Evaluations of Uncertainty and Sensitivity in Soil Moisture Modeling on the Tibetan Plateau

Figures & data

Table 1. Information about the study sites situated in the TP region.

Sites	Locations	Study periods	Elevations	Land surface types
Anduo	(91.625°E, 32.241°N)	6/16-6/22 (1998)	4700 m	Alpine meadow
Ms3478	(91.715°E, 31.926°N)	9/1-9/16 (1998)	5063 m	Alpine meadow
Ms3637	(91.657°E, 31.017°N)	8/1-8/31 (1998)	4533 m	Alpine meadow
GaiZe	(84.050°E, 32.300°N)	5/1-5/31 (1998)	4420 m	Alpine desert
ShiQuanHe	(80.080°E, 32.500°N)	7/1-7/31 (1998)	4278 m	Alpine desert

Fig. 1. Sensitivity analysis framework for model parameter combinations based on the CNOP-P approach.

Table 2. The physical parameters and their default, minimum and maximum values for the TP sites in the CoLM. Def = Default value; Min = Minimum value; Max = Maximum Value.

Index	Parameter	Physical meaning	Category	Def	Min	Max
P01	porsl(up)	Porosity of upper soil, fraction of soil mass that is void	soil	0.43	0.25	0.75
P02	porsl(low)	Porosity of lower soil, fraction of soil mass that is void	soil	0.45	0.25	0.75
P03	phi0(up)	Minimum soil suction of upper soil (mm)	soil	207.35	50.0	500.0
P04	phi0(low)	Minimum soil suction of lower soil (mm)	soil	288.93	50.0	500.0
P05	bsw(up)	Clapp and Hornberger “b” parameter of upper soil	soil	5.77	2.5	10.0
P06	bsw(low)	Clapp and Hornberger “b” parameter of lower soil	soil	8.32	2.5	10.0
P07	hksati(up)	Saturated hydraulic conductivity of upper soil (mm/s)	soil	0.0042	0.001	1.0
P08	hksati(low)	Saturated hydraulic conductivity of lower soil (mm/s)	soil	0.0028	0.001	1.0
P09	sqrtdi	Inverse square root of the leaf dimension (m^-1/2)	canopy	5.0	2.5	7.5
P10	slti	Slope of the low temperature inhibition function	canopy	0.2	0.1	0.3
P11	shti	Slope of the high temperature inhibition function	canopy	0.3	0.15	0.45
P12	trda	Temperature coefficient of the conductance-photosynthesis model	canopy	1.3	0.65	1.95
P13	trdm	Temperature coefficient of the conductance-photosynthesis model	canopy	328.0	300.0	350.0
P14	trop	Temperature coefficient of the conductance-photosynthesis model	canopy	298.0	250.0	300.0
P15	extkn	Coefficient for leaf nitrogen allocation	canopy	0.5	0.5	0.75
P16	zlnd	Roughness length for the soil surface (m)	soil	0.01	0.005	0.015
P17	zsno	Roughness length for snow (m)	snow	0.0024	0.0012	0.0036
P18	csoilc	Drag coefficient for soil under the canopy	soil	0.004	0.002	0.006
P19	dewmx	Maximum ponding of the leaf area (mm)	canopy	0.1	0.05	0.15
P20	wtfact	Fraction of the shallow groundwater area	soil	0.3	0.15	0.45
P21	capr	Tuning factor of the soil surface temperature	soil	0.34	0.17	0.51
P22	cnfac	Crank Nicholson factor	soil	0.5	0.25	0.5
P23	ssi	Irreducible water saturation of snow	snow	0.0033	0.03	0.04
P24	wimp	Factor for controlling whether water is impermeable	soil	0.05	0.01	0.1
P25	pondmx	Maximum ponding depth for the soil surface (mm)	soil	10.0	5.0	15.0
P26	smpmax	Wilting point potential (mm)	canopy	−1.5e + 5	−2.0e + 5	−1.0e + 5
P27	smpmin	Restriction for the minimum soil potential (mm)	soil	−1.0e + 8	−1.0e + 8	−9.0e + 7
P28	trsmx0	Maximum transpiration for vegetation (mm/s)	canopy	0.0002	0.0001	0.01

Fig. 2. The reference states of SSM simulated by using the default parameter values (a) as well as the maximal uncertainties in the simulated SSM due to errors from all 28 selected parameters at different TP sites during different simulation periods in terms of absolute changes (b).

Fig. 3. The variations (left column) in ET (a, mm day⁻¹), R_sur (c, mm day⁻¹) and subsurface soil moisture (e, m³ m⁻³) relative to their respective reference states (right column) of ET (b, mm day⁻¹), R_sur (d, mm day⁻¹), subsurface soil moisture (f, m³ m⁻³) caused by the CNOP-P-type parameter errors associated with all 28 parameters at different TP sites during different simulation periods.

Fig. 4. The sensitivity ranks of all 28 parameters based on the single-parameter sensitivity analyses using the CNOP-P approach (a) and the OAT method (b) at different TP sites. Numbers “1”, “2”, “3”, and “4” on colored boxes label the 4 most sensitive parameters (in order of sensitivity) at each site.

Fig. 5. The maximal uncertainties in the simulated SSM due to errors from the most sensitive and important 4-parameter combinations identified by using the CNOP-P approach as well as errors from the 4 most sensitive parameters determined by using the OAT method: (a) in terms of absolute changes, (b) in terms of percentage changes.

Table 3. The most sensitive parameter combinations at different TP sites identified by using the sensitivity analysis framework based on the CNOP-P approach.

Site	The most sensitive parameter combination
Anduo	P01, P02, P03, P05
Ms3478	P01, P02, P03, P04
Ms3637	P01, P02, P03, P04
GaiZe	P01, P02, P03, P05
ShiQuanHe	P01, P02, P04, P05

Fig. 6. The reductions in the uncertainties of the simulated SSM at different TP sites, which are represented by $\tau ,$ caused by the different decreased extents (as indicated by $\alpha$) of three types of parameter error: CNOP, CNOP_Single, and OAT.