1
Supporting Information
2
Appendix 1. Climate data.
3 4
Table A1.1. Characteristics of the meteorological stations of the Spanish National Institute of
5
Meteorology (AEMET) used to derive the relationship between elevation and mean annual
6
air temperature (see Fig. S1), and to obtain three regional temperature datasets. Temperature
7
data were standardized, checked for homogeneity and finally combined into three regional
8
datasets corresponding to the four study treelines.
9 Treeline site
Station
Latitude (N)
Longitude (W, E)
Elevation (m a.s.l.)
Period with available temperature data
Ordesa
Canfranc-Los Arañones Candanchú Sallent de Gállego Panticosa Refugio de Góriz Torla Pic du Midi Benasque Estany Redó Vielha Caldes de Boí Artiés Tredòs-Baqueira Port de la Bonaigua Cabdella Estany Gento Alp
42º 43’ 42º 47’ 42º 46’ 42º 43’ 42º 40’ 42º 38’ 42º 50’ 42º 36’ 42º 38’ 42º 43’ 42º 33’ 42º 42’ 42º 42’ 42º 39’ 42º 28’ 42º 31’ 42º 22’
0º 52’ W 0º 31’ W 0º 20’ W 0º 17’ W 0º 01’ E 0º 07’ W 0º 08’E 0º 38’ E 0º 46’ E 0º 48’ E 0º 50’ E 0º 53’ E 0º 57’ E 0º 59’ E 1º 00’ E 1º 00’ E 1º 53’ E
1160 1600 1305 1184 2215 1053 2862 1138 2240 940 1280 1185 1880 2263 1422 2174
1910-2012 1951-1969 1953-1969 1940-1969 1982-2012 1964-2012 1910-1984 1935-1969 1950-1997 1964-2012 1962-1995 1964-1991 1968-1987 1935-1968 1934-1969 1935-1969
1158
1942-1951
La Molina
42º 20’
1º 56’ E
Puigcerdà
42º 26’
1º 56’ E
1704 1145
1955-1969 1932-1963
Tessó
E. Pera / Meranges
10 1
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
Ordesa
Tessó
Meranges
14
Mean annual temperature (ºC)
Treeline
12 10 8 6 4 2 0 -2 -4 1000
1500
2000
2500
3000
Elevation (m) 10
2
r=0.71 1
9 0
-1
8
-2
-3
6
-4
1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 9
1
r=0.53
Pic du Midi
8
0
7
-1
6 -2 5 -3 4 1900
27 28 29 30 31 32 33 34 35 36 37 38
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
Pic du Midi mean annual temperature (ºC)
CRU mean annual temperature (ºC)
7
E. Redó mean annual temperature (ºC)
CRU E. Redó
2010
Years
Figure A1.1. Spatial and temporal variability of temperature data. (a) Linear regression showing the changes in mean annual temperature as a function of elevation across the Spanish Pyrenees (excepting the “Pic du Midi” station located at ca. 2800 m in the French Pyrenees). The elevation range (2100-2400 m) where most Pyrenean mountain pine (Pinus uncinata) treelines are located is indicated as a grey box. The regression line (continuous black line (equation y = 16.10-0.0059x, r² = 0.92) is shown with 95% confidence intervals (dashed lines) and prediction intervals (continuous grey lines). The uppermost images show three Pyrenean treelines. (b) Correlations obtained by relating the mean annual air temperatures reconstructed in a Pyrenean alpine lake (Estany Redó) or recorded at the highelevation Pic du Midi observatory (Bücher and Dessens, 1991) as compared with gridded interpolated data from the CRU dataset (Harris et al. 2014). 2
39
Appendix 2. Reproductive size threshold at treeline.
40 41
Table A2.1. Parameters and statistics (R2, explained variance; n, number of trees) of the
42
binary logistic equations fitted to cone presence or absence as a function of diameter at breast
43
height (dbh) to estimate the reproductive size threshold in four Pyrenean treelines. Equations
44
estimate the probability of cone production (Pc) as: ln(Pc / 1 − Pc) = a + b dbh, where a and
45
b are the fitted parameters.
46 Treeline site
Parameters
R2 (%)
n
a
b
Ordesa
-2.16
0.44
46.26
636
Tessó del Son
-1.30
0.15
53.04
196
Estanys de la Pera
-0.99
0.18
25.60
170
Meranges
-2.06
0.63
73.13
225
47 48 49
3
50
Appendix 3. Reconstructions of tree establishment.
400 300
O
200 100 0
Density (individuals ha-1)
400
T
300 200 100 0
400
E
300 200 100 0
400
M
300 200 100 0 1700
1750
1800
1850
1900
1950
2000
Year of establishment
51 52
Figure A3.1. Reconstructions of tree establishment for the four studied Pyrenean treelines
53
(O, Ordesa; T, Tessó; E, Estanys de la Pera; M. Meranges). The tick of each bar indicates the
54
last year of the decadal age class considered. For instance, the bar located over the 2000 tick
55
shows the number of individuals recruited between the years 1991 and 2000. 4
56
Appendix 4. Mortality model. 1000
M = 2.02 * (BA) - 7.06 R2 = 0.21; p<0.001
(a) -1 Mortality (trees ha )
800
600
400
200
0 0
20
60
Stand Basal Area (m2ha-1)
(b)
57
40
300
Young stand‐living trees (disturbed treeline recolonizing grassland) Young stand ‐dead trees Middle‐aged stand ‐living trees (densifying undisturbed treeline)
250
Density (stems / ha)
Middle‐aged stand ‐dead trees Old stand ‐living trees (static undisturbed treeline)
200
Old stand ‐dead trees 150
100
50
0 10
15
20
58
25
30
35
40
45
50
55
60
65
70
Dbh (cm)
59
Figure A4.1. Relationships observed between Pinus uncinata total stand basal area and the
60
number of dead trees (A). Frequency distribution of P. uncinata living and dead trees in
61
different size class (B). Data were obtained from 810 Pinus uncinata field plots. We used
62
data of mountain pine forests located in Catalonia (NE Spain), where 65% of the Iberian
63
distribution area of this species is located (Villaescusa and Diaz 1998, Villanueva 2004;
64
Martín-Alcón et al. 2012). 5
65
Appendix 5. Growth models and tree populations forecast.
25 BAI raw Age effect model (variance explained 28.35%)
Ordesa
-1
BAI (cm year )
20
2
15
10
5
0 0
20
40
60
80
100
120
Age at coring height (years)
66 67
Figure A5.1. Changes of basal area increment (BAI) as a function of tree age at coring height
68
(1.3 m) in Ordesa site.
69 70
Table A5.1. Linear mixed-effects models of the annual basal area increment (BAI) residuals
71
from the age model as a function of different monthly temperatures in Ordesa. Abbreviations:
72
K, number of parameters included in the model, including the fixed effect variables, i.e.
73
number of monthly temperature with significant support, plus the constant term, plus the
74
error (see also equation 1 in the main text); Δi, difference of Akaike information criterion
75
with respect to the best model; Wi, relative probability that the model i is the best model
76
given the observed data. Values in bold correspond to models with substantial support.
77
Months abbreviated with a “p” subscript correspond to the previous year.
Ordesa 52 trees n= 2949 Random: tree
Fixed effects sepp+novp+may sepp+novp+may+jul sepp+octp+novp+may+jul sepp+ octp+novp+may+jul+sep sepp+ octp+novp+may+jun+jul+sep sepp+ octp+novp+may+jun+jul+aug+sep sepp+octp+novp+may+jun+jul+aug+sep+oct null model
K 5 6 7 8 9 10 11 2
Δi 0.0 3.0 5.1 9.0 14.8 21.6 30.0 37.8
Wi 75.9 17.2 6.0 0.9 0.0 0.0 0.0 0.0
6
78 79
Table A5.2. Statistical parameters obtained by linear mixed-effects models of basal area
80
increment (BAI) residuals from the age model and monthly temperature for Ordesa. Linear
81
regression coefficients of the fixed factors, standard error of the regression and partial
82
explained variance for each selected variable are noted. Months abbreviated with a “p”
83
subscript correspond to the previous year. Variable
Value
Std.Error
Explained variance (%)
sepp
-0.089
0.0173
13.68
novp
0.074
0.0168
11.38
may
0.094
0.0167
14.42
84 85 86
Table A5.3. Explained variance by the full growth model used for Ordesa. Model includes
87
the effect of three components of tree growth: (i) cambial age at coring height, (ii) monthly
88
temperature time series and (iii) autocorrelation structure of tree growth. Explained variance by age model (%) Explained variance by climate model (%) Explained variance by auto-correlation model* (%) Total explained variance (%)
89
28.35 39.47 11.20 79.02
*The autocorrelation model value for BAI of the previous year was 0.45.
90
7
30
Tessó
BAI raw Age effect model (variance explained 36.61%)
-1
20
2
BAI (cm year )
25
15 10 5 0 0
50
100
150
200
Age at coring height (years)
91 92
Figure A5.2. Changes of basal area increment (BAI) as a function of tree age at coring height
93
(1.3 m) in Tessó site.
94 95
Table A5.4. Linear mixed-effects models of the annual basal area increment (BAI) residuals
96
from the age model as a function of different monthly temperature variables in Tessó del Son
97
Abbreviations: K, number of parameters included in the model, including the fixed effect
98
variables, i.e. number of monthly temperature with significant support, plus the constant
99
term, plus the error (see also equation 1 in the main text); Δi, difference of Akaike
100
information criterion with respect to the best model; Wi, relative probability that the model i
101
is the best model given the observed data. Values in bold correspond to models with
102
substantial support. Months abbreviated with a “p” subscript correspond to the previous year.
Tesso 107 trees n=5587 Random effect: tree
Fixed effects sepp+novp+jun+aug sepp+novp+may+jun+aug sepp+jun+aug sepp+octp+novp+may+jun+aug sepp+octp+novp+may+jun+jul+aug sepp+octp+novp+may+jun+jul+aug+oct sepp+octp+novp+may+jun+jul+aug+oct+nov sepp+octp+novp+may+jun+jul+aug+sep+oct+nov
K 6 7 5 8 9 10 11 12
Δi 0.0 3.6 8.4 10.4 15.9 23.2 30.3 38.1
Wi 84.2 14.1 1.2 0.5 0.0 0.0 0.0 0.0
103
8
104 105
Table A5.5. Statistical parameters obtained by linear mixed effect model of basal area
106
increment (BAI) residuals from the age model and monthly temperature for Tessó del Son.
107
Linear regression coefficients of the fixed factors, standard error of the regression and partial
108
explained variance for each selected variable are noted. Months abbreviated with a “p”
109
subscript correspond to the previous year. Variable
Value
Std.Error
Explained variance (%)
sepp
-0.087
0.020
7.17
novp
0.079
0.019
6.49
jun
0.117
0.025
9.66
aug
0.165
0.024
13.59
110 111 112
Table A5.6. Explained variance by the full growth model used for Tessó del Son. Model
113
includes the effect of three components of tree growth: (i) cambial age at coring height, (ii)
114
monthly temperature time series and (iii) autocorrelation structure of tree growth. Explained variance by age model (%) Explained variance by climate model (%) Explained variance by auto-correlation model* (%) Total explained variance (%)
115
36.61 36.90 9.49 83.00
*The autocorrelation model value for BAI of the previous year was 0.47.
116
9
20
E. Pera
15
2
-1
BAI (cm year )
BAI raw Age effect model (variance explained 28.55%)
10
5
0 0
50
100
150
200
250
Age at coring height (years)
117 118
Figure A5.3 Changes of basal area increment (BAI) as a function of tree age at coring height
119
(1.3 m) in Estanys de la Pera site.
120 121
Table A5.7. Linear mixed-effects models of the annual basal area increment (BAI) residuals
122
from the age model as a function of different monthly temperature variables in Estanys de la
123
Pera. Abbreviations: K, number of parameters included in the model, including the fixed
124
effect variables, i.e. number of monthly temperature with significant support, plus the
125
constant term, plus the error (see also equation 1 in the main text); Δi, difference of Akaike
126
information criterion with respect to the best model; Wi, relative probability that the model i
127
is the best model given the observed data. Values in bold correspond to models with
128
substantial support. Months abbreviated with a “p” subscript correspond to the previous year.
E Pera 80 trees n=2847 Random effect: tree
Fixed effects sepp+novp+may+aug+sep+oct sepp+octp+novp+may+jun+jul+aug+sep+oct sepp+novp+may+aug+oct sepp+octp+novp+may+aug+sep+oct novp+may+aug+oct sepp+octp+novp+may+jun+aug+sep+oct novp+may+aug null model
K 8 11 7 9 6 10 5 2
Δi 0.0 0.0 1.0 2.8 5.0 9.7 24.5 169.4
Wi 34.1 33.9 20.6 8.4 2.7 0.3 0.0 0.0
129
10
130 131
Table A5.8. Statistical parameters obtained by linear mixed effect model of basal area
132
increment (BAI) residuals from the age model and monthly temperature for Estanys de la
133
Pera. Linear regression coefficients of the fixed factors, standard error of the regression and
134
partial explained variance for each selected variable are noted. Months abbreviated with a “p”
135
subscript correspond to the previous year. Variable
Value
Std.Error
Explained variance (%)
sepp
-0.078
0.022
5.16
novp
0.158
0.021
9.67
may
0.091
0.023
5.91
aug
0.215
0.027
12.86
oct
0.100
0.018
6.39
136 137 138
Table A5.9. Explained variance by the full growth model used for Estanys de la Pera. Model
139
includes the effect of three components of tree growth: (i) cambial age at coring height, (ii)
140
monthly temperature time series and (iii) autocorrelation structure of tree growth. Explained variance by age model (%) Explained variance by climate model (%) Explained variance by auto-correlation model* (%) Total explained variance (%)
141
28.54 39.98 17.11 85.63
*The autocorrelation model value for BAI of the previous year was 0.55.
142
11
20
Meranges
15
2
1
BAI (cm year- )
BAI raw Age effect model (variance explained 13.97%)
10
5
0 0
20
40
60
80
100
120
140
Age at coring height (years)
143 144
Figure A5.4. Changes of basal area increment (BAI) as a function of tree age at coring height
145
(1.3 m) in Meranges site.
146 147
Table A5.10. Linear mixed-effects models of the annual basal area increment (BAI) residuals
148
from the age model as a function of different monthly temperature variables in Meranges
149
Abbreviations: K, number of parameters included in the model, including the fixed effect
150
variables, i.e. number of monthly temperature with significant support, plus the constant
151
term, plus the error (see also equation 1 in the main text); Δi, difference of Akaike
152
information criterion with respect to the best model; Wi, relative probability that the model i
153
is the best model given the observed data. Values in bold correspond to models with
154
substantial support. Months abbreviated with a “p” subscript correspond to the previous year.
Meranges 62 trees n=2865 Random effect: tree
Fixed effects octp+novp+may+jun+jul+oct sepp+octp+novp+may+jun+jul+oct octp+novp+may+jun+jul sepp+octp+novp+may+jun+jul octp+may+jun+jul sepp+octp+novp+may+jun+jul+aug+ oct sepp+octp+novp+may+jun+jul+aug+sep+oct null model
K 7 9 6 8 5 10 11 2
Δi 0.0 0.4 1.6 1.6 4.1 5.8 13.3 102.8
Wi 34.6 27.7 15.7 15.5 4.4 1.9 0.0 0.0
155 12
156
Table A5.11. Statistical parameters obtained by linear mixed effect model of basal area
157
increment (BAI) residuals from the age model and monthly temperature for Meranges. Linear
158
regression coefficients of the fixed factors, standard error of the regression and partial
159
explained variance for each selected variable are noted. Months abbreviated with a “p”
160
subscript correspond to the previous year. Variable
Value
Std.Error
Explained variance (%)
octp
0.087
0.018
6.53
novp
0.066
0.020
4.97
may
0.097
0.021
7.28
jun
-0.155
0.028
11.67
jul
0.211
0.030
15.90
161 162 163
Table A5.12. Explained variance by the full growth model used for Meranges. Model
164
includes the effect of three components of tree growth: (i) cambial age at coring height, (ii)
165
monthly temperature time series and (iii) autocorrelation structure of tree growth. Explained variance by age model (%) Explained variance by climate model (%) Explained variance by auto-correlation model* (%) Total explained variance (%)
166
13.97 46.35 23.91 84.23
*The autocorrelation model value for BAI of the previous year was 0.60.
167
13
20
O
15
15
10
10
5
5
T
2
BAI (cm )
20
0
0
168
1900
1920
1940
1960
1980
2000
2020
2040
2060
2080
2100
1900
1920
1940
1960
1980
2000
2020
2040
2060
2080
2100
1920
1940
1960
1980
2000
2020
2040
2060
2080
2100
20
E
15
15
10
10
5
5
M
2
BAI (cm )
20
0
169
1900
0 1920
1940
1960
1980
2000
Year
2020
2040
2060
2080
2100
1900
Year
170 171
Figure A5.5. Predicted basal area increment (BAI) of 1000 Mountain pine individuals per in
172
each treeline site based on climate-growth linear mixed effects models. For each study site,
173
we simulated the fate of a population of 1000 individuals, with an age structure of the
174
respective population, as sampled in 2010 (see Fig. A3.1). Note that changes in stand density
175
are only related to density-dependent mortality beyond the year 2040.
176 177 178 179 180 181 182 183 184 185 186 187 188 189
References Bücher, A. and Dessens, J. 1991. Secular trend of surface temperature at an elevated observatory in the Pyrenees. Journal of Climate 4: 859–868. Harris, I., Jones, P.D., Osborn, T.J. & Lister, D.H. 2014. Updated high-resolution grids of monthly climatic observations. International Journal of Climatology 34: 623–642. Martín-Alcón S., Coll L., Aunós A. 2012. A broad-scale analysis of the main factors determining the current structure and understory composition of Catalonian sub-alpine (Pinus uncinata Ram.) forest. Forestry 85: 225–236. Villaescusa, R. & Diaz, R. 1998. Segundo Inventario Forestal Nacional (1986–1996). Ministerio de Medio Ambiente, ICONA, Madrid, Spain. Villanueva, J.A. 2004. Tercer Inventario Forestal Nacional (1997-2007). Comunidad de Madrid, Madrid, Ministerio de Medio Ambiente, Madrid, Spain. 14
190 191
Appendix 6. Seed viability as a function of growth form.
192 193 194 195 196 197 198 199 200 201 202
Figure A6.1. Seed viability (means ± SD) measured as the percentage of germinated seeds
203
from cones taken in Mountain pine (Pinus uncinata) individuals sampled at the Ordesa
204
treeline site. Three growth forms were considered: shrubby individuals (krummholz), adults
205
(arborescent individuals at least 2-m tall presenting vertical or erect stems) and intermediate
206
individuals forming both shrubby and erect stems.
207
15
