Representing the Integrated Water Cycle
in Community Earth System Model
1
Hong-Yi Li, L. Ruby Leung, Maoyi Huang, Nathalie Voisin,
Teklu Tesfa, Mohamad Hejazi, and Lu Liu
Pacific Northwest National Laboratory
Water in the human-Earth system
2
Water underlies and influences many important climate processes and
feedbacks – a leading cause of uncertainty in projecting future climate
Water vapor and
cloud feedback Snow-albedo feedback Aerosol-cloud
interactions Carbon-water
interactions
Water is essential for energy systems, ecosystem services, and a
wide range of life sustaining and other critical human activities
Global and regional water cycles are influenced by natural processes
as well as the human systems – how will they co-evolve in the future?
Processes in the integrated water cycle
Challenge: How to represent the multi-scale, dynamic interactions
among the atmosphere, terrestrial, and human systems
3
Modeling the integrated water cycle
Objectives
Represent the dynamic interactions between the human
and earth systems and their influence on the water
cycle
Use the models to investigate the nexus of climate,
energy, water, and land under climatic and societal
changes for sustainable energy and water in the future
Approach
Improve model scalability to address the multi-scale
atmospheric and terrestrial water cycle processes
Add human components (water management, water
use, water demand) of water cycle processes in CESM 4
5 5
Model Structure
R-GCAM
River Routing/
Water Management
Land Use
Electric Infrastructure/
Electricity Operations
Weather Weather
Atmosphere and ocean
boundary conditions
ROMS CLM
Flux Coupler
Atmospheric forcing
Surface fluxes
RESM WRF
POP CLM
Flux Coupler
Atmospheric forcing
Surface fluxes
CESM CAM
GCAM
CLM coupled with river routing and water
management
6
River Routing Model Coupled CLM-Routing Simulation
Evaporation
Melt
Sublimation
Throughfall
Infiltration Surface
runoff
Evaporation
Transpiration
Precipitation CLM Hydrology
Aquifer recharge
Water table
Soil
Saturated fraction
Water Management Model
Irrigation water demand Reservoir operations
Improve and add new capabilities in Community Land Model (CLM) to represent
hydrology and human – water cycle interactions at multiple time and space scales
7 7
CLM4 •Based on TOPMODEL
•Assumptions:
High-resolution topographic data are available;
Subsurface flow is topographic driven.
A quasi-steady state to approximate saturated zone dynamics;
Recharge to ground water is spatially uniform;
These assumptions are invalid, e.g., over flat terrain or arid regions
VIC • Conceptual
• Limited assumptions: – land surface, and therefore
surface runoff generation, is heterogeneous;
– Subsurface flow is a nonlinear function of deep-layer water availability
• Calibration of parameters are
recommended
Runoff parameterizations in LSMs
8 8
VIC as an hydrologic option in CLM To be released in Summer 2013
Li et al., 2011
9
Comparison of observed and simulated annual LH over
MOPEX basins: CLM, CLMVIC, and other NLDAS models
9
0.0
20.0
40.0
60.0
80.0
100.0
120.0
0.0 20.0 40.0 60.0 80.0 100.0 120.0
Late
nt
hea
t (W
/m2)
MODIS latent heat (W/m2)
Mean latent heat (2000-2007)
CLM (50.6)
CLMVIC (48.1)
NOAH (57.0)
VIC (62.1)
MOSAIC (80.1)
1:1 line MODIS (46.2)
Runoff schemes affect C cycle modeling through
interactions among water, energy, and C cycles
10
Relative difference: -20.4% (CLM4VIC – CLM4)
MODIS: 112 Pg C/year
CLMVIC: 114 Pg C/year
CLM4.0: 143 Pg C/year
Lei, H., M. Huang, L.R. Leung, et al. under review in JGR-Biogeosciences
Both structural and parameter uncertainties in the runoff generation schemes can lead to large uncertainty in carbon modeling, highlighting the significant interactions among the water, energy, and carbon cycles and the need for improving hydrologic parameterizations in land surface models.
11
Enhancing the CLM spatial structure
Grid cell/Subbasin
Landunits
Glacier Wetland Vegetated Lake Urban
Columns
PFTs
Permanent
channel
Floodplain Floodplain Upland Upland
Subbasins vs lat/lon grids as computational units
Subgrid structure: PFT/elevation, permanent channel/floodplain
Subgrid atmospheric forcing
3D radiative transfer in mountains
PFTs/Ele
vation
Subbasin
representation
Comparison of grid vs subbasin representations
Land surface heterogeneity such as topography has a dominant
influence on hydrological processes
Using subbasins as the computational units eliminates the needs to
represent the redistribution of soil moisture between units and
improve the accuracy for estimating the topographic index used in the
TOPMODEL parameterizations of surface and subsurface runoff
Grid-based representation (CLM) Subbasin-based representation (DCLM)
1o
0.5o
0.25o
0.125o
12
The subbasin representation improves scalability
13
Subbasin (DCLM) Grid (CLM)
Runoff
Latent
heat
flux
Model skills (MAE)
at different
resolutions are
more strongly
correlated and
model skill
increases
systematically with
resolution in the
subbasin
representation but
not the grid
representation
14 14
14
Hillslope routing to account for event dynamics and impacts of overland flow on soil erosion, nutrient loading etc.;
Sub-network routing: scale adaptive across different resolutions to reduce scale dependence;
Main channel routing: explicit estimation of in-stream status (velocity, water depth etc).
(Li et al., JHM, 2013, in press)
Conceptualized network
Model for Scale Adaptive River Transport
(MOSART)
Case Study: Columbia River Basin
15
Daily runoff generation from Variable Infiltration Capacity model (VIC) at 1/16 degree resolution
Off-line evaluation against monthly naturalized streamflow data at selected major stations
Model for Scale Adaptive River Transport
(MOSART)
16
MOSART is more skillful in simulating streamflow compared to RTM
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
DALLE ICEHA PRIRA CHIEF BROWN WANET CORRA ARROW
NS coeff. for monthly mean streamflow -- subbasin based representation
RTM(1/2)
MOSART_grid(1/8)
MOSART_subbasin
VIC(1/16)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
DALLE ICEHA PRIRA CHIEF BROWN WANET CORRA ARROW
NS coeff. for monthly mean streamflow – grid based representation Q_RTM(1/2)
Q_RTM(1/4)
Q_RTM(1/8)
Q_RTM(1/16)
Q_MOSART(1/2)
Q_MOSART(1/4)
Q_MOSART(1/8)
Q_MOSART(1/16)
Q_VIC(1/16)
Large drainage area Small drainage area
Model for Scale Adaptive River Transport
(MOSART)
17
MOSART reproduces monthly variation of channel velocity with minimum calibration
Model for Scale Adaptive River Transport
(MOSART)
Coupling MOSART to CLM4
Grid-based representation: replacing RTM with MOSART, keeping
remapping and parallel algorithms
Subbasin-based representation: one-one mapping between CLM and
MOSART grids
Supporting global multi-resolution database
Hydrography parameters (flow direction, channel length and slope etc.) directly derived from high resolution DEM (1km) at 1/16, 1/8, ¼, ½, 1 and 2 degree resolutions
Other parameters (channel geometry and Manning’s coefficients) derived based on landcover and empirical Hydraulic Geometry relationships
Global testing using NCAR benchmarking case (I-2000 )
Global run done for 1948-2004 period (spatial resolution of CLM4 0.9*1.25 degree, MOSART 0.5 degree)
19
1129 GRDC stations were geo-referenced to the new river network (drainage area error no more than 10%)
NSC > 0 for 470 out of 1129 GRDC stations for monthly streamflow
Global testing/evaluation of CLM4-MOSART
-- Evaluation with GRDC observed streamflow
1.0E+04
1.0E+05
1.0E+06
1.0E+07
1.0E+04 1.0E+05 1.0E+06 1.0E+07
DR
T e
sti
mate
d u
pstr
eam
are
a
(km
2)
GRDC provided upstream area (km2)
Preservation of upstream area
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
1.0E+06
1.0E+00 1.0E+01 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06
Mean
of
sim
ula
ted
flo
w (
m3/s
)
Mean of observed flow (m3/s)
Preservation of water balance
20
CLM Water
column
Active layer
Water column
River bank
Active layer
River bank
Inundated area
Water
Sediment
Nutrient
Main channel routing Sub-network
routing Hillslope routing
Couple riverine biogeochemistry into MOSART
Water management model (Voisin et al.
2013, HESS, submitted)
Designed for full coupling in an earth system models
Assume no knowledge of future inflow
Use generic operating rules
21
Water management model
22
Total water demand Reservoir dependency
CLM
Routing model Natural flow in each units
WRM
Regulated flow
Natural and regulated flow
23
Combining flood control and irrigation objectives in operating rules
best capture the observed regulated flow in the Columbia river basin
Reservoir storage and supply deficit
Reservoir storage is only reproduced using operating rules that
combine flood control and irrigation priorities
At the American Falls, supply deficit is related to groundwater use 24
Integrating with IAM: Water market and
water demand
Water demand for various
sectors driven by energy
demand, GDP, and
agricultural land demand is
simulated by IAM (GCAM)
GCAM also solves the
energy market, land market,
and water market
simultaneously to establish
prices and shares by source
25
CLM
Routing model Natural flow in each units
WRM
Regulated flow
Water demand from
sectors other than
agriculture
Land use
Summary
On hydrologic modeling and model scalability:
Subbasin representation offers some advantage in scalability
The new river routing model (MOSART) represents hillslope,
tributary, and channel routing and works well across scales
A new CLM subgrid structure is being developed to account
for subgrid PFT, elevation, and inundation
On representing the dynamic interactions between
human and earth systems:
Developed a coupled system including CLM, river routing,
and water management to represent irrigation water use and
water management
Ongoing research to represent water demand, water use, and
water market using the coupled CESM - GCAM
Global implementation and evaluation underway 26