Info

100 150 200 250 300 Julian day

_ simulated observed soil temperature (°C)

Figure. 8. Comparison of soil temperature as simulated with HTSVS and observed at Yakutsk, Siberia. The time step used in (a) is three times smaller than that used in (b). Observation data from Levine (2007; pers. communication).

Table 2. Typical mean values and standard deviations (in brackets) of soil characteristics. The symbols ks, ns, b, ys, ps stand for the hydraulic conductivity at saturation, porosity, soil-pore distribution index, and density of the dry soil material. References are (a) Meyer et al. (1997), (b) Mohanty and Mousli (2000), (c) Schwartz et al. (2000), (e) Mendoza and Steenhuis (2003), (f) Kvaerno and Deelstra (2002), (g) Smith et al. (2003), (h) Parson (2001), (i) Wallace laboratories (2003), (j) Perfect et al. (2002),

(k) Carey and Woo (1999), (l) Schlotzhauer and Price (1999), (m) Pielke (2001), (n) Grunwald et al. (2001), (o) Landsberg et al. (2003), (p) Calhoun et al. (2001), (q) Lauren and Heiskannen (1997), (r) Clapp and Hornberger (1978). Note that Cosby et al. (1984) provide slightly different values than Clapp and Hornberger (1978).

Table 2. Typical mean values and standard deviations (in brackets) of soil characteristics. The symbols ks, ns, b, ys, ps stand for the hydraulic conductivity at saturation, porosity, soil-pore distribution index, and density of the dry soil material. References are (a) Meyer et al. (1997), (b) Mohanty and Mousli (2000), (c) Schwartz et al. (2000), (e) Mendoza and Steenhuis (2003), (f) Kvaerno and Deelstra (2002), (g) Smith et al. (2003), (h) Parson (2001), (i) Wallace laboratories (2003), (j) Perfect et al. (2002),

(k) Carey and Woo (1999), (l) Schlotzhauer and Price (1999), (m) Pielke (2001), (n) Grunwald et al. (2001), (o) Landsberg et al. (2003), (p) Calhoun et al. (2001), (q) Lauren and Heiskannen (1997), (r) Clapp and Hornberger (1978). Note that Cosby et al. (1984) provide slightly different values than Clapp and Hornberger (1978).

Soil-type

10"6 m/s

ns m3/m3

b

y s m

Ps

Sand

176r(43.9)a

0.395(0.056) r

4.05(1.78) r

-0.121(0.143) r

1580(90)p

Loamy sand

156.3r(31.7)a

0.410(0.068) r

4.38(1.47) r

-0.090(0.124) r

1610(100)n

Sandy loam

34.1r(13.7)a

0.435(0.086) r

4.90(1.75) r

-0.218(0.310) r

1520(140)p

Silt loam

7.2r(6.2)b

0.485(0.059) r

5.30(1.96) r

-0.786(0.512) r

1400(90)n

Silt

2.81r(1.325)c

0.476

5.33

-0.759

1420(70)k

Loam

7.0r(3.028)b

0.451(0.078) r

5.39(1.87) r

-0.478(0.512) r

1350(110)n

Sandy clay loam

6.3r(3.056)e

0.420(0.059) r

7.12(2.43) r

-0.299(0.378) r

1520(40)n

Silty clay loam

1.7r(0.806)f

0.477(0.057) r

7.75(2.77) r

-0.356(0.378) r

1410(60)n

Clay loam

2.5r(0.25)g

0.476(0.053) r

8.52(3.44) r

-0.630(0.510) r

1420(80)n

Sandy clay

2.2r(8.333)h

0.426(0.057) r

10.40(1.64) r

-0.153(0.173) r

1570(120)

Silty clay

1.0r(0.4)j

0.492(0.064) r

10.40(4.45) r

-0.490(0.621) r

1480(110)

Clay

1.3r(0.569)i

0.482(0.050) r

11.40(3.70) r

-0.405(0.397) r

1470(140)p

Humus

Peat

1.736 (0.938)1

0.923(0.342)

4.00(1.75)

-0.165(0.31)

106(243)

Moss

150 (400)k

0.900 (0.040)

1.00(1.75)

-0.120(0.310)

100(100)

Lichen

3356.5 (200)q

0.95 (0.060)

0.50(1.75)

-0.085(0.310)

120(30)

Various investigations using stand-alone versions of LSMs (e.g., Gao et al., 1996), NWPMs (e.g., Douville and Chauvin 2000), and GCMs (e.g., Wang and Kumar 1998) showed that initializing soil-moisture and temperature distributions is a huge source for errors in predicting the soil conditions correctly. Adjoint models and data-assimilation techniques can be applied for minimizing errors in initial soil conditions (e.g., van den Hurk et al., 1997, Callies et al., 1998, Reichle et al., 2001). Using this technique, however, is not possible for NWPMs, CTMs, GCMs or ESMs initialization due to lack of spatially continuous data.

The Project for Intercomparison of Land Surface Parameterization Schemes (PILPS) showed that LSMs strongly differ in accuracy because of, among other things, the choice of empirical parameters needed in parameterizations (e.g., Shao and Henderson-Sellers 1996, Slater et al., 1998). Typically, soil properties within a grid-cell or patch are expressed by assigning a mean value derived from laboratory or/and field studies thereby ignoring any variability. Consequently, predicted soil state variables and fluxes can differ over wide ranges in dependence of the parameter choice. Various parameter variation studies to assess whether slightly different parameters result in significant perturbations of soil temperature and moisture states. Such parameter-variation studies are subject to parameter interaction meaning that the parameter choice also affects simulated quantities that do not directly depend on the parameter. This fact makes optimal parameter choice difficult. Henderson-Sellers (1993), for instance, by using factorial experiments found that porosity is one of the most ecologically important parameters. Enhancing thermal diffusivities or volumetric heat capacities, for instance, may cool the soil and atmospheric boundary layer (locally more than 5 K and 1 K, respectively); enhancing volumetric heat capacities or thermal diffusivities may also affect atmospheric variables especially specific humidity, cloud and precipitation particles and may result in decreased maximum precipitation (Molders 2001). Errors may also stem from incorrectly assigned soil types. An about 5 % change in soil-type distribution may alter daily averages of the soil-moisture fraction by 29 % with respect to the reference case, and surface temperature by 2.3 K (Molders et al., 1997).

Besides systematic errors due to parameter choice, initialization, discretization, assumptions and physical parameterizations stochastic error is a source of uncertainty in predicted soil state variables and fluxes. As pointed out before, for describing soil heat and moisture transfer processes parameters have to be assigned that represent the soil characteristics. Herein stochastic errors result from the fact that the mean values of empirical soil parameters are in "error" by the amount of the standard deviation related to the natural (random) variability (Molders et al., 2005). For many soil parameters, this variability expressed, for instance, by the standard deviation is of the same order of magnitude as the parameter itself (cf. Table 2). Consequently, any soil state variable or flux predicted with these parameters is "error"-burdened too. Such uncertainty may even reduce the trust in predicting permafrost dynamics in GCMs and ESMs. For NWPMs and CTMs, it may limit the ability to simulate the evolution of active layer depth which is important information for agricultural purposes and assessment of river runoff. For GCMs and ESMs, this uncertainty may complicate climate impact assessment.

Errors in soil state variables and fluxes related to parameter uncertainty are of random kind for which they can be evaluated with statistical methods, for instance, Gaussian error-propagation (GEP) principles. This method permits researchers to investigate the relative importance of soil physical parameters (e.g., porosity) in producing prediction uncertainty at various potential conditions. Using GEP Molders et al., (2005), for instance, found that predicted distributions of soil temperature are less sensitive to uncertainty in thermal parameters than to uncertainty in hydraulic parameters. According to GEP results uncertainty in predicted soil-heat fluxes is within of the range as the typical errors in soil-heat flux measurements. They also found that the absolute value of soil-heat flux and its relative error decreases with increasing relative volumetric water content and concluded that soil-heat fluxes can be predicted with greater certainty after rain events or in the Tropics than under dry conditions or in dry regions.

Note that GEP can also be applied to examine how terms in the soil heat and moisture equations contribute to uncertainty in predicted soil temperature and moisture states. During phase transitions, the freeze-thaw term, for instance, can cause great uncertainty in volumetric water content and soil temperature (e.g., Molders et al., 2005). Similar was found using other methods by Molders and Romanovsky (2006).

Was this article helpful?

0 0

Post a comment