Davos Maposa 2014-IDRC14

289 views
225 views

Published on

5th International Disaster and Risk Conference IDRC 2014 Integrative Risk Management - The role of science, technology & practice 24-28 August 2014 in Davos, Switzerland

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
289
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Davos Maposa 2014-IDRC14

  1. 1. If the data is sufficiently tortured, it will confess. Estimating high quantiles of extreme flood heights in the lower Limpopo River basin of Mozambique using model based Bayesian 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org approach Daniel Maposa, Monash University, South Africa; James Cochran, University of Alabama, USA; ‘Maseka Lesaoana, University of Limpopo, South Africa.
  2. 2. 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Introduction • Floods have recently become a common natural disaster in Southern Africa and other parts of the world. • This paper pays attention to extreme floods in Mozambique; a developing and emerging nation, Limpopo River Basin where the February 2000 floods claimed the lives of more than 700 people and caused economic damages estimated at US$500million. • Limpopo River Basin is characterised by extreme natural hazards. • Recently in 2013 two women gave birth on rooftops in Chokwe district (Jackson, 2013)
  3. 3. 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Introduction • Source: AFP/Ussene Mamudo, January 2013
  4. 4. 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Motivation • A substantial amount of money that has been originally designated for development get diverted to relief and rehabilitation assistance to disaster affected people each year a disaster occurs. • IFRC claims that aid money buys 4 times as much humanitarian impact if used before a disaster than on post-disaster relief operations. • It is hoped that this study will help reduce the associated risk & mitigate the deleterious impacts of these floods on humans and property.
  5. 5. 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Materials and Methods • The data used in this study was obtained from the Mozambique National Directorate of Water (DNA). • Daily flood height data series (in metres) over the period 1951-2010 for the lower Limpopo River at Chokwe hydrometric station was used. The raw data consist of instantaneous daily river flows recorded at least once a day. • Sequential steps were taken to select the highest peak flood in each hydrological year, resulting in a sample of size 60 allowing the use of block maxima approach. • Rainfall cycle: October-April; Dry season, May-September
  6. 6. F D(G )  2  and corresponding normalisation sequences of 3    That is, the M b    k , m m   ( ), as    P G x m 9 (Fisher and Tippett, 1928; Coles, a  12       . 5th International Disaster and Risk Conference IDRC 2014   ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Probability framework of block maxima We consider independent and identically 1 distributed (i.i.d.) random variables (Xi )i1 with common distribution function 0 constants ( 0) m a  and ()nb such that lim ( ) ( ), . m m m m F a b F x x   distribution function F satisfies the extreme value condition with index or equivalently F 4 belongs to the domain of attraction of G 5 (Fisher and Tippett, 1928; Coles, 2001; Dombry, 2013). We divide the sequence of i.i.d. random variables 1 ( )i i X  6 into blocks of length m 1 and we define the th k block maximum by , ( 1) 1 max( ,..., ), 1 k m k m km M X X k   7   (Coles, 2001; Dombry, 2013). For a fixed m 1, the variables , 1 ( ) k m k M  8 are i.i.d. with distribution function m F and 0 m 10 2001; Dombry, 2013; Maposa et al., 2014). The extreme value distribution with index  is 11 given in Coles (2001), Dombry (2013) and Maposa et al. (2014) as     1/ G (x) exp 1 x , , 1 x 0    
  7. 7. Time series plot of annual daily maximum flood heights at Chokwe hydrometric station (1951-2010) 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org
  8. 8. 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Parameter estimation • Table 1: MLE estimates of the GEV parameters Parameter Estimate Standard Error 95% *CI 4.26452 0.25374 (3.7574;4.7723) 1.78893 0.17725 (1.4343;2.1436) -0.08351 0.07273 (-0.2290;0.0620) • Table 2: Bayesian estimates of the GEV parameters Parameter Estimate Standard Error 95% *CI 4.27235 0.00602 (3.7636;4.7872) 1.90141 0.00484 (1.5565;2.4002) -0.06824 0.00184 (-0.2027;0.1046)      
  9. 9. In-sample evaluation of estimated tail quantiles at Return period ML estimate *(Exceedances) Bayesian estimate *(Exceedances) 95th 0.05 20 years 8.97 m (1) 9.38 m (1) 98th 0.02 50 years 10.22 m (1) 10.79 m (1) 99th 0.01 100 years 11.10 m (1) 11.78 m (1) 99.5th 200 years 250 years 11.92 m (1) 12.18 m (1) 12.72 m (1) 13.02 m (0) 99.8th 0.002 500 years 12.94 m (1) 13.90 m (0) 99.9th 0.001 1000 years 13.65 m (0) 14.74 m (0) 99.99th 0.0001 10 000 years 15.76 m (0) 17.27 m (0) 5th International Disaster and Risk Conference IDRC 2014 Quantiles Exceedance probability (p) 99.6th 0.005 0.004 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org different probabilities *Exceedances in brackets represent the number of sample observations above the estimated flood level (quantile).
  10. 10. Return level plot of posterior distribution with 95% Bayesian credible intervals (dashed lines) at Chokwe hydrometric station 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org
  11. 11. Added value for the Post 2015 Framework for 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org Disaster Risk Reduction • How did your work support the implementation of the Hyogo Framework for Action: – Helps reduce the associated risk & mitigate the deleterious impacts of floods on humans and properties e.g. bridges, houses, factories etc. – Contributes towards a reduction in the amount of aid money spent on post-disaster relief operations. – Advancement of research in developing countries: Mozambique warrants attention as one of the least developed flood-prone countries. • From your perspective what are the main gaps, needs and further steps to be addressed in the Post 2015 Framework for Disaster Risk Reduction in – Research: Forecasting and monitoring of natural hazards – Education & Training: More training workshops, seminars and conferences on natural hazards – Implementation & Practice: Recommendations from natural disaster experts should be taken seriously by national and local governments. – Policy: Introduce more Disaster Risk Reduction, Resilience and Management courses in colleges and universities
  12. 12. 5th International Disaster and Risk Conference IDRC 2014 ‘Integrative Risk Management - The role of science, technology & practice‘ • 24-28 August 2014 • Davos • Switzerland www.grforum.org References • Maposa, D., Cochran, J.J. and Lesaoana, M. (2014). Estimating high quantiles of extreme floods in the lower Limpopo River of Mozambique using model based Bayesian approach. Natural Hazards and Earth System Sciences, Discussions, 2:5401-5425. doi:10.5194/nhessd-2-5401-2014. • Maposa, D., Cochran, J.J. and Lesaoana, M. (2014). Investigating the goodness-of-fit of ten candidate distributions and estimating high quantiles of extreme floods in the lower Limpopo River basin, Mozambique. Journal of Statistics and Management Systems (forthcoming). www.tandfonline.com/tsms • Dombry, C. (2013). Maximum likelihood estimators for the extreme value index based on the block maxima method. ArXiv: 1301.5611v1 [math.PR], available at: http://arxiv.org/pdf/1301.5611.pdf (last access: 17 August 2014). • Ferreira, A. and de Haan, L. (2013). On the block maxima method in extreme value theory. ArXiv: 1310.3222v1[math.ST], available at: http://arxiv.org/pdf/1310.3222.pdf (last access: 17 August 2014).

×