Modeling Holocene Surface Processes

1
Modeling Holocene Surface Processes
Helena Mitasova Department of Marine,
Earth and Atmospheric Sciences, NCSU,
Raleigh http//www.skagit.meas.ncsu.edu/helena/
2
Surface processes water
- surface water flow overland flow, channels,
rivers - subsurface flow, groundwater -
simplified steady state models can be used -
stochastic, event-based representation of storms
and flooding should be considered, - evidence of
floods and their extent would be extremely
valuable
3
Surface processes sediment

• Geomorphology
• - hillslope erosion/deposition sheet, rill,
  gullies
• - fluvial processes channel evolution
  meandering, ...
• - debris flows and land slides
• - anthropogenic modification of topography
  terraces, ...
• - wind erosion and sediment transport?
• - tectonic uplift? earthquakes?

4
Evolution of topography
potential elevation surfaces reconstructed from
existing data will be needed to perform flow
routing and estimate initial topographic
parameters needed in the models. information
indicating types, spatial pattern and rates of
relevant mass flows will be needed to select
and/or develop the most appropriate models of
surface processes.
5
Inputs for modeling
Climate intensity, duration and frequency of
rainfall. Soils properties such as erodibility,
hydraulic conductivity and their evolution due to
the change in climate and land use. Land cover /
land use / conservation practices type and
density of vegetation, land cover properties
influencing rainfall intercept, infiltration,
surface roughness (influencing flow velocity),
soil detachment and sediment transport capacity.
Evolution of land cover due to change in climate,
human settlements and agriculture and soil
properties.
6
Models
What models to use? Do we need to develop new
models? Ideal collaboration with CSMDS
initiative, on hold now Explore existing models
and modify them for the project. Models in
GRASS Simple, easy to use models for averaged
estimates of soil loss, useful as input for
decision making in agriculture but not suitable
for terrain change RUSLE3D hillslope soil
detachment, location of gullies, averaged soil
loss in watersheds USPED hillslope erosion and
deposition, location of gullies
7
Models in GRASS
More complex, physics-based models that simulate
wider range of effects, but will need more work
to adapt to the scale of our project r.sim.water/r
.sim.sediment r.terradyn script that uses
r.sim.water and r.sim.sediment to estimate
erosion and deposition and modifies terrain
accordingly. GRASS does not have fluvial
processes model land slide/ debris flow model,
if these processes are important, external
models have to be used.
8
Models in GRASS
erosion onlyUSLE/RUSLE2D
USLE/RUSLE3D
ARKLSCP slope length
ARKLSCP upslope area
erosion/deposition USPED
physics-based SIMWE
9
Erosion/deposition 2D flow
Observed depth of colluvial deposits
USLE/RUSLE erosion based on slope length
erosion deposition
USPED1D change in sediment flow USPED
2D change in sediment flow
10
Erosion/deposition USPED
deposition erosion
11
Short term terrain evolution
Chris Thaxton r.terradyn
12
External Models
Landscape evolution SIBERIA (G.
Willgoose) Landscape evolution with coupled
hillslope and fluvial processes CHILD (G.
Tucker) CHILD is the most complete model. MIT
may have a newer version that supports spatially
variable land use http//hydrology.mit.edu/researc
h/geomorphology/geoarcheology/index.html http//ww
w.geo.oregonstate.edu/lancasts/research.html SAND
BOX GUI for CHILD (python and C, available at
sourceforge.net) http//oregonstate.edu/penningj/
programming/child/ Land slides SHALSTAB,
SINMAP Debris flow University of Buffalo ITR
project
13
Data formats
Standard 2D raster time series readily available.
Conversion to TIN may require breaklines (e.g.
streams) in vector format. CHILD uses
TIN. Raster time series may become large and
hard to manage investigate the possibility to
use g3d format for storing time series of 2D
raster data. If we use external models (e.g.
CHILD, SIBERIA) they handle the time series using
their own formats, only elevation surfaces at
time intervals needed by other models would be
imported into GIS or other modeling environment.
14
Spatial and temporal scale
Spatial scale data driven - at least 30m
resolution (preferably 10m) is needed for
realistic, spatially distributed erosion
modeling. - spatially averaged modeling can be
performed at subwatershed level at lower
resolutions (100s m) Temporal scale needs to be
taken into account at 2 levels - interaction
with other models this can be years, driven by
time needed for a significant change in
topography to occur to have impact on human
decisions - numerical simulations require much
shorter time interval to ensure stable solution
15
Spatial and temporal scale
It is important to take into account combined
effect of many small events and few large events,
therefore a model could be run using stochastic
events to develop typical year(s) or decades
evolution and then simulate the longer periods
using averaged annual or decadal effects, again
with some stochastic component. If we have
enough computational power and the underlying
geomorphological model is simple enough, long
periods of time could be simulated using
stochastic events.
16
Interface with models of human decisions
Outputs of surface process modeling useful for
modeling human decisions - basic terrain
parameters, such as slope and aspect, - soil
loss and sediment deposition - flooding - other
terrain parameters solar radiation, wetness
index, ... Feedback into surface process
modeling - changes in land cover and soil
properties - terrain modification. Data
exchange raster format.
17
Impact of change in land use pattern
Original LU
sediment flow
net erosion/deposition Optimized
LU
18
Modeling impact of hedges
Seth M. Dabney et al. 1999 Lanscape Benching from
Tillage Erosion Between Grass Hedges
19
Evolution of hedge slope without tillage
Evolution of erosion and deposition pattern
Seth M. Dabney et al. 1999 Lanscape Benching from
Tillage Erosion Between Grass Hedges
20
Impact of clearing on water flow
current 49 forest
construction 24 forest
0.7m3/s
0.002m3/s
discharge m3/s
Spatial distribution of overland water discharge
for 1hr steady rainfall at 46mm/hr (2 year
design storm)
21
Net erosion/deposition
0.01-6.0kg/ms erosion968kg/s
0.001kg/ms erosion87kg/s
/m2s
erosion deposition
erosion deposition
main impact of disturbance is outside
construction site stream erosion within
protective buffer
22
Path sampling method
- based on duality between particle and field
representation - path
samples represent water or sediment evolving
according to the shallow-water
bivariate continuity equation
- solved by
operator inversion Green's function
representing short time propagation of the
sample points (drift diffusion)
Land cover red - disturbed areas green - forest
Water depth particles continuous
fieldparticle density
Rainfall excess, land cover and topography are
spatially variable
23
Sediment flow for different soils
sand ?1
clay ?0.1
detachment capacity limited case DCL
transport capacity limited case TCL
24
Multiscale simulations
Simulation of overland water flow 10m
resolution combined with 2m resolution
25
Conclusions
- landscape evolution can be best modeled by
spatially distributed, process-based models
integrated or coupled with GIS - hillslope
erosion and simple channel transport can be
modeled using GRASS modules - external models
need to be explored for fluvial transport -
CHILD most complete landscape evolution model?
- flood data and estimates of erosion and
sedimentation rates are needed to develop the
models - estimates of potential elevation
surfaces, climate, soil properties and land use /
land cover needed for inputs
H. Mitasova
26
Tangible GIS
GIS excellent for analysis, and modeling but
design of alternative scenarios is often a
tedious task. New type of tangible
communication environment between a computer
model and users can improve understanding
of landscape response to human induced changes
and has a potential to lead to better land
management solutions.
Collaborative with C. Ratti MIT Media lab and
SENSEABLE City Lab
27
Water flow modeling
Shallow overland flow - St. Venant equation
dynamic kinematic wave
steady state approx. diffusive wave
h - water depth, ie - rainfall excess, v -
velocity given by Mannings eq., ? - diffusion
const
kinematic wave
kinematic wave with predefined channel in the
depression
diffusive wave
H. Mitasova
28
Sediment transport
Sediment transport and net erosion/deposition
model is based on 2D generalization of a 1D
hillslope erosion model used in WEPP.
where D (r) ?(r)T (r)-qs(r) is
sources-sinks term and D (r)/Dc(r)
qs(r)/T (r)1 (Foster and Meyer, 1972)
c - sediment concentration, qs - sediment flow,
T - transport capacity, Dc- detachment
capacity, ? first order reaction coef.
Net erosion and deposition is computed as a
divergence of sediment flow.