Multimodel Ensemble Reconstruction of Drought over the Continental U.S

1
Multimodel Ensemble Reconstruction of Drought
over the Continental U.S

• Aihui Wang1, Dennis P. Lettenmaier1,
• Sarith Mahanama2, and Randal D. Koster2
• for presentation at
• Climate Diagnostics and Prediction Workshop
• Tallahassee, FL
• October 24 2007
• 1Department of Civil and Environmental
  Engineering,
• University of Washington, Seattle, WA 98195
• 2NASA Goddard Space Flight Center, Greenbelt MD
  20771

2
Outline

• Motivation
• Models
• Methodology
• Results
• Summary

3
Motivation

• Widely used, but link to direct observations
  (e.g., of soil moisture) is weak hence reliance
  on indirect methods, such as PDSI.
• Need for reproducible basis for identifying
  drought-affected regions.
• Land surface model representations of soil
  moisture (and runoff) offer an alternative means
  for estimating severity, frequency, duration, and
  variability of current droughts, and linking them
  to the climatology of observed droughts.

4
Models

• VIC Variable Infiltration Capacity Model
• (Liang et al. 1994)
• CLM3.5 Community Land Model version 3.5
• (Oleson et al. 2007)
• NOAH LSM NCEP, OSU, Air Force, Hydrol. research
  lab
• (Mitchell et al. 1994, Chen
  and Mitchell 1996)
• Catchment LSM NASA Seasonal-to-Interannual
  Prediction Project (NSIPP) LSM
• (Koster et
  al. 2000 Ducharne et al. 2000)

5
Model Schematic
6
Data

• Daily precipitation and temperature max-min,
  other land surface variables (downward solar and
  longwave radiation, near-surface humidity, and
  wind) derived via index methods. Methods as
  described in Maurer et al. (2002). Data duration
  is from 1915-2003, and period of analysis is
  1920-2003 . Spatial resolution 0.5 ? (3322 land
  grid cells), domain conterminous United States.
• Soil and vegetation parameters are from
  different sources for different models (generally
  NLDAS), as provided by model developers. Other
  parameters are model standard setup.

7
The challenge Different land schemes have
different soil moisture dynamics
Model simulated soil moisture at cell (40.25?N,
112.25?W)
8
Solution Normalized total column soil moisture

• For each model, total column soil moisture was
  expressed as percentiles (hence by construct,
  uniformly distributed from zero to one).
• Percentiles were estimated for each model by
  month, using simulated total column soil moisture
  for the period 1920-2003.
• Percentiles were computed using the Weibull
  plotting position formula.

9
Ensemble methods

• Two ensemble methods were used
• Ensemble-1 averaged 4 modeled soil moisture
  percentiles of each grid cell on monthly
  scale.
• Ensemble 2 first, normalized column total soil
  moisture
• modeled by 4 models individually second,
  averaged those normalized soil
  moisture of each grid cell in 4
  models third, calculated
  percentiles of those averaged values .

10
Areas for spatially averaged soil moisture
percentiles
Box sizes are 5 x 5 degrees
11
NW
12
NE
13
SW
14
SE
15
Multimodel comparison soil moisture as
percentiles
July 1934
VIC
CLM3.5
NOAH
Catchment
Ensemble-2
Ensemble-1
16
November 1952
VIC
CLM3.5
NOAH
Catchment
Ensemble-1
Ensemble-2
17
October 1963
CLM3.5
VIC
NOAH
Catchment
Ensemble-1
Ensemble-2
18
February 1977
CLM3.5
VIC
NOAH
Catchment
Ensemble-1
Ensemble-2
19
June 1988
CLM3.5
VIC
NOAH
Catchment
Ensemble-2
Ensemble-1
20
June 2002
CLM3.5
VIC
NOAH
Catchment
Ensemble-2
Ensemble-1
21
Severity-area-duration for the 1930s drought(s)
22
Conclusions

• Current drought products suffer from the absence
  of reproducible, objective methods for
  identifying drought extent and severity.
• Although widely used, PDSI has well-known
  shortcomings, especially the absence of a strong
  link to physical processes
• Land surface parameterizations, such as the
  family of NLDAS models, avoid these shortcomings.
  However, soil moisture, a key drought-related
  variable, is model-dependent
• Multi-model estimates of soil moisture,
  appropriately normalized, address all of the
  above shortcomings. When forced with common
  observations, major drought events appear to be
  plausibly reproduced by the individual models,
  and two methods of combining results into a
  multi-model ensemble.