Evaluation and Improvement of a MacroScale Land Surface Hydrology Model for a Stream Flow Trend Attr

1

• Evaluation and Improvement of a Macro-Scale Land
  Surface Hydrology Model for a Stream Flow Trend
  Attribution Study in the Northern High Latitudes
• Jennifer C. Adam1, Fengge Su1, Laura C. Bowling2,
  and Dennis P. Lettenmaier1
• Department of Civil and Environmental
  Engineering, Box 352700, University of
  Washington, Seattle, WA 98195
• 2. Department of Agronomy, Purdue University,
  West Lafayette, IN 47907
• FWI All-Hands Meeting, Estes Park, Colorado, May
  31 June 2, 2006

Photo http//www.globalcarbonproject.org/
Improvement of Frozen Soil Simulations
Evaluation of Simulated Streamflow Trends
ABSTRACT Combined annual streamflow volumes from
the six largest Eurasian basins have increased by
approximately 7 between 1936 and 1999. The
export of freshwater to the Arctic Ocean plays a
key role in both regional and global climates
(e.g. via effects on the strength of the North
Atlantic Deep Water (NADW) formation that drives
the thermohaline circulation), therefore it is
important to understand how river discharge to
the Arctic Ocean will respond to predicted
climatic changes. Incorporation and validation of
the key physical processes into the modeling
framework is crucial for the accurate prediction
of future river discharge rates into the Arctic
Ocean (and feedbacks to the climate system)
because these processes respond differently to
climatic changes. For example, increased
precipitation due to an accelerated hydrologic
cycle will likely continue to provide additional
freshwater to the system, whereas warming-induced
melting of permafrost provides freshwater only
until the excess ground ice in Arctic permafrost
is melted. We report a 70-year (1930-2000) run of
the Variable Infiltration Capacity (VIC)
macroscale hydrology model over the Eurasian
Arctic land domain. We evaluate the ability of
the model to reproduce the climatologies and
trends of observed and reconstructed streamflow.
Whereas the model reasonably captures streamflow
trends in regions of seasonally frozen soil, the
model underestimates trends in permafrost
regions, and we owe this partially to simulated
soil temperatures that are too cold. We focus
initial improvement of the streamflow simulations
on the parameterization of the finite-difference
frozen soils model. In particular, we use a
zero-flux bottom boundary instead of a constant
temperature bottom boundary, which allows the
model to solve for the bottom boundary
temperature we provide an observation-based
initialization of the bottom boundary
temperature and we apply differing bottom
boundary depths and node distributions.

• Cherkauer et al. (1999) finite difference
  algorithm
• solving of thermal fluxes through soil column
• infiltration/runoff response adjusted to account
  for effects of soil ice content
• parameterization for frost spatial distribution
• tracks multiple freeze/thaw layers
• can use either no flux or constant
  temperature bottom boundary

• Development 1 Bottom Boundary Initialization
  (Soil Temperature)
• Calculate long-term annual average soil
  temperature at 3.2m depth (near annual damping
  depth) from Frauenfield et al. (2004) station
  data fig. A
• Interpolate to the 100 km EASE grid (with
  lapsing of temperature with elevation) fig. B
• Spin-up model for 60 years with forcings that
  are held at 1930s climatology
• Check for drift

• Spatial Resolution
• 100km by 100km EASE (Brodzik 1997)
• Temporal Resolution
• 3-hourly
• 1930 to 2000
• VIC Features
• Two-layer energy balance snow model
• Frozen soil/permafrost algorithm
• Lakes and wetlands model
• Blowing snow algorithm

A) Frauenfield et al. 2004 station data

• Calibration Su et al. (2005)
• Trends Evaluation (see below)
• permafrost regions underestimated
• seasonally frozen soil captured

• Current Set-Up (Su et al. 2005)
• constant T bottom boundary damping depth of
  4m, Tb defined as annual average air temperature,
  15 nodes utilized
• spatial frost turned on

B) Interpolated station data
Yenisey
Ob
Lena

• Development 2 NOFLUX Bottom Boundary
• Use NOFLUX bottom boundary (fig. B) versus
  constant temperature bottom boundary (fig. A).
  Model solves for Tb.
• Use boundary depth equal to 3 x annual damping
  depth (15m) and initialize at more realistic soil
  temperature (-3 C) fig C.

Constant T BB Dp 4m Tb_init -12 C
NOFLUX BB Dp 4m Tb_init -12 C
NOFLUX BB Dp 15m Tb_init -3 C
Stream Flow (103 m3/s)
Yenisey
Ob
Lena
Observed (R-ArcticNet v3.0)
STATEMENT OF PURPOSE To evaluate and improve
the performance of a land surface hydrology model
to predict long-term trends in streamflow from
the Eurasian Arctic land areas.
Study Domain
Month
We focus on Northern Eurasian basins (stream flow
has been shown to be increasing and longer
records exist for these basins). We chose three
primary and nine secondary smaller and sub-basins
with varying extents of permafrost.
400 periods between 1936 and 2000 were tested
for 99 significance using the seasonal
Mann-Kendall test (Hirsch et al. 1982) for gauged
and reconstructed streamflow. Trend slopes were
calculated for the periods for which observed
streamflow trends passed 99 significance
observed (gauged and reconstructed) (left panel)
and simulated (right panel).
0

• Development 3 Node Distribution
• Increase number of nodes (e.g. 18 in this
  example)
• Use exponential node distribution to capture
  greater variability in surface layers (fig. B),
  versus linear node distribution (fig. A).
• For a cell in discontinuous permafrost, linear
  distribution predicts seasonally frozen soil,
  while exponential distribution predicts
  permafrost.

1
Simulated Trends
Observed Trends
2
3
Lena
4
5
6
Lena
Depth, m
A) Linear Node Distribution
7
0
Yenisey
1
2
3
Q-Observed Trend
4
5
Ob
6
B) Exponential Node Distribution
7
1930 1931 1932
1933 1934 1935
1936 1937
Soil Temperature, C
Yenisey

• CONCLUDING REMARKS
• We are able to simulate streamflow climatology
  fairly well for all primary basins, but
  streamflow trend is only reasonably captured for
  the basins outside permafrost regions (e.g. Ob).
  In permafrost regions, whereas observed trends
  show a large spread in trends with period,
  simulated trends are nearly always negligible for
  all periods. This indicates that soil moisture
  dynamics are not simulated reasonably in
  permafrost regions, partially due to simulated
  soil temperatures that are too cold.
• We focus improvements on the frozen soils model,
  adjusting the bottom boundary parameterization,
  initialization, and depth as well as number and
  distribution of nodes. We plan to diagnose
  further problems using observed ground data, such
  as soil temperature and soil moisture.

Ob
1940 1960 1980 2000
1940 1960 1980 2000
G Gauged (R-ArcticNET v3.0)
M Naturalized (McClelland et al. 2004)
Streamflow Trend, mm year-2
Q-Simulated Trend