Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Supporting Information
Appendix S1: projected climate change by the three GCMs
We conducted cell-by-cell temporal trend analysis on the major climate variables
from the bias-corrected results of the three GCMs using the non-parametric Mann-
Kendall test. The results show considerable differences between the projected
climatologies by the three GCMs (Fig. S1). All three predict significant (P < 0.01),
region-wide warming trends over the 21st century (Fig. S1a); however, the magnitudes of
warming trends differ: HadCM3 has the strongest warming trend followed by CCSM3
and PCM. Although the magnitudes of warming trends in CCSM3 and HadCM3 differ,
both models predict that the eastern and southern Amazon will experience the highest
rates of warming during the 21st century (Fig. S1a), and that these regions will also
experience significant upward trends in the vapor pressure deficit (VPD) (Fig. S1b). In
contrast, PCM predicts that most of the basin will experience relatively mild changes in
temperature, and little or no changes in VPD.
With regard to rainfall, the HadCM3 projection indicates that approximately half
(53%) of the region, mainly the eastern and southeastern Amazon, will suffer significant
reductions in precipitation during the 21st century (Fig. S1c). In contrast, PCM and
CCSM3 predict that significant portions of the basin (47% and 62% of the region,
respectively), located mainly in southern and western portion of the basin, will
experience significant increases in precipitation. Comparison of these predictions against
the nineteen GCM projections examined by Malhi et al. (2009) indicate that they span the
range of precipitation predictions for the Amazon region: the PCM projection presents a
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
slightly warmer but wetter future climate, while the HadCM3 projection represents an
extremely hot and dry scenario and the CCSM3 projection falls in between.
Fig. S1. Maps of temporal trends from 2009 to 2100 in (a) annual air temperature, (b)
vapor pressure deficit and (c) precipitation from the bias-corrected projections of three
GCMs (i.e. PCM, CCSM3, and HadCM3); grey areas denote non-significant trends with
90% confidence.
23
24
25
26
27
28
29
Appendix S2: spatial patterns of the Business-As-Usual and Governance land-use
scenarios in Amazonia
The Global Land-Use dataset (GLU) incorporates the SAGE-HYDE 3.3.1 dataset
and provides global land-use transitions on a 1° grid from 1700 to 1999 (Hurtt et al.,
2006). The Business-As-Usual (BAU) and Governance (GOV) datasets provide yearly 1-
km horizontal resolution land-use maps from 2002 to 2050. Consistent land-use transition
data for the entire period (1700–2100) were produced by following the procedure: (1) the
1-km BAU and GOV land-use maps were used to calculate the fractions of the three
land-use types for each 1° grid cell; (2) linear interpolation was used to connect the 1999
GLU land-use map to the 2008 GOV land-use map to produce continuous land-use maps
for the 2000–2008 period; (3) the land-use transition rates between 2000 and 2008 were
computed from the interpolated land-use maps; (4) the BAU and GOV land-use maps for
the 2009–2050 period were converted to their respective land-use transition rates for the
same period; and (5) BAU and GOV land-use transition rates were extrapolated from
2050 to 2100 by assuming the continuation of the same transition rates in 2050 until no
forest left in the grid cells.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
Fig. S2. Spatial patterns of land-use composition in 2100 under the (a) GOV and (b)
BAU scenarios; the projected rates of land-use transformation were derived and extended
from Soares-Filho et al. (2006).
46
47
48
49
Appendix S3: evaluation of the biosphere models' performance
In this study, we assessed the three biosphere models' predictive capabilities by
comparing modeled above-ground live biomass (AGB) and percent tree cover against
two sets of satellite remote sensing (RS) derived AGB estimates (Baccini et al., 2012,
Saatchi et al., 2011) and the MODIS Collection 5 MOD44B percent tree cover product
(DiMiceli et al., 2011), respectively. The Baccini et al. (2012), Saatchi et al. (2011), and
MOD44B data have nominal spatial resolutions of 500m, 1km, 500m, respectively. We
first aggregated these data to 1-degree and then compared them with our model results.
Both of the two RS biomass products were produced using a combination of data from
spaceborne LiDAR, optical and microwave imagery, and in situ inventory plots. An
evaluation of models' performance against site-level measurements of carbon fluxes and
aboveground biomass dynamics can be found in an ongoing study (Levine et al., The role
of short-term climate variability in governing Amazonian biomass dynamics, in
preparation) and Powell et al. (2013).
Although there are non-negligible discrepancies between the two RS products,
across the models, and between the models and RS products, the model predictions of
AGB show a similar spatial gradient to the satellite-derived estimates of regional AGB,
with AGB increasing from the southern and southeastern dry savanna zones to the
western and northeastern dense, moist forest regions (Fig. S3). ED2 AGB agrees well
with the RS AGB in these high biomass regions, but tends to underestimate AGB in low
biomass regions (Fig. S3a,d-f). IBIS AGB estimates are systematically lower than the RS
estimates in areas of high biomass but higher in areas of low biomass (Fig. S3b,d-f),
while JULES AGB is systematically higher than the RS AGB (Fig. S3c-f). A recent study
5
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
by Levine et al. (2014) compared site-level AGB estimates from the ED2 model and the
above two RS biomass products with the measurements of the RAINFOR network(Malhi
et al., 2002), and found that these model and RS estimates are qualitatively consistent
with that observed in the RAINFOR plots.
To more quantitatively describe the vegetation structure, we calculate the percent
tree cover (f treecover ¿ using the fully projected tree foliage cover by following Kucharik et
al. (2000): f treecover=1−exp (−0.5 ∙ LAI tree), where 0.5 is the canopy extinction coefficient,
and LAI tree is the total leaf area index of all tree plant function types. Percent tree cover
from all models generally compares well with the RS based estimate (Fig. S4). f treecover
shows the similar spatial gradient as AGB in models and RS estimates (Fig. S3,S4); the
higher biomass regions have higher f treecover, while the lower biomass regions have lower
f treecover. The inset quantile-quantile (Q-Q) plot (Fig. S4) shows that IBIS and JULES have
generally higher tree cover fractions than ED2 and the MODIS product. Although ED2
predicts lower tree cover in the lower tree cover region relative to the MODIS product,
ED2's predictions of tree cover fraction are close to the MODIS values in the areas with
middle to high tree cover fractions (Fig. S4).
Despite the uncertainty in the AGB and f treecover estimates of models and remote
sensing products and some discrepancies between the model and remote sensing results,
the overall similar spatial patterns and gradients from the model and remote sensing
results suggest that all three biosphere models are able to reasonably capture the present-
day composition and spatial variability of Amazonian ecosystems.
6
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
Fig. S3. Spatial patterns of present-day (2000~2008) above-ground biomass across
Amazonia from model estimates of (a) ED2, (b) IBIS, and (c) JULES, and remote
sensing based estimates of (d) Baccini et al. (2012) and (e) Saatchi et al. (2011), and (f)
the quantile-quantile plots of model estimates against remote sensing (RS) based
estimates.
7
94
95
96
97
98
99
Fig. S4. Spatial patterns of present-day (2000~2008) percent tree cover across Amazonia
from (a) ED2, (b) IBIS, (c) JULES, and (d) MODIS collection 5 MOD44B product. The
inset graph shows the quantile-quantile plot of model estimates against remote sensing
based estimates.
8
100
101
102
103
104
Appendix S4: evaluation of association between water stress regime and AGB from
model simulations and remote sensing estimates across the Amazon.
Table S1. Summary of the strength of association between water stress (MCWD) and
AGB from model simulations and remote sensing estimates across the Amazon; the
strength of association is quantified by Pearson's simply linear correlations and Kendall's
Tau (i.e. the rank correlation).
Water Stress Regime
Statistical Test AGBED2 AGBIBIS AGBJULES AGBBaccini AGBSaatchi
MCWDPearson's r 0.75*** 0.67*** 0.61*** 0.65*** 0.63***
Kendall's τ 0.61*** 0.45*** 0.45*** 0.48*** 0.42***
*** P<0.01
References
Baccini A, Goetz SJ, Walker WS et al. (2012) Estimated carbon dioxide emissions from
tropical deforestation improved by carbon-density maps. Nature Climate Change, 2, 182-
185.
Dimiceli CM, Carroll ML, Sohlberg RA, Huang C, Hansen MC, Townshend JRG (2011)
Vegetation Continuous Fields MOD44b. College Park, Maryland, University of
Maryland.
Kucharik CJ, Foley JA, Delire C et al. (2000) Testing the performance of a Dynamic
Global Ecosystem Model: Water balance, carbon balance, and vegetation structure.
Global Biogeochemical Cycles, 14, 795-825.
Levine NM, Zhang K, Longo M et al. (2014) Ecosystem heterogeneity determines the
resilience of the Amazon to climate change. Nature Climate Change, In revision.
9
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
Malhi Y, Aragao LEOC, Galbraith D et al. (2009) Exploring the likelihood and
mechanism of a climate-change-induced dieback of the Amazon rainforest. Proceedings
of the National Academy of Sciences of the United States of America, 106, 20610-20615.
Malhi Y, Phillips OL, Lloyd J et al. (2002) An international network to monitor the
structure, composition and dynamics of Amazonian forests (RAINFOR). Journal of
Vegetation Science, 13, 439-450.
Powell TL, Galbraith DR, Christoffersen BO et al. (2013) Confronting model predictions
of carbon fluxes with measurements of Amazon forests subjected to experimental
drought. New Phytologist, 200, 350-364.
Saatchi SS, Harris NL, Brown S et al. (2011) Benchmark map of forest carbon stocks in
tropical regions across three continents. Proceedings of the National Academy of Sciences
of the United States of America, 108, 9899-9904.
Soares-Filho BS, Nepstad DC, Curran LM et al. (2006) Modelling conservation in the
Amazon basin. Nature, 440, 520-523.
10
125
126
127
128
129
130
131
132
133
134
135
136
137
138