Strong correspondence

1
Satellites, Swamp Gas, and Siberia Using Models
and Remote Sensing to Estimate Greenhouse Gas
Emissions from Siberian Bogs Theodore J. Bohn and
Dennis P. Lettenmaier, University of Washington,
Seattle, Washington, USA UW Water Center Annual
Review of Research, Seattle, WA, USA, Feb 14, 2008
Background
Study Site Bakchar Bog

• Wetlands are largest natural source of methane
• Methane emissions depend on climate
• 30 of worlds wetlands are in Northern Eurasia
• High latitudes experiencing pronounced climate
  change
• How will wetland methane emisions respond across
  N. Eurasia?
• To answer this question, we will use a
  combination of modeling and remote sensing.

Idealized Bog Response to Climate Factors
100 X 100 km grid cell

• Higher Soil Temperature
• (via metabolic rates)
• Increases production of both CO2 and CH4 per unit
  volume

CO2
CH4
CO2
Strong correspondence between wetlands and high
topographic index
NPP
Living Biomass

• Higher Soil Temperature
• (via evapotanspiration)
• Lowers water table
• Shifts partitioning of Rh towards CO2

Organic carbon in soil
Aerobic Rh
Water Table
Map of Methane Emissions
Calibration at Site
Spatial Distribution

• Higher Precipitation
• Raises water table
• Shifts partitioning of Rh towards CH4

Partitioning of respiration (Rh) into CO2 and CH4
depends on water table position
Anaerobic Rh
The overall distribution of annual average
methane emissions is similar to the distribution
of wetlands. Exceptions include stream channels
(which can be masked out) and noise from the
DEMs low vertical resolution of 1m (which can be
addressed by smoothing).

• Increasing Temperature
• Shrinks saturated area
• Increases CH4 emissions from saturated area only
• Total emissions decrease

• These relationships are non-linear
• Water table depth not uniform across landscape
• We need to take account of heterogeneity of water
  table depth

• Increasing Precipitation
• Increases saturated area
• Total emissions increase

Modeling Framework
To cover all of N. Eurasia, we need to use a
large-scale hydrological model.

• Increasing Both T and P
• Maintains saturated area
• Increases CH4 emissions from saturated area
• Total emissions increase

• Large-scale Hydrological Model (VIC)
• Model landscape as large, flat grid cells (e.g.
  100 km)
• On sub-daily time step, simulate
• Moisture and energy fluxes through land surface
• Soil Temperature
• Net Primary Productivity (NPP)
• Water table depth
• Note these quantities are AVERAGE values over
  the grid cell

The hydrologic model was calibrated to match
observed water table depth in the pixel
containing the site. The methane emissions model
was then calibrated to match observed methane
fluxes at the site.
Wettest
Driest
How to simulate DISTRIBUTION of water table depth
within grid cell?
Conclusions
Comparison with Remote Sensing
Use Wetness Index concept (from TOPMODEL)
Beven and Kirkby, 1979
Start with DEM (e.g. SRTM30)

• Relate local water table depth Zwti to
• Mean water table depth Zwtmean
• Local slope
• Upslope contributing area
• Basically
• flat areas are wet
• steep areas are dry

In situ observations of water table depth and
methane emissions are too sparse in Siberia to
accurately portray landscape methane emissions
across the landscape. Remote sensing imagery can
help us monitor these vast areas.

• Large-scale hydrological models can reproduce the
  gross spatial distribution of water table depth
  in peatlands
• Temperature and precipitation exert opposing
  influences on peatland methane emissions, through
  their actions on the water table
• Trade-off ?T of 3 C for ?P of 5
• Lateral extent of saturated soil plays important
  role in total landscape methane emissions
• Under likely end-of-century climate conditions,
  methane emissions from Siberian peatlands could
  double

For each pixel, define topographic wetness index
?i ln(ai/tanßi) ai upslope contributing
area tanßi local slope
Here are preliminary results from comparison of
satellite-derived open water fraction (a) and
simulated water table depth (b). For each 900-m
pixel, a shallower average water table depth
should correlate with a greater open water
fraction. While the model simulates shallow
water tables in pixels that have little or no
open water, the overall distributions of open
water and shallow water table depths bear some
resemblance. We are currently investigating how
to improve the correlation. Simulated methane
emissions are shown in panel (c).
Local water table depth Zwti Zwtmean m(?i-
?mean) m calibration parameter
(a)
(b)
(c)
Flat? high ?i , shallow Zwti
Steep ? low ?i , deep Zwti
This is a gross first-order approximation of
water table dynamics, but it is more feasible
than a high-resolution model.
Process Flow
Future Work
Gridded Meteorological Forcings
Methane Emission Model (Walter and Heimann 2000)
NPP
VIC
CH4(x,y) f(Zwt(x,y), SoilT, NPP)
Soil T profile

• Investigate parameter uncertainty
• Refine techniques of using remote sensing to
  constrain/calibrate models
• Expand to all of Northern Eurasia

Topography(x,y) (SRTM30 DEM)
Zwtmean
Open Water Fraction (from JERS synthetic aperture
radar)
Simulated Water Table Depth (cm)
Simulated CH4 flux (g CH4/m2y)
TOPMODEL relationship
Wetness index ?(x,y) for all grid cells pixels
Zwt(x,y)