaly distribution (Figure 3). Hence we conclude that the impacts of 2010 drought on vegetation in the Amazon region were not only more widespread but also more severe and persisted well beyond the drought period, when compared to the 2005 drought. Finally, a simple analysis of the probability of occurrence of these droughts suggests that the 2010, rather than the 2005, drought to be the “once-in-a-century” drought (Table S1).
 We acknowledge funding from Earth science and Advanced Information Systems Technology programs at NASA. This research was performed using NASA Earth Exchange. NEX combines state-of-the-art supercomputing, Earth system modeling, remote sensing data from NASA and other agencies, and a scientific social networking platform to deliver a complete work environment in which users can explore and analyze large Earth science data sets, run modeling codes, collaborate on new or existing projects, and share results within and/or among communities. Arindam Samanta and Liang Xu contributed equally to this work.
 The Editor thanks Liana Anderson for her assistance in evaluating this paper.