Monin-Obukhov length. d was set to be 0 m for water surface, and z0 was set to be
0.0003 m based on the modeled z0 value for the East China Sea (Zhang et al. 2010).
R b is also parameterized after Erisman and Draaijers (1995) as
R b ¼ 2=ku Ã
ð
Þ S c =P r
ð
Þ
2=3
ð12:3Þ
where k is the von Karman constant, u à is the friction velocity, P r is the Prandtl
number, and S c is the Schmidt number.
R c for water surface is calculated according to Wesely (1989) as
R c ¼ R gs
ð12:4Þ
where R gs is the ground or water surface resistance. For NH 3 and HNO 3 , because
these two gases are easily dissolved in water, R gs is set to be 0 s m
À1 ; for NO 2, due to
its poor solubility in water, R gs is set to be 20,000 s m
À1 (Wesely 1989).
For particulate nitrogen species, the dry deposition velocity is parameterized
according to Slinn (1982) using the following equation:
V d ¼
1
R a þ R surf
þ V g
ð12:5Þ
where R a is the aerodynamic resistance, R surf is the surface resistance, and V g is the
gravitational settling (or sedimentation) velocity. Details for parameterization of
R surf and V g can be found in Zhang et al. (2010). Based on the results for the East
China Sea (Zhang et al. 2010), roughness length was set as 0.0003 m, and surface
resistance was set as 0 s m
À1 for NH 3 and HNO 3 but 20,000 s m
À1 for NO 2 (Wesely
1989).
12.2.3 Estimation of Riverine Inputs
Daily N inputs from river were estimated as a function of river discharge, by means
of load estimator (LOADEST) (Runkel et al. 2004):
ln L i
ð Þ ¼ a 0 þ a 1 ln Q þ a 2 ln Q
2
þ a 3 sin 2πdtime
ð
Þ þ
a 4 cos 2πdtime
ð
Þþa 5 dtime þ a 6 dtime
2
þ ε
ð12:6Þ
where Q is the daily river discharge, dtime is the decimal time, a 0 ~a 6 are the fitted
coefficients in the multiple regression model, and ε is the estimate error. Note that the
biweekly TN concentration was multiplied by the corresponding daily discharge so
that Eq. 12.6 is a function of load (L i ) instead of concentration. Within LOADEST,
the model to estimate N loads was set to be automatically selected from predefined
regression models (Runkel et al. 2004). To select the best one, LOADEST calculated
270
F. Zhou et al.
0.0003 m based on the modeled z0 value for the East China Sea (Zhang et al. 2010).
R b is also parameterized after Erisman and Draaijers (1995) as
R b ¼ 2=ku Ã
ð
Þ S c =P r
ð
Þ
2=3
ð12:3Þ
where k is the von Karman constant, u à is the friction velocity, P r is the Prandtl
number, and S c is the Schmidt number.
R c for water surface is calculated according to Wesely (1989) as
R c ¼ R gs
ð12:4Þ
where R gs is the ground or water surface resistance. For NH 3 and HNO 3 , because
these two gases are easily dissolved in water, R gs is set to be 0 s m
À1 ; for NO 2, due to
its poor solubility in water, R gs is set to be 20,000 s m
À1 (Wesely 1989).
For particulate nitrogen species, the dry deposition velocity is parameterized
according to Slinn (1982) using the following equation:
V d ¼
1
R a þ R surf
þ V g
ð12:5Þ
where R a is the aerodynamic resistance, R surf is the surface resistance, and V g is the
gravitational settling (or sedimentation) velocity. Details for parameterization of
R surf and V g can be found in Zhang et al. (2010). Based on the results for the East
China Sea (Zhang et al. 2010), roughness length was set as 0.0003 m, and surface
resistance was set as 0 s m
À1 for NH 3 and HNO 3 but 20,000 s m
À1 for NO 2 (Wesely
1989).
12.2.3 Estimation of Riverine Inputs
Daily N inputs from river were estimated as a function of river discharge, by means
of load estimator (LOADEST) (Runkel et al. 2004):
ln L i
ð Þ ¼ a 0 þ a 1 ln Q þ a 2 ln Q
2
þ a 3 sin 2πdtime
ð
Þ þ
a 4 cos 2πdtime
ð
Þþa 5 dtime þ a 6 dtime
2
þ ε
ð12:6Þ
where Q is the daily river discharge, dtime is the decimal time, a 0 ~a 6 are the fitted
coefficients in the multiple regression model, and ε is the estimate error. Note that the
biweekly TN concentration was multiplied by the corresponding daily discharge so
that Eq. 12.6 is a function of load (L i ) instead of concentration. Within LOADEST,
the model to estimate N loads was set to be automatically selected from predefined
regression models (Runkel et al. 2004). To select the best one, LOADEST calculated
270
F. Zhou et al.