The Amazon rainforest has recently experienced increasingly severe droughts. Currently, there is little known about the long-term effects of these droughts that result in tree dieback, altered rainforest canopy structure, and increased rainforest flammability in the Amazonia. Saatchi et al. (2013) use satellite microwave observations of rainfall from the Tropical Rainfall Measuring Mission and canopy backscatter data from the SeaWinds Scatterometer on board QuickSCAT to demonstrate that western Amazonia experienced severe water deficit during the dry season in 2005, and a subsequent disruption in canopy structure and decrease in canopy moisture. Furthermore, even with an increase in precipitation in the years following the drought in 2005, the decrease in canopy backscatter and thus alteration in the rainforest canopy characteristics and water content remained until the next major drought occurred in 2010. If droughts occur more frequently and severely in the Amazon rainforest due to climate change, this drought disturbance may cause considerable changes in the Amazonian rainforest canopy.—Hilary Haskell
Saatchi, S., Asefi-Najafabady, S., Malhi, Y., Aragao, L., Anderson, L., Myneni, R., Nemani, R. 2013. Persistent effects of a severe drought on Amazonian forest canopy. Proceedings of the National Academy of Sciences of the United States of America 110, 565–570.
Saatchi et al. conducted this study to understand the extent and severity of the long-term impacts of droughts on the Amazonian rainforest. Currently, only short-term effects of drought events have been studied through ground and satellite data. The long-term impacts of drought on Amazonian vegetation have only been studied on controlled, small-scale (1-ha plot) field experiments. Recent studies of rainforest structure and density show an increase in tree die off and a decrease in above ground biomass following drought periods. Saatchi et al. were able to conclude that the Amazon rainforest’s response to severe droughts includes the increased mortality of large trees’ leaves and branches in the upper canopy, when soil water availability declined below a critical threshold. This response has subsequent impacts on the Amazonian rainforest canopy
Two severe droughts have occurred in the Amazon over the past ten years, as demonstrated by the anomaly in the Rio Negro River’s water levels. These severe droughts are typically associated with El Nino–Southern Oscillation (ENSO) events, which result in decreased soil moisture. This decrease in soil moisture endangers Amazonian vegetation by crossing critical water availability thresholds for long periods of time. If water stress continues, it causes higher tree mortality rates and increased flammability across the Amazon rainforest. The two major Amazon droughts considered in this study occurred in 2005 and 2010. The 2005 drought was not considered an ENSO drought, due to its temporal and spatial extent. The peak of this drought hit during the dry season and mostly impacted south-western Amazonia, unlike droughts caused by ENSO events. River levels during the drought in 2005 were the lowest to date. The tropical North Atlantic Ocean’s sea surface temperature increase could be a major cause of the 2005 drought. Anomalies in precipitation leading up to the droughts were also major contributing factors to the severity of the droughts studied. In Southern Amazonia, rainfall decreased by almost 3.2% per year in 1970–1998.The region also experienced negative precipitation levels during the decade leading up to the severe drought in 2005. This pattern of decreased precipitation could worsen in the future based on climate model predictions of the effects of climate change. Therefore, if Amazonian droughts continue to be as frequent and severe as those considered in this study, the overall ecosystem function and health in the Amazon rainforest will be affected.
Satellite spectral observations are used to detect changes in the rainforest’s upper canopy characteristics, such as greenness and leaf area, and may be used to determine long-term impacts of drought. Due to clouds and atmospheric aerosols, after the drought in 2005, the data from these observations was largely contradictory. Therefore, Saatchi et al. used data from two microwave satellite sensors to measure precipitation and canopy water content, in order to quantify the severity of the Amazonian droughts in the years 2005 and 2010, and their subsequent impacts on canopy water content and structure. The authors analyzed Amazonian precipitation with three Tropical Rainfall Measuring Mission indices that measure monthly precipitation (TRMM; 1998–2010). These indices include the dry-season precipitation anomaly (DPA), dry season water deficit anomaly (DWDA), and maximum climatological water deficit (MCWD). For the DPA, DWDA and MCWD, more negative values indicate more severe water deficit. The information from these indices is complementary and provides spatial-specific indicators of the extent and severity of deficit in the Amazon rainforest.
To measure the impact of water deficit on the Amazon rainforest, the authors used observations from the SeaWinds Scatterometer onboard QuickSCAT (QSCAT: 2000–2009), which uses a microwave frequency to provide backscatter measurements that demonstrate temporal and spatial variations of water content and structure in the rainforest canopy. Backscatter in this study was the reflection of the microwave frequency back to the QSCAT Scatterometer. The QSCAT signal at 2.1 cm wavelength and incidence angles of 50° penetrates 1–5 m into the rainforest canopy, and then scatters from leaves and branches of the upper canopy of trees. Backscatter measurements demonstrate biophysical properties of forests, such as water content and canopy structure. Changes over time (diurnal and seasonal) of canopy water content and seasonal leaf phenology of the vegetation’s lifecycle events have the greatest impact on the radar backscatter. Large-scale rainforest degradation and deforestation result in fewer trees within the rainforest canopy. When fewer trees are present in the rainforest, this change affects the canopy by creating gaps in its structure and altering its water content or biomass. These structural and biophysical changes are thus reflected in the signal of the backscatter. To represent the upper-canopy rainforest structure and water content, the authors used QSCAT backscatter data from dawn orbits, which monitor vegetation at its least-water-stressed time of day. The time series for this study were on normalized monthly and seasonally, and demonstrated spatial variations in upper-canopy-forest structure over the Amazon.
The authors experienced failure in November of 2009 with the QSCAT sensor scanner. The scanner stopped collecting data globally, which made it impossible for them to analyze changes in the canopy during the 2010 drought. QSCAT was reliable and without bias or sensor degradation prior to this failure. In order to still analyze the 2010 drought, the authors used TRMM precipitation radar backscatter data, which responds to surface moisture deep in the canopy, and scatters across soil and understory vegetation below the canopy through gaps within the rainforest.
Based on three indices from the TRMM data, Saatchi et al. found a severe drought over southwestern Amazonia in the year 2005. Of the approximate 5.5 M/km2 of forested area in the Amazon basin, about 1.7 M/km2 experienced DWDA less than −1.0 σ in 2005. About 0.27 M/ km2 experienced severe drought with DWDA less than −2.0 σ. In 2010, as the drought persisted, the spatial extent and severity increased, resulting in 2.6 M/km2 of the area subject to DWDA less −1.0 σ and 1.1 M/km2of forest area with DWDA less than −2.0 σ. The southwestern part of the Amazon typically experiences a dry season in normal years, but with this drought, the severity of the water deficit, with MCWD less than −300 mm, became larger. The wetter forests in the central part of the Amazon that had the most negative DPA and DWDA experienced low to moderate deficit with MCWD less than −100 mm in 2005 and 2010. The time span of these decreased precipitation anomalies was fairly short over the past ten years. However, the spatial extent in the southwestern Amazon suggested a pattern consistent with the low water levels measured in the Rio Negro River and other rivers within the southwestern Amazon basin.
The impacts of the 2005 drought on the forest canopy were captured with QSCAT backscatter time series data. The authors found a strong spatial correlation with water deficit anomalies from droughts observed by TRMM data for the same period and the QSCAT backscatter data, indicating that drought caused the change in rainforest canopy properties. The dry season in 2005 demonstrated widespread decline in rainforest canopy backscatter of 2.1 M/km2with anomaly of less than −1.0 σ, in the southwest Amazon, and an area of 0.77 M/km2 experiencing a decline in backscatter with anomaly less than −2.0 σ. The areas affected by the QSCAT anomaly included old growth rainforests to terra firme(rainforests not inundated by flood waters), from south to north respectively.
The authors compared the QSCAT and TRMM anomalies, and found significant correlations, with a 1−3 month lag between the decreased precipitation anomaly and subsequent canopy characteristic changes. The patterns of these correlations depended on rainfall, and varied over the Amazon. The lag ranged from zero to three months. The longest lags occurred over the southwestern region, due to the coupled effects of the naturally-occurring dry season precipitation variation in this Amazon region and the 2005 drought. The spatial variations from QSCAT and TRMM in 2005 with very negative QSCAT values are larger than similar areas with less severe WDA. To explain this discrepancy, the authors considered areas with maximum water deficit MCWD in 2005 that exceeded 300 mm and whether there was severe water deficit during the entire dry quarter with DWDA less than −3.0 σ. QSCAT backscatter reduction in the southwest region could be associated with the WDA developing in the dry season in 2005.
In order to analyze the impacts of the drought in 2005 over time, the authors used time series data of TRMM and QSCAT anomalies for the drought-affected southwestern areas of Amazonia. They found that after the year 2005, even though the area recovered in total precipitation, WDA remained negative for the dry season in 2006 and 2007, before recovery from this water deficit started with an abnormally wet year in 2009 over the entire Amazon (except the northeastern region). The water deficit became most severe in the southwestern region in late 2009. Areas affected by water deficit in southwestern Amazonia had low values in QSCAT backscatter in 2005 through the end of November 2009. To analyze longer term data trends, the authors used autoregressive moving-average models, a statistical model used for predicting future patterns in time series, on the order of 5% of the data points. Saatchi et al. found that the anomalies in QSCAT over time imply that the 2005 drought caused a step change in the back-scatter properties of the canopy, and that there was little recovery in subsequent years following the drought. These step changes indicate abrupt changes in the mean level of a time series of data. The authors considered whether this response could be attributed to hypothetical changes in sensor performance such as backscatter signal and calibration, but found that the response was unique to certain regions in western Amazonia that had the greatest water deficit anomalies. The annual QSCAT spatial patterns for the dry season support the finding of long-term reduction in backscatter from 2005−2009, thus indicating a change in canopy structure with more gaps in the canopy.
The QSCAT anomaly was most negative at a value between −2 σ and −3.5 σ in 2005, and remained a negative anomaly compared to previous years. In the years before the 2005 drought, the most negative anomaly had been a value of 0 σ. (P −1.2 σ, and the most severe water deficit peaking in 2010 at −1.5 σ to −2.0 σ.
To quantify changes in QSCAT data in comparison to the TRMM water deficit data, the authors compared average monthly anomalies for the southwestern Amazonia, and calculated the difference between QSCAT backscatter anomaly and TRMM monthly WDA. Doing so indicated that the difference in anomalies was zero (slope of zero) before the drought in 2005, and had a negative slope after 2005. This further substantiates the finding that there is a lag in recovery of QSCAT anomaly relative to the TRMM WDA in southwestern Amazonia, meaning that the recovery of the rainforest canopy does not immediately respond to increased precipitation. The accumulation of negative anomalies during dry summer months resulted in the largest decline in QSCAT backscatter in September of 2005.
The authors were able to extrapolate their findings over the entire Amazon by mapping the spatial distribution of the pixels with negative anomalies (less than −1.0 σ) for both QSCAT and TRMM data. There were significantly (P < 0.01) negative slopes between the QSCAT anomaly and WDA after the 2005 drought. Larger negative slopes indicate areas with a longer time period between water deficit recovery and subsequent canopy recovery, which suggest that droughts have lasting effect on canopy characteristics. All of the regions studied had an abnormally high number of fires in 2005 and the years following, indicating that there was lower water content in the rainforest canopy and more flammable dead vegetation, arising from a lack of precipitation. The impacts of these fires and rainforest degradation over the years 2005−2009 did not affect the QSCAT negative anomaly. Still, over 35% of the fires between 2005 and 2009 took place in QSCAT pixels that demonstrated a strong negative anomaly of less than −1.0 σ. The relationship between fires and water deficit is further supported by the fact that more than 78% of the rainforest in 2010 occurred in pixels with large negative slopes less than −1.0 σ, indicating a very negative difference between QSCAT and monthly TRMM WDA data. The prevalence of fires and areas that recovered slowly from drought coincided with regions that had large seasonality of backscatter in QSCAT data, indicating that these events occurred in mostly transitional and seasonal rainforests in the Amazon.
From this study, Saatchi et al. concluded that the QSCAT anomalies in backscatter data for canopy structure demonstrate the effects of the 2005 drought on the Amazon rainforest, and that the QSCAT data patterns are consistent with areas that experienced the largest water deficit in the driest quarter. QSCAT backscatter variability is the result of changes in the top layer of the canopy, which are exposed to water-vapor pressure deficits, making them more sensitive to droughts. The anomalies in radar backscatter indicate a decrease in upper-canopy biomass, layering, water content, and change in canopy structure attributable to drought disturbance. The authors concluded that the severity of these declines in canopy characteristics and altered canopy structure in 2005 is significantly greater than natural cycles and normal seasonal declines in QSCAT backscatter resulting from phonological life cycles and canopy water content. The authors also concluded that because the QSCAT negative anomaly persisted over most of the western Amazon after the 2005 drought that droughts have lasting effects on the rainforest canopy. Due to the magnitude of the drought disturbance, there was a lag in recovery of the rainforest canopy in terms of biomass and roughness (canopy layering) after rainfall returned to normal levels.
Saatchi et al. discussed the need for more extensive surveys and airborne observations to better assess the decline of backscatter and its relation to rainforest disturbances evident in canopy biomass and layering. During the most severe periods of drought, tree leaves should wilt and shed, therefore causing a decline in net primary production of the rainforest canopy. Within a year of a drought event, recovery of the canopy to its original pre-drought status should occur. In central Amazonia, this recovery did occur where QSCAT backscatter anomalies recovered quickly after water deficits post-drought in 2005. However, the southwestern Amazon rainforest demonstrates a lasting decline in canopy structure, and a recovery time of three to four years, as indicated by delays in recovery of QSCAT backscatter data. Canopy structure in these instances is affected by branch dieback and tree fall creating holes in the canopy structure.
In order to verify the findings of this study, the authors attempted to find in situ observations over the region affected by the drought in 2005. From previous studies, there is consensus that large or emergent trees have higher mortality than small trees. The amount of light penetrating into the understory below the rainforest canopy is affected by the gaps in the canopy that occur from tree fall. After precipitation recovery occurs, tree fall from drought events can increase light availability, and therefore promote understory vegetation and pioneer species productivity. The authors found that in situmeasurements indicate that large trees die off more rapidly for a few years after drought. This finding suggests that canopy structure and biomass decline before recovering gradually with the emergence of more trees. However, this recovery process is slow, and may require a considerable amount of time to return to the pre-drought state. From their findings in this study, the authors suggest that the western Amazonia experienced a large-scale canopy disturbance due to the 2005 drought, resulting in the decline of and canopy structure and biomass that continued with slow recovery for the next few years. In the future, the authors suggest further research with direct examination of the effects of severe drought on forest plots in western Amazonia, to verify their conclusions.
Due to rainforest die-off from large-scale drought disturbance, there may be an impact on the rainforest’s carbon dioxide absorption and release processes. This impact is caused by the decay of wood from increased tree mortality. During the years 1995 through 2005, a decline in plant water availability created water stress before the onset of the 2005 drought, therefore playing a role in the canopy disturbance after the 2005 drought occurred. Because of the drought in 2005, and a local drought in 2007, much of the soil in the southwestern Amazonia may not have reached water content capacity that would encourage canopy recovery. Variability in the natural dry season over the southwestern Amazonia since the late 1980s and early 1990s may also have been contributing factors. Sea surface temperature is a major contributor to the negative anomalies and year-to-year variation.
Saatchi et al. concluded their study in 2009. From their findings, they suggest that the Amazon rainforest canopy response from the drought in 2005 was repeated in 2010. For the year 2010, there were no QSCAT data available. Therefore, further drought disturbances may have affected the rainforest canopy that was yet to recover from previous drought and decreased water availability. TRMM-PR backscatter anomalies indicate that surface moisture in southwestern Amazonia dropped significantly in 2010, and lasted longer than the dry season, further stressing the canopy that had not yet fully recovered. If this pattern of drought continues to occur on a 5–10 year time scale, or became even more frequent, the Amazonian rainforest canopy will be exposed to drought consistently, and the rainforest canopy will be slow to recover in structure and function. Southwestern Amazonia has been particularly subject to severe effects of rainfall variability during the last ten years, thus indicating that this region is perhaps the most vulnerable to large scale rainforest degradation based on climate change.