Zong-Liang Yang Guo-Yue Niu Robert E. Dickinson The University of Texas at Austin

Post on 14-Feb-2016

31 views 1 download

description

Modeling Surface and Subsurface Runoff in CLM. Zong-Liang Yang Guo-Yue Niu Robert E. Dickinson The University of Texas at Austin. Prepared for Land Model Working Group Meeting, March 14, 2005 Funded under NASA grant NAG5-12577. Outline. Introduction - PowerPoint PPT Presentation

transcript

Zong-Liang Yang Guo-Yue Niu

Robert E. Dickinson

The University of Texas at Austin

Modeling Surface and Subsurface Runoff in Modeling Surface and Subsurface Runoff in CLMCLM

Prepared for Land Model Working Group Meeting, March 14, 2005

Funded under NASA grant NAG5-12577

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

Performance of Baseline CLM (1)Performance of Baseline CLM (1)

Soil moisture (Sleepers River Catchment):Too lowOdd profile (9th layer driest)

Daily runoff (Sleepers River Catchment, Vermont, USA):Negative modeling efficiency because of large spikesSurface runoff (fast component) too high

Performance of Baseline CLM (2)Performance of Baseline CLM (2)

Monthly runoff (GSWP2 Project):Overestimated Surface runoff (fast component) too high Surface runoff is 80% of total runoff.

Parameterization of Runoff in Baseline CLMParameterization of Runoff in Baseline CLM

Guided by four considerations:

1) TOPMODEL:

topographic control on the growth and decay of saturated area and groundwater flow

2) 1-D 10-layer soil structure:

3) Topographic data availability:

a simple determination of the saturated area, allowing room for improvement when the topographic parameters are available globally.

4) BATS:

success in PILPS experiments, esp. PILPS 1c (The Red-Arkansas River Basin)

Parameterization of Runoff in Baseline CLMParameterization of Runoff in Baseline CLMRunoff = Surface runoff + Subsurface runoff

Surface runoff Rs = Fsat Qwat + (1 – Fsat) ws4 Qwat

TOPMODEL BATS

Qwat = Input of water at the soil surfaceFsat = Fractional saturated area = Fmax exp(–Dw)ws = averaged soil wetness in the top three soil layers

Subsurface Runoff Rsb = Fsat lb exp(–Dw) + (1 – Fsat) Kb wb2B+3

lb = maximum baseflow rate = 10-5 mm s-1 Kb = maximum drainage rate = 0.04 mm s-1

wb = averaged soil wetness in the bottom three soil layers

Ksat (z) = Ksat(0) exp(–f z )

Ksat(0) = saturated hydraulic conductivity at the soil surface, determined by soil texture following Cosby et al. (1984); f = 2 (tunable parameter)

Problems in the Baseline CLMProblems in the Baseline CLM1) The second term in surface runoff is redundant and too large.

Rs = Fsat Qwat + (1 – Fsat) ws4 Qwat

TOPMODEL BATS

2) The second term in subsurface runoff is redundant and too large.

Rsb = Fsat lb exp(-Dw) + (1 – Fsat) Kb wb2B+3

3) How to determine Ksat (0) and Ksat(z)? Following Cosby et al. (1984)? Allowing macorpores? How to account for vertical and horizontal Ksat?

4) How to compute Fsat?Constrained by a global constant? By topography?

5) How to determine the water table?By the total head equilibrium? The moving boundary? An explicit groundwater model?

Proposed Runoff Scheme in CLMProposed Runoff Scheme in CLM1) Surface runoff

Rs = Fsat Qwat + (1 – Fsat) max(0, Qwat – Imax)2) Subsurface runoff

Rsb = Rsb,max exp (-f zw) simplified from

Rsb = [ α Ksat (0) / f ] exp(- λm) exp(- f zw)

α= anisotropic factor for different Ksat in vertical and horizontal directionsλm= grid-cell averaged topographic indexzw= grid-cell mean water table depth

3) Ksat (0) = ksat exp (f Dc) Ksat (z) = Ksat(0) exp(–f z )ksat is determined by following Cosby et al. (1984).Allowing macropores.

4) Fsat = ∫λ ≥ (λm + f*zw) pdf(λ) dλ

5) The water table is diagnosed from an equilibrium relationshipψ(z) – z = ψsat – zw (i.e., the total head is equal across the soil column layers)

Topography-based Runoff SchemeRunoff production mechanism

Surface runoffSaturation excessInfiltration excess

Subsurface runoff Topographic control Bottom drainage “Over-saturated” water

recharged into upper unsaturated layers

Infiltration Excess

Wat

er T

able

Dep

th

Saturation Excess

Super-saturationTopography Bottomm

w

ef

KR

eRR

satsb

fzsbsb

)0(max,

max,

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

Maximum Fractional Saturated Area (FMaximum Fractional Saturated Area (Fsat,maxsat,max))

Using 1 km × 1 km topographic index (λ)

Using Γ-distribution fit to the 1 km data

Differences of (Middle – Top)

Fsat = ∫λ ≥ (λm + f*zw) pdf (λ) dλ when the water table is at the surface (zw = 0)

Defining the Maximum Fractional Saturated AreaDefining the Maximum Fractional Saturated Area

Fsat = ∫λ ≥ (λm + f*zw) pdf (λ) dλ

Fsat,max results when the water table is at or above the surface (zw ≤ 0)

Topographic Index λ

Simulations over the Sleepers River BasinSimulations over the Sleepers River Basin

TOPMODEL: Fsat = ∫λ ≥ (λm + f*zw) pdf (λ) dλ

SIMTOP: Fsat = Fsat,max exp (–0.5 f zw) Fsat,max = 0.42

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

Ksat, macropores and anisotropic factor

ksat depends on soil type (Cosby et al., 1984)

Stiglietz et al. (1997) :Ksat(0) = 1000 × ksat α=1, f=3.26

Chen and Kumar (2001):Ksat(0) = exp(f Dc) × ksat

= 6 × ksat α=2000, f=1.8

This study: Ksat(0) = exp(f Dc) × ksat

= 6 × ksat α=20, f=2 (global); =3.26 (Sleepers River)

or Rsb,max = 1.45×10–7m/s

fzsatsat eKzK )0()(

10–7 m/s 10–3 m/s0 m

1 m

2 m

3 m

10–10 m/s

Baseline CLM

Stiglietz et al.

mef

KR satsb

)0(

max,

Chen & Kumar

Ksat, macropores and anisotropic factor

fzsatsat eKzK )0()(

mef

KR satsb

)0(

max,10–7 m/s 10–3 m/s

0 m

1 m

2 m

3 m

10–10 m/s

Baseline CLM

Stiglietz et al.

Chen & Kumarα = 1

α = 20

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

Simulations over Various Regional BasinsSimulations over Various Regional Basins

West SiberiaEast Siberia

NW Canada

CongoAmazon India

E USAW USAC Europe

S AfricaSahara Australia

N America

EurasiaS Hemisphere

Simulations over the Sleepers River BasinSimulations over the Sleepers River Basin

TOPMODEL: Fsat = ∫λ ≥ (λm + f*zw) pdf (λ) dλ

Rsb,max = 1.45 ×10–7 m/s

Chen & Kumar

10–7 m/s 10–3 m/s0 m

1 m

2 m

3 m

10–10 m/s

Baseline CLM

Stiglietz et al.

Bottom sealed

Bottom NOT sealed

Simulations over the Sleepers River BasinSimulations over the Sleepers River Basin

10–7 m/s 10–3 m/s0 m

1 m

2 m

3 m

10–10 m/s

Baseline CLM

Stiglietz et al.

Bottom NOT sealed

Bottom sealed

Chen & Kumar

TOPMODEL: Fsat = ∫λ ≥ (λm + f*zw) pdf (λ) dλ

Rsb,max = 1.45 ×10–7 m/s

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

Comparison of Comparison of Simulated Water TableSimulated Water Table with with MeasurementsMeasurements in Illinois in Illinois

OutlineOutline Introduction

Current treatment of runoff in CLM and problems Saturation area

Surface runoff Ksat, macropores and anisotropic factor

Subsurface runoff Constant versus exponential Ksat

Continental-scale simulations Water table

Regional-scale simulations Comparison with observations

Sensitivity to parameters f Rsub,max

Sensitivity to f:Sensitivity to f:Simulations over the Sleepers River Simulations over the Sleepers River

Sensitivity to RSensitivity to Rsb,maxsb,max

Simulations over the Sleepers River Simulations over the Sleepers River

Simulations over the Sleepers RiverSimulations over the Sleepers River

Simulations over the Amazon BasinSimulations over the Amazon Basin

Coupled CAM2-CLM2 Results in AmazonSu

rfac

e ru

noff

(m

m/d

)

Soil

Moi

stur

e (m

m/d

)

ET

(mm

/d)

Prec

ipita

tion

(mm

/d)

Simplified TOPMODEL produced less surface runoff, allowing more water to infiltrate into deeper soil and to increase soil moisture. Transpiration increases significantly, more than compensating the decrease in the interception loss. As a result, both ET and precipitation show favorable increases.

1-2mm/d

Conclusions1) Based on offline tests for a small catchment or global continents, the

proposed runoff scheme is shown to be robust for a wide range of assumptions including

a) Different methods of Fsat,• Based on 1-km topographic parameters• Assuming a global constant

b) Constant versus exponential Ksat• In the constant profile case, results depend on whether the bottom

is sealed or not c) Different methods of water table.

2) The simulations of soil moisture and runoff are all improved over the baseline version.

3) In the Amazon region, canopy evaporation and surface runoff are reduced, soil is wetter, and both ET and precipitation are increased.

Future Work1) Increase the total soil thickness to ~10 m and make it

a geographic variablea) Need bedrock data,b) Adjust root depth and distribution,c) Collect the water table data,d) Compare with the GRACE data.

2) Global optimization of two calibration parameters (f and Rsub,max).

3) Include (unconfined) aquifer into CLM to study groundwater recharge, discharge, and climate-groundwater interactions.

Land Surface, Surface Water and

Groundwater

Can be detected by GRACE