Forecast Sensitivity Tests with New Observations

1
Forecast Sensitivity Tests with New Observations

• Mario Majcen and Jan Paegle
• University of Utah
• Salt Lake City
• Utah

2
Outline

• Background
• Motivation
• Goals
• Tools
• Results
• Summary and conclusions
• Questions

3
Background

• Mesoscale convective system observed on January
  17th 2003

January 17 at 1745 UTC
January 17 at 2345 UTC
January 18 at 0545 UTC
January 18 at 1745 UTC
4
Motivation

• Model forecasts intercomparison by Dr. Saulo
  (University of Buenos Aires)
• Models
• Global and regional model at CPTEC/INPE (Brazil)
• ETA model at CPTEC/INPE (Brazil)
• ETA model at the University of Maryland (USA)
• RAMS model at University of Sao Paolo (Brazil)
• RAMS model at University of Buenos Aires
  (Argentina)
• MM5 model at University of Chile (Chile)
• MM5 model at CIMA (Argentina)
• All models were initialized with the GDAS
  analysis at 00 UTC January 17.

5
Motivation

• Mean forecasted precipitation accumulation

6
Motivation

• Dispersion between forecasts

7
Motivation

• Forecasted and observed wind profiles (at
  Mariscal Estigarribia)

8
Outline

• Background
• Motivation
• Goals
• Tools
• Results
• Summary and conclusions
• Questions

9
Goals

• Goals of this study
• Examine if a new approach with a global variable
  resolution Euler model could predict this event
  better
• Examine forecast sensitivity to analysis
• Examine forecast sensitivity to inclusion of
  special observations
• Compare forecasts and analyses to the radiosonde
  observations taken at Resistencia, Santiago del
  Estero and Santa Cruz

10
Tools

• Characteristics of models and analyses used in
  previous studies
• Most of the models used in previous studies have
  primitive equations dynamical core
• All but one are limited area models (lateral
  boundary conditions may not represent well
  overturning divergent circulation associated with
  strong convection and precipitation, J. Paegle
  1978,Wang et al 1999)
• A global (CPTEC) model had coarse resolution
• All models were initialized with the GDAS analysis

11
Tools

• Characteristics of the model and analyses used in
  this study
• Euler equations dynamical core (retains rapidly
  propagating sound and gravity waves)
• A global model (no lateral boundary conditions
  imposed)
• Better resolution in the area of interest
• Model was initialized with the GDAS analysis, the
  CPTEC analysis (with and without special
  observations)

12
Tools

• Uniform resolution grid in Euler model

13
Tools

• Mathematical pole rotation and variable
  resolution in Euler model

14
Tools

• Increased resolution over the area of interest in
  Euler model

15
Outline

• Background
• Motivation
• Goals
• Tools
• Results
• Summary and conclusions
• Questions

16
Results

• Case description GDAS analysis 500 mb wind and
  geopotential height

• 00 UTC 01/17
  18 UTC 01/17.

17
Results

• Case description GDAS analysis 850 mb wind and
  temperature

• 00 UTC 01/17
  18 UTC 01/17

18
Results

• Case description 850 mb wind and mixing ratio

00 UTC 01/17
18 UTC 01/17
19
Results

• Case description Convective Available Potential
  Energy (CAPE)

00 UTC 01/17
18 UTC 01/17
20
Results

• Experiments performed in this study

21
Results

• Upper air observation sites

22
Results

• Experiment 1 500mb height anomaly correlation
  coefficients black global, blue 90S-20S, red
  20N-90N

20 N 90 N
global
90 S 20 S
23
Results

• One-day precipitation accumulation valid at 00
  UTC, January 18

Experiment 1
Observed
24
Results

• One-day precipitation accumulation valid at 00
  UTC, January 19

Experiment 1
Observed
25
Results

• Noise/signal ratio of RMS difference between
  Experiment 1 and the GDAS analysis wind.

850 mb
500 mb
26
Results

• Vertical profiles of wind at Santa Cruz 06 UTC
  January 18. Zonal wind (squares) and meridional
  wind (circles), solid lines denote observations.

GDAS analysis
Experiment 1
Obs. u
Obs. u
Obs. v
Obs. v
GDAS
FCST
27
Results

• Difference between Experiment 2 and Experiment 1
  initialization at 00 UTC January 17.

• 500mb wind and geopotential height
  850mb wind and temperature

28
Results

• Comparison of Exp. 1 and Exp. 2 500mb height
  anomaly correlation coefficients black- global,
  blue- 90S-20S, red- 20N-90N

Experiment 1
Experiment 2
29
Results

• One-day precipitation accumulation valid at 00
  UTC, January 18

Experiment 2
Observed
30
Results

• One-day precipitation accumulation valid at 00
  UTC, January 19

Experiment 2
Observed
31
Results

• One-day precipitation accumulation valid at 00
  UTC, January 18

Experiment 1
Experiment 2
32
Results

• One-day precipitation accumulation valid at 00
  UTC, January 19

Experiment 1
Experiment 2
33
Results

• Noise/signal ratio of RMS difference between
  Experiment 2 and Experiment 1 wind at 500mb
  (circles) and 850 mb (squares).

850 mb
500 mb
34
Results

• Difference between Experiment 2 and Experiment 3
  initialization at 00 UTC January 17

• 500mb wind and geopotential height
  850mb wind and temperature

35
Results

• One-day precipitation accumulation valid at 00
  UTC, January 18

Experiment 2
Experiment 3
36
Results

• One-day precipitation accumulation valid at 00
  UTC, January 19

Experiment 2
Experiment 3
37
Results

• Noise/signal ratio of RMS difference between
  Experiment 2 and Experiment 3 wind at 500mb
  (circles) and 850 mb (squares).

850 mb
500 mb
38
Results

• Vertical profile of temperature at Resistencia at
  06 UTC January 17 observed (solid, circles),
  GDAS analysis (dashed, asterisk), CPTEC analysis
  (dashed, squares), CPTEC analysis with SALLJEX
  (dashed, circles).

39
Results

• Vertical profile of dew point temperature at
  Resistencia at 06 UTC January 17 observed
  (solid, circles), GDAS analysis (dashed,
  asterisk), CPTEC analysis (dashed, squares),
  CPTEC analysis with SALLJEX (dashed, circles).

40
Results

• Vertical profile of wind at Resistencia at 06 UTC
  January 17 observed (solid, circles), GDAS
  analysis (dashed, asterisk), CPTEC analysis
  (dashed, squares), CPTEC analysis with SALLJEX
  (dashed, circles).

41
Outline

• Background
• Motivation
• Goals
• Tools
• Results
• Summary and conclusions
• Questions

42
Summary and conclusions

• The model retains a good predictability in the
  first 48 hours (500mb height anomaly correlation
  coefficients greater than 0.8)
• The 500mb height anomaly correlation coefficients
  are greater than 0.6 until about forecast hour
  100.
• Precipitation is not forecasted well.
• Sensitivity to analysis is evident in
  precipitation prediction.
• Sensitivity to additional observations is small.
  This could reflect limited impact of local
  observations, or the possibility that the
  observations were poorly assimilated.
• Additional observations have a small impact on
  these analyses at this time.
• None of the three studied analyses represent the
  atmosphere well at the three locations
  (Resistencia, Santiago del Estero, Santa Cruz)
  where special observations were taken during this
  event.
• More research is needed regarding the inclusion
  of special observations into objective analyses
  to make certain that good observations are not
  discarded.

43
Future Work

• Repeat the experiments using forthcoming GDAS
  analyses that are being performed at NCEP for the
  SALLJEX period.
• Study other cases.
• Check impact of model treatments of precipitation
  and turbulent diffusion.
• Check impact of global-scale uncertainty.

44
Questions?