Predictability of the Moisture Regime During the Pre-onset Period of Sahelian Rains

1
Predictability of the Moisture Regime During the
Pre-onset Period of Sahelian Rains

• Robert J. Mera
• Marine, Earth and Atmospheric Sciences
• North Carolina State University
• Seminar, April 3rd 2009

2
Motivation

• Why is the moisture regime important?

Prediction of Monsoon rainfall
African Easterly Waves
Agriculture
Public health Meningitis Outbreaks
3
Outline

• The Application
• Background
• Health-climate link
• Our Study
• Importance of Downscaling
• Predictability of Pre-onset Conditions
• Ensemble Prediction and Evaluation of Model Skill

4
The Application

• Meningitis is a serious infectious disease
  affecting 21 countries
• 300 million people at risk across the Sahel
• 700,000 cases in the past 10 years
• 10-50 fatality rate
• 256,000 people lost to the disease in 1996

SAHEL
5
Meningitis-Climate link

• Outbreaks coincide with dry, dusty conditions
  over the Sahel due to the Harmattan winds flowing
  south from the Sahara (Jan-May)
• Largest correlation occurs between low humidity
  and disease outbreaks (Molesworth et al., 2006)
• Disease occurrence drops dramatically with the
  onset of humidity

SH
SHL
Harmattan
ITCZ
Moisture
ITCZ
January
July
6
Meningitis-Climate link

• The most actionable case involves the link
  between humidity onset and cessation of disease

1998
2004
Pink of cases
Orange Relative Humidity ()
7
Current Efforts

• University Corporation for Atmospheric Research
  (UCAR) and the Google Foundation are funding
  efforts to explore climate-meningitis dynamics
• Global scale models will be employed for
  operational purposes

8
Our study Importance of Downscaling
65
60
55
50
45
40
35
Ghana
30
Ghana
25
20
WRF at 30km resolution
NCEP/NCAR Reanalysis at 2.5
Relative Humidity ()
9
The Scientific Question Predictability of
Moisture

• What are the dynamics governing the northward
  progression of the moisture regime?
• How well does the model represent the physical
  processes?
• What is the skill of the model in predicting the
  dynamics and statistics of the physical processes?

10
In the literature

• The West Africa summer monsoon is characterized
  by two steps preonset and onset (Sultan and
  Janicot, 2003)
• The preonset stage corresponds to the arrival of
  the Inter Tropical Front (ITF) at 15N

ITF
Rain (mm/day)
From Sultan and Janicot (2003)
11
Schematic Cross Section of the West African
Monsoon
200 hPa
Deep moist convection
AEJ
600 hPa
Deep dry convection
1000 hPa
10 N
20 N
Equator
Sahel
ITCZ
Sahara
Slide from John Marsham, U. of Leeds
12
Our Study

• The northward progression of moisture is related
  to the preonset stage of the monsoon and the
  position of the ITF
• Two important factors at work
• Interannual variability is dictated by fluxes in
  sea surface temperatures (SST), interaction with
  mid-latitude systems (teleconnections)
• Intraseasonal variability is related to east-west
  transient disturbances, African Easterly Jet

13
Data and Methods

• NCEP/NCAR, ECMWF Reanalysis, In-situ observations
  satellite data Statistics of Relative
  Humidity, etc
• We use the Advanced Research WRF (WRF-ARW) Model
  for downscaling of reanalysis and operational
  forecasts, sensitivity analyses

NCEP National Centers for Environmental
Prediction NCAR National Center for Atmospheric
Research WRF Weather Research and Forecasting
Model ECMWF European Centre for Medium-Range
Weather Forecasts
14
Preliminary analysis and results
15
Historical Data Reanalysis
JUN 24
JUN 14
APR 15
Two distinct slopes

• Mean 20002008 relative humidity time series ()
  computed on the grid points located between 10W
  and 10E longitude, 14.5N and 15.5N latitude

16
Model simulations
April 1, 2006 relative humidity () at the
surface, 925mb winds and u component at 0 to
delineate ITF
700 mb
AEJ
Cross section along the prime meridian from 0 to
20 N Relative humidity (shaded) and u
component at 0
20N
EQ
17
Ensemble Prediction

• We will use the ensemble prediction approach to
  generate probabilistic forecasts that will also
  allow us to analyze model skill

18
An ensemble forecast run was tested against
interpolated observations
Interpolated Observations
Ensemble Simulation
19
An ensemble forecast run was tested against
interpolated observations
-10 -8 -6 -4 -2 2 4 6 8 10
Relative Humidity Anomaly ()
The error (anomaly) is much smaller than the
signal
20
Analyzing Model Skill
Observations
No Yes
No No cost (?) Miss (?)
Yes False Alarm (?) Hit (?)
EPS Forecast
21
The Relative Operating Characteristic (ROC)

• The ROC method is widely used for estimating the
  skill of ensemble prediction systems (EPS)
  (Marzban, 2004)
• A perfect forecast system would have a ROC area
  (ROCA) of 1

22
An Extended ROC Procedure

• ROC plots model skill only for an optimum user
• We developed an extended (EROC) procedure that
  caters to a particular users needs

Shift in baselines According to user
Semazzi Mera, 2006
23
Model Skill for End-user

• Additional analysis through EROC can help with
  current health efforts and the incurred costs
• Transportation of Supplies
• Inoculation
• Personnel

24
Looking Forward

• Understanding the moisture regime statistics
  variance of 40 RH date and changes in slope of
  humidity trends
• Sensitivity studies using SSTs, land cover,
  meridional transient distrubances,
  teleconnections with mid-latitude systems
• Application of EROC for surface conditions
  pertinent to health efforts

25
Acknowledgements

• Dr Semazzi
• CML crew
• Google/UCAR group
• NOAA ISET
• Dr Arlene Laing, Dr Tom Hopson

26
Questions?
27
Auxiliary slides
28
Historical Data Reanalysis

• Mean 20002008 relative humidity time series ()
  computed on the grid points located between 10W
  and 10E longitude, 14.5N and 15.5N latitude

29
Large scale Climatology
30
Large Scale Climatology
31
(No Transcript)
32
Criteria for Issuing a forecast

• Decision to issue a forecast of an event (E) to
  occur is probabilistically based on the criteria

Where (N) size of the ensemble (n) number of
the runs in the ensemble for which (E) actually
occurs (p) probability given by the ratio
(n/N) This is the threshold fraction above
which the event (E) is predicted to occur based
on the model forecast
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)