Environmental Engineering Reference
In-Depth Information
Table 2.4
Results of model selection conducted with the stepwise AIC method using as
predictor annual and sub annual averages of meteorological variables
R
2
adj
PFT
Predictors selected
P
AIC
AIC (ann.
climate)
AIC (ann.
climate
+
LAI
and CUP)
N
ENF
CUP
,
LAI
,
Rg
,
PREC_CUP
,
TA_JJA
,
TA_MAM
0.98
<0.001
59.73
108.6
101.8
12
CUP
,
Rg
,
LAI
,
TA_JJA
,
TA_MAM
,
PREC_JJA
0.47
<0.05
200.23
NS
203.4
14
DBF
TA_MAM
,
Rg
,
PREC_CUP
0.91
<0.001
88.65
88.99
88.99
12
EBF
CRO
VPD_MAM
,
PREC_CUP
0.44
<0.05
147.85
NS
NS
15
GRA
CUP
,
TA_MAM
,
Rg
,
LAI
0.95
<0.001
101.71
NS
103.12
14
PREC
Precipitation,
TA
Air Temperature,
Rg
Incoming Shortwave Radiation,
VPD
Vapor
Pressure Deficit, LAI and CUP. The averages have been calculated for groups of months
(
MAM
March-April-May,
JJA
June-July-August) of for the whole Carbon Uptake Period (e.g.
for PREC_CUP). For each PTF the list of selected predictors are reported. In the column 'predic-
tors selected'
bold
variables indicate a positive correlation with NEE positive coefficient) while
variables in
italics
indicate negative correlation (negative coefficient). The AIC for the stepwise
analysis with annual meteorological variable (Table
2.1
) and meteorological variable, LAI and
CUP (Table
2.2
) are also reported for comparison.
AIK
Akaike's Information Criterion
This has been observed for all the PFTs even if for EBF the signal is weaker and
more variable. For CRO the management might be an important confounding fac-
tor hampering the identification of the critical period.
Finally, the results of the stepwise AIC analysis is reported in Table
2.4
includ-
ing as predictor also sub-annual aggregation of climate drivers. These results
emphasize that studies focused on the relationship between climate and NEE
should investigate the climate variability in particular periods rather than the aver-
age climate (Le Maire et al.
2010
).
The results showed that for almost all the PFT, the average spring tempera-
ture (March, April and May, TA_MAM) is selected as predictor of year-to-year
and spatial variations of annual NEE. Cumulative precipitation in summer (June,
July and August) or during the entire CUP (PREC_JJA or PREC_CUP) are also
selected as predictor of the variability in many PFTs, thus, indicating the water
control of NEE in many ecosystems and sites selected.
Moreover, LAI and CUP are also often selected as additional predictors except
for EBF and CRO indicating a strong control of LAI, explaining the spatial variabil-
ity (i.e. across sites) of annual NEE, and phenology, which is one of the main factors
controlling the temporal variability of NEE at site level (e.g. Marcolla et al.
2011
).
Among the different PFTs, EBF shows the strongest climatic control of the
inter annual and spatial variability of NEE. In particular, the variables selected by
