Transcript Slide 1

Water Balance at the Land Surface
Control volume
Win
water storage
Basic water balance equation:
8 February 2005
Lecture 3 - Koster
Wout
WinDt = Dstorage + WoutDt
CLIM 714
Land-Climate Interactions
Water Balance for a Single Land Surface Slab, Without Snow
(e.g., standard bucket model)
Terms on
LHS come from
the climate model.
Strongly dependent
on cloudiness, water
vapor, etc.
P
E
R
w
P
Terms on
RHS come are
determined by
the land surface
model.
= E + R + CwDw/Dt + miscellaneous
where
P
= Precipitation
E = Evaporation
R = Runoff (effectively consisting of surface runoff and baseflow)
Cw = Water holding capacity of surface slab
Dw = Change in the degree of saturation of the surface slab
Dt = time step length
miscellaneous = conversion to plant sugars, human consumption, etc.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Usually, a combination of water balances is considered. For example:
Water balance associated with canopy
interception reservoir
P
DWc
P = Eint + Dc + Dt
Eint = interception loss
Dc = drainage through canopy
(“throughfall”)
DWc = change in canopy
interception storage
Eint
Wc
Dc
Water balance in a snowpack
P (snow)
M
8 February 2005
DWsnow
P = Esnow + M + Dt
Esnow
Esnow = sublimation rate
M = snowmelt
DWsnow = change in snow
amount (“infinite”
capacity possible)
Wsnow
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Water balance in a surface layer
M+Dc
Ebs + Etr1
Water
storage
w1
Rs
Q12
M + Dc =
Ebs + Etr1 + Rs + Q12 + CW1DW1/Dt
Ebs = evaporation from bare soil
Etr1 = evapotranspiration from layer 1
Q12 = water transport from layer 1 to layer 2
CW1 = water holding capacity of layer 1
DW1 = change in degree of saturation of layer 1
w2
w3
Water balance in a subsurface layer (e.g., 2nd layer down)
Note: some models
may include an
additional, lateral
subsurface runoff
term
Etr
2
water
storage
w1
Q12
w2
Q23
w3
8 February 2005
Q12 = Q23 + Etr2 + CW2DW2/Dt
Lecture 3 - Koster
Etr2 = evapotranspiration from layer 2
Q23 = water transport from layer 2 to layer 3
CW2 = water holding capacity of layer 2
DW2 = change in degree of saturation of layer 2
CLIM 714
Land-Climate Interactions
Water balance in the lowest layer
Qn,n-1 = QD + Etr-n + CWnDWn/Dt
Etrn
Etr-n = evapotranspiration from layer n, if allowed
QD = Drainage out of the soil column (baseflow)
Qn,n-1
water
storage
wn
QD
A model may compute all of these water balances, taking care to ensure
consistency between connecting fluxes (in analogy with the energy balance
calculation).
P
Eint
Esnow
P
Dc
Ebs
Etr1 Etr2 Etr3
M
W1
Q12
W2
Q23
W3
8 February 2005
Lecture 3 - Koster
QD
CLIM 714
Land-Climate Interactions
Rs
Precipitation, P
Getting the land surface hydrology right in a climate model is difficult
largely because of the precipitation term. At least three aspects of
precipitation must be handled accurately:
a. Spatially-averaged precipitation amounts (along
with annual means and seasonal totals)
b. Subgrid distribution.
c. Temporal variability and temporal correlations.
Otherwise, even with a perfect land surface model,
Garbage
in
8 February 2005
Lecture 3 - Koster
Perfect land
surface model
CLIM 714
Land-Climate Interactions
Garbage
out
Accurate precipitation measurements are limited by availability
of rain gauges….
Each box is ~250 km
on a side
… and by inherent inaccuracies in
satellite-derived precipitation data
How Good is the Estimated SSM/I Rain Rate Climatology Data?
Over oceans, no “truth” data available for
validations
Estimated Nonsystematic Error
60
Nonsystematic error includes sampling and
random
Percent Error (%)
50
40
Sampling error dominates
30
F-13 and F-14 SSM/I, with similar sampling
have similar error
20
TMI has a slightly less, nonsystematic error
10
Figure compliments
of Al Chang,
NASA/GSFC
0
0
2
6
8
10
12
14
Rain Rate (mm/day)
F-14
8 February 2005
4
F-13
Lecture 3 - Koster
TMI
F-13+F-14
GPM
F-13 AM
CLIM 714
Land-Climate Interactions
Combining F-13 & F-14 almost satisfy the
TRMM 1 mm/day and 10% for heavy rain
GPM with 8 satellites will have 50% less error
than combining F-13 & F-14
Technical
notes for
figure
Precipitation: subgrid variability (1)
The bottom storm is more
evenly distributed over the
catchment than the top storm.
Intuitively, the top storm will
produce more runoff, even
though the average storm
depth over the catchment
(E(Yo)) is smaller.
Key points:
-- Specifying subgrid
variability of precipitation
is critical to an accurate
modeling of surface hydrology.
-- A GCM is typically unable
to specify the spatial structure
of a given storm. The LSM
typically has to “guess” it.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
From Fennessey, Eagleson,
Qinliang, and Rodriguez-Iturbe,
1986.
Precipitation: subgrid variability (2)
Here, the two storms have
similar spatial structure
and total precipitation
amounts. The locations of
the storms, however, are
different. If the top storm
fell on more mountainous
terrain than the bottom
storm, the top storm might
produce more runoff
Key point: A GCM is
typically unable to specify
the subgrid location of a
given storm. The LSM
typically has to “guess” it.
From Fennessey, Eagleson,
Qinliang, and Rodriguez-Iturbe,
1986.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
GPM Mission Design
OBJECTIVES
• Understand horizontal & vertical
Core
Constellation
structure of rainfall, its macro- &
micro-physical nature, & its
associated latent heating
• Train & calibrate retrieval
algorithms for constellation
radiometers
OBJECTIVES
• Provide sufficient global
sampling to significantly
reduce uncertainties in shortterm rainfall accumulations
• Extend scientific and societal
applications
Core Satellite
• TRMM-like spacecraft (NASA)
• H2-A rocket launch (NASDA)
• Non-sun-synchronous orbit
~ 65° inclination
~400 km altitude
• Dual frequency radar (NASDA)
K-Ka Bands (13.6-35 GHz)
~ 4 km horizontal resolution
~250 m vertical resolution
• Multifrequency radiometer (NASA)
10.7, 19, 22, 37, 85, (150/183 ?) GHz V&H
Precipitation Processing Center
• Produces global precipitation products
• Products defined by GPM partners
8 February 2005
Lecture 3 - Koster
Constellation Satellites
• Pre-existing operational-
experimental & dedicated satellites
with PMW radiometers
• Revisit time
3-hour goal at ~90% of time
• Sun-synch & non-sun- synch orbits
600-900 km altitudes
Precipitation Validation Sites for Error Characterization
• Select/globally distributed ground validation “Supersites” (research
quality radar, up looking radiometer-radar-profiler system, raingagedisdrometer network, & T-q soundings)
• Dense & frequently reporting regional raingage networks
CLIM 714
Land-Climate Interactions
Precipitation: temporal correlations
Temporal correlations are very important -- but are largely ignored -in GCM formulations that assume subgrid precipitation distributions.
This is especially true when the time step for the land calculation is of
the order of minutes. Why are temporal correlations important?
Consider three consecutive time steps at a GCM land surface grid cell:
time step 1
time step 2
time step 3
Case 1: No temporal
correlation in storm
position -- the storm is
placed randomly with the
grid cell at each time step.
Case 2: Strong temporal
correlation in storm
position between time steps.
Case 2 should produce, for
example, stronger runoff.
CLIM 714
8 February 2005
Lecture 3 - Koster
Land-Climate Interactions
Throughfall
Simplest approach: represent the interception reservoir as a bucket that gets filled during
precipitation events and emptied during subsequent evaporation. Throughfall occurs
when the precipitation water “spills over” the top of the bucket.
Capacity of bucket is
typically a function
of leaf area index, a
measure of how many
leaves are present.
8 February 2005
Lecture 3 - Koster
This works, but because it ignores subgrid precipitation
variability (e.g., fractional wetting), it is overly simple.
CLIM 714
Land-Climate Interactions
Spatial precipitation variability and interception loss
SiB’s approach
(Seller’s et al, 1986)
Area above line
is considered
throughfall
Precipitation assumed
to fall according
to some prescribed
distribution
Capacity of
reservoir
Original water in reservoir
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Note: SiB allows
some of the
precipitation to fall
to the ground without
touching the canopy.
Temporal precipitation variability and interception loss
Mosaic LSM’s approach:
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Evaporation
See notes from energy balance lecture. Note, though, locations of
moisture sinks for bare soil evaporation and transpiration:
Bare soil evaporation water is usually taken from the top soil layer.
Transpiration water is usually taken from the soil layers that
comprise the root zone. Different amounts may be taken from
different layers depending on:
-- layer thickness
-- assumed root density profile
e.g., transpiration water
taken from these layers…
but not this layer
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Runoff
a. Overland flow:
(i) flow generated over permanently saturated zones near a river
channel system: “Dunne” runoff
(ii) flow generated because precipitation rate exceeds the infiltration
capacity of the soil (a function of soil permeability, soil water
content, etc.): “Hortonian” runoff
b. Interflow (rapid lateral subsurface flow through macropores and seepage
zones in the soil
c. Baseflow (return flow to stream system from groundwater)
Runoff (streamflow) is affected by such things as:
-- Spatial and temporal distributions of precipitation
-- Evaporation sinks
-- Infiltration characteristics of the soil
-- Watershed topography
-- Presence of lakes and reservoirs
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Modeling runoff: basin scale
When variations in precipitation, topography, soil characteristics, etc., can
be explicitly accounted for (as in so-called “spatially distributed” hydrological
basin models), runoff can be predicted fairly accurately. The TOPMODEL
approach uses the statistics of topography to characterize the spatial
distribution of water table depth in a basin, with consequent impacts on
runoff generation.
simulated
observed
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
From Beven, K., “Spatially distributed
modeling; conceptual approach to
runoff prediction”, in Recent Advances
in the Modeling of Hydrologic Systems,
ed. By Bowles and O’Connell, p. 373-387,
Kluwer Academeic Pub., 1991.
Modeling runoff: GCM scale
Surface runoff formulations in GCMs are generally very crude, for at least
two reasons:
(i) Developers of GCM precipitation schemes have focused on producing
accurate precipitation means, not on producing accurate subgrid spatial
and temporal variability.
(ii) GCM land surface models generally represent the hydrological state of
the grid cell with grid-cell average soil moistures -- the time evolution of
subgrid soil moisture distributions is not monitored.
At best, we can expect first-order success with these
runoff formulations
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Controls in nature
Framework of typical LSM
SCALE: HUNDREDS OF KILOMETERS
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Note that because of the inherent inconsistency between nature and
the typical LSM’s soil layer framework, no “best” approach for modeling
runoff exists. Current LSM approaches are “all over the place”. Typically,
though, runoff is a function of the amount of moisture in the top soil layer.
Bucket model: total runoff = P + M - E if this is positive and the bucket is full.
total runoff = 0 otherwise.
GISS Model II: total runoff = max( 0.5 P
SiB: surface runoff =
excess over
infiltration capacity,
assuming subgrid
distribution of
throughfall.
W1
W1max
Dt, excess over capacity ).
This part of the throughfall
(above the line) runs off;
the rest infiltrates
Depth =infiltration capacity * Dt
Other approaches will be discussed later
in the course.
CLIM 714
8 February 2005
Lecture 3 - Koster
Land-Climate Interactions
throughfall distribution
Where does GCM
runoff go once it’s
produced? It may
“disappear” (i.e., since
it eventually ends up
in the ocean anyway, it
may be no longer
considered by the land
model -- it may be
effectively removed
and forgotten), or it
may be routed, using
routing networks like
this. The routed
runoff can be
compared to
streamgauge
measurements.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
(Implied) Global Annual River Discharge (kg/yr)
8.00E+16
7.00E+16
6.00E+16
Historical estimates of
river discharge are all
over the place -- more
evidence of uncertainty
in our estimates of the
global water cycle.
5.00E+16
4.00E+16
3.00E+16
2.00E+16
Legacy of estimates
CLIM 714
Schlosser and Houser, 2004
(submitted)
8 February 2005 Lecture 3 - Koster
Land-Climate Interactions
B0
P1
P2
P3
Satellite measurements may
provide valuable runoff data.
(The methodology is still in its
infancy.)
Observations: River and Lake Stage
Jehil Reservoir,
Afghanistan
TOPEX/POSEIDON-Radar Backscatter
Jehil
Reservoir
1990
Application of
satellite radar altimetry
and imagery to drought
investigation.
C. Birkett/ESSIC
8 February 2005
Lecture 3 - Koster
Jehil
Reservoir
1999
Jehil
Reservoir
1998
CLIM 714
Land-Climate Interactions
Jehil
Reservoir
2001 (dry)
Soil Moisture Transport, Baseflow
First, some useful definitions:
Porosity (n): The ratio of the volume of pore space in the soil to the total volume of the soil. When a soil
with a porosity of 0.5 is completely dry, it is 50% rock by volume and 50% air by volume.
Volumetric moisture content (q): The ratio of the volume of water in the soil to the total volume of soil.
When the soil is fully saturated, q = n.
Degree of saturation (w): The ratio of the volume of water in the soil to the volume of water at
saturation. By definition, w= q /n.
Pressure head (y): A measure of the degree to which the soil holds on to its water through tension
forces. More specifically, y =p/rg, where r is the density of water, g is gravitational acceleration, and p
is the fluid pressure.
Elevation head (z): The height of soil element above an arbitrary baseline.
Hydraulic head (h): The sum of the pressure head and the elevation head.
Wilting point: The soil moisture content (measured either in degree of saturation or pressure head) at
which plants can no longer draw the moisture from the soil. When modeling the root zone, this is often
considered to be the lowest moisture content possible.
Field capacity: The water content obtained when a saturated soil drains to the point where the surface
tension on the soil particles balances the gravitational forces causing drainage.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Estimating water transport in the saturated zone (i.e., below water table)
Darcy’s Law states that
h2 - h1
Q/A = flow per unit normal area = - K
L
where K = hydraulic conductivity
h = hydraulic head
L = separation distance
L
h2
More generally,
q=-K
h
h1
q = specific discharge = Q/A
Generalized Darcy’s Law:
relates flow to gravitational
and pressure forces.
(Recall: h = y + z)
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Hydraulic conductivity, K, is related to the soil’s specific permability:
krg
K=
m
Where r is the fluid’s density and m is its dynamic viscosity. K is thus a
function of soil and fluid properties.
K varies tremendously with
soil type. Small variations in
soil type, say across a field site,
could lead to orders of magnitude
difference in the ability to
transport moisture.
From Freeze and Cherry
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Moisture transport in the unsaturated zone (e.g., in the soil near
the surface) can also be computed with Darcy’s law, if appropriate
corrections are made to pressure head and hydraulic conductivity.
Z
qr = residual moisture
“specific retention”
If atmospheric
pressure defined
to be 0.
qr
Soil moisture
profile
capillary
fringe
p<0
p=0
p>0
Water table
Unsaturated zone
equations (from
Clapp and Hornberger)
8 February 2005
Lecture 3 - Koster
Recall: q = ratio of water volume
to soil volume,
n = porosity
q=n
q
Recall: w = degree of saturation,
= q/n
y(w) = ysaturated w -b
K(w) = Ksaturated w 2b+3
b = empirical coefficient
CLIM 714
Land-Climate Interactions
Soil parameter values used in GSWP, from Cosby
m
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
m/s
Three things that complicate moisture transport in the unsaturated zone:
1. Extreme nonlinearity.
b may have values between 4 and 10. If b=10, then K(q) = Ksaturated w 23
2. Hysteresis
Values of parameters not really a
unique function of moisture state; they
depend in part whether the soil has
previously been wet or previously
been dry -- whether the soil is
wetting up or drying down.
From Freeze and Cherry
3. Anisotropy.
Hydraulic conductivity may vary
with the direction of flow.
8 February 2005
Lecture 3 - Koster
For a given head gradient,
flow in this direction
may be easier
than flow in
this direction
CLIM 714
Land-Climate Interactions
GCM approaches to modeling subsurface flow
Typically,
-- Assume homogeneity of soil
constant Ksat, ysat
-- Ignore hysteresis
-- Concentrate on vertical transports only
-- Concentrate only on unsaturated zone and
determination of moisture drainage to water table
Discretization of Darcy’s law (e.g., SiB)
Darcy’s law for vertical flow can be written:
q=-K
h
qz = - K
=-K
h
z
(z + y)
z
y
=-K(
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
z
+ 1)
One possible discretization of Darcy’s law (continued)
Characterize the soil as stacked layers
(d = thickness)
Compute for each layer i:
yi = ysat wi -b
Ki = Ksat wi 2b+3
w1
d1
surface layer
w2
d2
root layer
w3
d3
recharge layer
Compute flow from layer i to layer i+1:
qz i,i+1 = K
yi - y i+1
d
+1
K = “average” K across distance = (diKi + di+1Ki+1)/(di+di+1)
d = effective depth for computing gradient = 0.5 (di+di+1)
For drainage out the bottom of the soil column (QD), one might equate it
to the hydraulic conductivity in the lowest layer. SiB, for example, goes
beyond this by also applying a “mean slope angle” term, sin x: QD = K3 sin x
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
It’s important to keep in mind that the different LSMs use very
different discretizations of the soil column -- there is no one “right way”
to do it.
Discretizations and moisture
transport paths for a wide
variety of LSMs, as outlined
by Wetzel and Chang (1996)
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Energy balance versus water balance
Energy balance:
Implicit solution usually necessary
Results in updated temperature prognostics
Water balance:
Implicit solution usually not necessary
Results in updated water storage prognostics
How are the energy and water budgets connected?
1. Evaporation appears in both.
2. Albedo varies with soil moisture content.
3. Thermal conductivity varies with soil moisture content.
4. Thermal emissivity varies with soil moisture content.
Question: Can we address how the energy and water budgets
together control evaporation rates?
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Budyko’s analysis of energy and water controls over evaporation
These assymptotes
act as barriers to
evaporation.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
The equation in fact characterizes the combined energy and
water balance behavior of GCMs in general...
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
… and can thus be used to explain, in part, differences in
GCM behavior.
Each letter
corresponds to
a different GCM
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
What determines the shape of Budyko’s curve?
If only annual means mattered,
the observed curve should look
like this:
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Seasonality, however, is important.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Note that if these seasonal effects
alone were considered, the
observed curve would actually
look like this:
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
This effect can bring the curve in line with the observed curve. Note, though,
that other effects also contribute to a region’s evaporation rate, including land
surface properties and the temporal variation of precipitation.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Budyko’s analysis: discussion
1. Annual precipitation and net radiation control, to first order, annual
evaporation rates.
2. The spread of points around the Budyko curve is large, though, due to
various additional factors:
-- phasing of seasonal P and Rnet cycles
-- interseasonal storage of moisture
-- Other land surface or meteorological effects (vegetation type and
resistance, topography, rainfall statistics, …)
3. Note also:
-- Land surface processes affect the precipitation and net radiation
forcing -- there’s not truly a clean separation between land
and atmospheric effects.
-- The land’s effects on hourly, daily and monthly evaporation are
relatively much more important than they are on annual
evaporation.
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Budyko’s equation for mean annual evaporation
Modification for interannual variability:
See Koster and Suarez, 1999
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
E
P
sE
sP
Rnet / lP
8 February 2005
Lecture 3 - Koster
CLIM 714
Land-Climate Interactions
Equation works well when tested with GCM data:
curve derived from
Budyko equation
sE
sP
RCLIM
/ lP
net 714
8 February 2005
Lecture 3 - Koster
Land-Climate Interactions