Carbon, Water and Energy Fluxes of Terrestrial Ecosystems in Italy - The Greenhouse Gas Balance of Italy - page 38

Table 2.2 Results of model selection conducted with the stepwise AIC method using as predic-
tor annual averages of meteorological variables

R 2 adj PFT Ta Precip Intercept P AIC N

13.08 (8.52) − 0.33 (0.15) − 456.3 0.36 <0.001 108.6 12

ENF

DBF NS NS 14

95.04 (25.02) − 0.16 (0.10) − 1,783.59 0.66 <0.01 88.99 12

EBF

CRO NS NS 15

GRA NS NS 14

Coefficients, their standard errors and the statistics of the best model selected are reported. Ta
Air Temperature, Precip Precipitation, AIK Akaike's Information Criterion, NS not significant

Table 2.3 Results of model selection conducted with the stepwise AIC method using as predic-
tor annual averages of meteorological variables, LAI and CUP

R 2 adj

PFT Rg Ta Precip LAI CUP Intercept P AIC N

− 17.32
(10.77)

− 116.23
(31.06)

276.21 0.64 <0.01 101.8 12

ENF

− 79.48
(40.64)

− 52.76
(41.37)

− 5.39
(1.98)

1,732.9 0.32 <0.05 203.4 14

DBF

EBF 95.04

(25.02)

− 0.16
(0.10)

− 1,783.6 0.66 <0.01 88.99 12

CRO NS NS NS 15

GRA − 28.53

(16.67)

20.19
(8.85)

− 64.51
(30.51)

− 4.25
(0.37)

824.96 0.95 <0.001 103.12 14

Coefficients, their standard errors and the statistics of the best model selected are reported.
Rg Shortwave Incoming Radiation, Ta Air Temperature, Precip Precipitation, AIK Akaike's
Information Criterion, NS not significant

characteristics and phenology in determining the spatial and temporal variability
of annual NEE (Table 2.3 ).
The results of the DStdev computed for each site and grouped for PFT are
reported in Fig. 2.6 . The time series of DStDev highlights the period that mostly
contribute to the inter annual variability of NEE at each site (hereafter referred as
the critical period).
As an example, the DStDev computed for IT-MBo site (Fig. 2.5 e) shows an
abrupt increase in DStDev at the onset of the growing season indicating that this
period has an important role in the IAV definition. Marcolla et al. 2011 concluded
that climatic conditions and snow cover in late spring largely influence the inter-
annual variability of NEE of this alpine grassland, and that the first half of the
growing season at IT-MBo mitigates the IAV thanks to the negative correlation
(R = − 0.77) between the timing of snowmelt and the rate of carbon uptake at the
beginning of the growing season. A second peak of DStDev due to the variable
timing of the meadow cut has been observed, emphasizing the important role of
management in controlling the inter annual variability.
The results show that for almost all the PFTs the critical period for the IAV
occurs during the spring transient of NEE, when carbon uptake begins after winter,
and in some water limited sites at the end of the growing season in late summer.