2. Integración de sistemas de
Observación y Modelación
para predicir desastres
“naturales”
y diseñar estrategias de
Adaptación al cambio
Ana P. Barros
Duke University
Lima, March 21, 2013
3. gestión sostenible del Agua
Desastres Naturales
Cuando
Cuanta Lluvia
Donde
Como
Variabilidad
Incertidumbre
11. FlashFlood QPE Operational Demo
Tropical Storm Fay
QPE and QPF [NFDB]
Challenges
Model Skill
Lack of Data
Data Latency
12. #2 – Nonlinearity – SE USA
Precipitation
Moisture Convergence
Li, Li and Barros, Climate Dynamics 2013
13. Predición de Sequía a Término Longo
• Measure of Drought
Standardized Precipitation Index
• Data
Physics
Statistical Relevance
• Demonstration
Data Independence, Length of Record
Non-stationarity
Material described in Barros and Bowden, 2008, J. Hydrology
15. Goal
12 month lead-time areal mean SPI12
Data Driven ANN Model
F1 F2
1 2 3 4 5 6 7 8 9 10 11 12 ………..24
***Length of Record
***Non-stationarity
16. Basin Average
ANN OUTPUTS
RMSE/R RMSE/R
12-month
Calibration Validation
SPI PC
Set Set
4.86 4.65
1
0.76 0.74
1.57 4.63
2
0.60 0.05
1.22 1.13
3
0.77 0.37
18. Adapatación
( la base científica)
Sistemas de Observación
Integración: COMO Funciona el Paisaje?
Evaluación dinamica de recursos
Gestión de riesgos
21. Desastres Naturales Cambios Antropogénicos
Monitoreo de Riesgo
Gestión
Variabilidad Climática de Hoy Análisis de Riesgo
Y Predicción de Eventos
Cambios Climáticos
Projectiones de Cambio Climático Adaptación
Adaptación ≡ Preparación
22. Processos de Formación de Lluvia
Clausius-Clapeyron
Variabilidad Interanual
vapor pressure saturation vapor pressure Year
September Rainfall Totals by
Appalachian Mountains
350.0
Clouds Large-Scale Moisture Convergence
300.0
Microphysics Heating
(2) Local
250.0 Evapotranspiration / CCN
(3) e
Rainfall Total (mm)
200.0 ~
Td Temperature na
La Ni
(1) T AMO +
150.0
Cooling / Lifting
Fog, Snow & Rainfall 100.0
50.0
0.0
2008 2009
Year
25. Ciclos Climáticos…
Dos series con Nilometer 622-1284 A.D.
media y variancia H=0.91
estadísticas iguales
The idea that
persistent
(0.5<H<1.0)
H=0.5 “Ruído Blanco”
movements in a
time series tend to
be part of larger
trends and cycles
more often than they
are completely
random.
From Koutsoyannis, 2004
(Mandelbrot and Wallis 1977, Hurst 1951, Barros and Evans 1996)
Notes de l'éditeur
Homogenous correlation maps for the first three VARIMAX rotated principal components of the 12-month standardized precipitation index (SPI) for the study region (1973-2002).
12-month lead-time self-organizing linear output (SOLO) forecasts of the first three VARIMAX rotated 12-month SPI principal component scores (Jul-1981 to Dec-2002). 95% confidence bounds derived from bootstrap resampling are also included for the SPI12 forecasts. Independent validation forecasts are shown for the period Jul-1990 to Jun-1993.
The spatial distribution in the study region of (a) RMSE, (b) the number of precipitation gauges, and (c) the standard deviation of precipitation (mm).