Problems in diagnosing Precipitation Trends in South America

1
Problems in diagnosing Precipitation Trends in
South America
Brant Liebmann Vicente Barros Carolina S.
Vera Julián Báez Leila M.V. Carvalho
Anji Seth Inés Camilloni Gil
Compo Marty P. Hoerling José A. Marengo
Prashant Sardeshmukh Dave Allured Mario Bidegain

2
Data We wish to thank the following agencies for
providing the data used in this study Agência
Nacional de Águas (Brasil) Agência Nacional
Energia Elétrica (Brasil) U.T.E. Uruguay C.T.M.
Salto Grande Servicio Meteorologico Nacional
(Argentina, Paraguay, Uruguay) FUNCEME (Ceará,
Brasil) IAPAR - SIMEPAR (Paraná, Brasil) DAEE
(São Paulo, Brasil) Minesterio del Ambiente y los
Recursos Naturales (Venezuela) Meteorogische
Dienst Suriname METEO-France ASANA (Bolivia)
3
Observed Precipitation Trend
mm/season
4
Statistical Relevance of Observed Trends
5
SACZ
Southern Brazil
1948-1999 Southern Brazil (2.5 mm/yr 21
variance explained by trend) SACZ (0.1 mm/yr 0
variance explained by trend)
6
Tendencies for 24 year moving segments Stations
with at least 50 years of data
7
(No Transcript)
8
Correlation of JFM precipitation with JFM
with Missing days (noted on abcissa) 27 seasons
at 49W, 15.3S
Correlation between complete and incomplete
record
Days missing from 90 day record
9
Trends calculated from missing seasons
Trend per 24 years
Trend explains 18 of variance over 24 years
10
Station pair correlation with distance
Each pair correlation
average
Daily, including annual cycle
11
Stations with fewer than 22 years
Stations with at least 22 years
(larger dots)
12
Stations with fewer than 22 years
Stations with at least 22 years data
19 seasons required
24 seasons required
13
Previous work Barros, Castaneda, Doyle (2000)
Increase in Annual total precipitation over most
of Argentina from 1956-1991. Castaneda and
Barros (1994) Humid Pampa (Argentina) increase
in rain, mainly after 1960. Robertson and
Mechoso (1998) Decadal variability and
non-linear upward trend in southeast South
America rivers. Marengo (2004) Northern Amazon
Basin decrease related to El Nino?
14
Observed Precipitation Trend
mm/season
15
Argentina 36 year segments
At least 34 years of 328 days per year
16
Argentina 24 year segments
At least 22 years of 328 days per year
17
Argentina 15 year segments
At least 14 years of 328 days per year
18
Observations versus AMIP runs January March
1976-1999
mm/season
19
From Compo and Sardeshmukh (2004)
20
Observations University of Delaware
monthly Models ARPEGE 8 runs CAM2 - 15
runs NSIPP - (14 9) 23 runs ECHAM4 -
24 runs ECHAM3 - 10 runs
21
mm/season
22
Observed Trend
mm/season
23
mm/season
24
mm/season
25
Running trends 24 year segments July
September Central Africa index
Observations
5 model average
26
Running trends 15 year segments July
September Central Africa index
Observations
5 model average
27
Central African Rainfall Anomaly
July - September
5 model average
observed
28
Complete records versus dense coverage Problems
if values are missing
All these problems are relatively minor Compared
to choosing appropriate period For practical
purposes, what constitutes a trend?
Should I be more careful?