© Crown copyright Met Office Extended-range forecasts for onset of the African rainy seasons examples and ideas for future work Michael Vellinga, Richard Graham and Alberto Arribas, Met Office
© Crown copyright Met Office
Extended-range forecasts for onset of the African rainy seasonsexamples and ideas for future work
Michael Vellinga, Richard Graham and Alberto Arribas, Met Office
© Crown copyright Met Office
Outline
1. Motivation
2. Onset forecasts
• Criteria and definitions
• Longrange forecast skill: level and sources
• Example forecasts
3. What next: thoughts and discussion
© Crown copyright Met Office
Motivation
© Crown copyright Met Office
Sahelian rainfall
© Crown copyright Met Office
Motivation
• Rain-fed agriculture in Africa vulnerable to fluctuations in timing of arrival time of rains, as well as rainfall amounts:
a) Long-range decision window:
• choice for slow or fast cropping varieties/crops
• agricultural logistics (field preparation, mobilisation of work force,..)
b) Medium-range to-monthly range decision window:
• optimal sowing time
• A strong information need among regional stake-holders, not well addressed at present (also non-agricultural applications)
© Crown copyright Met Office
Observed onset W Africa (JAS) GPCP mean 1996-2009
Calendar date (month/day)
© Crown copyright Met Office
Observed onset W Africa (JAS) GPCP standard deviation
St Dev (days)
© Crown copyright Met Office
Observed onset (SOND) GPCP standard deviation – Substantial!
St Dev (days)
© Crown copyright Met Office
Onset forecastsDefinitions, skill and examples
© Crown copyright Met Office
Onset definitions:two classes
1. Local criteria:• time of local rainfall threshold exceedance (e.g.
Sivakumar 1988, AGRHYMET)
2. Regional criteria:• Large-scale key changes to regional meteorology
(wind field, convection, OLR, MSE, etc)
• We have developed trial longrange forecasts using such definitions, presented at African RCOFs in 2011/12• Formulated as tercile probabilities (‘earlier than
average’, ‘average’ or ‘later than average’)
© Crown copyright Met Office
Local indicator used in our trial forecasts:• Describe temporal evolution with rainfall
accumulations between 21 May-27 October• Scale by long-term average season total
accumulation. Example: 1 member
Time
Percentage of season total rainfall
onset=20%
Average time of onset
•Heavy line: accumulated precip. from climatology
•Thin line: accumulated precip. for individual year
example: early onset in individual year
© Crown copyright Met Office
Forecast skill for onset
• ROC score for earlier-than-average onset• 25 April hindcasts, i.e. ~2 months’ lead
indicator 1)
© Crown copyright Met Office
Skill scores for onset over W Africa in ENSEMBLES
•Using standard definition for ‘regional’ onset (Fontaine Louvet 2006)
•ENSEMBLES hindcasts (1979-2005) in red circle
From Vellinga et al 2012 Clim Dyn
© Crown copyright Met Office
HadiSST and GPCP-derived onset
Observed teleconnection
Source of predictability: June SST
GloSea4 teleconnection
For
ecas
t sk
ill (
anom
. co
rr.)
For more details see: Vellinga et al 2012 Clim Dyn
© Crown copyright Met Office
Other regions…
GHA (SON)
SE Asia (MJJ)
S Africa (OND)
© Crown copyright Met Office
Seasonal forecasts for onsetExamples from 2011/12
© Crown copyright Met Office
2012 Onset W Africa (JAS)
Glosea4 6 May forecast: probability for early (lower tercile) onset
Observed onset anomaly 2012 (CPC/FEWS)
© Crown copyright Met Office
2011 Short Rains onset GHA (SON)
Observed (CPC/FEWS) anomaly
Glosea4 18 July forecast: probability for early (lower tercile) onset
© Crown copyright Met Office
2012 Short Rains onset GHA (SON)
Observed (CPC/FEWS) onset anomaly (days)
Glosea4 22 July forecast: probability for early onset
© Crown copyright Met Office
Discussion points:Possible interaction with S2S• Strong demand for this type of onset forecast information
• Not only at seasonal lead times, but also monthly
• Could set up multi-model prediction page for onset diagnostics: East/West/southern Africa
• What detail is possible at monthly lead time (range of dates)?
• What is the role of MJO/other tropical systems in delaying/advancing onset? How predictable are these?
• Is there multi-model agreement about years with high vs. low predictability for onset etc?
• Other aspects of intraseasonal variability - how accurately can a dry spell or wet spell be predicted in week 3?